Are We Slaves to our Genes? 1108426336, 9781108426336

Citation preview

ARE WE SLAVES TO OUR GENES? There is a common misconception that our genomes all unique, except for those in identical twins have the upper hand in controlling our destiny. The latest genetic discoveries, however, do not support that view. Although genetic variation does influence differences in various human behaviours to a greater or lesser degree, most of the time this does not undermine our genuine free will. Genetic determinism comes into play only in various medical conditions, notably some psychiatric syndromes. Denis R. Alexander here demonstrates that we are not slaves to our genes. He shows how a predisposition to behave in certain ways is influenced at a molecular level by particular genes. Yet a far greater influence on our behaviours are our world views that lie beyond science and that have an impact on how we think the latest genetic discoveries should, or should not, be applied. Written in an engaging style, Alexander’s book offers tools for understanding and assessing the latest genetic discoveries critically. denis r. alexander is the Founding Director (Emeritus) of The Faraday Institute for Science and Religion and Emeritus Fellow of St Edmund’s College, Cambridge. The former chair of the Molecular Immunology Programme at The Babraham Institute in Cambridge, he helped to establish the National Unit of Human Genetics at the American University Hospital in Beirut, Lebanon. He is the author of Genes, Determinism, and God.

Published online by Cambridge University Press

Published online by Cambridge University Press

Are We Slaves to Our Genes? denis r. alexander University of Cambridge

Published online by Cambridge University Press

University Printing House, Cambridge c b 2 8bs, United Kingdom One Liberty Plaza, 20th Floor, New York, n y 10006, USA 477 Williamstown Road, Port Melbourne, vi c 3207, Australia 314 321, 3rd Floor, Plot 3, Splendor Forum, Jasola District Centre, New Delhi 110025, India 79 Anson Road, #06 04/06, Singapore 079906 Cambridge University Press is part of the University of Cambridge. It furthers the University’s mission by disseminating knowledge in the pursuit of education, learning, and research at the highest international levels of excellence. www.cambridge.org Information on this title: www.cambridge.org/9781108426336 doi : 10.1017/9781108566520 © Denis R. Alexander 2020 This publication is in copyright. Subject to statutory exception and to the provisions of relevant collective licensing agreements, no reproduction of any part may take place without the written permission of Cambridge University Press. First published 2020 A catalogue record for this publication is available from the British Library. Library of Congress Cataloging in Publication Data n a m es: Alexander, Denis, author. t it l e: Are we slaves to our genes? / Denis R. Alexander. d es c r ip t i o n: Cambridge, United Kingdom ; New York, NY : Cambridge University Press, 2021. | Includes bibliographical references and index. i de nt i f ie rs : l cc n 2020019423 (print) | l ccn 2020019424 (ebook) | i sbn 9781108426336 (hardback) | i sbn 9781108445054 (paperback) | i sbn 9781108566520 (epub) s u b j e c ts : m e s h: Genetics, Behavioral | Genetic Phenomena | Genetic Predisposition to Disease c l a s s i f i c at i o n : l c c q h4 42 (print) | l c c q h4 42 (ebook) | nl m q u 450 | d dc 572.8 dc23 LC record available at https://lccn.loc.gov/2020019423 LC ebook record available at https://lccn.loc.gov/2020019424 isb n 978 1 108 42633 6 Hardback i sbn 978 1 108 44505 4 Paperback Cambridge University Press has no responsibility for the persistence or accuracy of URLs for external or third party internet websites referred to in this publication and does not guarantee that any content on such websites is, or will remain, accurate or appropriate.

Published online by Cambridge University Press

Contents

List of Figures Preface

page vi vii

1. Genetic Confusion

1

2. Genetic Information and How It Flows

17

3. Genes and Environments in Human Development

36

4. What Is Behavioural Genetics?

55

5. Genes and Mental Health

74

6. Genes, Education and Intelligence

94

7. Genes, Personality and Personality Disorders

111

8. Genes, Food, Exercise and Weight

130

9. Genes, Religiosity and Political Commitment

143

10. Gay Genes? Genetics and Sexual Orientation

155

11. Are We Slaves to Our Genes?

178

12. Genes and Human Identity

196

Notes Definitions of Technical Terms References Index

215 221 225 239

v

Published online by Cambridge University Press

Figures

1.1. Pedigree of a Lebanese family showing the inheritance of fructose 1,6-diphosphatase deficiency. From Alexander et al. (1985). page 3 2.1. The DNA double helix. Reproduced by permission of the Templeton Press. 20 2.2. The genetic code. Reproduced from OpenStax, Biology, OpenStax CNX. 22 2.3. Reading DNA. 25 2.4. How to make messenger RNA (mRNA). 25 2.5. Transcription and translation. Reproduced by permission of the Templeton Press. 26 5.1. The role of the UBE gene in autism and Angelman syndrome. Based on data from Yi et al. (2015). 80 7.1. Genetic correlations between 23andMe individuals and psychiatric disorders. From Lo et al. (2017). 116 7.2. A Dutch family in which a mutation in the monoamine oxidase A gene causes a complete lack of the monoamine oxidase protein. From Brunner et al. (1993a). 126 9.1. A comparison between identical and non-identical twins in terms of (9.1a) their frequency of contact and (9.1b) their emotional closeness. Adapted from Figure 1 of Neyer (2002). 147 10.1. The development of same-sex attraction (SSA) in a US demographic sample. Adapted, with permission, from Pew Research Center (2013) A survey of LGBT Americans: Attitudes, Experiences and Values in Changing Times. Washington, DC: Pew Research Center. 158

vi

Published online by Cambridge University Press

Preface

The aim of this book is to investigate whether the field of behavioural genetics provides any threats or challenges to our sense of human freedom. Are we really slaves to our genes? The book started life as the Gifford Lectures given at St Andrews University in 2012, published under the title Genes, Determinism and God (Cambridge University Press) in 2017. Out of that came the request to publish a popular version of similar material more accessible to the general reader without, necessarily, any background in genetics. This is that version. Those who have already read Genes, Determinism and God will recognise some similarities. However, this book contains several altogether new chapters, such as Chapter 1, which provides a basic introduction to DNA and genetics, Chapter 5, which describes the latest findings in the field of genetics and mental health, and Chapter 8, which looks at the role of genetic variation in differences in our physical size and shape. The science has also been updated as hundreds more papers and books continue to be published in this rapidly moving field. References are given to the relevant literature for those who wish to follow up further. The use of some technical language is inevitable when explaining some of the concepts involved, but technical terms are explained when first mentioned and are defined in a separate handy list. A book on popular genetics touches deeply on our own human identity, so the book covers the influence, or claimed influence, of human genetic variation on differences in our mental health, intelligence, educational attainment, personality, size, weight, levels of religious and political commitments, sexual orientation and much else besides, so hopefully there is something here for everybody. And after surveying these various aspects of our well-being – or otherwise – we then tackle the philosophical question as to whether our deep sense of free will is at all threatened by genetic vii

https://doi.org/10.1017/9781108566520.001 Published online by Cambridge University Press

viii

PREFACE

determinism, finally outlining some competing world views that make a huge difference to our concept of human identity in light of the latest findings in behavioural genetics. My thanks are due to the many friends and colleagues who have made helpful suggestions on earlier versions of the present text; in particular, my thanks go to Keith Fox, Julian Rivers and Leland Taylor. As always, the author is fully responsible for any errors that remain. I would also like to thank Chris Akhurst for his subediting and Beatrice Rehl, Caroline Morley and Gayathri Tamilselvan at Cambridge University Press for their helpful support and advice.

https://doi.org/10.1017/9781108566520.001 Published online by Cambridge University Press

1

Genetic Confusion

T

he patient was a tiny baby, just 2 days old, breathing abnormally fast. Eventually things settled down and the baby was sent home. But in the months that followed, her parents kept bringing her back to A&E with severe breathing difficulties and blood that was abnormally acidic. Each time, a tiny sample of blood was taken for tests. What might be going on? The year was 1984 and this was the National Unit of Human Genetics at the American University Hospital in Beirut, Lebanon. I had gone there a few years earlier to set up a new laboratory of biochemical genetics as part of the Unit. This was a country where consanguineous1 marriages between first cousins were common. But this little baby came from non-relatives and there was no clue from the family history as to what might be happening. There can be a myriad of reasons for abnormally acidic blood (‘lactic acidosis’). The first two times the baby was brought to A&E, the enzymes we tested turned out to be normal. This was followed by more hours in the library searching the literature (no online digital resources in those days!). Could it possibly involve a very rare deficiency of fructose 1,6-diphosphatase? This is an enzyme required for breaking down fructose – a sugar found especially in honey and mature fruit – essential for making cellular energy from the fructose. Without it, fructose is converted to lactic acid, so acidifying the blood. We set up the test using leucocytes (white blood cells) from control blood, ready for the next opportunity. Sure enough, the baby was soon back in A&E again and this time we nailed it: the fructose 1,6-diphosphatase levels in the baby’s blood were barely detectable – problem solved, only the thirty-ninth reported case in the world (Alexander et al., 1985). It turned out that, due to her failure to thrive, anxious relatives had been dosing the little girl with honey –

1

https://doi.org/10.1017/9781108566520.002 Published online by Cambridge University Press

2

1

GENETIC CONFUSION

unwittingly nearly killing her in the process. All that needed to happen was for the girl to be placed on a low-fructose diet and all would be well. I have sometimes wondered what happened to that little girl, who must now be a woman in her mid-30s. Did she get married and have a family? Has she kept to her diet and so maintained her good health? Of course, her samples were (quite properly) all anonymous when they arrived in the laboratory, so I will never know. She was from a Sunni Muslim family. Less fortunate was a baby boy aged 18 months who presented at the hospital around the same time with convulsions followed by irreversible coma and death on his sixth day in hospital. He was the first child of consanguineous parents from the Lebanese Druze community. His fructose 1,6-diphosphatase was also deficient (Alexander et al., 1985). Two of his first cousins, also the product of a consanguineous marriage, had previously died at the age of 2 years. Had it been possible to detect their deficiency within the first few months of life, they would all be alive today. Outside the hospital walls, the Lebanese civil war continued to rage and hundreds were dying. But genetically there were winners and losers as well. And in some cases, at least, if only the consequences of the genetic defect could be identified early enough – in this case an absent enzyme – then it meant life rather than death.

1.1 Mendel’s Inheritance To understand how the enzyme deficiency detected in those Lebanese babies is passed on through families, we need to go back to an Augustinian Moravian monk named Gregor Mendel (1822–1884). In the sheltered space of St Thomas’s Abbey in Brno (now in the Czech Republic) where he was friar and abbot, Mendel carried out a painstaking series of breeding experiments in which he bred nearly 30,000 pea plants of carefully selected varieties. Mendel’s experiments revealed several key findings. The varieties of pea plants that he started with bred true for many generations. Today, we would say that they were genetically pure lines. This was an important factor in his success. When Mendel cross-hybridised these different varieties, the traits inherited by the next generation of peas (the ‘hybrids’) were ‘particulate’ – their seeds were either wrinkled or smooth, or the plants were either tall or short. The hybrids showed only one of the two possible character traits present in the parents, inconsistent with the idea of ‘blending inheritance’ in which different traits merged with each other. Mendel also noticed that some traits were ‘dominant’ and some were ‘recessive’. When he crossed the tall pea

https://doi.org/10.1017/9781108566520.002 Published online by Cambridge University Press

1.1

MENDEL’S INHERITANCE

3

plants with the short pea plants, the ratio of tall to short plants after two generations came to approximately 3:1 – tall was a dominant trait and short was a recessive trait. But if he crossed tall with tall, he got only tall plants, and likewise short with short yielded only short plants. Experiments with peas having multiple different characters suggested that each trait (e.g. height, colour, texture) was inherited independently through subsequent generations. Mendel’s ‘particles’ that led to the inheritance of discrete characteristics in his pea plants are what we now call genes, and it is what we now call the ‘Mendelian Laws of Inheritance’ that allow us to understand what was going on in the Lebanese family just described. Figure 1.1 shows the pattern of inheritance of the defective gene in this family. Today, there are around 7,000 known ‘Mendelian’ genetic disorders,2 meaning medical conditions that are caused by a defect in a single gene with a pattern of inheritance like that shown in Figure 1.1. But they are mostly rare, many extremely rare, and taken together represent only a few per cent of the diseases that afflict humanity. In practice, the development of all the major diseases that impact our lives, such as cardiovascular disease, psychiatric disorders and some cancers, is influenced by hundreds of variant genes that operate together to generate higher or lower levels of risk. Each variant gene looked at individually is

Figure 1.1 Pedigree of a Lebanese family showing the inheritance of fructose 1,6 diphosphatase deficiency. Roman numerals I IV refer to generation number. Circles represent females and squares males. A symbol with a thick black vertical line indicates heterozygotes (carriers) and the solid black circle indicates the homozygous (two defective genes) patient with the deficiency. A dot in the middle of a square or circle indicates that this individual was tested and found to have normal levels of fructose 1,6 diphosphatase. A line through a symbol means ‘deceased’. From Alexander et al. (1985).

https://doi.org/10.1017/9781108566520.002 Published online by Cambridge University Press

4

1

GENETIC CONFUSION

inherited just as Mendel described, but in practice they are coordinating together as one big system to bring about various effects in our bodies. The system as a whole is now referred to as our ‘genome’. This term refers to the sum total of all the information encoded in our DNA. We’ll consider how the coding takes place in Chapter 2. But for the moment it’s worth noting that our genomes are like huge recipes and, just as in the recipe for a cake, all the ingredients have to coordinate together to produce the final product. We wouldn’t say that one particular component in the recipe causes the cake to be either wrinkled or smooth, just to pick up on Mendel’s language; we would say that it was the recipe as a whole, together with the particular oven temperature, that was the cause. This illustrates some of the problems that can arise from learning Mendel’s laws in school biology. If we start our education in genetics by thinking that ‘one gene leads to one characteristic’, this can spill over to how we think about genetics in general. Such an idea is reinforced by precisely the kind of medical situation illustrated in Figure 1.1. An error in a single gene causes a potentially lethal disease, which arises from not being able to digest fructose properly. This sounds like one gene leading to one particular fault in the system – which is true, although in fact the disease system in this case is quite complicated with many steps. Here again the cake metaphor might help: if, for example, we mistakenly leave the baking soda out of the recipe, then the post-oven result will be a dense mass with a heavy texture, not a cake. So a single recipe error leads to a complex developmental process with an unfortunate outcome. This happens with genes as well – an error in a single gene can result in a complicated chain of events that leads to a big difference in the eventual outcome. But the key word here is ‘difference’. The single variant gene does not encode the whole characteristic – the final outcome – but it does make a big difference to the outcome. As we’ll see later, genes as ‘difference makers’ is a really important concept when it comes to thinking about the role of different genes in variant human behaviours. The unfortunate fact is that the way biology is taught in schools spills over into the public understanding of genetics, and the idea that one gene causes a particular human characteristic, even in quite a deterministic kind of way, is still all over the place in the media, as the following section illustrates.

1.2 The Media Portrayal of Genetics One common media confusion is the idea that there is a ‘gene for’ some complex human characteristic. There are mean genes, gluttony genes,

https://doi.org/10.1017/9781108566520.002 Published online by Cambridge University Press

1.2

THE MEDIA PORTRAYAL OF GENETICS

5

gangster genes, liberal genes that cause you to read The Guardian and even the whimsical suggestion of a ‘geneticism gene’ that predisposes some people to think that behaviour is caused by genes. Some sample media headlines illustrate the point: ‘Reason to be cheerful: happiness gene is in Britain’s DNA’ (The Times front page3); ‘“Binge-drinking gene” discovered’ (BBC News4); ‘Study links spread of religion with “believer gene”’ (Huffington Post5); ‘Study shows how to tell if that man in your life has caring genes’ (Digital Journal6); ‘Teen survey reveals gene for happiness’ (New Scientist7); ‘The science of stress – does your child have the “worrier” gene?’ (The Times8), ‘Exam success may be due to a handful of genes’ (The Times9) and so forth. An interview with the singer Sinead O’Connor was headlined with a quotation from the singer: ‘I have no shame. I don’t have an embarrassed gene’ (The Times10). In 2006, an Australian Associated Press article began by stating that ‘New Zealand Maori carry a “warrior” gene which makes them more prone to violence, criminal acts and risky behaviour, a scientist has controversially claimed’ (Kowal and Frederic, 2012). Even sober academic journals such as Nature can seemingly not resist the temptation to compress a complex genetic finding into such attention-grabbing headlines as ‘“Ruthlessness gene” discovered’ (Hopkin, 2008) or ‘A gene for impulsivity’ (Kelsoe, 2010) even though the authors of the scientific papers whose work is being publicised studiously avoid such language. Discussing the tendency that many people drink alcohol at times of stress, Newsweek reassured readers that ‘if this is you, don’t blame yourself. Blame your DNA.’11 Another widely read newspaper asks: ‘Could it be that binge eaters really can’t help themselves? A new study says that weak genes – not weak willpower – may be the reason some people compulsively overeat.’12 The general impression given is that it’s the genes that run the show and so there’s not much you can do about it. Although science journals are generally more careful in their language, their news reporters occasionally slip up and give a similar impression. A news feature in the top scientific journal Nature illustrates this point well, entitled ‘The anatomy of politics – from genes to hormone levels, biology may help to shape political behaviour’ (Buchen, 2012). The author writes that ‘An increasing number of studies suggest that biology can exert a significant influence on political beliefs and behaviours’, reporting that ‘genes could exert a pull on attitudes concerning topics such as abortion, immigration, the death penalty and pacifism’. In the article, John Hibbing, a political scientist at the University of Nebraska-Lincoln, is quoted as saying that ‘it is difficult to change someone’s mind about political issues because their reactions are rooted in their physiology’. In this report, genes

https://doi.org/10.1017/9781108566520.002 Published online by Cambridge University Press

6

1

GENETIC CONFUSION

and physiology are seen as something different from ‘us’ and ‘our mind’, and they seem to be controlling us, so we cannot even change our mind. Political commentators and historians appear to find genetic explanations for cultural and political differences particularly alluring, perhaps because their grasp of the genetics does not match their expertise in other academic disciplines. In his book A Farewell to Arms (2007), the economic historian Gregory Clark argued that the English came to rule the world because the rich outbred the poor, so contributing more of their ‘superior’ genes to the conquering nation. In 2014, A Troublesome Inheritance – Genes, Race and Human History by Nicholas Wade stirred up a hornets’ nest with the suggestion that genetic differences between ‘the three major races’ help to explain economic differences between races and ‘the rise of the West’.13 But even experts in the field of genetics can inadvertently stir up a minor hornets’ nest with the kind of language used in their popular books. Leading behavioural geneticist Robert Plomin from London’s Institute of Psychiatry (where I did my PhD incidentally) faced a minor storm with his book Blueprint published in 2019, writing that ‘DNA is the major systemic force, the blueprint, that makes us who we are. The implications for our lives – for parenting, education and society – are enormous’ (Plomin, 2018). The social implications are barely spelt out in Plomin’s book in any detail, except in a rather speculative and futuristic kind of way, but the ‘blueprint’ metaphor is a powerful one and does seem to imply a rather deterministic role for our genetic endowment. Such an impression is certainly reinforced by comments such as ‘Nice parents have nice children because they are all nice genetically’ and ‘DNA isn’t all that matters but it matters more than everything else put together.’ All this led a reviewer of Blueprint in the journal Nature to claim that ‘It’s never a good time for another bout of genetic determinism, but it’s hard to imagine a worse one than this’ (Comfort, 2018). Strong words indeed, but an indication of how passions run deep in this particular field. Another recent article in Nature comments: ‘The DNA-as-blueprint model is outdated, almost quaint’ (Comfort, 2019). My own choice above of the ‘recipe’ metaphor could easily be misinterpreted as leading to a type of genetic determinism, although its aim is precisely the opposite: to communicate the many ways in which small variations in the composition of the recipe and the environmental conditions involved in the development of the cake (known as ‘cooking’) lead to very major changes in the final product. But metaphors and images are powerful and can exert a strong influence over the ways in which we think about things.

https://doi.org/10.1017/9781108566520.002 Published online by Cambridge University Press

1.2

THE MEDIA PORTRAYAL OF GENETICS

7

Another book that came out in 2019 also uses a title that readily lends itself to viewing the influences of genes through a deterministic lens. The Science of Fate by Hannah Critchlow (Critchlow, 2019) surveys the role of genetic variation in influencing our futures, with a focus on health and disease. Nothing wrong with that – this book will do the same in the context of human behaviour – but the problem comes with the way the material is slanted in a fatalistic direction. Small wonder that a review of the book in The Times was headlined ‘Relax, you have no free will’ with the subtitle ‘Science shows that everything from your flabby tummy to your political views is preordained.’14 The first announcements of the complete sequencing of human DNA in the early 2000s provide a fertile hunting ground for other powerfully influential metaphors. Descriptions such as ‘the Holy Grail’, ‘the Book of Life’ and ‘the Code of Codes’ were all used. Walter Gilbert, who first used the phrase ‘Holy Grail’ to describe the genome at a conference at Los Alamos in 1986, and who was one of the foremost promoters of the Human Genome Project, described its potential with this graphic image: ‘[O]ne will be able to pull a CD out of one’s pocket and say, “Here is a human being; it’s me!” . . . To recognize that we are determined, in a certain sense, by a finite collection of information that is knowable will change our view of ourselves. It is the closing of an intellectual frontier, with which we will have to come to terms’ (Gilbert, 1992). No equivocation there. In 2012, the first wave of thirty papers reporting the results of the ENCODE project were published. ENCODE stands for the ‘Encyclopedia of DNA elements’ and aims to map all the functional sequences of the human genome. The main introductory paper in this series begins its Abstract by emphasising that the ‘human genome encodes the blueprint of life’ (Dunham et al., 2012), again the same powerful metaphor describing how DNA works. The genome in popular scientific literature is often referred to as ‘an instruction manual’, giving the impression that the human body is assembled from the manual much as you might put together a piece of furniture from the kit supplied. We also note the ways in which the phrase ‘it’s in his/her DNA’ has come into common usage in all kinds of contexts, some rather odd. As Brad Pitt once told the Daily Mail while discussing US gun control: ‘America is a country founded on guns. It’s in our DNA.’15 ‘Diamonds and Antwerp – it’s in our DNA’ declares a website from Antwerp wishing to sell diamonds.16 The cloud computing service provider Oxygen assures us that ‘for Oxygen, security is in our DNA. The security of you and your company’s data will always be our priority.’17 In commenting on a new TV drama series, the Director-General of the BBC was quoted as saying that ‘Drama is something

https://doi.org/10.1017/9781108566520.002 Published online by Cambridge University Press

8

1

GENETIC CONFUSION

that is in the lifeblood of this country and in the DNA of the BBC too.’18 The presumed implications of such language are clear: what is in the DNA must be immutable and unchangeable – somewhat missing the point that our DNA is undergoing a constant process of change and diversification.

1.3 Genetic Testing and Genetic Determinism The proliferation of direct-to-consumer (DTC) genetic testing companies has also contributed to the idea that it is our genes that are pulling the strings of human destiny. The front page of The Guardian in 2019 proclaimed that ‘IVF couples could be able to choose the “smartest” embryo: US scientist says it will be possible to rank embryos by “potential IQ” within 10 years.’19 This was based on comments made by Stephen Hsu, Senior Vice President for Research at Michigan State University, but who is also co-founder of a company called ‘Genomic Prediction’, which – no surprise here – might well be offering such a service over the coming years. As we shall see later, the relationship between genetic variation and intelligence (which itself has many different definitions), is highly complex, and the claim made in The Guardian headline is highly dubious, but for the moment we simply note the deterministic framework within which the claims are being made. One in twenty-five Americans now receive personalised genetic test reports that predict their probabilities of developing various medical conditions over their lifetime.20 In 2017 alone, more people had genetic tests carried out than in all of the previous decade since they first became available. One might fondly imagine that when people are told that they have an increased probability of developing a certain disease, based on their genes, they will take extra precautions, such as a better diet and increased exercise, to avoid such an outcome. But surveys show that the opposite tends to be the case – once people learn that their chance of a disease, or a trait like overweight, is (supposedly) more based on genes than on the environment, they become more fatalistic (Dar-Nimrod et al., 2014, Persky et al., 2017). The genes seem to be more in control of the situation than they are. This presumably explains why many people experience more negative emotions and distress when informed about the higher genetic risk of developing a medical condition (Green et al., 2009, Bloss et al., 2011, Dar-Nimrod et al., 2013). Statements on genetics in relation to the environment are generally made on DTC company websites in a reasonably judicious way. But occasionally claims are made with distinctly deterministic overtones. As the Map My Gene website assures us: ‘Your DNA is the blueprint of life. It determines

https://doi.org/10.1017/9781108566520.002 Published online by Cambridge University Press

1.3

GENETIC TESTING AND GENETIC DETERMINISM

9

everything from how you look to how you behave . . . . In MapMyGene, our goal is to unravel that secret to you.’21 The DNA testing kit from the sequencing company 23andMe says ‘Welcome to you.’ The idea of a genetically determined destiny is reinforced by sperm banks that suggest that prospective users should consider the donor’s educational record, his athletic prowess, hobbies and favourite foods, as if these were somehow written into the genetic script provided by the sperm. Human eggs can likewise be purchased online with accompanying details about the donors. One of the problems with DTC genetic testing results is that the data can be passed on to third party app providers that then extrapolate even more wildly from the data than the original company that generated the data. For example, in 2019, a US entrepreneur called Joel Bellenson living in Kampala, Uganda, released an app that supposedly estimated a person’s level of attraction to other people of the same sex (Maxmen, 2019). It is noteworthy that gay sex in Uganda is liable to lead to prison if the person is caught. According to Bellenson, he put together the app over a weekend based on the finding (discussed further in Chapter 10) that several variant genes correlate with those who experience same-sex attraction, despite the fact that the authors of the paper in question took pains to emphasise that a person’s genes cannot predict their sexuality (Ganna et al., 2019). Bellenson posted his app on GenePlaza, an online marketplace for DNA interpretation tools, but after a few weeks of concerted opposition, GenePlaza removed the app. Up to 62 per cent of customers upload their genetic data on to third-party websites, seeking more interpretations of the data (Moscarello et al., 2019). The scope for misunderstanding the data is increasing, often leading to unnecessary scares and concerns for those who do so. Particularly striking is the finding from a psychology research group at Stanford University that even telling people that they are more likely to develop a medical condition due to their genetic constitution causes people to display precisely the kind of risk factors for that condition (Turnwald et al., 2019). For example, merely receiving genetic risk information was enough to increase the heart rate, change how running perseverance was perceived during exercise and change how fullness was perceived after eating. So the genetic information changed the mindset of the people being studied in such a way that it increased the risk of developing precisely the syndromes for which they had been told they had a greater genetic risk. In fact, in some cases, the risk from being told was greater than the actual genetically predicted risk, so presumably in such cases it would be better not to tell people that they had an increased genetic risk at all! The situation is similar to the consequences of telling some people about the side-effects of

https://doi.org/10.1017/9781108566520.002 Published online by Cambridge University Press

10

1

GENETIC CONFUSION

medications – when there is a greater prevalence of the side-effects compared with people who had not been told. We humans are highly suggestible. Does all the outpouring of the language of DNA in popular culture and via the current enthusiasm for genetic testing contribute to the idea that we are really slaves to our genes? It’s hard to say. But at the least it should act as a reminder of the way in which the language of science can be absorbed into public discourse and be deployed in ways that lie well beyond science. Given the long history of the ideological abuse of genetics, one cannot necessarily assume that such misuse of language is merely benign. Cultural osmosis is a powerful process in shaping attitudes, be they expressed in the context of politics, social attitudes, economics, sport or religion. It is only a century ago that we found Samuel J. Holmes, Professor of Zoology at the University of California at Berkeley, informing his readers in his book Studies in Evolution and Eugenics (1923) that anyone familiar with genetics could in a few generations ‘breed a race of idiots, a race of dwarfs, a race of giants, an albino race, an insane race, a race of moral imbeciles . . . a race of preeminent mental ability, or a race of unusual artistic talent’. There was no excuse, declared Holmes, to allow ‘degenerate human beings’ to reproduce (Paul, 1995). Although the main aim of this book is to investigate the role of genetic diversity in differential human behaviours and whether, as a matter of fact, purported roles are validated by the available data, the considerable ideological investments often made in the outcomes of such assessments should act as a warning that in this branch of science more than others the scope for use and abuse remains particularly large. More examples illustrating this point will be given as different topics are addressed throughout the book, including intelligence testing, aggression, sexual orientation, religiosity and politics. The investigators who entitle their publication ‘The Heritability of Foreign Policy Preferences’ (Cranmer and Dawes, 2012) cannot seriously expect that their paper will be treated as ‘pure science’. Overall, therefore, ‘genetic determinism’ with all its various shades of meaning remains a lively topic in public discourse and the outcome of the discussion is not merely academic. Genetically deterministic beliefs often correlate with non-egalitarian attitudes and there is abundant evidence that beliefs concerning the fixity of human identity, be it for perceived genetic or environmental reasons, have a remarkably negative impact on human flourishing. Of course, the truth or falsity of beliefs does not hinge upon their consequences, even though those may be negative. However, given the history of ideological abuse of genetics, it is as well to be very sure about scientific claims and judicious in their public dissemination.

https://doi.org/10.1017/9781108566520.002 Published online by Cambridge University Press

1.4

THE TEACHING OF GENETICS IN SCHOOLS

11

1.4 The Teaching of Genetics in Schools Most people begin to understand something about genetics during their early education. We have already flagged up the problems that can come with starting out with Mendel’s laws as a way of introducing the class to genetics. But the ways in which genetics is taught in schools can often go much further than this in terms of portraying the genes within a framework of genetic determinism. A comparative study of fifty biology textbooks from sixteen different countries produced some interesting results (Castera et al., 2008). The age of the students at which they were taught human genetics ranged all the way from 11 years (in Malta) up to 18/19 years (in many countries). An assessment of how much the textbooks tended towards genetic determinism was made by counting up the number of times phrases like ‘genetic programme’ appeared, and the extent to which identical twins were invariably pictured wearing the same clothes, hair styles and so on. The term ‘genetic programme’ has traditionally been used in many different languages (such as French) to highlight the controlling effects of the genes. In fact, the researchers found that this phrase is still often used in textbooks from France, Morocco, Lebanon and Finland. In German textbooks, however, the term was absent, no doubt due to the powerful historical and cultural linkage between Nazi eugenics and genetics, a linkage that impacts German attitudes and policies on genetics to the present day. As the authors of the study conclude: ‘. . . the contents of the textbooks are not just scientific knowledge, but could convey some implicit messages related to values’, the values they have in mind being the belief, for example, that a person’s identity is inherited from their parents (Castera et al., 2008). An independent study on the four main Finnish biology textbooks agreed with these authors, noting that ‘The textbooks expressed sometimes even strong genetic determinism . . . ’ (Aivelo and Uitto, 2015). A US review of the state of teaching genetics in US schools also revealed some matters for concern (Dougherty, 2009). The author points out that a focus on the inheritance of single-gene disorders, like the one already illustrated in Figure 1.1, ‘may inadvertently contribute to a poor understanding of genetics and encourage genetic determinism’. Analysis of the misunderstandings of genetics in student essays revealed that ‘many of the misconceptions have their roots in deterministic thinking and in an overly simplified view of patterns of inheritance’. Overall, the author concludes that ‘the predominant mode of genetics instruction primes many students to think deterministically and with a confused understanding of risk’.

https://doi.org/10.1017/9781108566520.002 Published online by Cambridge University Press

12

1

GENETIC CONFUSION

What happens if a very different approach is used to teach genetics to undergraduate students? The experiment has already been done (Radick, 2016). A research group at Leeds University, UK, taught a course in genetics starting not with Mendel’s laws of genetics but with the ideas of someone that nobody has heard of (well, not many anyway): Raphael Weldon, a professor at Oxford during the early part of the twentieth century. His views brought him into grumpy conflict with the Mendelian enthusiasts who flourished at the time, who put all their emphasis on the role of the newly discovered genes in controlling everything that was going on. No, said Weldon, how things develop and in what environments is just as important as the genes. Today, such an idea is a well-established biological perspective and indeed central to the ideas in this book. Sadly, Weldon died suddenly from pneumonia in 1906, otherwise it’s possible that eugenics might not have been such a powerful narrative in the early twentieth century. So the Leeds University research group, led by Greg Radick, decided to teach a student group genetics, starting not with the kind of Mendelian inheritance shown in Figure 1.1, but by showing how hundreds of variant genes are involved in the development of major diseases and how their influence is completely integrated with the environment and with the way that people develop, especially in their early years. Perhaps not surprisingly, as compared with a group of undergraduates taught in the traditional way by starting with the laws of Mendel, the ‘experimental group’ ended up with a less deterministic understanding of genetics and a more nuanced understanding as to the role of genetic variants in influencing human differences (Radick, 2016).

1.5 Does It Really Matter? Does it really matter if people think that they’re slaves to their genes? Well clearly it matters in terms of communicating today’s biology in an accurate way. Apart from the beliefs of a few mavericks, and apart from some interesting and striking exceptions that we’ll consider later, the idea that we’re slaves to our genes in terms of our human behaviour just isn’t part of contemporary biology. But some might still argue that it doesn’t really matter anyway if people believe that they’re slaves to their genes, even when in reality they’re not. The problem with this argument is that there is considerable evidence showing that the possession of deterministic beliefs has some seriously negative social consequences.

https://doi.org/10.1017/9781108566520.002 Published online by Cambridge University Press

1.5

DOES IT REALLY MATTER?

13

Psychologists have carried out many experiments in which subjects are influenced in subtle or not so subtle ways to disbelieve in free will, and the social consequences are then measured under controlled conditions. In one study, subjects started by reading either a text promoting the idea of determinism, claiming that scientists now realise that free will is an illusion, or a neutral text, before being tested for either passive or active cheating (Vohs and Schooler, 2008). The passive cheating opportunity arose from an experimental set-up in which the correct answer to a mathematical question appeared on a computer screen. Active cheating was measured by means of subjects paying themselves financial rewards on a difficult cognitive test when no one was looking. In both cases, a higher level of cheating was measured among those with prior exposure to the ‘free will is an illusion’ text. In a separate study, three different experimental protocols revealed that subjects previously exposed to pro-determinism texts rather than pro-free will texts were less likely to be prosocial and more likely to engage in antisocial behaviour (Baumeister et al., 2009, Stillman and Baumeister, 2010). In a further study, brain recordings were carried out on two groups of subjects who had previously been exposed to either pro-determinism or pro-free will texts in order to measure the readiness potential that is associated with the decision-making process (Rigoni et al., 2011). The early phase of the readiness potential was found to be reduced in those subjects exposed to the ‘free will is an illusion’ text, consistent with the measured behavioural changes. Furthermore, undermining free will can degrade self-control, perhaps helping to explain the increase in antisocial behaviour (Rigoni et al., 2012). It could be argued that controlled laboratory experiments do not accurately reflect the social consequences of beliefs about determinism and free will found in normal everyday life. However, studies investigating beliefs about free will have tended to confirm the conclusion that they correlate with generally positive social effects in daily life. For example, possessing a belief in free will, but not in several other key social beliefs, predicted better career attitudes and actual job performance among workers as assessed by their supervisors (Stillman et al., 2011). Others found that increased belief in free will moderately correlated with belief in a just world, intrinsic religiosity and more moral-based attitudes towards judging both self and others (Baumeister and Brewer, 2012, Carey and Paulhus, 2013). Parents who believe that parenting does not make much difference to their children’s outcomes tend to have children with worse outcomes (Baumrind, 1993). Whereas such observations carry the usual caveat that correlation does not entail causation, a reasonable interpretation is that belief in free will supports convictions

https://doi.org/10.1017/9781108566520.002 Published online by Cambridge University Press

14

1

GENETIC CONFUSION

about nonconformity and personal moral responsibility, reflected in increased conscientiousness in the workplace and in parenting, together with a heightened sense of justice. If people really believed that their decisions were determined by their genetics, their environment, or both, not just saying that but actually acting on their beliefs, then clearly much of the social cohesion that holds societies together would vaporise. The moral responsibility assumed by our legal systems plays a central role in assessing whether or not someone is guilty of a crime. The concept of love, which is central to human relationships, would make little sense if we were just determined to love whomsoever by our genes and our environment. Our choice to be an atheist, or an agnostic, or a theist, or wherever we may end up on the religious or non-religious spectrum, has little meaning unless we chose where to be on that spectrum. Asking the question as to whether we are truly slaves to our genes is of more than theoretical interest.

1.6 A Better Way Forward In this book, the aim is to map out a more helpful framing of this discussion – one that fits better with contemporary biology and which seeks to avoid emotive phrases and ‘either/or’ dichotomies. People like Raphael Weldon were not the only ones in earlier years trying to steer the discussion about human genetics in a more helpful direction, emphasising the ways in which the information provided by the genes was thoroughly integrated with other types of information input during development. One of these was Leonard Carmichael whose academic career in psychology and biology included becoming President of Tufts University in the USA (1938–1952). Nearly a century ago, Carmichael (1925) wrote that: In man, from the first environmental stimulation of the fertilized ovum until, it may be, well past three score and ten years, the human individual is not made up of two substances: one acquired, the other innate. The human organism and personality, rather, is a unity produced by both of these forces. The unique resulting totality cannot be . . . dichotomized as part native and part acquired . . . the question of how to separate the native from the acquired in the responses of man does not seem likely to be answered because the question is unintelligible.

There are two key ideas here that require some further unpacking. The first is the vital importance of development when thinking about genetic information. From the moment the egg is fertilised by the sperm until the

https://doi.org/10.1017/9781108566520.002 Published online by Cambridge University Press

1.6

A BETTER WAY FORWARD

15

moment of death, from a biological perspective we are in a continual process of development. During that process, our genetic information is woven together with millions of environmental inputs, some within our bodies, some from without, the developing system as a whole constrained to exist in a certain way by living on a planet with gravity, with 21 per cent oxygen in the air, and with regular light and darkness. Just how that development takes place, we shall consider more in Chapter 3. But for the moment it’s worth noting that in terms of differences in human behaviours, the influence of the genes in our daily lives is not for the most part exerted at the time when the behaviour takes place but rather is a reflection of our early development. Genetics and personal biographical histories are here tightly woven together. The second point that Carmichael makes is that in reality we cannot carve up our being as if we were composed of two substances, one innate (genetic) and one acquired (environmental). All living organisms are made up of a complex system of millions of components that all have to operate together in an integrated way if that organism is to flourish. Clearly, when doing biological research, we have to investigate how all the different components make their varying contributions to the whole. We then make numerical estimates as to how important the various components might be – a major topic in the rest of this book – but in reality it’s the interactions themselves between all the different components that make the system work. That’s why Carmichael asserts that separating the genetic from the environmental is ‘unintelligible’, not because we cannot assign proportional numbers to those aspects of our being for research purposes – we do that all the time – but because those different aspects are so thoroughly integrated (Keller, 2010). A simple illustration might help. The A380 Airbus is currently the largest airplane in civilian use, holding up to 853 passengers. It is made up of about 4 million individual parts produced by 1,500 companies from 30 countries.22 It takes months to assemble (the current record time is 80 days). Clearly, the quality of the individual components is of critical importance to its flying capability. Its correct assembly is likewise critical – however good the quality of a component might be, if installed incorrectly then the plane won’t fly properly. But what’s really important is how those 4 million individual parts interact with each other to make the complex system function. It’s the interactions that make the plane fly, and unless the actions of the components are thoroughly integrated, it won’t fly. We can press the illustration a little further. Looking now at the Airbus system as an operating aircraft, its overall direction and ‘behaviour’ is governed not by its component parts, important as they are, but by the system operating as a whole. What it can do – for example, fly at

https://doi.org/10.1017/9781108566520.002 Published online by Cambridge University Press

16

1

GENETIC CONFUSION

a maximum speed of 1,185 km/hour and go for 15,000 km without landing – is defined entirely by the system. But its schedule is determined by a level of discourse, largely economic, that goes way beyond a discussion of the strength of its wings and its braking capabilities. Major environmental factors – the airline company’s decisions – impinge upon where it ends up every day, but where it goes is absolutely constrained by the kind of plane it is. If the runway is too short, then it just can’t go there. At the same time, its development is continuing ever onwards, with regular servicing and the replacement of parts. Illustrations shouldn’t be pressed too far, but there are some parallels here with our own human identities. Compared with an Airbus, we are vastly more complex, with around 37 trillion cells in our bodies. With us, it’s also our development that is the key – how well we are put together – but also the way in which all those millions of components interact with each other to form an integrated system. It’s the interaction and the integration that make us who we are. And at the higher functional level, where we go and what we do is our choice as agents (we are the airline company executive!). Yes, those choices are heavily constrained by our biological existence – we cannot walk on water, we need oxygen, we need our sleep – but it’s how the complex system operates as a whole that is ultimately the most important question. Just where the genetic information comes from to make all this possible is the topic of the next chapter.

https://doi.org/10.1017/9781108566520.002 Published online by Cambridge University Press

2

Genetic Information and How It Flows

W

e started chapter 1 with a story from the national unit of Human Genetics in Beirut. This one will start there as well, but a bit earlier – back to late 1981 when I first started working there. It illustrates the way in which genetic information flows into a bodily situation where the measurement was quite easy but the interpretation more tricky. There is a disease that is well known in the Jewish community known as Tay–Sachs disease in which the cells die in the brain and spinal cord, so the affected child usually dies when they are 3–5 years old. It’s caused by a defect in a gene that encodes an enzyme that is part of the cell’s waste disposal unit known as the lysosomes. These are little organelles inside the cell that contain more than sixty different enzymes, all of which have the job of breaking down chemicals inside the cell that are no longer needed. Once broken down, the products are then ejected from the cell and end up in the urine. Each enzyme is encoded by a different gene. The problem comes if there is a defect in one of the genes that encodes one of the enzymes. The specific waste products then build up inside cells because there is no enzyme made to degrade them, and this in turn kills the cells. Nasty. In the Lebanese population, the clinically identical disease is called Sandhoff disease, caused by a defect in a very similar gene to the one that causes Tay–Sachs disease. One of the early signs of both diseases is a characteristic cherry-red spot visible in the eye. So there I was in late 1981, having trained up my excellent new Lebanese technician to measure the normal level of these lysosomal enzymes in the plasma (blood minus cells), where they are generally found at low levels. And literally just a couple of weeks after her initial training, a clinician in the hospital thought that he had seen a sick male 8-day-old baby with a cherry-red spot in his eye. So along came our very first blood sample to the laboratory to see if it might be Sandhoff disease. My technician measured the level of the relevant enzyme 17

https://doi.org/10.1017/9781108566520.003 Published online by Cambridge University Press

18

2 GENETIC INFORMATION AND HOW IT FLOWS

and came back looking puzzled. It was not deficient at all but way higher than the level in the control plasma that was always included for comparison. Well, what would you do in such a situation? Ahem, I said politely, I think we might need to try and measure that again. So she did, and again the result came back way too high. So then she measured seven different lysosomal enzymes from the same plasma sample and they all came back way higher than the control values. Clearly something interesting was going on, but what? The next obvious step was to go out and measure lysosomal enzyme levels from family members. Fortunately for us, the Lebanese tend to have big families and this Palestinian family was no exception. They all lived in the Sabra Palestinian refugee camp where just months later there was to be the most terrible massacre. We found the grandmother of the big extended family from three generations all living in the camp. The grandmother is the key person to get our side on such occasions. She happily gave us a small blood sample herself and then ordered the rest of the family to do likewise, which of course they did, and from that single visit we ended up with thirteen blood samples from different generations of the same family, of which five turned out to have super-high levels of all of the many lysosomal enzymes we measured, one of them being the grandmother. Yet all those with this abnormality in the blood were healthy. To cut a very long story short, it turned out in the end that this was all due to a novel gene mutation that had never been described before (Alexander et al., 1984, 1986). And the hospital clinician who thought that he’d seen a cherry-red spot said in the end that he was mistaken, and certainly the baby did not have Sandhoff disease. What was happening is that only one of the two copies of a critical gene was defective in these five individuals and so their single good copy was enough to keep the disease at bay but was not enough by itself to prevent excess lysosomal enzymes leaking into the blood. If first cousins from this family ever got married, then there would be a one in four chance that one of their offspring would have a double dose of the defective gene, in which case the baby would almost certainly die due to enzyme deficiency inside their lysosomes. But massacre and then the civil war took over the Sabra camp and so we never had the chance to help this family further with genetic counselling. But I’ve always been grateful to that hospital clinician who made an error in seeing a cherry-red spot that wasn’t really there in the eye of a baby. These things happen, and in this case it led to the discovery of a totally new human mutation, as far as I know never since found in any other family in the world.

https://doi.org/10.1017/9781108566520.003 Published online by Cambridge University Press

2.1 WHAT IS A GENE?

19

Why tell that story here? Because it is a good illustration of the way that genetic information flows through individuals and families, often hidden from view when it causes differences but with the results suddenly becoming visible when the appropriate measurements are made. The pathway from gene to a visible difference is a long and winding one leading through lots of rather complex biochemistry, as in the present case. Eventually, the tiny difference in a single gene is revealed in some difference in the whole living organism. But how does this information flow actually happen and what do we really mean by a ‘gene’?

2.1 What is a Gene? The definition of a gene used to be nice and simple – a stretch of DNA that encodes a protein. The DNA is the chemical that encodes the genes and it’s found in the nucleus of cells, a membrane-enclosed package that keeps this precious little bundle of information away from the cell’s other main duties, which include energy production and waste disposal. DNA is composed of subunits called nucleotides, which each comprise a sugar molecule, a phosphate group and one of four chemical groups known as bases; these four bases are adenine (represented as the letter A), thymine (T), guanine (G) and cytosine (C) (we will refer to these as the ‘genetic letters’). G binds exclusively to C, and A binds exclusively to T, which is how the DNA exists as a double helix, as Figure 2.1 illustrates. If you find mnemonics helpful, then remember a Good Cup of Asian Tea, and that’s all you need to know to recall the structure of DNA – G pairs with C, and A pairs with T. Each molecule of DNA is packaged together with some proteins, crucial for its function, to form a chromosome. Fortunately, the structure of DNA is simple, elegant and quite easy to explain, which is how Watson and Crick were able to present it in 1953 in just a single page of the journal Nature. Imagine that you are at the bottom of a spiral staircase. Each step has a height of 3.4 feet, a little more than a yard, so you have to stretch up a bit to make the steps. Each step is rather wide – in fact, 20 feet wide – and each time you step up, you start to twist around the centre by 34° (we’ll see why the steps are this wide and this high in a moment). As you go up the steps, you notice that each one is labelled ‘hydrogen bond’, so they seem pretty uniform – that’s the name of the chemical bond that holds the steps in place. But then you notice a pattern. If the step is joined to a G on the right side, then it’s always joined to C on the left. Go up a step and it may be joined to an A on the right side, in which case the step joins up with a T on the left.

https://doi.org/10.1017/9781108566520.003 Published online by Cambridge University Press

20

2 GENETIC INFORMATION AND HOW IT FLOWS

Figure 2.1 The DNA double helix. Four chemical genetic letters adenine (A), thymine (T), guanine (G) and cytosine (C) serve as the basis for the genetic code. A binds exclusively with T, and C with G, allowing faithful replication of DNA. Reproduced by permission of the Templeton Press.

Now take out a notepad and jot down the labels on the different genetic letters as you climb upwards. Starting from the bottom of the staircase, those from the right read ATGTACAAGGATGTGCTATTGTAA and onward, so those on the left read TACATGTTCCTACACGATAACATT. The sequence of letters seems to make no rhyme or reason as you jot them down. After exactly ten steps, or 34 feet precisely, the spiral makes a complete turn. In other words, as you look down from the right-hand end of step number ten, you can see someone else standing below on the right-hand end of step number one. By now, you should be getting the picture of the double-helical structure of DNA illustrated in Figure 2.1. The only difference, to make picturing the mental spiral stairs a bit easier, is that we have made one angstrom (1 Å) equal to one foot. An angstrom is actually quite small, only one-hundredmillionth of a centimetre, or 0.000000004 inch. So in reality, the vertical

https://doi.org/10.1017/9781108566520.003 Published online by Cambridge University Press

2.1 WHAT IS A GENE?

21

distance between the genetic letters that make up the double helix on each side is just 3.4 Å, and the hydrogen bonds join them together (like the steps) across are a total diameter of 20 Å. To keep that distance exactly equivalent at each step, a G always has to bind to a C, and an A always to a T. The shapes of G and A don’t allow mutual binding to each other, and neither do the shapes of C and T. We humans have twenty-three pairs of chromosomes, the twenty-third pair encoding the genes for males (XY) or females (XX). Each chromosome contains one spiral staircase, one long double-helical DNA molecule. The fact that we have pairs of chromosomes is really important because that means that we have two copies of all our genes encoded by the two molecules of DNA, one molecule in each chromosome pair. If one gene is defective, we still have the backup. This explains why the five family members from Beirut who had the abnormally high levels of lysosomal enzymes in their blood seemed healthy – their backup functional gene was still making sure that enough enzymes were getting packaged into their lysosomes. Without that backup, they would indeed have been in trouble. By convention, the human chromosomes are numbered from one to twenty-two according to size, with chromosome 1 being the largest and twenty-two the smallest. Chromosome one contains about 248 million genetic letters or 248 million steps in our staircase analogy. That’s a very long way up if you’re climbing a stairway. Based on 3.4 Å per step, this comes to a bit more than 3 inches long, or more than a yard long if you add all the chromosomes together, totalling 3.2 billion genetic letters, the same number of letters as 2,000 copies of War and Peace. You might wonder how such a huge collection of chromosomes could fit into the tiny nucleus of a cell, and the answer is a marvel of packing. Instead of being stretched out lengthwise in the way pictured here, in reality the chromosomes are tightly folded, making your vacation packing of too many items in one crammed vehicle look trivial by comparison. In the DNA double helix, the hydrogen bonding is of exactly the right strength for the molecule. It’s like a zipper on a jacket – fixed enough to hold the garment together but not so fixed that unzipping becomes too difficult. As Watson and Crick pointed out in their 1953 paper in Nature, DNA has exactly the right structure to facilitate its own replication. It simply unzips down the middle and then each strand of the helix is used to generate a complementary strand in which the information is replicated by keeping to the exact same GC and AT pairing rules. The outcome is two identical daughter molecules of DNA. That’s how inheritance works.

https://doi.org/10.1017/9781108566520.003 Published online by Cambridge University Press

2 GENETIC INFORMATION AND HOW IT FLOWS

22

It is the precise sequence of the genetic letters that encodes the genetic information. The sequence of genetic letters that you jotted down in your notebook as you climbed the spiral staircase does indeed encode the real sequence of part of a protein. Twenty different amino acids are used to build different proteins, and the specific sequence of these amino acids gives the protein its particular properties. Each amino acid is encoded by a triplet of genetic letters in the DNA sequence known as a ‘codon’ and the code manual is shown in Figure 2.2. Using the manual (which is based on the RNA sequence, which uses uracil (U) in place of thymine (T)), we can take the sequence and now separate it out into triplets: ATG TAC AAG GAT GTG CTA TTG TAA

U

U

UUU UUC UUA UUG

C Phe Leu

Pro

CAU His CAC CAA GIn CAG

CGU CGC CGA CGG

Arg

U C A G

AAU AAC AAA AAG

AGU Ser AGC AGA Arg AGG

U C A G

GGU GGC GGA GGG

U C A G

A

AUU AUC AUA AUG

IIe

ACU ACC ACA ACG

Thr

GCU GCC GCA GCG

Ala

First letter

CCU CCC CCA CCG

Val

U C A G

Ser

Leu

GUU GUC GUA GUG

UGU Cys UGC UGA Stop UGG Trp

UAU Tyr UAC UAA Stop UAG Stop

CUU CUC CUA CUG

G

G

UCU UCC UCA UCG

C

Met

A

Asn Lys

GAU Asp GAC GAA Glu GAG

Gly

Third letter

Second letter

Figure 2.2 The genetic code. Each codon of three ‘letters’ encodes an amino acid. There are four different letters, known as bases, in the DNA: T = thymine (transcribed into RNA as U = uracil), C = cytosine, A = adenine and G = guanine. In the double stranded DNA helix, a G binds to a C and an A binds to a T (or U in RNA). Protein coding genes in the DNA are transcribed into messenger RNA (mRNA) and it is the triplet codon sequences in the mRNA that are shown here. RNA uses the same nucleotides as DNA except that uracil is used in place of thymine. Abbreviated forms of the twenty amino acids are shown, e.g. Tyr = Tyrosine; Ser = Serine; Gly = Glycine etc. Reproduced from OpenStax, Biology, OpenStax CNX.

https://doi.org/10.1017/9781108566520.003 Published online by Cambridge University Press

2.1 WHAT IS A GENE?

23

We can then immediately translate that sequence into: START tyrosine lysine aspartate valine leucine leucine STOP

where each of the names shown refers to a particular amino acid, and the START and STOP signals show the code-copying machinery where to start and stop reading. The fact that both CTA and TTG encode the amino acid leucine illustrates the fact that the code contains redundancy: in some cases, several triplet codons encode the same amino acid. Only six amino acids are shown in this sequence, far fewer than would be found in a typical protein, which might typically contain hundreds of amino acids or more. But the example illustrates an important principle: how the sequence of DNA is converted into the sequence of amino acids in a protein. As proteins give the structure and function to living organisms, we can now see how DNA can operate as a storehouse of information that helps to build bodies. I first started looking at the triplet codon manual illustrated in Figure 2.2 way back in the mid-1960s, but it still fills me with a buzz of excitement and admiration to the present day – maybe not everyone’s cup of tea but hopefully this book will help you capture some of the buzz. And we are now in a better position to ask again what we really understand by the word ‘gene’. When human DNA (known as the ‘human genome’) was first fully sequenced in the early 2000s, everyone thought that counting up the number of protein-coding genes would be quite easy, but in fact it has proved to be remarkably difficult for a number of technical reasons that needn’t detain us here. There are four main online databases of current genetic information that scientists use all the time, and they still don’t agree on the precise number, but at present the number of protein-coding genes is believed to be in the range 20,000–20,500.1 This is only about 1.5 per cent of the 3.2 billion genetic letters found in human DNA taken as a whole. But the rest of the DNA has plenty of other functions. One of these functions is to make another type of gene: RNA genes. In fact, DNA is used as a template to make various forms of RNA, a chemical very similar in structure to DNA except that it uses the genetic letter U (for uracil) in place of the T found in DNA. In the next section, we’ll see how information flows out from DNA via a type of RNA called messenger RNA (mRNA), which is involved in the synthesis of proteins. Here we note that DNA encodes not only proteins but also other RNA molecules that act as regulators and communicators in their own right, so these sequences are now called RNA genes. The full number of this type of gene is not yet known, but at the time of writing, the Gencode database lists 25,528 ‘non-coding RNA genes’.2 So there are now known to be more RNA genes than protein-coding

https://doi.org/10.1017/9781108566520.003 Published online by Cambridge University Press

24

2 GENETIC INFORMATION AND HOW IT FLOWS

genes and we need to come up with a new definition of a gene: a DNA sequence that is transcribed into a functional RNA molecule or that is transcribed into RNA and then translated into a functional protein (Salzberg, 2018). We note that the DNA sequence still remains the primary source of information in both types of gene. And what ‘transcribed’ and ‘translate’ mean we are now about to find out.

2.2 The Genetic Information Flow The way that the information flows out of DNA into the finely tuned structures of the proteins that make up our bodies is really fascinating and we will just provide a brief summary as to how this works. Here, we are referring to the information from protein-coding genes. As we have noted, RNA can contain information just like DNA by means of a specific sequence of genetic letters. RNA can also form a double helix like DNA, but in the present context only a single strand of RNA is involved. What happens is that the specific sequence of genetic letters on one of the two strands of double-helical DNA in the nucleus of the cell is ‘transcribed’, a process known as ‘transcription’, into a molecule of single-stranded RNA, known as messenger RNA, or mRNA for short. The message encoded in the RNA is then ‘translated’ into the amino acids of a particular protein. Note how the terminology of language is used so often in talking about the way that information flows from the genes. ‘Transcription’ sounds as if there were a very small DNA clerk sitting inside the nucleus and calling out the sequence for the RNA clerk to copy down. The reality is a bit more sophisticated than that. Think back to the DNA helical staircase. The right-hand strand going up is known as the ‘sense strand’, in which the genetic code makes sense as it’s read from the bottom upwards. The other strand is known as the ‘antisense strand’. The challenge now for the DNA is how best to convey its information out of the nucleus to the cell cytoplasm, where the information is needed to synthesise new proteins. The DNA accomplishes this task in an elegant and essentially simple way using an enzyme that ‘reads’ the antisense strand. As Figure 2.3 illustrates, in front of the gene is a ‘promoter region’ that tells the enzyme when to begin reading. Once certain regulating factors, known as ‘transcription factors’, do or do not bind to the promoter region, the gene is then ready to be transcribed; the enzyme copies the DNA into mRNA starting at a specific recognition sequence. This creates a complementary genetic letter sequence of mRNA, using the same pair-wise rules already outlined: a G in the DNA produces a C

https://doi.org/10.1017/9781108566520.003 Published online by Cambridge University Press

2.2 THE GENETIC INFORMATION FLOW

25

Figure 2.3 Reading DNA.

Figure 2.4 How to make messenger RNA (mRNA).

in the mRNA, and an A produces a U (not a T because RNA uses U in place of T). Once the enzyme reaches another recognition sequence, it falls off and trundles off to work on another gene. Using the same example sequence as shown above, we can now see what our mRNA looks like in Figure 2.4. Inspection of the mRNA shows that it now has exactly the same sequence as the sense strand of the DNA, except that it has a U in place of each T in the DNA. Once an mRNA transcript is made in this way, the gene is said to be ‘expressed’. The transcript is now ready for work – to use its information to produce the amino acid sequence of a protein in a process called ‘translation’, described below. First, though, notice that the example shown in Figure 2.4 is not really a gene, just a small portion of a gene used for illustration. In reality, genes encode proteins that are several hundred – or sometimes more than 1,000 – amino acids long. A human muscle protein called Titin is an astounding 34,350 amino acids long, so it takes 3 × 34,350 = 103,050 nucleotides to encode it. At any given time, hundreds of different mRNA molecules of all different sizes are being generated in the nucleus. The transcription machinery is incredibly fast; on average, the enzyme adds forty new genetic letters to a given mRNA every second. This means that a gene with 1,000 genetic letters is transcribed into a new molecule of mRNA in only 25 seconds, which is pretty impressive. The next challenge, as Figure 2.5 illustrates, is translating the mRNA sequence into the precise sequence of amino acids that make up the protein

https://doi.org/10.1017/9781108566520.003 Published online by Cambridge University Press

26

2 GENETIC INFORMATION AND HOW IT FLOWS

Figure 2.5 Transcription and translation. Transcription is the process by which mRNA is synthesised using a DNA template. Transcription occurs in the nucleus of the cell, and mRNA molecules then move into the cytoplasm, which occupies the main body of the cell, where they are translated into proteins. Translation is the process by which proteins are synthesised from individual amino acids using mRNA as a template. Translation occurs on ribosomes, which bring together the mRNA and amino acids bound to transfer RNAs (tRNAs). Reproduced by permission of the Templeton Press.

being synthesised. This involves a really cool little molecular production line. First, the completed mRNA moves out of the nucleus through little holes into what is known as the ‘cytoplasm’ of the cell – all of the cell interior outside of the nucleus. There it is used as a template to construct the new protein with the help of sophisticated pieces of molecular machinery called ribosomes. Clearly for a translation process to work properly, a translator is required, and in this case a collection of clever adaptors called transfer RNAs (tRNAs) carry out the work. The tRNAs contain the same four genetic letters as the other types of RNA, such as mRNA. But they are different from mRNAs in being much shorter and of a standard length, usually seventy-four to ninetyfive genetic letters in length. Each tRNA binds to a particular amino acid and

https://doi.org/10.1017/9781108566520.003 Published online by Cambridge University Press

2.3 EDITING THE FLOW OF GENETIC INFORMATION

27

at the same time contains a triplet sequence of genetic letters known as the anticodon. The anticodon corresponds exactly to the codon in the mRNA sequence, in the same way that a plug fits into a complementary socket, and so the tRNA brings exactly the right amino acid to the assembly line, the amino acid found in the triplet codon back in the DNA, just as specified by the genetic code shown in Figure 2.2. The clever ribosome now plugs in the next tRNA, which has the next anticodon corresponding to the next codon in the mRNA. So this brings the next amino acid along. The first two amino acids of our new protein are now next to each other, brought into position like two boats moored side by side. An enzyme then catalyses a bond between the two amino acids, which glues them together quite firmly, and the ribosome then shuttles along to pick up the next amino acid in the growing chain until a STOP signal is reached and the protein is complete. By this time, the protein will already have folded up into a shape that its precise amino acid sequence specifies, and this specified conformation gives the protein its particular properties; it may be an enzyme, a structural protein, a regulatory protein, a specialised muscle protein or whatever else the body might need at that particular moment. The same basic process of transcription followed by translation, and virtually the same genetic code, are common to all living things: bacteria, plants and animals. Bacteria are different from plants and animals in that they don’t have a nucleus, so the mRNA does not travel from the nucleus into the cell. They often have just one main chromosome, compared with our twenty-three pairs of chromosomes, which helps simplify the process. Other minor differences occur in the way the system works in cells with or without a nucleus, but the general idea is the same.

2.3 Editing the Flow of Genetic Information If that was the whole process, then it does rather look as if there’s a one-way traffic of information that flows from DNA to RNA to proteins, which then make our bodies what they are. Indeed, in the earlier understanding of the genome as a ‘blueprint’, the products of the genes were ‘read off’ as if each gene had one fixed and determinate product. That can easily look like a rather deterministic narrative. But we now know that this is not the case: most genes can generate several different proteins and their amino acid sequence is not invariably specified by the sequence of genetic letters in the DNA. Our understanding of the genome has now changed radically from a ‘read-only’ concept, in which the genome was perceived as a static

https://doi.org/10.1017/9781108566520.003 Published online by Cambridge University Press

28

2 GENETIC INFORMATION AND HOW IT FLOWS

repository of information, to a ‘read–write’ model in which the genome is under constant dynamic regulation and modification (Shapiro, 2013). In fact it has been estimated that more than a million different proteins are made from our modest 20,500 protein-coding genes due to all the editing involved. Here, we will not attempt a complete description of how the editing takes place – that would take a whole book3 – but rather highlight a few key examples.

2.3.1 One Gene Can Give Rise to Many Different Proteins For those who remember their school biology, the mantra ‘one gene – one protein’ is probably easy to remember. The problem is that the idea is quite false. And one reason it’s false is called ‘alternative splicing’, something that takes place in the nucleus of the cell (Sulakhe et al., 2018). A protein-coding gene encodes both introns (for intragenic regions) and exons (for expressed regions). The exons refer to the DNA sequences within the gene that end up encoding the protein, whereas the introns are ‘spliced out’ and their sequence does not encode any of the protein. So by shuffling around the exons, one gene can generate many different types of mRNA and each in turn will encode a different protein. Around 95 per cent of our protein-coding genes undergo alternative splicing, generating perhaps around 100,000 different proteins from our 20,500 genes. In fact, one research group has estimated that there are as many as 205,000 different mRNAs made from our 20,500 genes, with the potential to make proteins, although as yet only around 100,000 are thought to actually do that (Hu et al., 2015). Splicing varies among individuals, thereby increasing the amount of genetic variation in the human population. It can also affect which transcription factors are made. So we can already begin to see here how complex networks of plus/minus genetic signals can be built up using these various splicing products. Not surprisingly, abnormal splicing is involved in many different genetic diseases and in cancer. In fact, around one-third of all disease-causing mutations are due to splicing errors, one example being the mutation that causes the blood disease thalassaemia in the Chinese population. The genome contains a code within a code in the sense that certain features identify the introns and exons where splicing occurs. But this code is a lot more complicated than the sixty-four triplet nucleotides that encode the amino acids. Instead, it consists of a complex algorithm that includes more than 200 different features of DNA structure that predict where splicing will occur (Barash et al., 2010, Ledford, 2010). Furthermore, there

https://doi.org/10.1017/9781108566520.003 Published online by Cambridge University Press

2.3 EDITING THE FLOW OF GENETIC INFORMATION

29

is a complicated bit of molecular machinery that actually carries out the splicing, and the composition of this machinery varies depending on the particular environment of the cell. This variation in turn impacts on whether the splicing occurs or not. So this is one example among many where the environment influences which protein is finally made.

2.3.2 RNA Editing Whereas alternative splicing works by re-assembling and swapping around segments of DNA sequence, RNA editing is in a sense more radical in that it changes the actual sequence of the mRNA during or after its transcription, and this can also happen in many other types of RNA. This happens by the insertion, deletion or substitution of one of the four genetic letters at selected sites in RNA. The RNA produced in this way is called a ‘cryptogene’, so-called because the ‘image’ in the DNA of the resulting products is now no longer recognisable in the DNA. RNA editing, although less well known than alternative splicing, is by no means a rare esoteric mechanism but rather reflects the normal functioning of both plant and animal cells. More than 100 million editing sites have been reported in human RNA (Bazak et al., 2014), although only a relatively few of these cause a change in the amino acid sequence following translation (Ulbricht and Emeson, 2014). But some of the changes can have big potential impacts on protein functions, including those found in the brain and immune system (the system that defends us against invasion by bacteria, viruses, etc.). The real winners in the RNA editing stakes are animals like squid and octopuses (Liscovitch-Brauer et al., 2017). Their RNA editing is so common, especially in the way it affects their nervous systems, that octopuses don’t have nearly so much variation in their protein-coding genes as we do, simply because they can get lots of variation from RNA editing. Maybe that’s why they’re so clever – they have lots of different ways of boosting their brain power. Bumblebees are also really good at RNA editing where it seems to be regulated by their brood-care and foraging behaviours (Porath et al., 2019). We humans don’t use RNA editing so much in generating variation between us, but it’s still critically important. Some people on the autistic spectrum (discussed further in Chapter 5) have widespread dysfunction of RNA editing in their brains, which in itself highlights the importance of this mechanism in brain development (Tran et al., 2019). Imagine an author writing a book (the DNA) that then goes to a very vigorous editor who changes quite a bit of the author’s text (the RNA editing). The reader will never know what the author wrote but will just see the final text produced by

https://doi.org/10.1017/9781108566520.003 Published online by Cambridge University Press

30

2 GENETIC INFORMATION AND HOW IT FLOWS

the team. In reality that’s what every newspaper reader is doing. So it’s good to remember that not all the information is found in the DNA. We are reading an edited text.

2.3.3 Transcription Factors The proteins that regulate gene expression – the switching on and off of genes – are known as ‘transcription factors’, as already mentioned. Their name depends on the fact that they regulate which genes are transcribed into mRNA. There are up to 2,600 different transcription factors in all our different cells, and at any one time there are around 300,000 transcription factor molecules in every one of them. Every aspect of our existence is regulated by complex gene regulatory networks within which transcription factors play a vital part, be it development, brain function, reproduction, digestion, muscle action or functioning of the immune system. So there is a constant traffic of information coming from the environment and impacting on the expression of either a few highly selective genes or hundreds of genes depending on the signals involved. When we choose our body state, whether it be by lack of sleep, too much exercise or too little, overindulgence in food or drink, starvation, risking infection by unclean habits or choosing to live angry lives, we are ipso facto influencing the status of our genomic system. For example, when participants in a study were exposed to just 1 week of insufficient sleep (average 5.7 hours per night) and compared with a control group who slept an average of 8.5 hours of sleep per night, no fewer than 711 genes were found to be expressed at either higher or lower levels in the sleep-deprived compared with the control group (MöllerLevet et al., 2013). The genes affected included many involved in circadian rhythms, metabolism, stress and immune responses. Not all the regulation of gene expression in such a study can be ascribed to transcription factors alone – other types of molecule are also involved in the complex web comprising the regulatory network that switches genes on and off. A second example will also help to illustrate the way in which the environment ‘talks’ to the genome by means of transcription factors. Let us imagine that you choose to take a high-flying job in the city that is highly paid but intensely stressful. As a result, your adrenal cortex, located just above the kidneys, is very likely to start secreting more of a steroid hormone called cortisol. A region of the brain known as the hypothalamus controls the level of cortisol secretion, increasing it in response to stress. Cortisol has numerous effects on the body, including raising blood sugar, modulating metabolism and suppressing the immune system, useful from an

https://doi.org/10.1017/9781108566520.003 Published online by Cambridge University Press

2.3 EDITING THE FLOW OF GENETIC INFORMATION

31

evolutionary perspective in terms of preparing the body for stressful situations in which a good supply of energy is important but less useful perhaps in preparing the body for a day in the office, especially considering that cortisol is implicated in increased blood pressure. Cortisol enters its target cells and there binds to a receptor that is itself a transcription factor, thereby activating it and causing it to bind to a whole range of gene promoters – hence the many effects. So taking that stressful job in the city can have a major impact on the regulation of your genome, in turn leading to changes taking place in billions of your body’s cells. Such examples could be multiplied a 100-fold, illustrating the many different ways in which environmental information is conveyed to the genome, with combinatorial control by myriad transcription factors and other types of molecule regulating the fine-tuning of gene expression. Vast interacting communication networks operate to regulate the genome, and it is the system as a whole that functions like an orchestra to produce the ‘right’ music. Going back to our newspaper analogy, with the authors of all the articles representing the genome, the various editors of the different sections – the transcription factors – now decide which articles are actually going to be published. As a result, most of what the poor journalists write is not published that day. So again when you’re reading the genomic newspaper, you’re not really reading the genome, just a section of it. And that selection is different between people – and different for you at different times of day, depending on what you’re doing. It’s not the case that ‘DNA rules’ – it’s the whole molecular system that is involved, tightly integrated with input information from the environment.

2.3.4 Epigenetic Control of the Genetic Information Probably the most important input information of all is that which leads to ‘epigenetic’ changes – the process whereby DNA itself, or the proteins with which it’s tightly packed in the chromosome, is chemically modified in such a way that, again, genes are switched on or off. The word ‘epigenetics’ derives from combining the Greek epi, meaning ‘over’ or ‘above’, with ‘genetics’, because the chemical changes involved don’t change the DNA genetic letters sequence. So the epigenetic changes are called ‘marks’, as if the editor were making marks on the author’s text to indicate what actual text should appear in the newspaper that day. Or if the DNA sequence is the musical score on the page, then epigenetics is all those Italian words (fortissimo and the like) that give instructions as to how the notes should be played.

https://doi.org/10.1017/9781108566520.003 Published online by Cambridge University Press

32

2 GENETIC INFORMATION AND HOW IT FLOWS

Unlike the situation with transcription factors, epigenetic changes can be inherited. In other words, when the DNA replicates and two cells are formed from one cell, each of the new cells contains the same DNA molecules – complete with their epigenetic marks. During early development, as we’ll be thinking about more in Chapter 3, the different cells of the body become specialised to carry out different functions. All the cells contain exactly the same DNA, the specialisation taking place by great swathes of the DNA being switched on/off as a result of epigenetic changes. And it’s really fortunate for us that the epigenetic changes are passed on from cell to cell during DNA replication, otherwise the cell specialisation would be reversible. We might start seeing liver cells growing on our arm or brain cells appearing on the end of our nose, which would be inconvenient, to say the least. Having said that, it’s worth also keeping in mind that, whereas there is an error rate of around one in a million genetic letters when the DNA is copying itself ready to make two cells in place of one, the error rate is about one in a thousand for the transfer of epigenetic marks through a cell division. That’s remarkably accurate and certainly good enough to keep our cells doing the same job nearly all the time as they keep dividing within our livers, brains, kidneys or wherever. Given the role of epigenetic ‘marks’ to maintain cell specialisation, even through many cycles of DNA and cell replication, it might not at first seem obvious that epigenetic regulation of gene expression is also a dynamic environmentally influenced process, but this is indeed the case. First, the development of the fetus involves waves of epigenetic modifications that can be modulated by environmental inputs in ways that impact on the health and well-being of the individual for the rest of their life. Some examples will be provided in Chapter 3 when discussing early human development. Second, dynamic epigenetic regulation of gene expression is important in our daily lives in response to environmental inputs. Third, environmentally induced epigenetic modifications in an individual organism can be transmitted transgenerationally, for many generations in the case of certain animals and plants, and in a limited way at least in the case of humans. One animal model that well illustrates the way in which postnatal environmental differences lead to permanent behavioural changes in the adult via epigenetic regulation is provided by the maternal care of rat pups (Meaney, 2010). Rat pups like lots of licking and grooming by their mothers during the first week of life. More surprising is the fact that the effects of that one week of postnatal experience have an impact on the rat that remains for the rest of its life. Some mothers are naturally good at such care for all their pups, others less so. When the rats are later exposed in adulthood to mildly stressful

https://doi.org/10.1017/9781108566520.003 Published online by Cambridge University Press

2.3 EDITING THE FLOW OF GENETIC INFORMATION

33

experiences, the ones who received the most licking and grooming during the first week of life remain the most calm, with a lower startle response and less cortisol production (increased cortisol is a sign of stress in rats as in humans). Conversely, the ones who received poor maternal care overreact to such stressful stimuli. By swapping litters of pups between the ‘caring’ and ‘uncaring’ mothers on postnatal day one, it has been shown that such behavioural traits in adulthood are predicted by the licking and grooming received during week one, not by the genetic inheritance received from the parents. And by week two, the maternal care effects disappear – it is grooming and licking in week one that counts, not during week two. The caring or non-caring traits persist into the next generation, because the females demonstrate either caring or non-caring responses to their own new litters in turn. Epigenetics appears to play a key role in connecting the maternally received care or indifference of postnatal week one with the rat’s responses to stress in later life. But it needs to be emphasised that there are lots of changes in the expression of key genes involved here – there is no one ‘licking response gene’ that changes the behaviour of the rat pups! In fact, it has been shown that the expression of hundreds of genes is different between high-care and low-care rat offspring, with measurable epigenetic changes at multiple regulation sites in the DNA (McGowan et al., 2011, Bagot et al., 2012). What about humans? Here also there are some fascinating studies suggesting that environmental impacts on the parents in one generation can affect the health outcomes of not only their own children but also perhaps their grandchildren as well. One of the most famous examples involves the children born of parents who suffered malnutrition in utero due to the Dutch Hunger Winter of 1944/45. These children were found to be fatter during their early years of life and suffered a greater measure of ill health in later life compared with controls (Painter et al., 2008). The offspring of prenatally undernourished fathers, but not mothers, were heavier and more obese than the offspring of fathers and mothers who had not been undernourished prenatally (Veenendaal et al., 2013). Individuals who were prenatally exposed to famine had, six decades later, changes in epigenetic marks on specific genes, relevant to health, compared with their unexposed, same-sex siblings (Heijmans et al., 2008, Tobi et al., 2009), but it is not (as yet) known whether such differences are transmitted to the next generation. Another study has examined how the offspring of American Civil War (1861–1865) veterans, who suffered severe famine and deprivation as prisoners of war (POWs), fared in their later lives compared with those with

https://doi.org/10.1017/9781108566520.003 Published online by Cambridge University Press

34

2 GENETIC INFORMATION AND HOW IT FLOWS

parents who suffered no such deprivation (Costa et al., 2018). The results are based on detailed information on the socio-economic and family structure for 4,593 children of 1,407 former POWs and 15,310 children of 4,960 non-POW veterans – so a thorough examination! Among sons born in the second quarter of pregnancy, when maternal nutrition was inadequate, those of ex-POWs who experienced severe hardship were 1.2 times more likely to die than those of non-POWs and ex-POWs who fared better in captivity. That means that for every 100 sons in the control group who died in a certain year, there were 120 offspring of the ex-POWs who died in the same year. There was no difference in the age when the daughters died. This particular study had no direct way of looking at epigenetic differences in the offspring, but infers by excluding other explanations that epigenetic differences are the most likely reason for the increased death rate. Why was it only the males whose lifespan was affected and not the females? Males, but not females, have a Y chromosome, and it’s known, as the authors of this study point out, that in other studies alterations have been observed in the sperm of adult men as a result of age, diet, smoking, alcohol and exposure to toxins (poisons) (Bromfield, 2014). So one possibility is that epigenetic changes were introduced into the sperm of the POWs by their deprivation and lack of adequate nutrition, which later impacted on the health of their sons. There are plenty of other similar studies in which environmental impacts on parents have been shown to have health consequences for their children and even grandchildren. For example, extensive studies have been carried out on the ‘Överkalix cohort’, a group of three-generation families from northern Sweden in which the grandparents experienced alternating periods of feast and famine, with distinct effects on the health of their children and even their grandchildren (Alexander, 2017, Vagero et al., 2018). But although in some of these examples, such as the Dutch Hunger Winter cohort, epigenetic differences have been measured in the offspring of the pregnant mothers who suffered famine, it is not completely clear as yet what the link is between these changes and the health problems faced by the offspring. In contrast to mice and rats, which reproduce quickly – hence their major use in research laboratories – humans are a bit slow in this respect, making such studies challenging. Some media reports have gone rather over the top in claiming too much for such results – declaring that epigenetic changes in us can influence our grandchildren and their children in turn. This may turn out to be true, but the data are currently not that clear, and it’s plants and animals that really give us convincing data showing that this can indeed happen in many other species around us.

https://doi.org/10.1017/9781108566520.003 Published online by Cambridge University Press

2.4 INFORMATION IN ALL DIRECTIONS

35

2.4 Information in All Directions Enough has been considered so far to see that there is no linear ‘DNA command post’ telling the rest of the body what to do and how to do it. Yes, the primary source of information is in the DNA. But it’s no blueprint like the architect’s plans for a new building. Rather, it’s a key source of information that requires complex molecular machinery to interpret the messages. DNA without all that machinery would be as useful as a CD or memory stick containing a computer program or app without any computer or mobile to run it on. And as the DNA messages flow, so the information is edited and sometimes changed altogether. Environmental information comes in all the time to switch the genes on or off, sometimes with longterm consequences, sometimes to keep our body states right up to date, ready for the next challenge. In Chapter 3, we will see how the information flow is involved in our development to make us who we are – development that continues all the way from the fertilised egg to the moment of death.

https://doi.org/10.1017/9781108566520.003 Published online by Cambridge University Press

3

Genes and Environments in Human Development

O

ne of the problems that happens when we read about ‘information flow’ from the genes is that it gives the impression that the information is flowing right now, today. As we’ve already seen, that’s certainly true for how the body reacts to different environments today – too much stress, too little sleep, lots of exercise and so forth. But when it comes to the role of genetic involvement in the many differences that we observe in human behaviour, then most of the time what we should be thinking about is our early development, especially our 9 months as a fetus and then the first few years of postnatal life. The way that genetic information is integrated with all the other bits of information flowing over that time leaves us with a lifetime of personality and other differences from those around us. Our early development describes the ways in which the ‘genotype’ – all that information stored in the DNA – makes a vital contribution to the ‘phenotype’ – what we look like and who we are right now.

3.1 Fetal Development Development begins with fertilisation, which in turn begins with a huge race. A few hundred million sperm start swimming vigorously towards the uterus. Each one has a head 5 μm across (0.0002 inch), with a wiggly tail 50 μm long (0.002 inch), the head being around twenty times smaller than a human egg, and the wriggling sperm looking a bit like a tadpole, except that a tadpole is approximately 1,000 times bigger than a sperm. You can easily see a tadpole, or even a human egg, with the naked eye, but you can’t see an individual sperm without a microscope. The attrition rate in the big race is remarkably high. Out of 200 million sperm that start the race, only around 1 million make it into the uterus (womb), most being killed along the way by acid secretions, or by a flow 36

https://doi.org/10.1017/9781108566520.004 Published online by Cambridge University Press

3.1 FETAL DEVELOPMENT

37

that pushes them backwards. Of these, only 10,000 make it to the top of the uterus, many having been attacked by the white blood cells that defend our bodies against foreign invaders. The sperm then have the agonising choice as to which of the two fallopian tubes to swim up, because an egg has been released only at the end of one and not the other, and only 1,000 sperm swim up each, with around 200 actually reaching the egg in one of the tubes. The sperm can smell, and molecules with an alluring smell coming from the egg give them motivation as they swim against the flow. Upon their arrival upstream, they furiously seek to impregnate the egg, but only one can win. The egg is protected by a jelly-like membrane, but the sperm inject enzymes that help them to break their way in. Once there’s a winner, the egg immediately creates a hard outer capsule so that no more sperm can enter. Inside the egg, the sperm delivers its precious cargo of twentythree chromosomes from the father where their destiny is to pair up with the twenty-three chromosomes from the mother. The egg, meanwhile, represents a huge packet of information compared with the tiny sperm, and that’s not just genetic information. The human egg just prior to fertilisation contains at least 3,000 different proteins, 7,500 different mRNA molecules and many thousands more small non-coding RNA molecules involved in regulating gene expression. It represents a hugely complex system – that’s what we inherit from our mothers. By itself, DNA would be as useless as a piece of software without a computer to run it on. The egg’s X chromosome DNA also contains far more genetic information than the little sperm, around 800–900 protein-coding genes in all. The sperm contains either an X or a Y chromosome. If the egg is fertilised by a sperm containing a Y chromosome, then a male embryo develops (XY), but if the sperm contains an X chromosome, then it’s a female embryo that will develop (XX). The Y chromosome contains only 50–60 protein-coding genes, but one of them, the SRY gene, is rather important as it triggers male development. The fertilised egg is known as a ‘zygote’, and in the early zygote, it’s not the DNA that causes development to begin but rather the proteins inherited from the mother’s egg, which regulate which genes in the DNA are switched on and off. It takes a couple of days before the new DNA inherited from both parents begins to generate the proteins that ensure its own continuing regulation. Hundreds of genes and proteins are involved in this transition from the life of the egg to the life of the zygote, with rapid cell replication now organised by the new genomic system generated by the pairing of the chromosomes from the mother and father. Proteins are the players in the

https://doi.org/10.1017/9781108566520.004 Published online by Cambridge University Press

38

3 GENES AND ENVIRONMENTS IN HUMAN DEVELOPMENT

DNA orchestra that cause the genes to play an integrated symphony of life. Causal networks operate in all directions. Around 6–12 days after fertilisation, the growing embryo, now known as a blastocyst, implants in the wall of the uterus. The blastocyst is a hollow ball of cells, with a special group of cells, known as the inner cell mass, at one end. These cells, known as stem cells, form most of the tissues that will generate the baby; the other cells in the blastocyst are largely devoted to producing the special support tissues that allow placental mammals such as humans to attach to the mother’s uterus. Once implanted, the cells derived from the inner cell mass flatten to form the embryonic disc, around 1 mm in diameter, with its outer surface layer of cells known as the ectoderm. Over the next few days, the ectoderm begins to receive chemical signals from nearby cells in the embryo that cause it to form the ‘neural plate’. As the embryo continues to grow, the edges of the neural plate curl around to form a tube, which ultimately becomes the nervous system with the brain at one end and the spinal cord at the other. At this stage, the very early nervous system contains around 125,000 cells, but these are not neurons, mature brain cells, but immature precursor cells that have the potential to develop into either neurons or glial cells (a crucial cellular supporting cast within the central nervous system). Once the tiny embryo is hooked up to the mother’s bloodstream via the placenta, all the nutrients and oxygen in the mother’s blood are made available to this rapidly growing bundle of cells. At this time, negative environmental inputs, such as alcohol, nicotine, drugs or the wrong medications, can have a profound effect on the growing embryo in ways that can give the eventual child lifelong learning difficulties or cause other more subtle behavioural differences. At this stage, it is worth pausing to jump ahead to consider the structure of the fully developed adult brain, as we are then in a better position to appreciate the dramatic developmental transitions that lead from a mere 125,000 precursor cells to the 100 billion neurons (1011) and 1 trillion glial cells (1012). If we take the current world population as 7.5 billion, this means that you have enough neurons and glial cells to give each living human around 13 and 130 of them, respectively. Neurons are the brain cells that make up the brain’s signalling networks. This is what people are referring to when they talk of the brain’s ‘hard wiring’, not a great metaphor because it implies fixed outputs from a hard-wired system, which is very far from being the case. Neurons communicate with other neurons by means of small connecting units known as ‘synapses’. The synapses involve a small gap, so each neuron doesn’t actually touch the next neuron but rather

https://doi.org/10.1017/9781108566520.004 Published online by Cambridge University Press

3.1 FETAL DEVELOPMENT

39

communicates using chemicals (‘neurotransmitters’) that diffuse across the gap. Glial cells provide support and renewal for the neurons, giving our neurons a good clean-up every day, mostly when we’re asleep, but they also engage in some forms of signalling. Each neuron in the adult brain receives, on average, 5,000 synaptic connections, although the range is very broad, and some neurons connect up to other neurons by as many as 200,000 synapses. As there are around 100 billion neurons in the adult brain, this implies that the brain contains a staggering 500 trillion synapses (5 × 1014), making the human brain the most complex entity in the known universe. The transition from a 1 mm diameter neural plate to a brain with 500 trillion connections provides plenty of opportunities for the integration of genetic and environmental information. In fact, most of these 500 trillion connections are made in the immediate postnatal years in response to environmental inputs, as will be described below. But in the early stages, how do neurons in the developing fetal brain ‘know’ how to connect up with the millions of other neurons being brought into existence each day of a pregnancy? As already noted, the human genome encodes a paltry 20,500 proteins, although the final number of functionally distinct proteins may well be more than 1 million, based on the mechanisms already described in Chapter 2. Even so, there is clearly insufficient information in the genome per se to specify the billions of specific synaptic connections that characterise the architecture of the mature brain. Many details have yet to be worked out, but the broad outlines as to how this happens are now clear. In the early months of fetal development, genetic information, in coordination with input from the fetal microenvironments, is dominant in terms of establishing the broad structures and areas of the brain. But increasingly during the final 3 months of pregnancy, at a time when the fetus can hear and feel, and even more so in the immediate postnatal years, it is the immediate environment of the newborn that plays a critical role in constructing the developing brain, particularly in specifying its synaptic architecture. The early-to-middle months of fetal brain development lay down intrinsic ‘learning structures’ that are then primed to respond to environmental inputs at certain later sensitive periods. These learning structures in turn then lead to the next stage in the development of specific areas. Over the 18–23-week gestational period, 76 per cent of our protein-coding genes are expressed in the fetal brain and 44 per cent are dynamically regulated – not just permanently in the ‘on’ position (Johnson et al., 2009). Genetic information and environmental information are integrated at every stage.

https://doi.org/10.1017/9781108566520.004 Published online by Cambridge University Press

40

3 GENES AND ENVIRONMENTS IN HUMAN DEVELOPMENT

So the focus during embryonic brain development is on getting the general structures and the links between them set up correctly, but with genetic variation among individuals making a difference all the way along. In fact, so great are the individual differences in brain structure and function by adulthood that a research team was able to identify individuals out of a group with more than 90 per cent accuracy simply by inspection of a map of brain regions that are active during mental activity, giving a new meaning to the term ‘fingerprint’ (Finn et al., 2015). By around 1 month after fertilisation, the nascent embryonic nervous system is growing at breathless speed: about 250,000 new cells are being generated every minute during the first half of the pregnancy. Even at 6 months of gestation, the surface of the fetal brain is quite smooth, but in the final trimester the surface begins to wrinkle in order to pack in a larger surface area – the cerebral cortex that we use for thinking, learning and the higher functions that make us who we are. If you spread out the surface of the adult brain, it would cover one or two pages of a broadsheet newspaper, much bigger than our heads, but the wrinkles (‘folds’ and ‘grooves’) allow it to pack its computing power into the constrained dimensions of our skulls. Sadly, it’s when bad things happen during pregnancy that we often learn a lot about how environmental inputs can influence behavioural traits in the growing postnatal human. In general, disruption of neural development in the first trimester is more deleterious than in the third trimester. Fetuses in their tenth to twentieth week of gestation suffered severe brain development problems following the dropping of the atom bomb on Hiroshima, whereas such problems were less prevalent in fetuses that were at an earlier or later stage of development (Klingberg, 2013). Ingestion of certain drugs, excessive alcohol or smoking by the pregnant mother, or severe inflammatory responses, can all impact negatively on the emerging fetal neuronal system in ways that affect later behavioural development. As is true for radiation, there are windows of sensitivity during which exposure to such agents, called teratogens, is particularly damaging to the fetus. Drugs easily cross the placental barrier and are present also in the maternal milk, so can affect the development of the child pre- as well as postnatally. The effects of prenatal drug exposure are long lasting and persist until adulthood (Šlamberová, 2012). In fact, three generations have to be taken into consideration when considering drugs taken by pregnant women: first the woman herself, second her offspring, and third the fetal germ cells that may be modified by DNA mutations or epigenetically in such a way as to affect her grandchildren (Escher and Robotti, 2019).

https://doi.org/10.1017/9781108566520.004 Published online by Cambridge University Press

3.1 FETAL DEVELOPMENT

41

Pregnancies and early postnatal development in poverty have measurable effects on children’s brain structure and function. Here, we are not referring to poverty as defined in higher-income nations, when there is nearly always some adequate nutrition during pregnancy, but to the kind of grinding poverty that one finds in the slums of Dhaka, Bangladesh, where children play around open canals of sewage and where an ongoing study funded by the Bill & Melinda Gates Foundation is making some striking observations (Storrs, 2017). Babies with stunted growth are the focus of attention, having smaller volumes of grey matter than non-stunted babies, a condition associated with worse scores on language and visual-memory tests at 6 months old. In turn, these changes correlate with differences in brain imaging results (Jensen et al., 2019a). Sorting out cause and effect is notoriously challenging in such studies, but in the Dhaka research, malnutrition, inflammation and caregiving have all been identified as playing important environmental roles (Jensen et al., 2019b). The possible effects of ‘maternal distress’, a general term that includes a wide range of psychological stressors such as anxiety and depression, on the developing fetal nervous system has received extensive attention, with many thousands of publications describing such effects. The Avon Longitudinal Study of Parents and Children (ALSPAC) continues to be of particular relevance in such studies (Pearson, 2012) and has already led to more than 200 research publications. In 1990, researchers started collecting specimens together with social and medical data from 14,500 pregnant women in the Avon area of the western UK, tracking the subsequent health and sociological characteristics of their offspring, a study that remains ongoing. One of the many findings from the ALSPAC study was a higher incidence of antisocial behaviour and higher anxiety at ages 4 and 7 years for prenatally stressed children after controlling for obstetric risks, psychosocial disadvantage, and postnatal anxiety and depression (O’Connor et al., 2003). The fact that the effect persisted between the ages of 4 and 7 years suggests that the influence remains unaltered by changing life stresses on the child, as this period includes the beginning of formal schooling. An obvious explanation for such findings could be a genetic profile in which mothers genetically predisposed to anxiety or depression give birth to offspring with similar characteristics. It is therefore significant that increased anxiety and antisocial behaviour at age 4–10 years has also been noted in prenatally stressed children born via in vitro fertilisation (IVF) treatment (Rice et al., 2010). This IVF study examined children carried by their genetic mother and compared them with children born via egg and embryo donation, who share a uterine environment but no genetic link with the mother.

https://doi.org/10.1017/9781108566520.004 Published online by Cambridge University Press

42

3 GENES AND ENVIRONMENTS IN HUMAN DEVELOPMENT

There was no difference in the association between maternal stress and child outcome depending on whether or not the child was genetically related to the mother, suggesting that it is the environmental stress during fetal development that plays a key role in this instance. In all such studies, there is always the proviso that correlation does not equate with causation, and even when other confounding factors have been excluded by statistical analysis of the data, it is difficult to exclude the possibility that other causal factors, known or unknown, may be involved. For example, malnutrition is known to affect fetal brain structure and function in a manner analogous to prenatal stress, and maternal malnutrition during pregnancy is often associated with anxiety and depression, so in some maternal stress studies the stress may be acting as a proxy for malnutrition (Monk et al., 2013). As always, there is a danger that the media will exaggerate the risks of negative postnatal outcomes as a consequence of the status of the pregnant mother, forgetting the myriad other factors that impact on the status of the newborn: paternal sperm quality, genetic variation, social expectations and traditional practices, poverty and much else besides. The following media headlines on this topic have all been recorded: ‘Mother’s diet during pregnancy alters baby’s DNA’,1 ‘Grandma’s experiences leave a mark on your genes’2 and ‘Pregnant 9/11 survivors transmitted trauma to their children.’3 The relatively modest impact of the great majority of prenatal environmental influences therefore needs to be highlighted. But overall, the impact of environmental factors on the developing fetal neuronal system provides many good illustrations of the ways in which environmental and genetic inputs integrate to form the human individual. Many different complementary narratives can be told to describe this process. There is the narrative of maternal choice: whether to smoke and drink excess alcohol during pregnancy, or not. There is a narrative in which a couple splits during the pregnancy, leading in turn to maternal distress and a greater probability of negative behavioural and psychiatric outcomes in the later life of the offspring. There is a narrative of war or natural disaster in which maternal stress is again mediated to the fetus through no fault of the mother. And another complementary narrative, to be considered further below, includes the array of epigenetic changes, switching genes on and off in response to the latest maternal messages, often in ways that lead to long-term epigenetic differences in the offspring. The important point is that none of these narratives, and indeed the many others that could be told, is in any kind of rivalry with any other, even though the narratives belong to very different disciplines that use different research methods in their investigations. At

https://doi.org/10.1017/9781108566520.004 Published online by Cambridge University Press

3.2 POSTNATAL DEVELOPMENT

43

every level of discourse – be it sociological, anatomical, hormonal, genetic or epigenetic – there is complete integration of signals and influences, so that differences in fetal development among individuals become like the thousand different nuanced ways in which a Beethoven symphony can be performed depending on how the notes on the page are interpreted using different instruments and conductors. Mentioning Beethoven is a reminder to say that by the third trimester, the fetus can both hear and feel. Development of the auditory system begins very early in gestation. All the major structures of the ear required for hearing are in place by 23–25 weeks and by 26 weeks the fetus can perceive sound and react to it. From 26–30 weeks onwards, the hair cells in the cochlea (the auditory portion of the inner ear) are fine-tuned for specific frequencies and can convert acoustic stimuli into neuronal signals that are sent to the brain. After 30 weeks, the auditory system is now developed sufficiently to respond to music and to distinguish between different speech sounds, presumably representing the very beginnings of language acquisition. Studies on preterm babies (gestational age in the range 28–32 weeks) have revealed that, even during this period, when neurons are still migrating to their assigned locations and the cortex construction is only just beginning, subtle differences in human speech syllables can already be discriminated, a finding that underlines the innate capacity of the brain to process auditory information before any significant opportunity for learning becomes available (Mahmoudzadeh et al., 2013). Playing music to preterm babies in neonatal intensive care units had beneficial effects on the development of a normal brain architecture, more similar to that of full-term newborns (Lordier et al., 2019). Shortly after birth, neonates are able to distinguish between the mother’s voice (which they prefer) and a strange female voice, between the mother’s voice and the father’s voice, and between the mother’s language and a foreign language, showing preference for the mother speaking her native language (McMahon et al., 2012). Unfortunately, however, earlier claims that playing Mozart or other uplifting music to the fetus might be beneficial for its subsequent postnatal IQ or general well-being have not been confirmed.

3.2 Postnatal Development At birth, the human newborn is bombarded with a whole new array of sights, sounds, smells, touch, language, people, dogs, cats and other fascinating environmental phenomena. The newborn also accumulates a wonderful range of bugs – billions of bacteria, viruses and fungi – that make quite a big difference to future health (Charbonneau et al., 2016). Some are

https://doi.org/10.1017/9781108566520.004 Published online by Cambridge University Press

44

3 GENES AND ENVIRONMENTS IN HUMAN DEVELOPMENT

encountered by the baby as it passes down the vagina during birth. Others come from the mother’s skin during breastfeeding and yet others from the mother’s milk, not to mention cuddles with the family’s cat or dog. Many of the bugs are beneficial for future health, and premature babies born by caesarean section therefore miss out, especially if no breastfeeding follows. Babies born prematurely are more liable to develop a nasty inflammation of the gut, in some cases leading to surgery to remove part of the intestines, an intervention that can have lifelong consequences. Breast milk appears to give some protection against this inflammation, possibly by containing chemicals that preferentially help the good bugs to thrive in competition with the ‘bad’ bugs (DeWeerdt, 2018). But leaving aside the rather special case of our bodily collections of bugs, different environments in interaction with the genetic information found in the newborn are as important postnatally as they are prenatally, especially when it comes to the development of the newborn’s brain. The infant brain is not a miniature version of the adult brain but a continuously self-organising system that only self-assembles correctly if the right environmental inputs are available at the right time. The physical development of the infant brain is critically dependent upon such exposure. As already mentioned, the adult brain contains around 1011 neurons, but in total twice as many brain cells are generated in the developing fetal brain, with half of these being pruned away during development. At birth, many of the fetal brain cells are still immature brain cells that have yet to develop into mature neurons. It is the experiences of the growing baby that help shape the synaptic architecture of its brain. This is reflected rather dramatically in the average weight of the human brain, which triples during the first year of life from 300 to 900 g. After this, the weight increases more slowly to reach 90 per cent of its adult weight (average 1400 g for men and 1250 g for women) by the age of 5 years. This increase is not mainly due to an increase in the number of cells but rather reflects the growth in number of synaptic connections and other changes in the brain cells as they mature. The synaptic branching process continues right into adolescence in some brain areas, albeit at a lower rate than in the newborn, and in fact never totally stops, even in the adult, depending on the part of the brain being examined (Dahl et al., 2018). The synaptic architecture of the brain develops according to the principle of ‘use it or lose it’. Without neuronal activation, brain cells die by a process known as ‘programmed cell death’. Around half of all the brain cells that are formed in the fetus are pruned away as the brain matures. The pruning involves the failure of these cells to generate efficient

https://doi.org/10.1017/9781108566520.004 Published online by Cambridge University Press

3.2 POSTNATAL DEVELOPMENT

45

synaptic connections in comparison with their neighbours. The neurons that continue to flourish are those that receive strong signals via their growing collection of synaptic connections, and among these signals are those that counteract the cell death programme. So the neuronal architecture develops by a process akin to natural selection in which the ‘successful neurons’ are selected in comparison with the ‘unsuccessful neurons’. The role of the environment in providing the signals that help construct the development of the postnatal brain is well illustrated by the existence of sensitive periods at which sensory inputs are vital for the development of specific brain areas. In early infancy, these affect sight, hearing and balance. Miss the environmental input at the critical time and there are long-lasting effects, some of which may remain with the growing individual for the rest of their lives. Face recognition and discrimination provide some of the earliest postnatal environmental inputs. Newborns prefer face-like stimuli to non-face-like stimuli, prefer familiar over unfamiliar faces, and attend more to attractive over unattractive faces. At 3–4 months of age, infants respond to gender information in human faces. Specifically, young infants display a visual preference towards female over male faces, if raised by females but towards male over female if raised by males. Early visual impairment due to cataracts leads to persistent deficits in facial recognition, even when the cataracts are removed in the first months of life. The window of opportunity for learning how to discriminate faces is a narrow one. For example, 6-month-old babies show recognition memory for different human faces as well as for different monkey faces. However, such recognition memory for individual monkey faces disappears by 9 months of age unless experience with such faces is provided (Pascalis et al., 2011). In fact, monkeys selectively exposed to human faces can only discriminate human faces, not monkey faces, and monkeys selectively exposed to monkey faces can only discriminate monkey faces, not human faces (Sugita, 2008). Taken together, the data suggest that experiences affect infants’ face processing by the age of 3 months, and it is these experiences that sculpt their subsequent face-recognition abilities. This early competence is then refined further during child development and on into adulthood (Pascalis et al., 2011). As a reviewer of this field suggests: ‘Although biological factors may initially play some role in biasing the newborn’s visual system toward faces in their environment, the existing evidence overwhelmingly suggests the important role of experience in the development of face-processing expertise’ (Pascalis et al., 2011).

https://doi.org/10.1017/9781108566520.004 Published online by Cambridge University Press

46

3 GENES AND ENVIRONMENTS IN HUMAN DEVELOPMENT

Overall, it is the construction of the synaptic architecture of the visual cortex at the back of the brain in response to visual inputs that has been most thoroughly investigated. If a baby is blind from birth, then the visual cortex does not develop normally and permanent visual impairment can occur. In fact, even with normal development, the size of the primary visual cortex in adults varies considerably. There are a whole series of ‘sensitive periods’ during which visual inputs are required for the neuronal construction of the developing visual cortex. In the human newborn, binocular vision is absent at birth and is switched on when a baby is 4 months old: visual experience is key for its onset (Birch, 2012). Visual acuity – the ability to discriminate between, for example, stripes of different sizes – is significantly reduced in newborns with cataracts in both eyes following their removal and optical correction of vision at various times during the first 9 months of life (Lewis and Maurer, 2005). However, even after 1 hour of patterned visual input, there was an improvement in acuity, which improved with further visual input, but acuity still fell below the normal range by 2 years of age, and again when measured in the age range of 5–18 years. So a deficit in visual input during the critical first 9 months appears to cause permanent damage to normal visual development. However, if cataracts were removed by 10 days of age, then acuity was improved and, in some cases, normal acuity was eventually attained with increasing age. Those who are deaf from birth likewise undergo marked changes in their auditory cortex (located on the sides of the brain). Brain imaging and postmortem studies in adults who have been congenitally deaf from birth have revealed that neurons that would normally be restricted to the visual cortex can also be found in the auditory cortex (Linden, 2007). In normal development a few neurons may stray over into the auditory cortex, but they are gradually eliminated. But in the congenitally deaf, such neurons not only stay but also sprout new synaptic connections. It seems that if there is a lack of auditory inputs, then neighbouring sensory areas in the brain take advantage of the deficit, a process helped in this case by the withering away of the auditory neurons by disuse. Conversely, in those who are congenitally blind, much of the visual cortex responds strongly to auditory and tactile inputs rather than to visual stimuli. The sensory regions of the brain display plasticity – the ability to adapt in their development depending on what inputs are, or are not, being received. Eventually, however, this plasticity declines or is lost as the brain matures, although as we shall see below, in other respects the brain remains plastic – able to change – throughout adult life.

https://doi.org/10.1017/9781108566520.004 Published online by Cambridge University Press

3.2 POSTNATAL DEVELOPMENT

47

The importance of sensitive periods of brain development can be seen most strikingly, and at the same time most sadly, in infants who have suffered severe emotional and physical deprivation. The largest and most informative recent studies on such cohorts come from Romanian orphans. Romania’s orphan crisis began in 1965 when Nicolae Ceaușescu took over as leader. Over the next 24 years of his communist leadership, Ceaușescu deliberately adopted a policy of increasing the orphan population, banning contraception and abortions, with the aim of creating a population loyal to, and dependent upon, the State. By the time revolutionaries executed Ceaușescu and his wife by firing squad on Christmas Day 1989, an estimated 170,000 orphans were living in more than 700 state orphanages. In the year 2000, Charles Nelson, a neuroscientist from Harvard University, launched the Bucharest Early Intervention Project in which his research team enrolled 136 institutionalised children ranging in age from 6 to 31 months, of whom half were randomly placed in foster care and half remained in state orphanages, subsequently tracking the physical, psychological and neurological development in both groups (Nelson et al., 2013). A third group was also recruited who had never been institutionalised. As virtually no foster care was available in Bucharest when the study started, the research team had to recruit a whole new cohort of foster homes. Since that time, many papers have been published describing their findings. The results are clear. Provided that the children were placed in foster homes by the age of 2 years, significant gains could then be measured in psychological development, IQ and motor skills compared with the control group in the state orphanages. At 30, 40 and 52 months, the average IQ of the institutionalised group was in the low to middle 70s, whereas it was about 10 points higher for children in foster care and about 100, the standard average, for the group that had never been institutionalised (Nelson et al., 2013). At around the age of 8 years, the children who grew up in institutions were found to have less white matter in their brains (the myelinated axonal sheaths involved in connecting up neurons) compared with those in foster care. This may explain why electroencephalographic (EEG) brain measurements in 8–16-year-olds likewise revealed distinct traces in those institutionalised up to that age, whereas those placed in foster homes before the age of 2 years displayed EEG traces indistinguishable from the children who had never been institutionalised (Vanderwert et al., 2016, Debnath et al., 2019). Once again, as with prenatal findings, but now with postnatal development even more dominated by environmental inputs, the various narratives that can be told at different explanatory levels are all complementary in nature and thoroughly integrated in the developing infant. Political decisions

https://doi.org/10.1017/9781108566520.004 Published online by Cambridge University Press

48

3 GENES AND ENVIRONMENTS IN HUMAN DEVELOPMENT

by a communist leader in Romania, failure of care for young infants, cataracts in both eyes, failure of normal synaptic connectivity in key areas of the brain, genetic variation that contributes to more or less resistance to stress – these and many other levels of understanding all provide complementary accounts and all are required and important to gain a full appreciation of the complexity of human development.

3.3 Adult Development Development never ends but continues throughout adult life until it finally ceases in death. Genetic information is changing as mutations take place in different types of cell, some leading to disease. Human decisions change the environment of the growing individual in ways that impact on gene expression. Diet, smoking, exercise, drug taking, alcohol use, stress, choices about where we live, the daily choice of lonely environments or sociable environments, and many other factors epigenetically modify our genomes in ways that change not only us but potentially also our children and perhaps, as noted in Chapter 2, even our children’s children – a sobering thought. The human individual seen from a biological perspective is a dynamic ever-adapting, ever-changing organism. This is seen perhaps most strikingly in the brain. Time was when it was thought that at birth we were endowed with a fixed number of neurons and a predetermined set of synaptic connections, and that after that, ageing entailed an inevitable loss of neurons without any possibility of renewal. The past few decades of brain research have completely transformed such a picture. The hippocampus provides a striking example. It’s actually formed of two sizeable lobes tucked away on either side of our brain and is involved especially in learning, memory and emotion. Its early development is similar to other brain areas: an early prenatal increase in the number of neurons, and then a postnatal proliferation of synaptic connections, reaching a maximum by the age of 2 years, followed by a process of pruning completed by around the age of 5. The hippocampus is the one brain area where new neurons are now known to be produced throughout life, a process known as adult neurogenesis (Christian et al., 2014). For many years, the whole idea was controversial, but new techniques have now shown quite clearly that neurogenesis continues well into adulthood, even in people in their ninth decade, but drops off sharply in patients with Alzheimer’s disease (Boldrini et al., 2018, MorenoJiménez et al., 2019). Adult humans add around 700 new neurons to their hippocampus each day, corresponding to an annual turnover of 1.75 per cent of the renewable hippocampal cell population, followed by a modest decline

https://doi.org/10.1017/9781108566520.004 Published online by Cambridge University Press

3.3 ADULT DEVELOPMENT

49

during ageing (Spalding et al., 2013). The new neurons connect to those already present and appear to have special functions, distinct from those already present, that contribute to brain functions (Christian et al., 2014). Animal studies have implicated adult neurogenesis in long-term memory. Intensive learning in animals impacts on the survival of the newly formed neurons. An interesting theory suggests (with some experimental support) that newly formed neurons help to record memories in a way that distinguishes them from earlier memories already laid down (Kheirbek and Hen, 2014). Experience, mood, behavioural states and antidepressants all impact on the regulation of adult neurogenesis in the hippocampus. Learning and memory provide further striking examples of brain plasticity. Memory is a multi-stage process that initially requires consolidation of the memory in the hippocampus. Once this process is complete, memories are then downloaded to the cortex in a process known as systems consolidation. Central to long-term memory in the cortex is the potentiation of neuronal circuits by modulating the synaptic inputs that a neuron receives so that it becomes more likely to ‘fire’. So at any given moment during the day, some of our synaptic connections are physically weakening, whereas others are becoming stronger. The goal of every teacher and lecturer should be to change their students’ brains (for the good) by the end of class. It is then up to the student to reinforce those synaptic connections involved in memory by effective revision. Some striking differences have been reported in various brain regions in cohorts of adults whose hobbies or professions entail that they engage in various tasks, much of the data being obtained by means of various brain imaging techniques. A group of London taxi drivers was found to have enlargement of the posterior regions of their hippocampi in comparison with a relevant control group, an increase that correlated with the amount of time they had spent driving taxis (Maguire et al., 2000), a finding confirmed in a further study, which also showed that the changes were not observed in London bus drivers (Wang et al., 2015). Learning how to juggle also appears to impact on brain anatomy. A group of people who had never juggled before were divided into two groups and one group was taught how to juggle, involving daily practice over a period of 3 months, before having a rest period of a further 3 months during which they did no juggling. Compared with the control group, the jugglers displayed expansion of grey matter in two distinct brain areas, as revealed by brain imaging, but this expansion was already declining following the 3-month rest period, underlying the plasticity of brain structures in response to environmental inputs (Draganski et al., 2004). Playing a musical

https://doi.org/10.1017/9781108566520.004 Published online by Cambridge University Press

50

3 GENES AND ENVIRONMENTS IN HUMAN DEVELOPMENT

instrument has likewise been associated with specific changes in brain anatomy (Hudziak et al., 2014), as has learning a second language (Legault et al., 2019). Many other examples could be given, but the take-home message is clear: development continues throughout adulthood, and much of its direction is in our hands. By means of this brief survey of human development, it should now be clear why the dichotomous language of ‘nature’ and ‘nurture’ is completely inadequate as a way of understanding human identity. Instead, all the environmental and genetic signals are brought together to generate one integrated product – the human person. There is a constant interaction among all the various components of the complex systems involved. Many different stories can be told about these components, but they are complementary, not rival, stories, and we need them all to do justice to the complexity of human life. We will now see how genetic inputs are involved in providing important information as part of these complex systems throughout the process of development.

3.4 How Do Environmental Inputs Integrate with Genetics? Two of the gene information flow systems that are of particular relevance to human development – transcription factors and epigenetics – were introduced in Chapter 2. A common pattern is that environmental inputs regulate transcription factors expression via epigenetic mechanisms, and the transcription factors in turn switch on a complex array of proteins and signalling molecules that are used to generate new items, such as more synaptic connections. Right now, as you read this book – if you have made it this far – I am changing your ‘epigenome’, which is what we call the sum total of all those epigenetic signatures on your DNA. Some of those new signatures will be involved in building new synaptic connections, or in strengthening old ones, in order that memory traces remain once this reading session is over. To give a full account of all such changes that take place in epigenetic and transcription factor regulation during human development would take a whole book in itself, so here just a few examples will be provided from the world of epigenetics. It will be remembered that epigenetic modifications of the DNA can be used for permanent long-term changes that can be passed on from cell to cell during cell replication, or they can be used like a dimmer switch to cause a nearby gene to become more or less switched on. Very early on in fetal development, there are some rather dramatic epigenetic DNA modifications that last for life. For example, as we’ve already noted, if an egg (which

https://doi.org/10.1017/9781108566520.004 Published online by Cambridge University Press

3.4 HOW DO ENVIRONMENTAL INPUTS INTEGRATE WITH GENETICS? 51

contains the X chromosome) is fertilised by a sperm carrying an X chromosome, then the resulting cells that start dividing will be XX and will eventually become a female. But there is no need to have two X chromosomes – in fact, having two risks too much protein of the same type being made, due to doubling of the number of protein-coding X-linked genes. So when the early embryo has only about 32 cells, around 85 per cent of the genes in one of the X chromosomes in each cell are switched off – for life – never to be heard of again. This means that adult female bodies are a mixture of cells (a ‘mosaic’) containing either their mother’s or their father’s X chromosome. Most of the time that doesn’t matter a bit, but occasionally it leads to some less than dramatic outcomes, such as females having patches of skin without any sweat glands or the colour red not appearing as bright to them as in most people. Originally, it was thought that this so-called X chromosome inactivation was complete – all the genes on that chromosome were entirely switched off. However, it is now known that around 15 per cent of the protein-coding genes remain active to some extent, with important consequences for human health in some cases (Ainsworth, 2017). X-chromosome inactivation also explains the patchwork of fur colours in tortoiseshell cats. Some of the cat’s skin cells make fur pigments using the genetic instructions found on the X chromosome inherited from their father, while others use the instructions found on their mother’s X chromosome. So next time you see a tortoiseshell cat, say to yourself ‘epigenetics’, because epigenetic mechanisms play a key role in X chromosome inactivation, along with plenty of other molecular mechanisms (Gendrel and Heard, 2014). Some epigenetic changes can silence gene expression for life, and X-chromosome inactivation provides a good example. In passing, it’s worth highlighting the fact that it was the British geneticist Mary Lyon who first proposed X-chromosome inactivation in 1961 in order to explain her findings on the coat colour of mice carrying sex-linked colour genes (Lyon, 1961). A couple of years later, the process was dubbed ‘Lyonisation’ at a symposium on the topic, but after around the mid-1980s, this term began to fade in the literature and it’s now only used occasionally. Lyon carried out her undergraduate studies in Cambridge at a time when the university admitted only 500 female students, in contrast to more than 5,000 male students. Well into her 80s, Lyon was still active in research a few days a week in her old laboratory at Harwell near Oxford, and soon after her death in 2014, The Genetics Society started awarding the annual Mary Lyon Medal in her honour. I think it’s a pity that the word ‘Lyonisation’ is no longer commonly with us – agreed, the word is a bit of a mouthful, but it’s good to

https://doi.org/10.1017/9781108566520.004 Published online by Cambridge University Press

52

3 GENES AND ENVIRONMENTS IN HUMAN DEVELOPMENT

keep alive the names of those female scientist pioneers who have made so many vital contributions to the advancement of science. Coming back to early fetal development, the permanent epigenetic changes involved in changing stem cells into the specialist cells that make up our bodies have already been mentioned in Chapter 2. Here again, it is the stability of these changes that is the key. Every cell of our bodies, except for the red blood cells that carry oxygen from our lungs to our tissues, contains the whole human genome. So clearly for the more than 220 different cell types that are made during our development, great swathes of DNA information have to be switched off in order for cells take on their specialised functions. If all the genes stayed switched on all the time, then our bodies would be just one giant blob made up of stem cells, which wouldn’t be any fun at all. But the variation in epigenetic signatures is also very striking – not the permanent differences that make cell types what they are but rather the differences that change gene expression transiently or that make a difference to a particular individual during early development so that their future life will be different in some respects from that of their brother or sister. Each human sperm and egg appears to be epigenetically unique, displaying far more epigenetic variation than genetic variation (Petronis, 2010). It is therefore not surprising that epigenetic differences continue to amplify as the pregnancy continues, influenced by differing environmental inputs. A mother’s diet during pregnancy can be sufficient to change the ways in which her baby’s genes are expressed. In one study, the epigenetic changes in the DNA of 237 babies were investigated and found to contain 1,423 regions that were highly variable among individuals (Teh et al., 2014). The researchers estimated that 25 per cent of the differences in the variable regions could be explained by the underlying differences in DNA sequence, whereas 75 per cent of the variation was best explained by the interaction of the DNA with different in utero environments, including maternal smoking, maternal depression, maternal body mass index, infant birth weight, gestational age and birth order. An example illustrating the way in which a pregnant mother’s choices have long-term effects on the epigenetic status of the offspring comes from the ALSPAC study, mentioned earlier in this chapter. Using a cohort of 800 mother–offspring pairs, it was shown that maternal smoking during pregnancy influenced the epigenetic modification of genes involved in fundamental developmental processes (Richmond et al., 2015). The epigenetic differences remained when compared at the ages of 7 and 17 years. A comparison of paternal and maternal smoking with epigenetic changes

https://doi.org/10.1017/9781108566520.004 Published online by Cambridge University Press

3.4 HOW DO ENVIRONMENTAL INPUTS INTEGRATE WITH GENETICS? 53

in the DNA of their offspring showed consistently stronger maternal associations, providing further evidence for causal intrauterine mechanisms. This study represents but one in a large collection of publications in which prenatal smoke exposure has been associated with reduced birth weight, poor developmental and psychological outcomes, and increased risk for diseases and behavioural disorders later in life (Knopik et al., 2012). That the epigenetic differences are not intrinsically determined is made clear by the remarkable differences in epigenetic profiles between identical twins at birth, although the discordance is greater between non-identical than between identical twins (Ollikainen et al., 2010, Gordon et al., 2012). Some of the differences are presumably due to competition for resources within the uterus. Around two-thirds of all identical twins are formed when the embryo is split into two between 5 and 9 days after fertilisation, so the twins end up sharing the same placenta. But around one-third of identical twins are formed when the embryo splits into two before 5 days postfertilisation, so these twins end up each having their own placenta, just like non-identical twins. Different twin pairs will therefore have different experiences in the uterus, such as differences in infection, blood supply and space to grow. By the time of birth, let alone in later life, there are many reasons why identical twins are not truly identical. As individuals, even from the same family, continue to grow, their epigenomes grow apart as they continue to be exposed to different environments. Again, many sad examples come from studies such as those on Romanian orphans. Childhood institutionalisation is associated with a different epigenetic profile. One well-known epigenetic ‘mark’ involves the transfer of chemical methyl groups on to one of the DNA genetic letters, known as cytosine. This epigenetic process is called ‘methylation’ and is often associated with the dimming of expression of particular genes. In children from institutions compared with children who lived with their biological parents, more than 800 differentially methylated genes were detected (Naumova et al., 2012). Individuals who experienced childhood maltreatment and who then subsequently committed suicide in later life have also been reported to have different gene methylation profiles compared with a control group who experienced no childhood maltreatment and who died suddenly from other causes (Lutz and Turecki, 2014). There are many other studies of this kind. Translating such epigenetic differences into a detailed account of relevant gene regulation is an immense challenge, and is helped greatly by animal model studies. Men and women display significant differences in the epigenetic modifications of their genomes, as has been measured in four different large

https://doi.org/10.1017/9781108566520.004 Published online by Cambridge University Press

54

3 GENES AND ENVIRONMENTS IN HUMAN DEVELOPMENT

European cohorts. In fact, 1184 different sites in the genome were shown to be epigenetically different between men and women in this large study (Singmann et al., 2015). What does this mean? Almost certainly, some changes will be causal whereas other will be the result of life-style preferences and other influences. What’s interesting in our present context is just how much our genomes are dynamic systems under constant epigenetic regulation. For example, some rather small changes in diet can have a significant impact on the epigenome. One individual was given a dose of vitamin D3 every 28 days, sufficient to raise his blood level by a modest amount, but this alone was sufficient to change the epigenome at hundreds of sites in the DNA extracted from samples of his white blood cells (Carlberg et al., 2018). This doesn’t mean that we should all rush out and start taking vitamin D pills (although some in less sunny climates, people should think seriously about doing that), but it does provide a vivid example of the way in which the body responds with detailed epigenetic changes to very small changes of environment (Martin and Fry, 2018). Throughout life, the epigenetic clock is ticking, and a fascinating observation is that epigenetic changes in specific regions of the human genome show a 96 per cent correlation with the chronological age of the individual under study. This means that if you’re just given an anonymous small sample of someone’s tissues, you can estimate their age rather accurately simply by measuring the epigenetic modification status of parts of their DNA (Simpkin et al., 2017). But most impressive of all is the sheer diversity of the epigenetic differences among adult human individuals. For example, one study (among many) that examined the human epigenome took 18 different post-mortem tissues from four different individuals and found that out of 26,474,560 DNA sites that are epigenetically modified, no fewer than 4,073,896 (15.4 per cent) were really different between the four individuals (more than 30 per cent different), and this is just the tip of the epigenetic iceberg (Schultz et al., 2015). And it’s human diversity that brings us to our next big question: what is the relationship between genetic variation and human behavioural differences, if any? And how do we find out? Fortunately, diversity at the DNA level isn’t nearly as great as at the epigenetic level, so the question can be answered, but it is a little trickier than might at first appear.

https://doi.org/10.1017/9781108566520.004 Published online by Cambridge University Press

4

What Is Behavioural Genetics?

A

t the dna level, we have 3.2 billion genetic letters in our genomes. How different are we at that level? The answer is: more than you might think. As we have just noted, we are even more different when we start comparing our epigenomes. But in this chapter we will focus on our DNA sequence. In terms of clarifying the DNA differences between us, it’s the dramatic decrease in the price of sequencing the human genome that is changing the scene so quickly. To obtain the first full human genome sequence cost nearly US$3 billion, but now the price is down to less than U$1,000 – and that’s all happened in less than 20 years. Of course, analysing the results so obtained costs a lot more than $1,000, but these comparisons are done by computers. ‘Bioinformatics’ is the name given to this particular discipline. So what have we learnt so far? The draft human genome sequence was first published back in 2001 (Lander et al., 2001) and since then further refinements and corrections have been carried on year after year, a process that continues to the present day. The result is a ‘standard human reference genome’ against which all other genome sequences in the world are then compared. The problem is that the ‘standard’ is based largely on the DNA sequence from one individual. Further work has shown that different major populations in the world have many differences not found in the reference genome. For example, sequencing of the DNA of 910 people of African descent has shown that this population taken together has an ‘extra’ 10 per cent of DNA genetic letter sequences not found in the reference genome (Sherman et al., 2019). The vast majority of these genetic letters don’t represent any protein-coding genes, but, as the authors of the study suggest, in the future it might be best to construct ‘reference genomes’ for every major population of the world. This idea is supported by the finding that in 50,000 non-European individuals, there are twenty-seven variant

55

https://doi.org/10.1017/9781108566520.005 Published online by Cambridge University Press

56

4 WHAT IS BEHAVIOURAL GENETICS?

genomic regions that are not found in the European DNA sequence database (Wojcik et al., 2019). At this moment, it might be good to head off any racist concerns about such studies. From a geographical perspective, people overall are more likely to marry someone from their own part of the world – no surprises there. But this also means that they are more likely to accumulate similar variants in their genomes, because these are passed down from generation to generation in various interesting new mixtures. If this wasn’t the case, then all those DNA companies that like to tell you about your geographical origins wouldn’t have a leg to stand on. By the start of 2019, it was estimated that 26 million people had already added their DNA to the four leading companies offering such information. Unfortunately, some companies tend to exaggerate and overinterpret some of the data (dubbed ‘genetic astrology’ by some1), but clearly some broad inferences are possible regarding your geographical ancestry. Also, it has long been known that certain gene variants relevant to medical genetics tend to be found more frequently in some populations than in others – again, a tendency to have children with someone from a similar geographical area and background is the simple explanation. So the brief summaries that follow should be seen as what they are – work in progress, with numbers that will become more refined as the number of human sequences worldwide goes up from the thousands to the millions. The most common type of variation among individuals is differences in individual genetic letters found at precise points in the genome. These are known as ‘single nucleotide polymorphisms’ – or SNPs, pronounced ‘snips’. ‘Nucleotide’ is the chemical name for what we have been calling a ‘genetic letter’ (see Chapter 2), whereas the genetic term ‘polymorphism’ simply refers to ‘more than one type’ of genetic letter at a particular spot in the genome. We all contain between 4 and 5 million different SNPs in our genome and vary between each other in around one in 1000 of our SNPs.2 There are around 100 million SNPs in people round the world, and the number continues to grow as genome sequence databases become more representative of different human populations. Most SNPs are common, whereas others are rare. The vast majority are not located in protein-coding or regulatory regions of the DNA and so it makes no difference to the phenotype having one genetic letter in a particular spot rather than another. A small minority, however, make a significant difference in causing disease, either the kind of genetic diseases that were introduced in Chapter 1 where a single gene mutation causes disease or, more commonly, when many SNPs

https://doi.org/10.1017/9781108566520.005 Published online by Cambridge University Press

4 WHAT IS BEHAVIOURAL GENETICS?

57

together contribute to the probability that a particular disease, like heart disease or diabetes, may develop. Furthermore, around half a million SNPs or more are found in known regulatory regions of the genome, so the potential to make a difference there is rather significant. Having around 5 million SNPs in our genome sounds like a lot – which it is, as that means 5 million genetic letters that, on average, are different from those of the person next door. But in fact that 5 million is overshadowed in terms of numbers of genetic letters by other types of variation. For example, there are many insertions or deletions of small segments of DNA, generally involving six genetic letters or fewer, known as ‘indels’ – standing for ‘insertions–deletions’. Analysis of the whole-genome sequences from 2,504 individuals from 26 different populations revealed 3.6 million different indels across them all (1000 Genomes Project Consortium, 2015). Another study investigated the amount of variation across more than 60,000 people in their ‘exomes’, meaning in their protein-coding genes that are converted into mRNA, and discovered more than 300,000 indels (Lek et al., 2016), which is a lot when we remember that the exome represents the 1.5 per cent of the human genome that encodes proteins. But all these differences together are dwarfed by so-called ‘structural variants’, meaning chunks of sequence difference that are larger, often much larger, than indels. For example, a typical human genome contains up to 2,500 structural variants, amounting to around 20 million different genetic letters (1000 Genomes Project Consortium, 2015), although of course some of these could be the same in different individuals. So if we add that to the 5 million different SNPs, plus some indels, it seems that each individual in the world varies, on average, from any other individual by roughly 0.5 per cent of their genome, much more than we realised until recently. This approximate figure of 0.5 per cent is also supported by the finding that, when both chromosome pairs of a single individual were sequenced (remembering that each set of twenty-three chromosomes contains the full genome, 3.2 billion genetic letters long), the two chromosomes were found to be about 0.5 per cent different in their sequence (Levy et al., 2007). In this context, it’s worth remembering that one chromosome of the pair is from the mother and one from the father, and therefore from different ancestors. It’s also the case that the mother’s and father’s chromosomes swap some DNA segments before they end up in the egg or sperm. So the 0.5 per cent difference is a somewhat approximate estimate of the differences between the genomes of the two parents. Where does all the genetic variation come from in the human population? The answer is: as a result of many different mechanisms. Most mutations

https://doi.org/10.1017/9781108566520.005 Published online by Cambridge University Press

58

4 WHAT IS BEHAVIOURAL GENETICS?

happen as a result of errors creeping in when the DNA replicates every time a cell divides, so much of our genetic variation is simply inherited from our ancestors. Over thousands of years, there have been mutations in the germline DNA of millions of people who have passed on that variation to us. If that wasn’t the case, we’d all exist like one giant genetically identical clone of humans and life would be very boring. In fact, through genome sequencing, we now know that every baby born has a whole set of new mutations that are not found in either of their parents. In an Icelandic study, it was estimated that a newborn contained on average 60 new mutations not found in either parent. The exact number varied depending on the age of the father. In this study, whereas the mother contributed on average fifteen new mutations, regardless of her age, a 20-year old father contributed on average twenty-five mutations to his child whereas a 40-year-old father contributed around sixty-five new mutations (Kondrashov, 2012). In a similar Dutch study, 250 family ‘trios’ – mother plus father plus newborn – had their DNA sequenced (Francioli et al., 2015). For each newborn baby in this study, there was an average of thirty-eight mutations. Overall, 78 per cent of the mutations came from the father and the rest from the mother. On average, the offspring born to 40-year-old Dutch fathers had twice as many mutations as those born to 20-year-old fathers. In fact, it has been estimated that paternal age explains about 95 per cent of the variation in the global mutation rate in the human population (Kong et al., 2012). The fact that more mutations come from the fathers than from the mothers is not unexpected because females are born with all the eggs they’ll ever have, whereas sperm is constantly being produced in men right into old age by a series of development steps. By the age of 70, a male’s sperm cells have undergone around 1,400 cell divisions. Each time a cell, along with its DNA, replicates (divides into two new cells), there is the possibility of errors creeping in, so many more mutations are due to replication errors than, for example, the perhaps better-known causes, such as radiation or nasty chemicals in the environment. Thirty-eight mutations in your precious newborn sounds like a lot, but don’t worry too much about that number because in the vast majority of cases those mutations occur in regions of the DNA that are non-functional. Having said that, if you want to be a dad, from a geneticist’s perspective it may be best to start your family earlier rather than later if circumstances allow you any choice in the matter. Having seen how much genetic variation there is among us all, the task of behavioural genetics is now to find out how much difference this makes in our behavioural differences. Some people feel instinctively allergic to this question, as if its very asking poses a threat to our feeling of human freedom.

https://doi.org/10.1017/9781108566520.005 Published online by Cambridge University Press

4.1 GENES AND BEHAVIOUR

59

But there’s really no reason to feel threatened. After all, at least half of our genes are involved in helping to build our brains during early development, as we were considering in the Chapter 3, so if we are 0.5 per cent different from the next person at the level of genetic variation, then it would be surprising if that made no difference at all to the myriad ways in which our brains help mediate our interactions with our environments. Here, we will focus on the main concepts and language involved in the field of behavioural genetics, leaving the details and the maths to the textbooks. In fact, the language used in this field is especially important because, as we’ll see in a moment, one key word in particular unfortunately leads to much confusion.

4.1 Genes and Behaviour The aim of behavioural genetics is to attribute the proportions of variance with regard to a behavioural trait to genetic variation and to environmental influences in a specified population. The mathematical approaches utilised are known as biometrics. A ‘trait’ may refer to any behaviour that varies within a population: a medical trait that results in behavioural differences, personality, intelligence, sexual orientation, religiosity, divorce, number of hours per week watching television – you name it, behavioural geneticists have measured it. Behavioural genetics operates like any other scientific approach by the use of methodological reductionism – if you want to find out how a system works, then start by breaking it down into its components, which can then be measured separately. In this case, the aim is to list all the factors involved in making the trait different and then to treat each factor mathematically to assess its contribution to the variance in a trait in a given population. So the total variance in a population with respect to a particular trait is first broken down into genetic and environmental influences. The environmental component can be further broken down into two subcomponents: common or shared environmental effects, and unique or non-shared environmental effects. The shared environment comprises events that make children raised in the same family more similar to each other and less similar to those who do not share it. Examples include socio-economic status, nutrition and parenting style. Non-shared effects describe influences that are unique to an individual, or which have differential effects on them, such as smoking and drug taking, accidents and psychological trauma, and which make individuals less similar to each other. Non-shared effects tend to affect just one family member.

https://doi.org/10.1017/9781108566520.005 Published online by Cambridge University Press

60

4 WHAT IS BEHAVIOURAL GENETICS?

One of the traditional aims of behavioural genetic studies is to calculate the ‘heritability’ of a particular trait. And here the linguistic challenges begin. For most people, the word ‘heritability’ means ‘what I inherit from my parents’, and the word can indeed mean that. But since 1940, it has been used in a more technical sense in the field of genetics to mean ‘the proportion of variance of a trait in a particular population that can be ascribed to genetic variation in that population’. So the word ‘heritability’ now has two meanings that are quite distinct: the first meaning is about me and my personal genetic inheritance from my ancestors; the second meaning is a population statistic that has no units but which can be expressed as a percentage or as a proportion of 1.0. Unfortunately these two distinct meanings are often confused in media reports, and sometimes scientists are not always astute enough when reporting their genetic discoveries to make clear which meaning they have in mind. This can lead to the ‘gene for’ type of language critiqued in Chapter 1. A scientific report may say that intelligence, for example, has a heritability of 50 per cent, meaning that 50 per cent of the variance of this trait (however measured) in a population can be attributed to genetic variance, but what people understand from the news report is that 50 per cent of their own personal intelligence is inherited from their parents, which is not at all the claim that is being made. The Cambridge biologist Patrick Bateson provides the following illustration to help us (Bateson and Gluckman, 2011). We notice that nearly all people have two legs. When they only have one leg, it is nearly always due to an accident, in other words due to the environment. So the variation in leggedness in the population is 100 per cent environmental and 0 per cent heritable. That sounds odd because everyone knows that you need genes of a certain kind in order to develop legs. It is just that in practice the genomic system responsible for leg building during development nearly always encounters the same environmental context, and so the system builds two legs. Likewise, possession of a brain is heritable in the sense that it is inherited, but in biometrical terms the heritability of having a brain is zero because there is no variation in a population in this particular trait. So we note that heritability tells us nothing about the complex interplay of components described in Chapter 3 that leads to our development as individuals. Heritability is a statistical construct referring to variation in a population, not the interplay between genes and environment in any particular individual. Imagine a population A in which all people can see. The heritability of sightedness in this population must therefore be zero as there is no variation in this population with respect to this trait. But now let us say that we

https://doi.org/10.1017/9781108566520.005 Published online by Cambridge University Press

4.1 GENES AND BEHAVIOUR

61

introduce a few congenitally blind individuals into the population, now called population B. The trait of sightedness must now have a positive heritability value due to the variation. Does this mean that genes are more involved in the development of vision in population B compared with population A? No! A high heritability value likewise does not necessarily imply a high level of genetic influence in a given individual with respect to a particular trait; it simply means that variant genes are playing a particularly significant role in how that trait varies in a particular population (Visscher et al., 2008). Heritability values can also change in a population over time, due, for example, to changing environments or to greater inbreeding, or the influx of other genetic variants due to outbreeding. The traditional methodology for a heritability analysis is a family-based design, although, as we shall see later on, other methods are now becoming more popular with the advent of genome sequencing. Family studies rely on the fact that relatives share both genetic and environmental influences on behavioural traits; the same influences are assumed to affect the trait in the relatives under study. The genome is assumed to be identical throughout life. The level of shared influence affects the proportion of the variance among relatives that can be assigned to either genetic or environmental effects. Relatives living in the same environment are assumed to share 100 per cent of common environmental effects, while those living apart are assumed to share no common effects. By definition, no relatives share unique environmental effects. Comparisons between trait variation in identical compared with non-identical twins remain central to such heritability analysis. In practice, non-identical twins, or any pair of full siblings, share 37–62 per cent of their genes in common (Visscher et al., 2006).3 A component of the variation is due to the different amounts of exchange of chromosomal material (‘crossing over’) that occurs in the parents during the generation of the sex cells – the sperm and the ova. And although identical twins are assumed to be genetically identical, the reality is more complicated as we have already noted when thinking about epigenetic differences between identical twins that are already present at birth. Behavioural genetics uses three main types of family-based design to calculate heritability. The first approach involves the study of monozygotic twins separated at birth and raised in different homes. However, this strategy is not that common, given that the number of twins separated at birth is happily rather small. We shall return later to the meaning of the phrase ‘separated at birth’.

https://doi.org/10.1017/9781108566520.005 Published online by Cambridge University Press

62

4 WHAT IS BEHAVIOURAL GENETICS?

The second more common approach is to take identical twins who have been raised together and compare them with non-identical twins who have also been raised together. There are about 11 million identical twins in the world and more than seventeen twin registries, each containing more than a million twins, so there is plenty of scope for such studies. To give some idea of the scope, one meta-analysis (meaning analysis of all the results published over a certain period) incorporated data published during the years 1958–2012 on the variation of 17,804 human traits based on more than 14 million different twin pairs drawn from thirty-nine different countries.4 The traits ranged from psychiatric and cognitive traits, and specific medical conditions, to social values, religion and spirituality. Virtually all twin studies within this 50-year period were included. Across all traits, the heritability was 49 per cent. The third approach involves adoption study designs, which work in one of two ways: either genetically related siblings living apart from each other are compared, thus providing an estimate of genetic effects, or genetically unrelated adoptee siblings living together are compared, thus providing an estimate of shared environmental effects. The adoption design can also be extended to compare adopted children with their biological and nonbiological parents. Generally, an adoption study compares multiple sets of relatives. Calculating the heritability from such studies is rather straightforward, although checking carefully on all possible complications, with their associated statistical concerns, can itself become quite tricky. Those who like maths are referred to standard textbooks on the subject, such as the book by Robert Plomin and colleagues (Knopik et al., 2017). As far as twin studies are concerned, it’s worth keeping in mind some of the assumptions involved. The basic assumption is that identical twins have identical genomes, which is often true, but there are still plenty of exceptions. There are hundreds of examples in the scientific literature reporting that one of an identical twin pair develops a certain genetic disease (e.g. Huntington’s disease or sickle cell anaemia) but not the other. One reason for this is that after the very early embryo splits to form twins in early fetal development, mutations can subsequently occur in one twin and not the other. Sometimes these differences involve whole DNA segments moving around between different chromosomes or other types of ‘structural variants’. In fact, the lack of concordance between twins with reference to a particular medical condition can be quite useful for researchers trying to identify specific genes involved in that condition, because if a new mutation is in one identical twin and not the other, and only that twin has the illness, then the gene variant

https://doi.org/10.1017/9781108566520.005 Published online by Cambridge University Press

4.1 GENES AND BEHAVIOUR

63

immediately becomes a candidate for further investigation (Vadgama et al., 2019). The epigenetic differences between twins provide another type of difference, as already mentioned. In one study investigating epigenetic differences in forty monozygotic twin pairs ranging in age from 3 to 74 years, closely similar profiles were noted in early life, but in older twins the epigenetic differences between the twins amounted to 35 per cent of the total measured (Fraga et al., 2005). Those twins who had spent less of their adult lives together and/or who had different medical histories were those who showed the greatest differences in epigenetic profiles. Differences in smoking habits, diet, exercise and life events have all been proposed to be involved in such epigenetic differences. Epigenetic differences imply that a gene might be permanently switched off from birth in one twin and not the other, in which case the person might as well not have the gene at all, because it is not expressed. A more likely scenario, however, is that a gene might be partially switched off in one twin and fully expressed in the other twin. Either way, from a functional perspective, the identical twins will not have identical genomes. Sophisticated statistical approaches have been developed to accommodate the finding that identical twin pairs can no longer be assumed to be completely identical at the genetic level, nor at the level of their epigenetic gene regulation. Furthermore, DNA-based statistical methods that do not require twins or any assumptions about them have reached conclusions very similar to those from the classical twin studies (Turkheimer, 2011). In any event, for the moment it remains the case that, taken overall and despite the many provisos, identical twins are more similar to each other at the genetic level than non-identical twins, and therefore their comparison remains a useful method for carrying out behavioural genetic studies, providing the results are assessed with a considerable degree of caution. Twins raised apart is now only of historical interest as, thankfully, that practice is now no longer with us. But data from such studies are still cited. In theory, it sounds like the ideal scenario for investigating the relative roles of genes and environment in behavioural differences. But in fact separation of twins for adoption in reality rarely occurred immediately after birth. For example, in the Minnesota registry of twins raised apart, the average time together before separation was 5 months, but this ranged up to 4 years; the time apart until first reunion ranged from 0.5 to 65 years, and the average total time spent together before the study was over 2 years (Bouchard et al., 1990). In another much-used twin registry of Swedish twins ‘raised apart’, in total 52 per cent of the twins were separated after the age of 1 year, and

https://doi.org/10.1017/9781108566520.005 Published online by Cambridge University Press

64

4 WHAT IS BEHAVIOURAL GENETICS?

18 per cent until not after the age of 5 (Pedersen et al., 1988). These results are representative of all studies of twins reared apart. So in practice, even socalled separated twins lived together first for some time before separation, and it is worth remembering the huge increase in brain synaptic density that occurs over the first 2 years of life – so both twins responded to experiences shared in common during the first year of life. Another challenge for behavioural genetic studies comes in the definition of what we actually mean by the word ‘environment’. Let us say that 30 per cent of the variation in a trait in a particular population is ascribed to the ‘shared’ or ‘non-shared’ environment – what exactly does this mean? It seems fairly obvious until you start thinking about how to define it precisely, and then things start getting difficult. Hanging over all these usages of the word ‘environment’ is the wider question as to where the environment starts and where it ends? The answer to this question depends on what kind of environment we’re thinking of outside of the genome sequence itself. We’ve already seen in Chapter 2 that there’s a type of ‘microenvironment’ surrounding the DNA inside cells, and the molecular editing machinery makes a difference to the DNA information in use, including epigenetic regulation. Now of course in the context of twin studies and the like, it’s the germ-line DNA that is of interest – the transgenerational information passed on in the egg and the sperm – but different cellular microenvironments will make a difference as to how that information is used during early development. Then there is the ‘environment’ between all the cells of the body and the skin that separates us from the outside world. But wait a minute: what about our microbiome, that great mass of around 4 × 1013 bacteria in our bodies, mostly in our guts, roughly equivalent to the number of our own cells (Sender et al., 2016), bacteria that, taken together, contain around 100 microbial genes for each human gene? One gram of human gut tissue contains 60 billion bacteria, around eightfold higher in number than the human population of this planet. It is now quite clear that the precise composition of our microbiomes makes a very significant difference to our health and well-being, and in mice at least, the ways in which gut bacteria influence the brain are now beginning to be understood (Kiraly, 2019). So is that great mass of bacteria inside us part of our ‘environment’? Well certainly from a technical perspective the answer is ‘yes’, because all those bacteria are external to our germ-line DNA, but then the use of language in this case seems a bit odd. And of course we have the world outside our skin and how we interact with it, which is what we usually think about when we think of our ‘environment’. From the biometric perspective, the environment is something different again: that which contributes statistically to a proportion of variance in a population in respect to

https://doi.org/10.1017/9781108566520.005 Published online by Cambridge University Press

4.1 GENES AND BEHAVIOUR

65

a particular trait. The problem is that there is no general ‘theory of the environment’, as there is with genetics, that will help in resolving this issue of definitions. As with many other investigations of human identity, we end up ‘thingifying’ a thing that can’t really be ‘thingified’ (‘reified’ is the more technical term). Talking about the ‘environment’ as a statistical construct contributing to variance in a population makes it sound like a clearly defined thing, but the reality is far more complex. The cynic might suggest that the ‘e’ for environment when measuring the proportion of variance of a trait in a population just stands for ‘everything else’, everything influencing trait variation that does not appear to be attributable to genetic variation. Like everyone else, we will continue using the ‘e word’ in the rest of the book, but it’s just worth remembering that the ‘environment’ is a slippery concept, with many different meanings according to context. Another tricky challenge in interpreting the results from behavioural genetics arises from epistasis – the interactions among genes located at different places in the genome. As we have already noted, the genome acts as a system with all its various parts interacting with each other all the time. The problem in behavioural genetics is that most of the time, such interactions that have longer-term effects in influencing differences in the trait being studied are simply not known. One variant gene may have a synergistic rather than an additive effect in trait development, whereas another may have an inhibitory effect. For example, there is a rare gene mutation present in up to 0.5 per cent of Icelandic and Scandinavian populations that protects against the development of Alzheimer’s disease. Compared with their countrymen who lack the mutant gene, Icelanders with even a single copy of the variant gene are more than five times more likely to reach the age of 85 without Alzheimer’s disease and with a 50 per cent better chance of reaching their eighty-fifth birthday (Callaway, 2012). Without prior knowledge of that valuable variant gene, it might readily have been assumed that epistasis was unimportant in the development of Alzheimer’s disease (this will be discussed more in Chapter 5). This example is more the rule than the exception: the ability of mutations to cause disease depends to a significant extent on their genomic context (Jordan et al., 2015). If gene–gene interactions make for complications, then interactions between variant genes and the environment, often referred to as gene-byenvironment interaction, or G × E, adds another challenge. The G × E acronym can actually include two aspects. The first refers to a gene– environment interaction in which a given environment may have a different influence on an individual depending on their genotype. This is

https://doi.org/10.1017/9781108566520.005 Published online by Cambridge University Press

66

4 WHAT IS BEHAVIOURAL GENETICS?

often observed in siblings. With one sibling, a trauma has a striking and longterm effect, whilst leaving the other relatively unscathed. As we shall note in Chapter 5, one of a pair of identical twins may develop some mental disorder, whereas the other does not. The second aspect of G × E refers to a gene– environment correlation in which those with particular genotypes that contribute to a particular trait are more likely to choose or be chosen for similar environments. Although it may seem counter-intuitive, most forms of G × E interaction increase identical twin similarity more than non-identical twin similarity, thereby increasing the genetic proportion of variance, and so the heritability value, in standard biometric approaches (Burt, 2011). G × E interactions include a genetic predisposition that then shapes the environment in such a way that it impacts in a different way on the individual had there been no predisposition in the first place. For example, a child with athletic ability will be picked out for coaching that might not be available to other children; likewise the musically minded child who plays the violin. So the idea here is that our genomes help shape our environments, which in turn then influence our genomes. In addition to the environment impacting the genome, the genome itself also affects the extent and outcome of that impact. Individuals differ by genotype in the extent to which they are affected, either positively or negatively, by exposure to certain environments. At the end of the day, all the different influences are integrated. But it will be clear by now that giving unambiguous assessment of the various statistically validated ‘proportions of variance’ with respect to a particular trait in a population is no easy matter. Some fancy statistical efforts are now in place to help with that challenge, but it’s not easy. This explains why heritability values are treated by some in the field with a certain degree of scepticism. Probably the most important finding is that the heritability for a certain behavioural trait is positive rather than zero, thereby pointing to a role for genetic variation in making a difference in the prevalence of a certain trait within a certain population. As we shall see in Chapter 5, this fact alone has significant pastoral implications for parents whose child is suffering from a syndrome such as autism. But the actual heritability value, a unit-less proportion of variance, is not that interesting. One reason for this is that the value depends on the particular population and the particular environment in which the trait is being measured. If a population has rather limited genetic variation, perhaps because it is located on an island in which the population has tended to breed with other islanders for many centuries, then the heritability may be relatively small, whereas if a genetically more diverse population was placed in an identical environment, then the heritability value might well be much higher.

https://doi.org/10.1017/9781108566520.005 Published online by Cambridge University Press

4.2 THE GENES THAT INFLUENCE BEHAVIOURAL DIFFERENCES

67

Likewise, if two populations had identical levels of genetic variation but were placed into two different environments, then again the heritability values pertaining to a particular trait might well be different. A higher heritability value does not necessarily mean that genomic contributions to the development of a trait are ‘more important’ than for lower values, although it might do. And a higher heritability value certainly does not mean that the trait in question is more or less amenable to change. It all depends. The conclusions of a very experienced practitioner in the field summarises the situation well: ‘Heritability is greater than zero for all individual differences, and takes a determinate value for none of them. Figuring out how “genetic” traits are, either in absolute terms or relative to each other, is a lost cause: everything is genetic to some extent and nothing is completely so. There is little more to be said’ (Turkheimer and Harden, 2014). Many of these ideas about the field of behavioural genetics will crop up again as we examine specific traits in the chapters that follow. For now, it’s worth noting that at first glance the field of behavioural genetics seems to support the old dichotomous language of ‘nature’ and ‘nurture’, as if on one side we have genetics contributing a certain fixed number to the population variance – heritability – while on the other side we have the environment. So it looks as if genes and the environment are in some kind of competition with bigger or smaller numbers for one or the other telling us which the winner might be. But hopefully by now, and taking into account the developmental biology described in Chapter 3, once we delve further, the nature–nurture or genes–environment dichotomies simply begin to melt away. In reality, everything is interacting with everything else all the time and it’s the interactions that are the most important aspect of the whole process, not the individual components, which are necessary but not sufficient for everything to work together. So all the interactions integrate to produce the final product – human identity. Another way of saying all this – at least when it comes to the major traits that we’ll be considering later, such as personality, intelligence and aggression – is that these are all 100 per cent genetic and 100 per cent environmental. As in any complex system, it is the interactions that count.

4.2 Hunting for the Genes that Influence Behavioural Differences When heritability studies became well established and were gaining respectability just a few decades ago, it was widely assumed that a relatively few

https://doi.org/10.1017/9781108566520.005 Published online by Cambridge University Press

68

4 WHAT IS BEHAVIOURAL GENETICS?

genes would explain the heritability of a given trait. For example, if the heritability for a trait was 50 per cent, then maybe just ten variant genes would explain this, each one counting for an average 5 per cent of the variance. So the hunt for these significant genes started with candidate gene studies. The approach is simple: a gene is identified that is thought to be important in some key aspect of brain function, such as aggression, and then variants of this gene are measured in a random population of individuals. If there is a significant correlation between the presence of the specific variant and the trait being measured, then this particular variant may be cited as contributing to the development of the trait. The choice of candidate genes was based largely on animal studies in which it was shown that certain neurotransmitter genes, for example, together with the genes that encode enzymes that help make these transmitters or break them down once used, play key roles in brain functions. So the approach was perfectly rational in the context of the knowledge available at the time, but it never really worked in terms of replication. And the reason it never worked, as we now know, is that there are hundreds or thousands of variant genes that contribute to a particular complex trait that varies within a population, each one contributing a tiny percentage of the total variation. So with rare exceptions (which we’ll come to later), there simply are no genes that exert a major dominating influence in any given behavioural trait. Sadly, a very large amount of time and research funding was spent on this wild goose chase, but that’s not uncommon in science: one can only proceed with the present knowledge in hand, and if that knowledge is very incomplete (as it was in the present case), then wrong theories will result. The whole field shifted with the introduction of genome-wide association studies (GWAS). This approach capitalises on the millions of SNPs that are present in the human genome. It is now routine to present at least a million DNA SNPs on a dense microarray to a sample of DNA taken from an individual in such a way that the person can be ‘genotyped’ for all those SNPs. So the SNPs act as ‘flags’ to mark different segments of the genome, which vary slightly among individuals. If a particular SNP keeps on associating with a particular trait at levels above chance, then the inference is made that there should be a variant gene nearby, either in a protein-coding or in a gene regulatory region, that contributes to the development of the trait in question. Most commonly, it is the nearest protein-coding gene to the SNP that is identified as being relevant, although this is not invariably the case. If you find it difficult to think about SNPs spread all over your genome, then maybe thinking about a Martian spacecraft landing in the USA might make things easier. Imagine that the USA is covered with hundreds of

https://doi.org/10.1017/9781108566520.005 Published online by Cambridge University Press

4.2 THE GENES THAT INFLUENCE BEHAVIOURAL DIFFERENCES

69

thousands of flags of different colours (the SNPs) scattered randomly over the landscape. You have just arrived from Mars. Your mission is to investigate the main population centres of the USA and you are told that your spaceships should land at the combinations of flags that have been chosen to designate cities. Unfortunately, when you land your vehicle it’s in the middle of a desert in New Mexico, and you have to use your Earth Rover to go off wandering looking for the nearest city. When you finally get to Las Vegas, you find that it only makes a tiny contribution to the whole population. Your Martian colleague, meanwhile, has the luck to land smack bang in the middle of Central Park in New York. The random distribution of flags means that while some are located right in the middle of New York City, none are present in Vegas. That’s not a perfect illustration for GWAS, but highlights the way in which gene hunters can land up in very different places in the genome when the SNP marker flags being used are randomly distributed. An interesting point to note about GWAS is that, unlike the earlier candidate gene studies, they are conducted without any idea as to what the particular genes might be that are involved in a particular trait. In other words, it is basically a fishing exercise. Furthermore, the statistical standards for significance are set extremely high in order to exclude random associations, although even then the overlap in terms of ‘SNP hits’ between different replication studies can be discouragingly low. Increasing the cohort for study into the tens or hundreds of thousands can certainly improve the statistics. There has been an extraordinary growth in GWAS over the past few years. Over the period 2005–2018, there were 3,639 studies on 3,508 different traits (Mills and Rahal, 2019). The ‘poster child’ for GWAS has been based on the fact that the heritability for human height is 80 per cent. This means that 80 per cent of the variation in a population with regard to height can be attributed to gene variation. Now of course measuring height is not part of behavioural genetics, although height does clearly play a role in people’s behaviour. For example, tall people have a greater tendency to end up in professional basketball. And clearly nutrition and other environmental factors play key roles in the average height of a population. For socio-economic reasons, preschool children raised in North Korea are up to 13 cm shorter and up to 7 kg lighter than children who were brought up in South Korea, despite the fact that these populations were one country until the early 1950s (Schwekendiek, 2009). Heritability values say nothing about the average height of a population – indeed, if there were no variance in a population with respect to height, then of course the heritability would be zero. The reason that height has become the ‘poster child’ for GWAS is simply that it’s easy to

https://doi.org/10.1017/9781108566520.005 Published online by Cambridge University Press

70

4 WHAT IS BEHAVIOURAL GENETICS?

measure in a very large number of people, each one of whom is then genotyped. One GWAS investigation of 253,288 individuals of European descent revealed that if the level of statistical significance was set at a very stringent level, then 697 genetic variants were identified that contributed around 20 per cent to the heritability (Wood et al., 2014). If the results of this study were pooled with others, then sufficient SNP variation has been identified that could explain about 50 per cent of the heritability. But it’s clear that the average contribution of each variant gene, considered individually, is tiny, around 0.001 per cent or less. The large number of gene variants per se should not be at all surprising: with increased or decreased height, everything has to change – the size of organs, the expansion of the skin, the construction of the nervous system, and so forth. What is more surprising is the so-called ‘missing heritability’ that is noted even in a study investigating a quarter of a million people. In fact, the value of 50 per cent contribution to variation in height measured by pooling several large studies is at the upper limit of what has generally been achieved so far in the GWAS approach to any trait, be it medical or behavioural. Typically, the relevant SNPs identified explain less than 10 per cent of the variation in a given population with respect to a particular trait. This has led to an extensive discussion in the literature as to what accounts for this so-called ‘missing heritability’. At least for height, the mystery is now being resolved. In one study5, whole-genome sequence data were obtained from 21,620 unrelated individuals of European ancestry and related to their height. It was discovered that around 50 per cent of the heritability could be explained by rare variants that contribute very tiny amounts to the overall trait variation – less than 0.1 per cent of the heritability and often much lower. Whole-genome sequencing enabled these investigators to capture genetic differences found in only one out of 500 people or even one in 5,000 people (Geddes, 2019). The problem with genotyping people, rather than obtaining their wholegenome sequence, is that the DNA chips currently used for genotyping represent only about half a million SNPs that are commonly found in the ‘reference genome’. But as has already been noted, the total number of SNPs in each person anywhere in the world is about tenfold higher than this, so using ‘standard chips’ can easily miss a lot of variants, many of them very rare, which nevertheless can make their contributions to the overall variation in height. Generally, the discovery of ‘rare variants’ seems to be resolving the ‘missing heritability’ problem in a multitude of GWAS data sets. At the same time,

https://doi.org/10.1017/9781108566520.005 Published online by Cambridge University Press

4.2 THE GENES THAT INFLUENCE BEHAVIOURAL DIFFERENCES

71

this is highlighting the incredible number of gene variants that apparently contribute to the variation in a population with respect to a particular human characteristic, be it medical, or a measurement of height or weight, or some behavioural trait such as aggression. Many examples will be provided in later chapters of specific GWAS in the context of behavioural genetics. But for the moment it’s worth pointing out that the whole approach used by GWAS in the study of, for example, disease may be based on the wrong assumptions (Boyle et al., 2017). For example, in contrast to the kind of diseases illustrated in Chapter 1 in which a single variant protein-coding gene causes a disease, the major non-infectious diseases that fill the wards of hospitals around the world are instead polygenic, influenced by hundreds of genetic variants. The assumption of GWAS is that the hundreds of variant genes identified must have something to do with the disease. But in practice many of the gene variants identified by GWAS are not in protein-coding genes but rather in variant segments of our DNA involved in gene regulation. So with rare exceptions, GWAS are not picking out significant biological pathways but rather selecting for a lot of gene variants that may be involved in multiple regulatory systems inside cells. The SNPs that contribute the bulk of the heritability tend to be spread across the genome and are not near genes with disease-specific functions (Boyle et al., 2017). None of this means that GWAS are without value, but it does mean that the idea that identified gene variants are necessarily directly involved in the trait being investigated should be treated with some caution. In reality, the variants might be regulating a whole range of other genes that in turn are involved in many cellular processes, perhaps one of which might have some influence on disease development. There is clearly a large functional overlap in the contributions being made by many of the variant genes being identified, as will be illustrated in Chapter 5 (Visscher et al., 2017). The results will become far more interesting once we understand how all the various cellular pathways are connected, but this is a long way off as yet. And if all this is the case for medical syndromes that are clinically rather well defined, then what, one imagines, must be the case for complex behavioural traits such as religiosity, or being an extrovert, or being very intelligent. More of that later. For the moment, it’s worth noting that the overall picture emerging from GWAS – of hundreds or thousands of gene variants contributing to the variation in specific traits in a population – is not at all surprising when one considers the complex interactions between the genome and environment taking place during human development as described in Chapter 3.

https://doi.org/10.1017/9781108566520.005 Published online by Cambridge University Press

72

4 WHAT IS BEHAVIOURAL GENETICS?

4.3 Polygenic Scores One of the spin-offs from GWAS has been the generation of ‘polygenic scores’. Sometimes one can obtain quite a good idea as to how quickly a new idea becomes popular in the scientific literature by simply counting the number of times the new terminology is used in the titles of published papers. In 2009, the term ‘polygenic score’ did not appear at all; then in 2010, there was a single example, and the numbers then grew gradually over the years to reach 17 by 2016 and 66 by 2019. Polygenic scores are widely used in animal and plant breeding and have only more recently become popular in human studies. So what is a ‘polygenic score’? As GWAS became more extensive, it soon became apparent that it was going to be difficult to render many of the variant gene ‘hits’ statistically significance in terms of their contribution to the heritability, simply because their contribution was so tiny and/or the number of people being studied was rather small. By 2017, there was already a GWAS database that contained the summary statistics for 173 traits based on 1.5 million individuals and 1.4 billion associations between SNPs and traits (Zheng et al., 2017). So if you add up all the contributions, based on all the SNPs that tend to associate with a particular trait until adding on more SNPs doesn’t make the score go any higher, then you obtain a ‘polygenic score’ (also called a ‘polygenic risk score’), which may well be statistically significant. In practice, it’s not simply adding up but rather weighting the different SNP scores according to how much they appear to contribute to the overall heritability value. So more formally we can say that a polygenic risk score is calculated by multiplying the number of independent risky gene variants a person carries by the effect size of each variant and then summing these products across variants. To get the idea, imagine that you are carrying out a survey on levels of aggression in a population and you have twenty questions that can have the answers ‘None’ = 0, ‘A little’ = 1 or ‘Much’ = 2, where the ‘Much’ scores are all pointing to more aggressive behaviour in different circumstances. Clearly, the top ‘most aggressive’ score will be 40 and the least aggressive score will be zero. Now let us say that you interview 1000 people in this way – you can now map out all the scores on a kind of score card so that whenever you want to test any new person in the future, you can give them a score of 0–40 and see how they compare with everyone else. The analogy with a polygenic score is not perfect, but it gives some idea of the process of adding up SNP contribution values. But what’s worth keeping in mind with polygenic scores is that they are statistical constructs based on

https://doi.org/10.1017/9781108566520.005 Published online by Cambridge University Press

4.3 POLYGENIC SCORES

73

populations. So a polygenic score is really a way of making a probabilistic prediction about what trait someone might develop based on their ‘SNP map’ and the way that a particular collection of SNPs has previously been shown to associate closely with that trait. Now if I see a newborn baby, I can make a reasonably accurate prediction that when fully grown its height will be, on average, halfway between the height of its parents, but in reality the spread of heights is rather broad. But if the baby is genotyped, then its polygenic score on height will enable me to make a rather precise prediction about its future height. Notice, however, that height is a very physical trait with 80 per cent heritability, and the polygenic score cannot predict more than the number 80 per cent allows because future environmental differences remain unknown. This becomes even more relevant when the heritability is 50 per cent or less, because the polygenic score is limited to estimating probabilities within that 50 per cent of variation. So let us say that I have a polygenic score that puts me in the top 10 per cent of the population who are likely to develop schizophrenia. That doesn’t mean that I will end up developing schizophrenia. The polygenic score is probabilistic, not deterministic. I can equally well say that being very tall is a risk factor for bumping your head. But then no doubt there are tall people who, because they are tall, take extra care not to bump their head, so they never do. In the case of schizophrenia, the reason for uncertainty is different – the syndrome is only 50 per cent heritable, as we shall see in Chapter 5, so in such a polygenic condition, in which the environment plays such a big role (50 per cent) in the variation in a population, the polygenic score is of limited value. This chapter has given at least some idea of the potential but also the complexities involved in interpreting data arising from the field of behavioural genetics. As we now apply the approach to a range of practical examples, the potential – but also the complexities – will hopefully become even clearer.

https://doi.org/10.1017/9781108566520.005 Published online by Cambridge University Press

5

Genes and Mental Health

I

t was the mid-1950s. my late elder brother, about 10 years older than I, had won an open scholarship to Trinity College, Oxford, to read history, and the academic world seemed to be at his feet. I can well remember visiting the college with my parents to ‘see’ my brother perform in a medieval mystery play held in the walled gardens round the back of the college. I say ‘see’ because in fact my brother, who had a loud voice, played the role of God, or at least the voice of God, which entailed standing precariously on the top of a ladder propped up on a parapet above the gardens invisibly behind some trees. I also remember the play because its single sheep became untethered and tore off round the garden in a sudden burst of freedom chased by obviously rugby-playing dons in gowns who were watching the play – much more fun and memorable than the play itself for a 10-year old. But then in his second year, my brother had quite a severe mental breakdown and was diagnosed with what we now know as bipolar disorder but what was then known as manic–depressive psychosis. I can remember the murmured and anxious conversations by my parents back home. Piles of books on the current understanding of manic–depressive psychosis piled up on the sideboard – an understanding that was pretty limited in the mid1950s. I could tell that my parents were going through a period of selfquestioning: what have we done wrong to cause such a terrible thing to happen to our son? I think in due course, as a 10-year old who didn’t really know what was going on, I unknowingly became part of the therapy – amusement on occasion during his ‘manic’ phase and help with physical exercise through endless games of table tennis, good therapy for depression. Today, there might well be good reason for parents to be upset in similar circumstances – just as any parent would be upset if their offspring suffers from some nasty illness or other – but genetics has simply dissolved the concern that it’s their fault. Apart from situations where the parents are 74

https://doi.org/10.1017/9781108566520.006 Published online by Cambridge University Press

5.1 AUTISM SPECTRUM DISORDERS

75

abusive or severely neglectful in some way, when their behaviour might well contribute to the onset of a mental disorder (definitely not relevant in the present example), the better understood and important role of genetic variation in the development of many types of mental illness has now lifted the load of guilt from the minds of parents where it often used to hover in previous decades. For my part, happily free of that particular set of genetic variants (I assume, as an incurable optimist), my brother’s experience, and my parent’s response, were both influences in my later decision to do a PhD in neurochemistry (study of the chemistry of the brain) at the Institute of Psychiatry in London, and then to spend the first part of my research career in neurochemistry. We clearly do not have the space in a single chapter to do full justice to the huge field of psychiatric genetics. Our aim here will be more modest: to provide a few illustrations of mental syndromes where there are clear differences in behavioural traits from other people and, at the same time, clear evidence for a role for genetic variation contributing to the variation in a population that we see in regard to that particular trait. As illustrations of the applications of behavioural genetics methods to the world of medicine, these examples are also useful in seeing how a (generally) well-defined medical entity can be profoundly polygenic in terms of the role of genes in its development.

5.1 Autism Spectrum Disorders Autism spectrum disorder (ASD) is an umbrella term that includes Asperger’s syndrome, with a prevalence of around one out of every 160 children, on average, around the world. This figure depends on exactly how the disorder is diagnosed and, like all the syndromes discussed in this chapter, there is a fuzzy boundary round the syndrome – diagnosis is not invariably clear cut. Particularly so with autism, the definition back in the 1940s started rather tight and has become increasingly inclusive of more facets of behaviour over the years – hence the ‘spectrum’ language – meaning that cross-generational comparisons become a bit tricky (Rødgaard et al., 2019). The diagnosis is frequently made between the age of 3 and 6 years, although sometimes later, following that critical period when the great mass of synaptic connections in the brain have been established. In some cases, the switch from normal to abnormal behaviour is quite sudden. The disorders are characterised by impairments in social interaction and communication of varying degrees of severity (hence the word ‘spectrum’) and by restricted,

https://doi.org/10.1017/9781108566520.006 Published online by Cambridge University Press

76

5 GENES AND MENTAL HEALTH

repetitive and stereotyped behaviour and interests. Severe forms of autism are also characterised by seizures and intellectual disability. Most current theories of autism revolve around the idea of the development of abnormal synaptic connectivity during early infancy. There is a higher level of autism among cohorts good at maths compared with those good at the humanities, and around four times as many males suffer from autism as females. Indeed, it has been suggested that autism represents an extreme form of the male brain (Baron-Cohen, 2010), a theory that has received some empirical support but also plenty of detractors. Nearly half of children diagnosed with autism have average or above average intelligence. It is unclear whether more intelligent children are developing the condition or whether they are being diagnosed at a higher rate than in the past (and the idea that vaccination is a cause of autism has been thoroughly debunked). Right up to the 1990s, autism was still being blamed on birth trauma, infections, poor parenting and even child abuse. Psychologists suggested that unemotional ‘refrigerator mothers’ (terrible language!) caused their infants to set up self-defence mechanisms that led to autism. But it was already noted back in 1991 that the relative risk of a child being diagnosed with autism is increased at least twenty-five-fold over the population prevalence in families in which a sibling is affected, suggesting a significant familial influence. Later work using thousands of twin pairs has since been used to derive the heritability of autism, generating values in the range of 56–95 per cent (Colvert et al., 2015). In a later study based on 37,570 twin pairs, a heritability of 83 per cent was estimated, with 17 per cent of the variation in the population being due to (unknown) non-shared environmental factors (Sandin et al., 2017). What is perhaps most informative is that the concordance of having autism is close to 100 per cent in identical twins, whereas it is only around 50 per cent in non-identical twins. In other words, where one of the twins is on the autism spectrum, the other twin is almost certain to develop autism, whereas if one of a pair of non-identical twins is autistic, then there’s only a roughly 50 per cent chance that the other twin will develop autism (Tick et al., 2016). Now of course this does not discount the role of the environment – far from it – as pairs of identical twins in particular are likely to share very similar environments – but overall, these and other findings point to important genetic components in explaining the variance in a population with regard to ASD. As the weight of evidence has grown demonstrating the importance of genetic variation, so a huge burden of unnecessary guilt has been lifted from the shoulders of the parents of autistic children who had previously blamed

https://doi.org/10.1017/9781108566520.006 Published online by Cambridge University Press

5.1 AUTISM SPECTRUM DISORDERS

77

themselves for their child’s condition. The positive pastoral impact of genetics upon families in which a child is suffering from a particular disorder can be very considerable. So which genes contribute to the development of ASD? The GWAS described in Chapter 4 have played a significant role here. Earlier results seemed to identify hundreds of genetic variants that are associated with autism; in fact, there are 800 genes in the autism database. But the problem is that autism is not like height. It may be remembered that it took more than a quarter of a million Europeans to find out that 697 of their genetic variants contributed to the variation in height seen in that population in a statistically significant manner. Obviously obtaining data from an equivalent number of people with autism is a huge enterprise and large cohorts can only be achieved by organising large international consortia. Fortunately, the international collaboration of scientists is being highly successful in this regard, and the Psychiatric Genomics Consortium, which now involves more than 800 researchers from thirty-six countries, is a sign of this success.1 The Consortium found it hard to replicate the earlier SNP ‘hits’ once they started putting together results from a bigger cohort. For example, a large international study obtained GWAS data from more than 16,000 individuals with autism (Autism Spectrum Disorders Working Group of The Psychiatric Genomics, 2017). Their study identified no genetic ‘hits’ at the strictest statistical level, but around 100 genetic variants were noted that were close to significant, and there are independent results suggesting that some of these could be important in the development of autism. Of particular interest was their finding that quite a sizeable proportion of these variants were the same as those identified in GWAS for schizophrenia, as discussed below. Rare mutations have also received particular attention in the autism field. In one study, genomic sequencing was carried out in around 2,500 families in which only one of two siblings was on the autistic spectrum (Iossifov et al., 2014). The total burden of novel mutations discovered in the autistic cohort was significant for up to 45 per cent of those with autism. The rate of novel mutations likely to disrupt gene function was nearly twofold higher in the autistic cohort compared with their unaffected siblings. Another study on a similar number of autistic individuals focused on mutations that occurred in both copies of the same gene so that the gene function was wiped out altogether (Werling et al., 2019). Those with autism were around three times more likely to contain such gene disruptions, and many of the genes involved were those known to play important roles in early brain development. There are also rare cases of ASDs that appear to be caused by a single gene disorder (Zoghbi and Bear, 2012). For example, fragile X syndrome is so

https://doi.org/10.1017/9781108566520.006 Published online by Cambridge University Press

78

5 GENES AND MENTAL HEALTH

called because the single mutant gene that causes the syndrome is located on the X chromosome and around one-third of the carriers of this mutation develop ASD (Ebert and Greenberg, 2013). As expected, differences in epigenetic regulation may also be involved in the development of autism, as shown by analysing the epigenetic status in identical twins either concordant or discordant for autism (Wong et al., 2013). A further point to emphasise is that there is no expectation that all those on the autism spectrum will finally be found to have precisely the same set of variant genes. The more likely scenario is that there will be a subset of variant genes that are often, but not always, found in those with autism, and then a broad range of other variants that are quite different among different individuals but which, taken together, take the risk of developing autism over a certain threshold. Gradually, the gene variants that cause the development of autism are being identified. Might this open the way to new therapies for autism? Maybe. If, for example, ten genetic variants are identified that all line up on the regulation of the same brain control pathway at a molecular level, this might provide clues as to which pathway to inhibit or amplify. But given that autism generally develops in very young children, and given the current understanding that it’s caused by errors in synaptic connections, this may not be realistic. However, drugs could be developed that help with the more negative behavioural characteristics (which are only present in some), even if they don’t get to the root cause. Plus it’s of course good to remember that those at the high-performance end of the autism spectrum are much sought after by companies in Silicon Valley for their IT and related skills. And let’s not forget the Swedish teenager Greta Thunberg who pioneered a worldwide campaign against global warming from 2018 onwards. Such examples raise the challenging question as to whether a ‘cure’ should be sought at all. The answer to this question no doubt depends very much on where on the spectrum someone is located. To look at all the genetic candidates would take up too much space and, in any case, I know that not everyone can wax lyrical about molecules and their interactions for long periods (I can, easily). But I cannot resist providing you with just two examples because they show how research on one syndrome – autism, in this case – can shed light on another syndrome and, furthermore, on how genes implicated in autism might exert their actions at the molecular level. As emphasised in the Chapter 4, everything is tangled up with everything else. The first candidate gene we’ll take as an example has the exciting name ‘UBE3A’. It encodes a protein with the same name2 and here we’ll just call it UBE for short. The protein has long been known to have important

https://doi.org/10.1017/9781108566520.006 Published online by Cambridge University Press

5.1 AUTISM SPECTRUM DISORDERS

79

roles in brain development, plus lots of other roles as well, including cancer. The reason the protein has so many roles is that it’s an enzyme involved in degrading other proteins. Such enzymes have to be kept on a tight leash, otherwise they can do terrible damage. In fact, the expression level of UBE is kept very carefully regulated during brain development. And when the UBE gene is abnormally duplicated so that the offspring have more than the normal two copies of the gene, then autism almost certainly results (Vatsa and Jana, 2018). So too much UBE protein leads to autism. But what happens when there is too little? Here, there is another fascinating part of the story. It may be remembered from Chapter 3 that one of the two copies of the X chromosome inherited from the mother is inactivated, thereby preventing too much of the proteins being made that are encoded by the genes found on the X chromosome. What we didn’t mention is that there is a similar kind of process called ‘imprinting’, which affects genes from either the mother or the father. We have about 75 genes that are ‘imprinted’, which means that either the paternally or maternally derived genes are switched off epigenetically, and this silencing is then maintained throughout the subsequent production of cells for the rest of our lives. Imprinting is nonrandom: the same sets of genes are silenced consistently on either paternally or maternally derived chromosomes. Now normally when you have a mutation in one of your two copies of a gene, there is still a good copy around to compensate for the defect. But the paternally inherited copy of UBE is an example of a silenced gene. This means that the level of its UBE protein product must be really critical for normal nerve cell development. And when there is a defect in the UBE gene on the maternally inherited chromosome, then there’s no UBE protein at all and the result is Angelman syndrome, a condition resulting from a failure of normal development of the nervous system, leading to jerky movements, frequent seizures, sleep disturbance and yet, despite all these challenges, often a happy and smiling outward demeanour. Those afflicted with the syndrome do not generally speak more than five to ten words, if any. Dr Harry Angelman was the physician from northern England who first encountered three children in his medical practice with the condition, although at first he was unsure whether it was a single condition or several. Only after he went on holiday to Italy and saw an oil painting in the Castelvecchio Museum in Verona of a smiling boy entitled ‘Boy with a Puppet’ did Dr Angelman realise that perhaps this condition went back centuries, in turn leading him to write up his description for a medical journal.

https://doi.org/10.1017/9781108566520.006 Published online by Cambridge University Press

5 GENES AND MENTAL HEALTH

80

So clearly regulating the level of the UBE protein for normal nerve cell development is critical: too much and you end up on the autism spectrum – too little and you get Angelman syndrome. And just to show that regulating the UBE enzyme activity is really important, there is another level of regulation (Yi et al., 2015). Back in Chapter 2, it was mentioned that many chemical modifications can take place that increase the range of functions of proteins. One of these involves the transfer of a chemical group called a ‘phosphate’ on to a specific amino acid within the protein. Believe it or not, the attachment of this tiny chemical group can either dramatically increase or decrease the activity of the enzyme. As Figure 5.1 illustrates, UBE is regulated just like that. There is an amino acid called threonine at precisely the 485th position in the amino acid sequence of UBE and when that threonine has a phosphate on it, then the enzyme is inactivated, so you might as well not have any UBE there at all, and if the phosphate level is kept high, then Angelman syndrome will

P Phosphorylation (a) UBE3A Active enzyme

UBE3A Dephosphorylation

Inactive enzyme

(b) UBE3A Threonine 485 mutant

UBE3A Loss-of-function mutant

Permanent activation AUTISM

Permanent inactivation ANGELMAN SYNDROME

Figure 5.1 The role of the UBE gene in autism and Angelman syndrome. The UBE3A enzyme is a protein encoded by the UBE gene. (a) When a phosphate group (P) is transferred to amino acid threonine 458 on the UBE3A enzyme (‘phosphorylation’), it is inactivated, and is reactivated when the phosphate is removed (‘dephosphorylation’). (b) A mutation at threonine 458 prevents phosphorylation, resulting in permanent activation of the enzyme, contributing to the development of autism. Other mutations in the UBE gene result in permanent inactivation of the encoded UBE enzyme, leading to Angelman syndrome. Based on data from Yi et al. (2015).

https://doi.org/10.1017/9781108566520.006 Published online by Cambridge University Press

5.1 AUTISM SPECTRUM DISORDERS

81

be the result. But if that threonine is mutated so that it can no longer have a phosphate stuck on it at that position, then UBE becomes permanently in the ‘on’ position and autism is the result. Indeed, a child with autism has been found with this specific mutation in his UBE gene (Yi et al., 2015). One of the really interesting aspects of this work was that when nerve cells were looked at in the laboratory, if they had a mutation at this precise position in the UBE gene, then they did not grow normally. The mutation really does make a difference. One could repeat similar kinds of stories for many of the other candidate genes that are thought to be involved in autism. In each case, there are many different layers of the story and many different ways in which a variant of the gene might end up contributing to autism. Looking at Figure 5.1, we might also ask about the regulation of the enzyme that puts phosphates on to UBE, and of the different enzyme that takes the phosphate groups off again. It doesn’t take much imagination to see how dozens of genes encoding proteins could be involved in the development of autism by impacting on the regulation of UBE in different ways. Multiply that up for all the candidate genes and one can see how finding out how autism develops is quite a challenge. Our second example of a candidate gene involved in autism also has a catchy name: PTCHD1.3 Like so many of the genes involved in autism, mutations in the gene are involved in only a tiny fraction of people with the syndrome, but more than 40 per cent of individuals with mutations in this gene develop autism-like behaviour. There is a cluster of syndromes where the symptoms overlap with those on the autism spectrum but which are nevertheless distinct, and the intellectual disability, sleep disruption and attention deficit hyperactivity disorder (ADHD) associated with PTCHD1 gene mutations provides an example of one of those clusters. One way to find out how a gene works is to generate a colony of mice that have the same mutation as that found in humans. When mice were genetically engineered not to have any PTCHD1 protein, then their behaviour was largely consistent with that seen in people with ASD, including sleep disruption, hyperaggression, 4 and deficits in attention and learning (Wells et al., 2016). But what the investigators also noticed was that PTCHD1 expression is particularly high during early brain development in a part of the mouse brain involved in the regulation of vision, attention and sleep.5 There are now cunning methods available to remove the expression of a gene from specific parts of the brain, so when PTCHD1 was totally removed from that tiny area of the brain, these mice displayed hyperactivity and deficits in sleep and attention but did not exhibit the learning deficits and hyperaggression observed in mice carrying the brain-wide mutation. So gradually it’s

https://doi.org/10.1017/9781108566520.006 Published online by Cambridge University Press

82

5 GENES AND MENTAL HEALTH

becoming possible to dissect out different parts of the brain and find out how mutations in specific genes relate to specific behaviours. The advances are remarkable, but there is still a long way to go in order to achieve the ultimate goal of seeing how genes and the proteins they encode relate to behavioural differences in disorders such as autism.

5.2 Schizophrenia Schizophrenia is a condition that develops in early adulthood and, like autism, is more common in males. The condition is perhaps better called ‘the schizophrenias’ as the word includes quite a range of clinical categories. It is associated with disorganised thinking, a lost sense of reality, paranoia and hallucinations. Many people with schizophrenia unfortunately end up becoming addicted to drugs or alcohol, and this can lead to problems in diagnosis, but the overall incidence in the world population is around 1 per cent. The amount of research on schizophrenia happening in different parts of the world is huge. To give an idea, 3,135 scientific articles were published in 2019 with the word ‘schizophrenia’ in their title. As with autism, it is not so long ago that psychoanalysts would speak of the ‘schizophrenogenic mother’ as if there were something about the mother’s personality that induced the onset of schizophrenia in her children, and it was not until the adoption studies of the 1960s that this idea was shown to be false. The risk for developing schizophrenia increases with genetic relatedness: if there is a second-degree relative who already has schizophrenia (sharing 25 per cent of their genes in common), then the risk rises from 1 to 4 per cent; for a first-degree relative (sharing 50 per cent of their genes in common), the risk is 9 per cent; between non-identical twins it is 17 per cent; and between identical twins, the risk rises to 48 per cent. Expressing this another way: if one member of an identical twin pair develops schizophrenia, then there is a 50:50 chance that their co-twin will also develop the syndrome. A meta-review of twin studies in relation to schizophrenia (meaning the averaging of the results from many independent investigations) suggested that the heritability of schizophrenia is 81 per cent (range 73–90 per cent), with 11 per cent of the variance being attributable to shared environmental influences and 8 per cent to non-shared (Sullivan et al., 2003). In a large population study carried out in Sweden, a heritability value for schizophrenia of 64 per cent was obtained, with shared environmental effects accounting for 5 per cent of the variance and so 31 per cent attributable to the nonshared influences (Lichtenstein et al., 2009). What is clear from all such

https://doi.org/10.1017/9781108566520.006 Published online by Cambridge University Press

5.2 SCHIZOPHRENIA

83

studies is that environmental variation plays an important role in explaining the variance in a population with respect to the trait of schizophrenia. As far as the search for the contributing variant genes is concerned, once again GWAS approaches have led to many interesting findings, whereas earlier candidate gene studies in which attempts were made to track variant genes through family studies have unfortunately proven unsuccessful (Johnson et al., 2017). In one major GWAS involving nearly 37,000 cases of schizophrenia and more than 113,000 controls, 128 different SNPs were reported that associated with the disease at greater than chance levels, identifying 108 different relevant locations in the genome (Schizophrenia Working Group of the Psychiatric Genomics Consortium, 2014). Of these, 75 per cent of locations included protein-coding genes, many thought to be biologically relevant to the development of the disease. One of the surprises from this study was the finding that a number of genetic variants involved in the immune system were found to be associated with schizophrenia. The immune system refers to all the many mechanisms in our bodies that defend us against attacks by bacteria, viruses and other ‘foreign invaders’. A further study identified variants in a gene that encodes a protein known to be important in both the immune system and in the ‘synaptic pruning’ mentioned in Chapter 3 whereby the brain’s synaptic architecture continues to be sculpted right into early adulthood (Sekar et al., 2016). In addition, it is now becoming clear that rare mutations are also involved, many of them de novo, meaning that they represent new mutations that have arisen in one or a few schizophrenic individuals. These mutations may either contribute to the development of schizophrenia – or in some cases – actually protect people against its development (Purcell et al., 2014, Marshall et al., 2017). Structural variants, involving changes in relatively large segments of DNA, have also been found to be more likely in DNA from those suffering from schizophrenia (Rees et al., 2014). In some studies, an overlap has been reported between the variant genes that contribute to the development of schizophrenia and those that contribute to autism and to individuals with intellectual disability. It is worth emphasising that, as has been the case with autism, the numbers of ‘really significant’ gene variants involved in schizophrenia has been declining as the statistical criteria for acceptance have become increasingly stringent. One study carried out a critical review of twenty-two separate GWAS publications and concluded that only nine gene variants could be judged to be really well supported at the present time as being involved in schizophrenia (Prata et al., 2019). But one should not conclude from this that nine will be the final number. Again, it is good to remember the GWAS of

https://doi.org/10.1017/9781108566520.006 Published online by Cambridge University Press

84

5 GENES AND MENTAL HEALTH

height example. At present, the cohorts of those with schizophrenia under study number well into the thousands. But what will happen when the number is up to more than a quarter of a million? Only then, perhaps, will the final number of significant genes involved in schizophrenia become clearer. The fact is that, in the case of identical twins, it is nearly certain that if one twin is autistic, then the other will be also, whereas in the case of schizophrenia, if one twin has it, then the chances of the other twin having it are only 50 per cent. So what are the environmental influences that may be involved? Many have been described, including prenatal infection, poor prenatal nutrition, obstetric complications at birth, social disadvantage and childhood trauma. Long-lasting epigenetic changes may result from such early traumas (Cattane et al., 2018). There is also a significant genetic correlation between cannabis use and schizophrenia (Arseneault et al., 2002), with some evidence that cannabis use is a risk factor for developing schizophrenia, whereas having schizophrenia is a strong risk factor for cannabis use (Pasman et al., 2018).6 If you have been diagnosed with schizophrenia, or have family members who have, then cannabis use is not a good idea. Yet our understanding of how environmental factors integrate with the genomic variation in the development of schizophrenia remains very limited. One common suggestion is that there is an accumulation of risk factors coming from both genes and the environment, and that just one or a few more mutations coordinate to take the system ‘over the edge’. The appropriate analogy from the world of physics would be a ‘phase transition’ in which a series of very small changes occur in a material until the critical point comes at which one further change causes the material to acquire a quite different state. This raises the provocative question as to whether the development of schizophrenia might be almost entirely due to genetic variation. If this were the case, the reason that identical twins are only 48 per cent concordant for the disease might be the fact that in one twin, but not the other, one or a few rare mutations have occurred against a background of high genetic risk, shared by both twins, and these few ‘extra’ changes are then sufficient to trigger the ‘phase transition’ leading to development of the disease. Following the discussion of ‘polygenic scores’ in Chapter 4, it will be no surprise to hear that many papers are coming out seeking to estimate the polygenic score for schizophrenia (Rammos et al., 2019). So far, the polygenic scores can predict around 8 per cent of the variance involved in populations with regard to somebody in that population becoming schizophrenic –

https://doi.org/10.1017/9781108566520.006 Published online by Cambridge University Press

5.3 BIPOLAR DISORDER

85

providing, of course, that they are genotyped to identify the relevant variant genes (Toulopoulou et al., 2019). Given that the heritability scores obtained for schizophrenia are above 64 per cent, it is clear that there is a long way to go before it might be possible to make firmer predictions about whether someone in childhood will later develop schizophrenia. And it is always good to remember that polygenic scores, based on many genetic variants associated with a complex syndrome such as schizophrenia, are always only going to be predictive in a probabilistic kind of way. This is where the situation is very different from the type of single-gene mutant disorders described in Chapter 1 where, if the mutation in both copies of a specific gene is identified at birth, resulting in a protein deficiency, it is 100 per cent certain that a certain disease will develop. With polygenic disorders, where the genetic understanding comes from population studies, and where the environment plays a key role, making firm predictions about whether a specific individual will develop the disorder is never likely to be possible. So if you were a teenager, would you want to know your polygenic score for schizophrenia? It’s hard to know why this would be useful. Even if you scored very high relative to the rest of the population, you wouldn’t know whether or not you would ever develop schizophrenia. Of course, it could provide an even stronger argument than the usual ones against taking mindaltering drugs. To end this section on a slightly more positive note, it is of interest that the parents or siblings of those who suffer from schizophrenia are more likely to be found in the artistic or scientific creative professions (Kyaga et al., 2013). Higher polygenic scores for schizophrenia in adults who do not suffer from schizophrenia have also been shown to be associated with artistic society membership or creative professions (Power et al., 2015). These data are consistent with the idea that some of the genetic variants involved in schizophrenia may contribute towards a predisposition to creativity of various forms. Every cloud has a silver lining.

5.3 Bipolar Disorder The illness that my elder brother suffered from, called manic–depressive psychosis in his day, is now known as bipolar disorder, with an average incidence of around 1 per cent in a population. It can occur at any age, although it often develops in the early 20s (as with my brother) and rarely over the age of 40. Diagnosis can initially be difficult as, if it starts with depression, then it won’t yet be known whether the polar opposite upswing will occur later. Unlike autism and schizophrenia, many studies found that

https://doi.org/10.1017/9781108566520.006 Published online by Cambridge University Press

86

5 GENES AND MENTAL HEALTH

men and women are equally likely to develop the disorder, although in other studies women were found to be more represented (Johansson et al., 2019). Suicide among sufferers from bipolar disorder is ten- to thirtyfold higher than in the general population (Dome et al., 2019). As the name suggests, the illness involves two ‘poles’ – clinical depression for some time, maybe many months, followed by a period of elation and hyperactivity for a few months, followed then once more by depression. These are not normal mood swings, which most of us experience, but rather swings from one clinical state to another, with a positive response to pharmaceutical treatment in many cases. At times, bipolar disorder may be associated with psychosis when people see or hear things that are not there, or become convinced of things that are not true. This aspect of the disorder in particular highlights its overlap with some of the symptoms of schizophrenia. The heritability of bipolar disorder is 55–60 per cent (Johansson et al., 2019), so less of the variation in a population can be attributed to genetic variation compared with autism and schizophrenia, with environmental factors playing more of a role. Episodes of bipolar disorder may be triggered by very stressful events, such as the breakdown of a relationship, by some form of abuse or by sleep disturbance. But the same stressful events may occur in others without a genetic predisposition, and no bipolar disorder will result. By now, reports of GWAS hunting will be familiar and what is of particular interest is that in a massive GWAS-driven genetic comparison of more than 20,000 people with bipolar disorder and more than 33,000 with schizophrenia, 114 genetic variants were identified that are shared in common between the two conditions (Bipolar Disorder and Schizophrenia Working Group of the Psychiatric Genomics Consortium, 2018). Many of these genes are involved in the construction and function of synaptic connections between brain cells (Prata et al., 2019). In addition, four genomic regions were identified that are different between the two disorders. So the genetics reflects the distinct clinical presentation of the two conditions – plenty of similarities but some distinct differences as well.

5.4 Major Depressive Disorder We all have ups and downs in life, and if someone has just lost a loved one or a relationship has broken up, then it’s very normal to feel rather down in our mood for a while. This is not major depressive disorder (MDD). When you get MDD, a dark mood just comes over you like a wave, as if coming more from within than without: normal sleep patterns are disrupted, appetite

https://doi.org/10.1017/9781108566520.006 Published online by Cambridge University Press

5.4 MAJOR DEPRESSIVE DISORDER

87

changes, life becomes dark and gloomy, you have low self-esteem and there is a strong conviction that these dark feelings will go on for ever. Winston Churchill called it a ‘black dog’. Individuals vary greatly in their symptom severity, treatment response and outcome. It might be triggered by some stressful event, but the trigger leads downwards into the pit without any apparent strength to bounce back, as most people do after a difficult time in life. There are more than 300 million people battling depression worldwide, and academics are at greater risk than many. It is no good telling someone with MDD to just cheer up, look on the bright side and so on, because that’s just what their condition doesn’t allow them to do. Depression is exhausting. In addition to continuing emotional support and friendship, what sufferers need is to see a psychiatrist at the earliest opportunity who will most likely prescribe some pills. Fortunately there are now some really good medications around, and it is important that someone with MDD continues with their pills as long as instructed, even when they don’t seem to be helping much at the beginning. Some types of medication take weeks to act. And once someone with MDD feels better, there is then the danger that, now they’ve recovered, they will feel that they just don’t need those pesky pills any more. But often if they stop taking them too soon, they’ll sink into depression once more. The brain chemistry is just messed up for a while and that’s what’s good to remember. It’s not your fault and it’s not something that you chose – it just came upon you, like dark thunder clouds rolling into your head unbidden. MDD affects 2–4 per cent of the population at any given time, and about 16 per cent of the population at some time or other during our lives (Kessler et al., 2003). Therefore, not surprisingly, the cohorts used to explore genetic contributions to the disorder have become larger and larger. Obtaining big numbers is not so difficult, given the prevalence of this disorder, but making sure that the diagnostic criteria being used across the group are consistent can be more of a challenge. The heritability of MDD is 31–42 per cent, so compared with the other three conditions considered so far in this chapter, the contribution of genetic variants to the variance seen in a population with respect to MDD seems quite a bit lower. Perhaps for this reason the hunt for genetic variants involved in MDD has been quite challenging. It was not until 2015 that the first two genetic variants were reliably associated with MDD (Ledford, 2015). Since that time, the discovery pace has been hotting up. Based on nearly a quarter of a million individuals with MDD, no fewer than 102 genetic variants have now been associated with the disorder by means of GWAS (Howard et al., 2019), some of which are known to be involved in

https://doi.org/10.1017/9781108566520.006 Published online by Cambridge University Press

88

5 GENES AND MENTAL HEALTH

synaptic pathways involved in amplifying (‘exciting’) signals in the brain (Howard et al., 2018), so it’s easy to see how a reduction in such pathways could lead to depression. As with schizophrenia, MDD is also associated with specific structural variants in the DNA (Kendall et al., 2019). So at the time of writing, the real hunt for relevant genetic variants is only just beginning, and the first 102 identified are likely to be the tip of the iceberg, with each variant making a tiny contribution to the overall risk of ever having MDD (Howard et al., 2019). The take-home message from such studies is that we all carry greater or lesser numbers of genetic risk factors for depression. When these numbers reach a certain critical point, and when certain environmental inputs reach a certain level, then MDD will result. The concordance between identical twins is around 22 per cent – in other words, in roughly three-quarters of twin pairs, only one member of the pair will suffer, while the other will not. A particular complex set of variants is not deterministic but represents an increased risk of developing MDD under certain circumstances. Polygenic score analysis is still in the early stages of development for MDD, but so far, if you have a polygenic score in the top 10 per cent of all scores, then you are a little more than twice as likely as the bottom 10 per cent to develop MDD at some stage during your life (Wray et al., 2018). That might sound like quite a bit, but remember that the overall incidence in the population is around 2–4 per cent, so this might just be telling you that in your particular population your chance of developing MDD sometime in your life has just gone up from 2 per cent (let’s say) to 5 per cent, which is still not too bad. A significant overlap between the gene variants identified in MDD and those in schizophrenia and bipolar disorder has been reported in several studies, pointing to some common biological pathways in the development of these conditions, with a focus on certain specific parts of the brain (Wray et al., 2018). So far, the results indicate important roles for gene regulatory processes rather than for genes encoding common proteins. For example, one mutation found in MDD, which is also found in schizophrenia, is involved in the alternative splicing of genes. As will be remembered from Chapter 2, alternative splicing can generate many different proteins from a single gene, so alteration of this mechanism could well lead to many downstream effects. But despite such findings, it is now perfectly clear that there is no ‘gene for depression’ but rather a multitude of genes, each contributing a tiny amount to the overall risk factor. So depression is not ‘inherited’, but its prevalence might be slightly higher in some multi-generational wider family histories than in others, simply because accumulations of relevant genetic variants are more likely.

https://doi.org/10.1017/9781108566520.006 Published online by Cambridge University Press

5.5 ALZHEIMER’S DISEASE

89

As the generations continue, this ‘increased risk’ set of genetic variants will soon disappear, based on the assumption, of course, that you marry someone with a sunny disposition and with no apparent tendency to develop MDD.

5.5 Alzheimer’s Disease All the conditions highlighted so far tend to develop in people younger than 30. Alzheimer’s disease (also referred to simply as Alzheimer’s) is the opposite, typically afflicting people in their 70s and 80s. In people over the age of 65, the risk of developing Alzheimer’s doubles every 5 years. Alzheimer’s is responsible for around two-thirds of all cases of adult dementia. In 2015, almost 50 million people around the world had dementia, and due to the ageing population, this number is expected to surpass 130 million people by 2050 (Drew, 2018). However, in the UK and other developed countries, there has actually been a modest proportional decline in Alzheimer’s since 1990, presumably due to improvements in education, nutrition and healthier lifestyles. Many studies have shown that the more education you have, the less chance you have of Alzheimer’s in later life. The incidence of Alzheimer’s is about twofold higher in women than in men, and this difference is not due to greater female life expectancy; the reasons for the difference are not really known. Alzheimer’s starts with subtle changes in behaviour and increasing memory loss. This is followed by a slow but relentless decline in cognition with increasingly striking behavioural changes until the person one knew is no longer really there. The disease is associated with various chemical and structural changes in the brain, leading to a progressive loss of brain cell connections, and in its final stages with an actual shrinking of the brain, with death as the eventual consequence. By the time the symptoms of Alzheimer’s begin to be noticeable, the changes in the brain have already begun, starting around 10–15 years earlier, in particular the accumulation of a tangled web of protein known as amyloid-β. There are many risk factors for Alzheimer’s which in principle can be modified or treated (Sohn, 2018). These include diabetes, obesity, depression, smoking and low educational attainment. Factors such as exercise and a Mediterranean diet – full of whole grains, fruit and vegetables, together with fish and olive oil – have all been promoted as means to help prevent the development of Alzheimer’s (Horder et al., 2018, Lourida et al., 2019). Being married also seems to make a difference. A review of fifteen different studies has revealed that being single raises the chances of dementia in later life by

https://doi.org/10.1017/9781108566520.006 Published online by Cambridge University Press

90

5 GENES AND MENTAL HEALTH

45 per cent compared with those who are married (Sommerlad et al., 2018). But our emphasis here is on the genetics. The heritability of the usual form of Alzheimer’s is in the region of 60–80 per cent, although this estimate should be treated with some caution as it rather depends on what approach is used to measure it. In one Swedish study using twins, the heritability was estimated at 74 per cent and the concordance for the condition among identical twins was 67 per cent, compared with 22 per cent between non-identical twins (Gatz et al., 1997). In other words, if you randomly picked elderly identical twin pairs out of a large sample of twins of whom at least one has Alzheimer’s, then two out of three twin pairs will both have the condition. Of course, you can see the problem in making that kind of estimate: maybe the non-concordant twin pairs will match up later once they get older? But other studies roughly agree with this one, so the estimates are likely to be in the right ballpark. Alzheimer’s is unlike the other conditions discussed so far in this chapter in that, way back in 1993, long before GWAS, a single genetic variant was identified called apolipoprotein E (ApoE), which was found to be strongly associated with Alzheimer’s disease. When Jim Watson, co-discoverer of the DNA double helix, had his genome sequenced and published in 2007 when he was aged 79 (‘the first genome to be sequenced for less than $1 million’), he was happy for the complete sequence to be published except for one small segment (Check, 2007). Which segment? That which encodes ApoE. Why? Well, one of the versions of the ApoE gene, known as variant 4, has a frequency of 40 per cent in those with Alzheimer’s disease compared with the rest of the population. Having two copies of ApoE4 increases the lifetime risk of Alzheimer’s disease right up to 80 per cent. As Jim Watson’s grandmother had suffered from Alzheimer’s, very sensibly he did not wish anyone to know his ApoE sequence.7 The Alzheimer’s Society recommends against genetic tests that will reveal whether someone is carrying one or two copies of the ApoE gene. Flora Gill, daughter of the late writer A. A. Gill and of politician Amber Rudd, had a grandfather with Alzheimer’s and took the test, but greatly regretted finding out that she has two copies of ApoE4.8 As Flora Gill writes: ‘Now I have my results, I’m particularly aware of any mention of Alzheimer’s, whether a character portrayed on television or a story recalled by a friend – every tale feels like a glimpse of my potential future.’ What does the ApoE gene do? Well it’s involved in cholesterol transport. Cholesterol is a fatty substance found in your blood that most people have heard of because they know that if you have too much of the wrong type, it causes atherosclerosis (fatty deposits on the walls of your arteries) that are a risk for

https://doi.org/10.1017/9781108566520.006 Published online by Cambridge University Press

5.5 ALZHEIMER’S DISEASE

91

stroke and heart disease. As you can imagine, given the identification of ApoE4 as such a risk factor in 1993, there has been an important amount of research since then to elucidate how it works in the development of Alzheimer’s. But like many other aspects of the condition, although a huge amount is known, it’s not yet enough to see clearly how it acts together with the many other genetic variants to contribute to the risk of developing the disease. As with the other conditions reviewed in this chapter, there are now plenty of other gene variants identified through GWAS that contribute to Alzheimer’s, but, unlike ApoE4, they all make very small contributions to the overall variance in the population. Dozens of significant variants have been reported and active research continues to clarify which ‘hits’ are genuinely important (Escott-Price et al., 2017, Giau et al., 2019, Lutz et al., 2019). The overall ‘pattern’ of the hits is giving some important clues about how Alzheimer’s disease first gets going. There are also other genetic variants that protect people against the development of this disease. As mentioned in Chapter 4, one of them is quite common among Icelanders, such that those with even a single copy of the variant gene are more than five times more likely to reach the age of 85 without the development of Alzheimer’s disease. There is also an early-onset form of Alzheimer’s disease, which is quite distinct and can be attributed to one of three genes. The mutant genes are ‘dominant’, that is, only a single copy, plus some other contributing factors (Lacour et al., 2019), is enough to cause the development of this form of the disease, generally when the person is in their 40s. This means that 50 per cent of their offspring on average will inherit the disease if one of the parents has the mutant gene, because 50 per cent of the offspring will receive the ‘bad’ copy and 50 per cent will receive the ‘good’ copy. People with this early-onset form are already developing the amyloid-β patches in their brains in their 20s and 30s, which are thought to play a critical role in the development of the disease. Fortunately, less than 1 per cent of all those diagnosed with Alzheimer’s disease has this early-onset form of the disease, and due to its dominant mode of transmission, mutations tend to accumulate in some very local populations. One of the early-onset mutant genes probably arrived in South America with the Spanish conquerors some 375 years ago and now affects about twenty-five extended families in Antioquia, a mountainous area of northwest Columbia full of coffee plantations. As the development of Alzheimer’s disease is so predictable in the individuals who have the mutant gene, the families in this region are collaborating with an intensive research project to study the onset in a way that will eventually benefit all those who suffer from this crippling disease (Reardon, 2018). The early-

https://doi.org/10.1017/9781108566520.006 Published online by Cambridge University Press

92

5 GENES AND MENTAL HEALTH

onset form of Alzheimer’s provides a unique opportunity to attempt pharmaceutical treatment to determine whether it will prevent disease onset, a goal that has so far remained out of reach of the pharmaceutical industry despite huge effort.

5.6 The Take-home Message This brief overview of the role of genetics in the development of five major mental disorders highlights certain key principles in psychiatric genetics. First, it will be noted that the role of genetic diversity in the heritability of the condition varies quite a bit. In the case of ASD, the heritability is very high: providing certain sets of gene variants are present, then autism will result. Identical twins nearly always both have autism, and the condition generally becomes apparent when they are very young. However, with conditions that develop later in life, such as schizophrenia, and even more so with bipolar disorder and MDD, the heritability is lower and the concordance between identical twins is lower – 50 per cent in the case of schizophrenia and around 25 per cent in MDD. When it comes to the very different Alzheimer’s disease, the concordance between identical twins is back up to 67 per cent. So there is quite a range among these conditions in the role of genetic variation in having the final say in what happens. Genes make big differences, but the size of these differences depends very much on other factors, and these factors accumulate throughout life, with environmentally induced epigenetic changes very likely to be critically involved. Second, with rare exceptions such as the ApoE4 gene in late-onset Alzheimer’s disease, and even more so the mutant genes involved in earlyonset Alzheimer’s, there are hundreds of genetic variants involved in contributing to the population variation in the various conditions, each one making only a tiny difference. Likewise, there are genetic variants that are protective. The brain is a highly complex organ, so it’s hardly surprising that hundreds of genes are involved in its dysregulation. Third, there are subsets of genetic variants that are shared between the different conditions, for example between autism, schizophrenia and bipolar disorder (Smeland et al., 2019). In fact, it has been quite a surprise to investigators over the past decade to find out just how many variants are in common. What this probably means is that there are certain ‘regulatory modules’ involved in the laying down of synaptic connections during early development, and some of these modules share the same dysfunctions in these different conditions. But equally, the total sets of genetic variants are clearly distinct, reflecting the clinical distinctiveness of the condition in each case.

https://doi.org/10.1017/9781108566520.006 Published online by Cambridge University Press

5.6 THE TAKE HOME MESSAGE

93

Fourth, the range of conditions described in this chapter has been chosen partly to make the point that we are to a large extent slaves to our genes in certain medical conditions within behavioural genetics but much less so in others. Autism just happens, and to the best of our knowledge at present nothing can be done to change that. But clearly at the other end of the spectrum (and of life), as in Alzheimer’s disease, lifelong physical and mental self-care in terms of diet, exercise and cognition do seem to make a difference in terms of prevention – involving personal choices throughout life. Equally, however, we can never ‘blame’ someone for developing the disease because we would never know in an individual case what role might have been played by such environmental factors. Genetics is based on population averages and probabilities. For the other conditions mentioned here, the choice to take certain mind-altering drugs does appear to play a role in some cases, as in the development of schizophrenia. We will be returning to the discussion on genetic determinism and how that relates to free will in Chapter 11. This has been a bit of a gloomy chapter, although things should generally cheer up a bit from now on. But medical behavioural genetics is hugely valuable in helping us to understand the role of genetic variation in the wide range of behaviours observed in human populations.

https://doi.org/10.1017/9781108566520.006 Published online by Cambridge University Press

6

Genes, Education and Intelligence

B

obby moore was sentenced to death in texas way back in 1980 after he fatally shot a 73-year-old clerk during a Houston robbery. In 2014, a Texas court determined under current medical standards that Moore was intellectually disabled – with evidence including low IQ scores and his inability to tell the time or days of the week as a teenager. In some US states, but not all, mental disability must include an IQ value of less than 70 to avoid capital punishment. Below an IQ of 70 and you live, above 70 and you die. So in some parts of the world accurate measures of IQ can be matters of life and death. In practice in the present case, the US Supreme Court battled with the courts in Texas over the issue and Moore’s death sentence was finally changed to a life sentence in late 2019 after decades of struggle, based on Moore’s intellectual disablement.1 Generally people are quite accepting of the aim to find out more about the role of genetics in the kind of medical disorders reviewed in Chapter 5. As noted there, the dominant role of genetics in some disorders is a relief to many parents as they realise that it really wasn’t their fault that their child totally unexpectedly had developed autism (for example) ‘out of the blue’. But come to the question of the role of genetics in intelligence and educational attainment, or measurements of IQ in cases of criminal justice, and suddenly the tone changes and the hackles go up. This is for very understandable reasons. The topic has a nasty eugenic history and even to the present day, as we will discuss later, there are those who would wish, mistakenly in my view, that genetic information should play a role in educational policy. But for the moment let us just focus on the science and see where that leads us.

94

https://doi.org/10.1017/9781108566520.007 Published online by Cambridge University Press

6.1 WHAT IS INTELLIGENCE?

95

6.1 What is Intelligence? ‘Few constructs are as mysterious and controversial as human intelligence. One mystery is why, even though the concept has existed for centuries, there is still little consensus on exactly what it means for someone to be intelligent or for one person to be more intelligent than another’ (Davidson and Kemp, 2011). Indeed, all definitions continue to remain disputed to the present day. ‘Emotional intelligence’ is not always top of the list of personal qualities when seeking to employ top scientists. Indeed, some people might unkindly suggest that there is an inverse relationship between academic and emotional intelligence – but surely that is going too far! More valued in South Korea might be the kind of intelligence associated with nunchi, which is the art of intuiting what people are thinking, and learning how to anticipate the needs of others.2 The idea came into Korea 2,500 years ago with the teaching of Confucius and is now a well-developed cultural practice. I suspect that low nunchi scores in Korea might correlate with low scores on an imaginary emotional intelligence scale in Western countries. Whilst recognising the difficulties involved in agreeing on some universal cross-cultural definition of ‘intelligence’, it’s simply a fact that the main focus on measuring intelligence in Western education systems involves assessing analytical skills such as verbal dexterity, logicalmathematical skills and problem-solving ability, precisely the set of skills that help in making good progress through Western education or in obtaining a good job in companies that use psychometric testing for their applicants (‘psychometric’ simply refers to the battery of tests now available to assess cognitive and other abilities). It is certainly the case that in Western countries, intelligence is largely related to one’s cognitive abilities. However, ‘[t]he overall picture is that intelligence is defined and perceived differently by people from different parts of the world, and that these differences are largely reflective of long-standing cultural traditions’ (Ang et al., 2011). This is all highly relevant to questions concerning the measurement of intelligence and, therefore, to the genetics of intelligence. In turn, this is part of a wider discussion as to what counts as a ‘trait’ and whether a ‘trait’ is a local cultural construct or one that is genuinely universal. A ruler cannot be used to measure an invisible entity of unknown dimensions. In reality, there are many different ‘intelligences’ – systems of abilities – only a few of which can be captured by the standard psychometric testing techniques discussed further below.

https://doi.org/10.1017/9781108566520.007 Published online by Cambridge University Press

96

6 GENES, EDUCATION AND INTELLIGENCE

6.2 IQ and Intelligence Testing The IQ acronym is now very dated but is still around, more in some parts of the world than in others. During the past century, hundreds of intelligence tests have been developed, both by theorists and by commercial testing companies. They are used, particularly in the USA, by schools and universities, military forces, governments and employers for a variety of reasons including clinical diagnosis, school admissions, performance assessment and testing job suitability. In the USA, such scores can have an enormous impact on life outcomes: a score can mean the difference between being offered a college place or a job. As already noted, it can even mean the difference between life and death. Despite a steady barrage of critiques, including those who question the whole rationale of IQ testing, a large number of countries have embraced intelligence testing with almost equal enthusiasm to the USA, particularly in the developing world, where there is a pressing need to use limited public educational and employment resources effectively. In the UK, IQ testing is largely past history, except in academic research studies and in the diagnosis of certain medical conditions entailing cognitive disability. What do IQ tests attempt to measure? IQ stands for ‘intelligence quotient’, but in reality IQ is not a quotient, which is a bit confusing. The history of the idea has been described elsewhere (Murdoch, 2007, Urbina, 2011). In brief, several different types of test, usually around a dozen verbal and non-verbal tests, are used to obtain a ‘mental cognition score’, which is matched against the average in a population for that particular score for that particular set of tests. This score is then divided by the person’s actual age and then multiplied by 100 to obtain the IQ score. So let us say that I’m 16 years old and that my mental age score is exactly average for someone aged 16: 16/16 × 100 = 100. In other words, someone with an IQ of 100 is exactly average for that population. But let us say that I’m still aged 16 but my mental age score now matches that of a 22-year-old – my IQ would now be above 130, which would put me in the top 2.5 per cent of the population and eligible to join the high-IQ society known as MENSA. Those with an IQ of less than 70 would be in the bottom 2.5 per cent. You can now see why IQ is not a quotient, which simply means what you get when you divide one number by another. Rather, it is a way in which the ability to do certain groups of tests can be compared with how good many other people of a certain age are at doing the same tests. Only a small number of test groups still refer to the measurement obtained as an IQ score from historical habit, although some of these are the most used, ensuring the current preservation of the IQ terminology.

https://doi.org/10.1017/9781108566520.007 Published online by Cambridge University Press

6.2 IQ AND INTELLIGENCE TESTING

97

In today’s intelligence testing, there is a wide range of batteries of tests to choose from, some designed for clinical use, others for general adult use, others for children and so forth. Each test battery is composed of between ten and twenty subtests, each of which contains multiple items (questions). Some items are timed, some are not. Each subtest measures a different facet of intelligence, such as visual–spatial reasoning or short-term memory, and subtests are often put together to yield scores for specific domains of intelligence, such as perceptual reasoning or comprehension. There is no agreed set of subtests that must be used in a battery. Instead, modern subtests are designed or reused based on the specific theoretical model being used by the test designers. The lack of agreement on a theory and structure of intelligence is the major reason for so many intelligence tests, although many share similar or overlapping components (Gottfredson and Saklofske, 2009). IQ scores are often misinterpreted or misrepresented, so it is worth summarising the most common misunderstandings. First, an IQ score is not a fixed or absolute score, and therefore has no intrinsic or absolute meaning – it is entirely based on comparison with the standardisation group. So IQ values change around every 10 years, depending on the recalibration of IQ based on the standardisation group, so that the average IQ remains at 100, which is of particular relevance if you happen to be in a US state prison in which the IQ value is used to decide the application of the death penalty.3 Second, the standardisation group must be representative of the test-taker, otherwise the result will make no sense. For example, a 15-year-old’s IQ score would be much lower if calculated using a standardisation group of 50-yearold college graduates rather than a standardisation group of other 15-yearolds. Third, an intelligence test score by itself describes only the reality of the test-taker’s cognitive ability at the time they took the test. Without more information, it cannot say anything about the degree to which the score is dependent on hereditary or environmental influences. Fourth, IQ scores cannot predict the potential future development of intelligence for a given individual. For example, someone whose score is affected by adverse environmental factors (e.g. childhood malnutrition) may receive a higher IQ score later in life if their environmental conditions improve. Fifth, because of the re-standardisation of scores for historical consistency, IQ scores appear to be directly comparable. However, comparisons between scores obtained on different tests should be made with care. Each test battery measures a slightly different facet of cognitive ability, although it has been shown

https://doi.org/10.1017/9781108566520.007 Published online by Cambridge University Press

98

6 GENES, EDUCATION AND INTELLIGENCE

that test batteries correlate very highly with each other, due to their overlapping nature (Urbina, 2011). Most now working in the neurodevelopmental field think that when trying to understand differences among children and the reasons why some children struggle at school, IQ is not a particularly helpful concept. Instead, most investigators tend to study particular cognitive skills such as working memory, executive control, attention, specific language skills and phonological skills. These provide a more mechanistic account of learning differences and in most cases are better predictors of specific aspects of learning, and there is a far better idea of the cognitive and neural processes underpinning these more specific skills, which ultimately therefore provide better targets for interventions.4

6.3 General Intelligence or ‘g’ One of the many controversial aspects of intelligence testing has been, and continues to be, the measurement of so-called ‘general intelligence’ known as ‘g’, now often referred to as ‘general cognitive ability’ (Plomin et al., 2013a, Bouchard, 2014). This arises from the fact that test results are correlated with each other, and g is a measure of how well different test results correlate. In fact, all tests correlate with each other positively, although measures of spatial and verbal ability (for example) correlate with each other more highly than other measures, such as nonverbal memory tests (Plomin et al., 2013a). Some tests contribute more to g than others in a way that is related to the cognitive ability being assessed; for example, abstract reasoning provides a better assessment of the value of g than less complex cognitive processes such as simple sensory discriminations. Imagine a matrix containing the values of a wide range of psychometric tests measuring differing aspects of cognitive ability; now measure the correlations between the values in all the different tests – it is the highest set of correlations that contributes most to the value of g. The validity of g has been promoted on the grounds that its value has long-term stability after childhood, reportedly greater than the stability of any other behavioural trait (Deary, 2012). Furthermore, use of different intelligence tests on various populations, including twins reared apart, has generated identical or nearly identical values for g. The value of g also correlates positively with many life outcomes, such as good health, occupational status and educational attainment, although on the face of it, the last factor is not very surprising, given that the Western educational system depends to a considerable degree on precisely the cognitive

https://doi.org/10.1017/9781108566520.007 Published online by Cambridge University Press

6.3 GENERAL INTELLIGENCE

99

abilities that psychometric testing measure, and that psychometric methods are influential upon the structure of the various exams necessary to gain entry to US universities. Furthermore, the degree of motivation makes a big difference to test scores in young people (Duckworth, 2011), so low values may reflect a kind of self-fulfilling prophecy in which those with low motivation score poorly in psychometric tests and thereby fail to gain entry to the best universities. If differential motivation is influenced by genetic variation, for which there is some evidence, then perhaps the heritability of intelligence test results is more to do with motivation than with intelligence per se. What exactly does g represent? No one really knows. It clearly represents a statistical construct: the correlation of abilities as assessed by certain scores obtained from psychometric intelligence tests. Some theories point to a single main mechanism underlying g, such as speed of brain information processing, whereas other theories suggest a combination of different cognitive abilities. There are plenty of reported correlations between values of g and various aspects of brain anatomy, although cause-and-effect relationships in such human studies are notoriously difficult to establish. For example, the precise thickness of the thin layer of brain cells and their synaptic connections on the top surface of our brains tends to vary a bit among individuals. Using brain scanning techniques, the thickness can now been measured – known as ‘cortical thickness’. There is a correlation between general intelligence and cortical thickness, and genetic variation is involved in that correlation (Schmitt et al., 2019). Others have pointed to specific features of brain cellular functioning that may contribute to levels of g (Geary, 2018) and to the relative speed of brain processing (Schubert et al., 2017). There are many other studies of this kind, but the problem is in knowing whether biological differences measured are the result of the person in question being a high achiever and therefore using their brain a lot at full cognitive capacity, or whether the brain differences emerge during early development in response to a high level of genetic informational input. Or maybe both. The current variation in the theories reflects the current lack of knowledge in the neurosciences as to how measured differences in cognitive abilities translate into variation in brain mechanisms and vice versa. But given the complexity of ‘intelligence’ as a trait (if it can be counted as one), and given the complexity of the brain, the most likely interpretation is that there is a huge array of brain mechanisms that underlie observed differences in cognitive ability.

https://doi.org/10.1017/9781108566520.007 Published online by Cambridge University Press

100

6 GENES, EDUCATION AND INTELLIGENCE

6.4 The Heritability of Intelligence The heritability of the main products of psychometric testing – IQ and general intelligence – has been the initial focus over many decades of the attempts to investigate the role of genetic variation in intelligence. As already discussed, to what extent such parameters can be equated with ‘intelligence’, broadly defined, is open to question, but as the literature is dominated by the genetics of the fruits of psychometric testing, that will remain the focus here. The results are full of surprises. First, let us make a comparison with the heritability of body weight. This is 95 per cent for children aged 5 years, but then heritability decreases into adulthood until it is around 60 per cent by 50 years of age. This is because people choose their food intake and their level of exercise as they grow older, so environmental factors become more important. Genetic factors certainly do not disappear, as we shall be considering further in Chapter 8, but proportionally they contribute to less of the variance in the population. Now compare this with IQ. Biometric tests can start being used on children as young as 4–5 years old at which age the heritability of IQ is around 22 per cent. By the age of 16, the heritability is 62 per cent and by the age of 50 it is around 80 per cent (Sauce and Matzel, 2018). So the trend, at least as far as heritability values are concerned, is exactly the opposite to body weight. What’s going on? A clue may come from the gene–environment (G × E) interactions mentioned in Chapter 3. It may be remembered that such interactions can have two aspects. The first involves a correlation between genes and environment. So if there are some small differences in intelligence in young children, as they grow up the more intelligent will tend to flourish in those environments that further stimulate their cognitive enhancement. Conversely, those less intelligent in very early life will end up in less cognitively enhancing environments. So the idea is that the genetic differences that make minor differences in early life become amplified by personal choice of environments in later life. We like what we like doing because of who we are, and who we are is partly dependent on our particular genome, and so as we choose what we like doing in life, that tends to amplify the genetic effects. The second G x E aspect involves different types of response to environmental changes as a child grows up, which reflect their own particular variant genome. So once again, these changes will tend to amplify the contribution of the genes to the heritability in a population as life goes on, as more and more G × E interactions amplify the role of these genetic differences. For example, different children will respond differently to friends at school (or a lack of

https://doi.org/10.1017/9781108566520.007 Published online by Cambridge University Press

6.4 THE HERITABILITY OF INTELLIGENCE

101

friends), or to their experience of school itself, or to the socio-economic status of their peers. So the direct actions of genes are not the only way in which genes are thought to impact on intelligence; the indirect actions of genes via the two types of G × E interactions are also important. The increase in IQ values with increasing age, together with the contrasting effects of different socio-economic environments and other environmental variables in different countries, no doubt play a significant role in the widely differing heritability values for IQ reported from different parts of the world. One report described a meta-analysis of more than 200 studies that estimated the heritability of IQ, including twin, family and adoption studies, and involving more than 50,000 pairs of relatives, concluding that the heritability of IQ is 34 per cent (Devlin et al., 1997). All such studies, as their authors generally emphasise, highlight the fact that environmental factors contribute a large slice of the variation in IQ as measured in a given population. For example, in the study by Devlin, modelling suggested that the in utero environment was an important factor in contributing to the observed variation in the populations being analysed (Devlin et al., 1997). Given the close statistical links between IQ measurements and ‘general intelligence’, it is no surprise to find that heritability estimates for g are roughly in the same range as for IQ, although the precise values depend once more on which particular method is being used. As emphasised in Chapter 3, in any event it is not the precise heritability values that are of most interest, but the fact that the heritability values are not zero does suggest that genetic variance is important in contributing to the variance that exists in populations with respect to IQ and g values generated via psychometric testing. This is, of course, of no surprise when considering such measurements in the light of the developmental biology discussed in Chapter 3. Given the fact that the majority of protein-coding genes in the genome are expressed at various stages of brain development, it would indeed be surprising if genomic variation among individuals made no difference at all to a person’s cognitive abilities. As a reminder, at the level of the individual, complex human behavioural traits, which include ‘intelligence’ however that may be defined, can be described as 100 per cent due to the environment and 100 per cent due to genomic information, integrated to generate unique human individuals with traits that vary when measured in populations, as discussed in Chapter 4. Measurements of the heritability of IQ and g have value in emphasising this fact. The powerful effects of the environment in changing IQ values are well illustrated by the many adoption studies that have been carried out on IQ. For example, in a French adoption study, a group of impoverished children

https://doi.org/10.1017/9781108566520.007 Published online by Cambridge University Press

102

6 GENES, EDUCATION AND INTELLIGENCE

with an average IQ of 77 were adopted at the age of 4–6 years old (Duyme et al., 1999). When their IQ was subsequently measured in adolescence, the individual values showed a significant correlation with their pre-adoption values, consistent with a role for their variant genomes. But of most interest was the finding that the average rise in IQ was dependent on the socioeconomic status of the adopting family – an average gain of 7.7 IQ points in the families of low socio-economic status compared with a gain of 19.5 IQ points in the adoptive families of high socio-economic status. This is a striking result. As has been pointed out (Sauce and Matzel, 2018), by comparison a successful college graduate in the USA is 15 IQ points above the average, so a gain of nearly 20 IQ points is highly significant. Similarly, a meta-analysis of sixty-two studies from a multitude of countries (totalling 18,000 adopted children) found an average increase in IQ of 17.6 points within several years of adoption (van Ijzendoorn et al., 2005). In terms of average differences in IQ, such studies (and there are many others) point to a striking role of environments in influencing IQ values, so the results of biometric intelligence tests should not be thought of as something fixed that stays with someone for life, although at the same time genetic variation in conjunction with early developmental biology seems to have quite a bit to do with the ‘likely range’ of IQ values expected in response to different environmental experiences throughout life. These results from adoption studies highlight one of the many challenges involved in interpreting data based on adoption. One complication arises from the so-called ‘range restriction’ (Joseph, 2010). Adoptive parents represent a specially selected and therefore non-representative population. For example, there is the selection of families who wish to adopt a child by adoptive agencies, including the strict criteria used in allowing a family to adopt, plus the decision of the family to adopt in the first place plus the further decision to allow their children to be included in an adoption research study. As far as the adoptees are concerned, many come from single teenage mothers who may be poorly placed economically and educationally to raise their child well in the early months or years of life prior to adoption, whose offspring may therefore receive ‘sub-optimal obstetric care’ (Rutter, 2006). Furthermore, adoptees as a population are often at a greater risk for being diagnosed with a psychiatric disorder compared with the general population. In a helpful illustration used by Eysenck and Kamin (1981), we may imagine the situation with boxers who have been organised into subdivisions according to weight. Given that fights can only take place between boxers of similar weight, the correlation between weight and boxing success is

https://doi.org/10.1017/9781108566520.007 Published online by Cambridge University Press

6.4 THE HERITABILITY OF INTELLIGENCE

103

therefore necessarily low. In terms of the cohorts of adoptive parents, it is as if they are all in the heavyweight division, because they have been selected to represent a rather special category of families, and so this will contribute to a lower parent–child IQ correlation observed in adoptive families. But in the unlikely event that poor families were selected to adopt more often, then the correlation would presumably be much higher. We might also erroneously infer from all this that there is no inherent relationship between weight and boxing success, but we would be quite wrong. So even though adopted children may correlate more individually with their biological parents than with their adoptive parents, as a group they might in fact be more similar to their adoptive parents than to their biological parents, which is precisely what the studies of the kind mentioned above have shown. ‘Thus, focusing on parent-child correlations at the expense of evaluating group mean IQ differences, as behavioural geneticists frequently do, can paint a misleading picture of the potential roles of genetic and environmental influences on intelligence’ (Joseph, 2010). Many other environmental variants have been found to correlate with differences in IQ. For example, a study led by Christopher Eppig at the University of New Mexico, Albuquerque, suggested that infections may be a critical factor affecting the maturation of the brain, thereby impacting on the results from intelligence testing (Eppig et al., 2010). Eppig assessed World Health Organization (WHO) statistics on the number of disease years caused by the twenty-eight most common infections using data from 192 countries in Africa, Asia and the Pacific region. Data on intelligence tests were available from 142 countries, allowing the investigators to estimate that 68 per cent of the differences in mean test scores among the countries could be attributed to their levels of childhood infection. This correlation was the strongest of all the variables tested, including educational level, malnourishment and gross domestic product. The usual proviso that correlation does not equate with causation is important in this context, but the findings represent but one striking example, out of dozens that could be cited, of correlations between IQ, g and environmental factors. A further observation, known as the ‘Flynn effect’, has also led to much discussion concerning the relative impact of genetic variation and varying environments on IQ measurements. The observation is that the average IQ in different countries has been rising steadily over the past half-century or more since IQ measurements started to become available (Flynn, 2012). For example, American average IQ has been increasing at a rate of 0.3 points per year over the past half-century. The same tendency has been observed around the world. South Korean children’s IQ increased at double the US

https://doi.org/10.1017/9781108566520.007 Published online by Cambridge University Press

104

6 GENES, EDUCATION AND INTELLIGENCE

rate over the period 1989–2002 (Flynn, 2012). Children in urban Brazil (1930–2002), Estonia (1935–1998) and Spain (1970–1999) made gains similar to the US rate of increase. As always in IQ testing, the results are only meaningful if the same or highly similar tests are used. Between 1952 and 1982, young Dutch males gained 20 IQ points on a test of forty items selected from Raven’s Progressive Matrices, a particular widely used intelligence test that measures what is termed ‘fluid intelligence’, the ability to solve nonverbal problems on the spot without a previously learned method for doing so (often contrasted with ‘crystallised intelligence’, the ability to apply acquired knowledge). This is of particular interest as the Raven’s test has been designed to be largely independent of cultural differences. The gains in the Dutch study could not be dismissed on the grounds that the cohort being tested was maturing earlier than their forebears. Many theories have been put forward to explain the Flynn effect (Hunt, 2011). What is quite clear is that the changes are taking place far too quickly for genetic variation to play a role: half a century is a very short period in relation to the timescale required for significant changes in population genetic diversification. James Flynn himself prefers to explain the changes by invoking the Industrial Revolution as an ‘ultimate cause’, with the intermediate causes being ‘probably its social consequences, such as more formal schooling, more cognitively demanding jobs, cognitively challenging leisure, a better ratio of adults to children, richer interaction between parent and child’ (Flynn, 2012). Flynn argues that the great gains noted upon use of the Raven’s test suggest that people have learnt how to tackle problems less concretely and more abstractly over the years. A massive analysis in order to address this question was carried out based on 271 independent IQ measurements obtained from 4 million participants in 31 countries spanning a century (Pietschnig and Voracek, 2015). The authors of this study concluded that the interactions between genes and environment of the type suggested by Flynn, together with improvements in education and nutrition, provide the most likely explanations. The main point in our present context is that there is nothing ‘fixed’ about the average values obtained from intelligence testing on populations – powerful environmental effects are at work. From a practical perspective, US courts have taken the Flynn effect seriously in terms of assessing the IQ values of those facing the death penalty (Flynn, 2012). Overall, it is impossible to deny that genetic variation in a population makes a difference to the variation in the values obtained from the biometric tests that lead to values such as IQ and general intelligence. But equally, the massive roles played by environmental differences, and especially by G ×

https://doi.org/10.1017/9781108566520.007 Published online by Cambridge University Press

6.5 THE MOLECULAR GENETICS OF INTELLIGENCE

105

E interactions, are vital in gaining a balanced picture. As always, it is a question of both/and, not of either/or.

6.5 The Molecular Genetics of Intelligence If the findings summarised in the previous section are correct, then it should be possible to identify genetic variants that are involved in the variation in biometrically measured intelligence in a given population. Only recently have such studies begun to achieve some degree of assurance that the gene variants so identified really do contribute to the overall heritability of intelligence (Sniekers et al., 2017, Savage et al., 2018). Earlier cohorts being studied, as so often has been the case in the field of behavioural genetics, were simply not large enough to pick up genetic variants that displayed a significant association with the variation in intelligence as measured in various ways, including IQ. This has now changed. For example, one study examined the variation in the general intelligence (g) in more than a quarter of a million genotyped people (Savage et al., 2018). This led to the discovery of 205 variant DNA regions containing SNPs that are associated with variation in intelligence among this large group of individuals. Gene mapping studies then enabled the identification of 1,016 genes involved in the population variation (remember that SNPs identify regions of the DNA that may contain several genes). This sounds like a large number, but it’s good to remember that at least half the protein-coding genes in our bodies are involved in brain development, so it wouldn’t be surprising if eventually thousands of variant genes are involved in contributing to the variation in intelligence in a population. But that would probably involve studying groups of several million or more to achieve such a data set. Apart from the large number of variant genes identified, several other interesting observations came out of this study (Savage et al., 2018). With such a large set of identified genes, it then becomes possible to map them to previously identified groups of genes involved in specific aspects of brain function. So it was shown that the big gene set includes sets of genes involved in brain cell replication, synapse formation, regulation of brain development and other aspects of the nervous system that make perfect sense when thinking about gene products that contribute to more efficient brain functioning. The big set of variant genes identified also contained subsets of genes that correlated strongly, either positively or negatively, with other behavioural

https://doi.org/10.1017/9781108566520.007 Published online by Cambridge University Press

106

6 GENES, EDUCATION AND INTELLIGENCE

traits. For example, there was a strong correlation, perhaps not surprisingly, with educational attainment. There was a strongly protective effect of intelligence on schizophrenia, meaning that those with a high number of gene variants for intelligence would be less likely to suffer from this disorder. Higher intelligence was also found to correlate with a risk for autism but had a protective effect on the development of Alzheimer’s disease. In another study, genes associated with high intelligence were also associated with the development of bipolar disorder (Smeland et al., 2019), which could have something to do with my own brother’s development of this disorder. So such studies are useful in helping in the challenging task of disentangling the different brain pathways involved in both health and disease.

6.6 The Genetics of Educational Attainment Educational attainment is defined in all studies as the number of years of schooling completed, including university. The heritability of educational attainment has been estimated at 40 per cent (Rietveld et al., 2013), and 57 per cent for university exam achievement in a UK population (SmithWoolley et al., 2018), so clearly genetic variation has something to do with the number of years that people spend in education. By this stage, it will not be surprising to hear that GWAS have been used on a cohort of more than a quarter of a million people to identify seventy-four different DNA regions that associate with educational attainment (Okbay et al., 2016). Many of these regions were found to be involved in regulating gene expression in the fetal brain. Candidate genes were preferentially expressed in neural tissue, especially during the prenatal period, and were particularly involved in biological pathways involved in neural development. As with the correlation studies mentioned above for general intelligence, genes associated with higher educational attainment were found to correlate highly with increased cognitive performance (no surprises there), an increased risk of bipolar disorder and a decreased risk of Alzheimer’s. By increasing the cohort under study to 1.1 million individuals, no fewer than 1,271 SNPs that associate with educational attainment have now been discovered in a more recent study, the results being virtually identical between men and women (Lee et al., 2018). Many of the new gene variants are relevant for all the various processes involved in communication between brain cells. By adding up all their weighted contributions, a polygenic score of 11 per cent was obtained, meaning that the 1,271 SNPs can, so far, explain 11 per cent of the variance in educational attainment in the population with European ancestry under study. However, as the authors point out, GWAS

https://doi.org/10.1017/9781108566520.007 Published online by Cambridge University Press

6.6 THE GENETICS OF EDUCATIONAL ATTAINMENT

107

scores can make estimates that are too high because they don’t include the G × E effects already discussed in which, for example, environmental influences amplify the genetic effects. Also, the SNPs that correlate with intelligence and educational enhancement vary a lot between different populations, so the polygenic score will vary a lot depending on where the SNP collection came from. It’s also good to remember in this context that the total heritability is 40 per cent – so if the numbers are correct, then there’s still quite a long way to go to get from a polygenic score of 11 per cent to one of 40 per cent, suggesting that eventually perhaps as many as around 5,000 associated SNPs will be necessary to obtain a complete genetic score. But as we’ve already noted, a big proportion of the genome is needed to provide information for brain development, so such large figures are not unexpected. Another interesting output of this particular study was that many of the SNPs identified overlapped with those identified by GWAS on other related traits already known to correlate with enhanced education. Giving the number of overlapping SNPs in brackets, these correlations included high performance in biometric intelligence tests (225) and self-reported ability in maths (618). Once again, underlying many of our distinctive phenotypes are common sets of genetic variants that influence a broad spectrum of traits. The ability to generate polygenic scores for educational attainment from the mass of SNPs now being identified has opened the door to an intriguing approach to start disentangling the complex relationship between genetic and environmental influences (Kong et al., 2018). One study involved genotyping Icelandic parents and children, and investigating the genetic variants in the parents (that correlated with higher education) that were not passed on to their offspring. The idea here is that the genetic variants in the parents that influence them in the direction of higher education in turn then help to create an environment for their children in which they are more likely to pursue higher education themselves in due course. In fact, the authors of this study calculated that the polygenic score estimated for the genetic variants in the parents that correlated with higher education and that were not transmitted to their offspring was 30 per cent of the polygenic score for the variants that did influence educational attainment in the offspring. In other words, the genetic influence in the parents that had an indirect environmental influence on their children was quite large. If all this sounds a bit of a convoluted argument, well it is – but is interesting because it illustrates the cake-cooking metaphor well. It’s not just the genetic recipe in the cake that counts but also the genetic influences that help the cook to perform well. It is of possible concern that attempts are being made to correlate geography with the gene variants that associate with educational attainment. In

https://doi.org/10.1017/9781108566520.007 Published online by Cambridge University Press

108

6 GENES, EDUCATION AND INTELLIGENCE

one study, thirty-three traits were analysed in around 450,000 individuals in the UK (Abdellaoui et al., 2019). Lower polygenic scores relating to educational attainment were reported to correlate with those living in what was once a coal-mining area, whereas higher educational attainment scores correlated with those who had recently migrated away from the area. The authors of the study were quick to point out that ‘it’s very difficult to say anything about directions of causality’ (Adam, 2019). Indeed. Such studies are open to the accusation of pointlessness, especially when they are readily open to eugenic misinterpretations.

6.7 So What? By this stage, it would not be unusual for the ‘So what?’ question to be posed. What difference does it really make to our lives if we now know that thousands of genetic variants have some influence on the differences that we observe among people in traits like intelligence and educational attainment? One author commented in the introduction to his paper on this topic that ‘Few discoveries would have greater impact than identifying some of the genes responsible for the heritability of cognitive abilities’ (Plomin et al., 2013b). That might depend on what kind of impact one is looking for. From a brain research perspective, as we have already noted, the overlap of large sets of identified genetic variants among different traits can certainly provide some important leads for different brain pathways and mechanisms to explore further. In this context, the large groups of relevant genetic variants identified are likely to be more important than investigating the roles of single genes. After all, if a single gene contributes 0.001 per cent influence to the overall variance in a trait, then that hardly gives one a great deal of confidence that its influence as a single gene is worth investigating. But if a large group of variant genes involved in a specific brain function make, for example, a 3 per cent contribution to the variance, then that might provide a more hopeful starting point for the investigation. Leaving the basic research questions on one side, other conclusions from the current data still remain. For example, the results make clear that enthusiastic transhumanists who think that genetic engineering might be used to generate a cohort of more intelligent humans are living in cloud cuckoo land. In this respect, the complexity of the genome remains its best defence against ill-advised manipulation. There are several points to keep in mind in this context. First, apart from the immense practical difficulties in genetically engineering hundreds or even thousands of genes scattered throughout the genome, even if this were feasible, the ‘optimal genetic variant collection’ for intelligence or

https://doi.org/10.1017/9781108566520.007 Published online by Cambridge University Press

6.7 SO WHAT?

109

educational attainment still wouldn’t guarantee that the person in question would score highly in these aspects of life. The role of the variants is probabilistic based on population studies – and these provide no guarantees for any given individual. Environmental factors also play a major role. For example, the genetically engineered person designed to be superintelligent might simply decide to be a stand-up comedian. Nothing wrong with that at all, just that maybe that wasn’t really the aim of the exercise. And in any case, if the experiment was successful, would one really want to generate a new class of ‘superintelligent’ children who would bring yet more fragmentation into social and educational life? We have enough of that already. It’s also worth remembering that some of the most intelligent people in the world have been the most evil. All this is relevant to the announcement by Stephen Hsu, Senior Vice President for Research and Innovation at Michigan State University and CEO of the company Genome Prediction, that during IVF it would soon be possible to rank embryos according to their relative genome scores for IQ and then implant just the embryos with the highest scores.5 But even if one thought this was a good idea on ethical grounds (personally I do not), there are also many practical difficulties that provide barriers against going in this direction. For example, even if there were lots of embryos to choose among, the range of polygenic scores is unlikely to be that great, for the simple reason that the embryos come from just a mother and a father, so the variant ‘SNP range’ would be restricted indeed. And in any case, as our discussion on genetic engineering indicates, there is no guarantee that the particular embryo choice made would have the outcome expected. As we have also noted, high intelligence correlates with conditions such as autism and bipolar disorder, so if you select for trait A, then you might also bring along with it traits B, C and D, which could be less what you want. Should our new ability to generate polygenic scores for enhanced education make any difference to educational policy or practice? It’s hard to see how or why this should be the case. Every child should be given the educational opportunities that will enable the child to achieve their full potential. Their full potential might well be influenced to some degree by their genetic variation, but the only way to find out that potential is to educate them as efficiently and as fairly as possible. We don’t need any knowledge of genetics (except for medical genetics on occasion) to achieve that goal. Discussing the role of genetic variation with respect to intelligence and educational attainment can too easily lend itself to arguments that involve inequality in social or educational practice, with the spectre of embryo

https://doi.org/10.1017/9781108566520.007 Published online by Cambridge University Press

110

6 GENES, EDUCATION AND INTELLIGENCE

choice hanging in the background. Eugenic ghosts haunt us. In Chapter 12, we will be picking up these concerns in more detail and thinking about how human identity and equality can be preserved in the face of such threats. But for the moment we need to pursue our tour through the labyrinth of genetic diversity that guarantees that we are all different in our personalities.

https://doi.org/10.1017/9781108566520.007 Published online by Cambridge University Press

7

Genes, Personality and Personality Disorders

I

n 2006, bradley waldroup, who lived in a trailer home in the mountains of Tennessee, USA, got into a violent argument with his wife and her female friend, as he believed they were having an affair. In a sudden escalation of violence, he shot and killed his wife’s friend, shot his wife in the back as she tried to flee, then dragged her inside and cut her with a knife and a machete, before she was saved by the intervention of the police. Waldroup was charged with first-degree murder, attempted first-degree murder and aggravated kidnap. During the penalty phase of the trial, the defence team introduced expert genetic testimony, seemingly without any objection from the court, citing data that was generally well accepted at the time (but now no longer accepted) claiming that a particular genetic variant contributed to aggression. The defence argued that Waldroup was not able to act as a ‘reasonable man’ might have acted in the same situation due to his genetic predisposition, making the act less volitional. The jury proved receptive to this argument. Waldroup was cleared of first-degree murder and found guilty of voluntary manslaughter of the friend, attempted murder of the wife and especially aggravated kidnap, meaning he was no longer eligible for the death penalty, but was sentenced to 32 years in prison. Comments made by the jurors afterwards showed that the genetic evidence was a crucial part of this decision. When one member of the jury was asked if their decision was swayed by Waldroup’s genetics, the juror responded, ‘Oh I’m sure . . . . And his background – nature vs. nurture’ (Denno, 2013). ‘One juror even remarked that the MAOA evidence suggested – [e]vidently it’s just something that doesn’t tick right . . . . Some people without this would react totally different than he would’ (Walker, 2013). It is perhaps not surprising that a later feature on the case by National Public Radio was headlined ‘Can your genes make you murder?’1

111

https://doi.org/10.1017/9781108566520.008 Published online by Cambridge University Press

112

7 GENES, PERSONALITY AND PERSONALITY DISORDERS

Our topic for this chapter is not merely theoretical – it has many implications for our families, our daily lives, our work experience and our legal systems. It may seem a little unfair to highlight personality disorders in a chapter about personality given that there are many other more positive aspects of personality that could be emphasised – such as kindness, agreeableness and sociability. But the fact of the matter is that not only are personality disorders, such as aggression, a worldwide social and political challenge, but also the literature on their genetics contains many fascinating stories that are highly relevant to our main aim: to explore the impact of genetic diversity on differences between people. But first to personality itself.

7.1 Personality and Heritability ‘Elizabeth is such a pleasant person’ we may say, ‘whereas you know her brother Robert? He’s just so shy, hardly says a word.’ We all know what the word ‘personality’ means because we use it all the time in our daily speech. But if asked to define the word, we might find that a little trickier. It’s not really a ‘thing’ but rather a complex list of average behaviours that make us describe people with different labels, none of the labels by themselves being sufficient to do justice to the whole picture. So research on the connections between genetic diversity and personality immediately raises the question as to how personality can best be broken down into a collection of traits that can then be measured and assessed. Fortunately, there is a vast literature from the field of psychological testing that provides a good starting point. Research has focused on five broad dimensions of personality called the ‘Five-Factor Model’ (FFM) (Goldberg, 1990). These include the best-studied traits such as neuroticism, involving moodiness, anxiety and irritability, and extraversion, which involves traits such as sociability, impulsiveness and liveliness. The five traits included in the FFM lead to the acronym OCEAN: Openness to experience (culture) captures intellectual curiosity and creativity; Conscientiousness (conformity, will to achieve, order and discipline versus undependability); Extraversion (versus introversion); Agreeableness (likeability and friendliness versus antagonism); and Neuroticism (versus emotional stability). Personality scores for these five different traits are measured using questionnaires generating graded quantitative scores for each trait. Extensive twin studies carried out in five different countries and involving 24,000 pairs of twins have reported heritability values of about 50 per cent for extraversion and 40 per cent for neuroticism (Loehlin, 1992). Overall heritability values in the range of 30–50 per cent are typical for measured

https://doi.org/10.1017/9781108566520.008 Published online by Cambridge University Press

7.1 PERSONALITY AND HERITABILITY

113

personality traits based on twin studies (Vukasovic and Bratko, 2015), although values from adoption studies tend to be lower for reasons discussed below. Many different traits have been studied, including agreeableness and ‘grit’, also known as conscientiousness. In all cases, the heritability is around 40 per cent and the environmental component of the variance is accounted for almost entirely by non-shared environmental effects (Loehlin, 1992, Vukasovic and Bratko, 2015). This, of course, does not mean that a shared environment is not vitally important in the development of personality in the life of an individual child in a specific home. The point in the present context is that the common home environment does not seem to account for the variance in personality traits observed among siblings. Instead around 30–50 per cent of the variance in a particular personality trait appears to be attributable to genetic variance and around 50–70 per cent to non-shared environmental influences outside the home, such as different schooling experiences, different peer group pressures, personal choice of environments, random life events and so forth. Furthermore, unlike the development of children’s cognitive abilities, which nearly all parents deem to be a ‘good thing’, personality traits have both positive and negative aspects, and only very stringent parenting would seek to change basic personality traits, unless at the extreme they proved disruptive to the social life of the family. This might also contribute to the lack of variance attributable to shared common environments (Turkheimer et al., 2014). It is also of interest that the extent of the contribution of genetic variation to the measured variance in traits such as neuroticism and extraversion remains remarkably stable over the lifetime of individuals, especially from the age of around 20 onwards (Turkheimer et al., 2014), although neuroticism has been reported to increase in old age. By contrast, measurements of the non-shared environment contribution to variance is less stable. After reviewing the extensive results on this point, Turkheimer et al. (2014) invites us to: Consider a hypothetical pair of identical twins whose personalities develop throughout the lifespan. In the usual absence of shared environmental effects, their pair average score on personality traits is highly, even per fectly, stable relative to the average of other pairs, largely as an expression of their genetic endowment. At the same time, within pair differences between twins show environmental influences that come and go, system atically but temporarily, from childhood through late middle age. At any given point in time, something might happen to make one member of the pair more extraverted or more neurotic; then, as time goes by, the within pair difference decays, and the twins return to their genetically influenced

https://doi.org/10.1017/9781108566520.008 Published online by Cambridge University Press

114

7 GENES, PERSONALITY AND PERSONALITY DISORDERS

mean. Finally, in old age, new differential processes appear to be established that make within pair differences much less stable over time.

Extensive tests on personality have been carried out using twins raised apart. As already mentioned, ‘twins raised apart’ doesn’t really imply that the twins have been separated from birth, but rather that they were (tragically!) separated sometime during the first 5 years of life (the exact average time depending on the cohort in question). Some of the main studies were carried out before the FFM system for measuring personality was introduced, so direct comparisons are not always easy in such studies. Nevertheless, the main take-home message remains the same as for the twin-study results summarised above. One well-known study involved the use of the Minnesota Study of Twins Reared Apart (MISTRA), a project initiated in 1979 that has made very significant contributions to the field of behavioural genetics (Segal, 2012). A 1988 study utilised 44 identical twin pairs raised apart in comparison with 27 non-identical twin pairs raised apart, plus many of both types raised together (Segal, 2012). Based on eleven different psychological inventories used for the self-reporting of personality, the heritability of the various personality traits ranged from 39 to 58 per cent, very similar to the later twin studies described above based on the FFM criteria for measurement. In addition, the results suggested that 36–56 per cent of the variance in personality traits could be attributed to the non-shared environment, whereas only 0–19 per cent was attributable to the shared environment, depending on the trait in question, so again these results are similar to those obtained in the later twin studies. One of the most surprising aspects of the genetic studies on twins and personality has been the finding that the heritabilities of all the OCEAN traits that have been studied so far seem to be in the 30–50 per cent range, pointing to similar contributions by variant genes for each trait. Before such studies began, the expectation was that different personality traits might display very different heritability values. So far, it seems not. In contrast to the consistent data arising from twin studies, the data on personality from adoption studies have been more ambiguous. In studies in which the parents assess the personalities of their young children using cohorts where the child has been adopted and is then assessed by the adopting parents, little or no evidence for any genetic contribution to the variation in personality traits has been obtained (Plomin et al., 1991, Schmitz, 1994). One contributing factor to this puzzling outcome is what is known as the ‘contrast effect’, an effect that applies to twin studies as much as it does to adopted children. It has been shown that parents often exaggerate the

https://doi.org/10.1017/9781108566520.008 Published online by Cambridge University Press

7.2 THE MOLECULAR GENETICS OF PERSONALITY

115

differences between their children (biological or adopted), for example reporting that one is very active in contrast to their inactive sibling, even though relative to other children of similar age they might not be that much different (Saudino et al., 1995, 2004). However, when the personalities of children are assessed by observers, there is a much closer match between the results obtained from twin and adoption studies (Plomin et al., 2008). So if, as it seems, genetic variation contributes to the great variety of personality traits that we see all around us in our daily lives, is it possible to identify the genetic variants involved?

7.2 The Molecular Genetics of Personality We have already highlighted the fact that candidate gene studies have an unfortunate history of lack of replication, and both personality and personality disorder studies have a particularly woeful history in this regard. In fact, one review of no less than 369 studies of candidate gene associations found no clear agreement on a single variant gene (Balestri et al., 2014). It will therefore be no surprise to hear that GWAS and related approaches have now taken over this field of study. Significant ‘genetic hits’ have been a long time coming, but now at last several studies have reported reproducible genetic variants that associate with different personality traits. For example, one study on several different cohorts of genotyped individuals identified six variant DNA regions that associate with the five main personality traits (Lo et al., 2017). As with other traits surveyed in previous chapters, it was only when the numbers being studied were greater than a quarter of a million, in this case using data provided by the DNA testing service 23andMe, that significant hits could be identified. Once reliable gene variants are identified in this way, this opens the door to seeing, at the genetic level, how much correlation there is between the different sets of genes. In this study, the gene set that correlated with neuroticism was inversely related to the gene sets for other personality traits, whereas gene sets for agreeableness, conscientiousness, extroversion and openness were all positively correlated. In turn, conscientiousness at the genetic level correlated with high academic performance. The sets of genetic variants identified for various personality traits were also compared with six different psychiatric disorders. Previously, at the phenotypic level (referring to the person’s health, in this instance), high levels of neuroticism, extraversion and openness (taken together) have been associated with bipolar disorder, and high neuroticism with major depression and anxiety. Low agreeableness has been associated with narcissism. So

https://doi.org/10.1017/9781108566520.008 Published online by Cambridge University Press

7 GENES, PERSONALITY AND PERSONALITY DISORDERS

116

II

–0.8

I

Extraversion Agreeableness

Principal component 2 (19% of total genetic variance) –0.6 –0.4 –0.2 0.0 0.2 0.4 0.6

0.8

it was of interest to see how the genetic variants identified in this study matched up with the genetic variants found to be associated with these different psychiatric disorders. The results are illustrated in Figure 7.1 which divides the space up into four quadrants (quarters), so that those traits found in the same quadrant are correlated, with arrows pointing in roughly the same direction, whereas there is a negative correlation when the arrows point in opposite directions. As can be seen, neuroticism was found to correlate with depression, and extraversion with attention deficit

Openness

Bipolar

ADHD

Schizophrenia Conscientiousness

Anorexia nervosa

Autism

Major depression

Neuroticism III –0.9

IV –0.7 –0.5 –0.3 –0.1 0.1 0.3 0.5 0.7 Principal component 1 (25% of total genetic variance)

0.9

Figure 7.1 Genetic correlations between 23andMe individuals and psychiatric disorders. This analysis is based on the overlapping genetic variants that associate with the personality trait (grey font) and a particular disorder (black font). The smaller the angle between the different arrows, the closer the correlation. Arrows pointing in opposite directions indicate a negative correlation. For example, there is no overlap between the genetic variants involved in conscientiousness and those involved in autism or major depression. From Lo et al. (2017).

https://doi.org/10.1017/9781108566520.008 Published online by Cambridge University Press

7.3 WHAT DOES IT ALL MEAN?

117

hyperactive disorder (ADHD), which we will consider further below. Openness, bipolar disorder and schizophrenia all cluster in the first quadrant, meaning that they tend to share genetic variants in common. Interestingly, all three have some similarities at the phenotypic level. For example, they have all been linked to heightened creativity. Most personality traits (conscientiousness, agreeableness and extraversion) cluster in the second quadrant. Neuroticism and depression are in the fourth quadrant. These findings provide additional support for shared genetic influences between personality traits and psychiatric disorders, and for the notion that they exist on a continuum in phenotypic and genomic space. Maladaptive or extreme variants of personality may contribute to the persistence of, or vulnerability to, psychiatric disorders.

7.3 What Does It All Mean? By this stage of the discussion, someone might well be asking the very reasonable question ‘Well, what exactly does this all mean?’ Envisaging the role of genetics in medical disorders is relatively easy: if an unfortunate combination of genetic variants is present in the same individual, then some medical disorder may result. This is a difference – it can be diagnosed. And even with intelligence, at least of the type measured using biometric techniques, envisaging how hundreds of genetic variants may be involved in helping to build a brain during early development that works a little more efficiently on some tasks rather than on others is also perhaps not so difficult to imagine. But agreeableness? Or extraversion/introversion? It’s maybe more difficult to see how having one precise genome over another is really going to make a difference. And clearly ‘agreeableness’ can be measured using questionnaires, or based on the opinions of others, but it’s not really a ‘thing’ like having blue eyes or being tall. And even very agreeable people can get grumpy on occasion. Yet the evidence that genetic variance is involved in personality traits, as we have seen, is very strong, and we have only touched the tip here of a big iceberg of results. So it’s worth emphasising that the number of genetic variants associated with personality traits will eventually turn out to be far more than six, and indeed others have already been put forward based on further GWAS that we have not mentioned here. As in any other complex behavioural trait, hundreds of genetic variants may well turn out to be the final answer. It is also no surprise that many of the relevant genetic variants that have been suggested involve various brain mechanisms and pathways that are likely to be of importance in early brain development (Kim et al.,

https://doi.org/10.1017/9781108566520.008 Published online by Cambridge University Press

118

7 GENES, PERSONALITY AND PERSONALITY DISORDERS

2015). So we are back to the early developmental biology again of Chapter 3. Myriad different pieces of genetic information are woven together with myriad different environmental inputs, coming from both within and without the body, to generate the wonderful richness and varieties of personalities that we experience all around us every day. If we all had identical genomes, then certainly our personalities would all be more similar to each other than they are at present. But the thought of living with billions of extroverts feels somewhat exhausting.

7.4 Personality Disorders The bigger the difference in the phenotype, the easier the genetic analysis becomes – at least most of the time. The easiest are the ‘one mutant gene – one disease’ types of situation that run in families. The most difficult are traits, such as intelligence and personality traits, that are clearly highly polygenic in terms of the genetic contribution to the variance, and where there is not always general agreement among researchers as to how exactly the trait itself should be defined. But in between, there is a whole other collection of conditions that are often called ‘disorders’ because they are seen to lie right up at one end of a normal range, thereby leading to social difficulty or disorder of some kind. And these disorders are easier for behavioural geneticists to investigate because they have, most of the time, a clear phenotype that can be labelled and used to recruit people with that particular phenotype for their studies. Here, we will discuss just two examples.

7.4.1 Attention Deficit Hyperactivity Disorder In 2019, there were 894 scientific publications with ADHD mentioned in their titles, so here we can give but a brief summary of a huge topic. ADHD is the most frequently diagnosed personality disorder in children. It refers to children who exhibit exceptionally high activity, have a poor attention span and display increased impulsivity. In some of its symptoms, it overlaps with obsessive–compulsion disorder (OCD). Obsessions are unwanted thoughts, ideas and impulses that occur more than once, while compulsions are repetitive behaviours that are driven by the obsessions. The precise diagnosis of ADHD remains a topic of active discussion. In the USA, the diagnostic criteria are broader than in Europe, meaning that as many as 11 per cent of children have received a diagnosis by adolescence, twofold more boys than girls2, whereas in Europe the percentage of children so diagnosed is more like

https://doi.org/10.1017/9781108566520.008 Published online by Cambridge University Press

7.4 PERSONALITY DISORDERS

119

5 per cent – still very high. In Europe, the diagnostic criteria tend to be more restrictive, with an emphasis on hyperactivity that is of early onset and not necessarily connected with high anxiety levels. Diagnosis is generally in the age range 5–7 years. Symptoms continue into adulthood in more than threequarters of cases. The heritability of ADHD is high. Estimates from more than thirty different twin studies have now established that the heritability is in the region of 70–80 per cent throughout life and that the environmental risks are non-shared with other siblings (Faraone and Biederman, 2005, Nikolas and Burt, 2010). It will be remembered that the heritability refers to a population, not to any particular individual within that population, but a high heritability does flag up the implication that genetic variation within a population has a lot to do with why some people have the trait whereas others do not. In the present case, this conclusion is supported by the observation that the concordance between identical twins, of whom one was diagnosed with ADHD, in a large German study was found to be 85 per cent for males and only slightly less for females, compared with 43 per cent for opposite-sex non-twin sibling pairs (Langner et al., 2013). So out of the various behavioural disorders that affect young children, only the autism spectrum disorders have higher heritability than ADHD, and it will be remembered that for autism the concordance for identical twins is close to 100 per cent. In fact, ADHD has been found to have considerable overlap with both autism and neuroticism both from a genetic perspective and based on the impact of unique environmental risk factors (Polderman et al., 2014, Park et al., 2017). A Swedish study based on nearly 18,000 adult twin pairs also discovered that symptoms of ADHD were associated with anxiety disorders, major depression, bipolar disorder, OCD and alcohol dependence (Friedrichs et al., 2012). A multitude of reports have also highlighted the association of ADHD with ‘increased risk of harmful outcomes, such as injuries, traffic accidents, increased healthcare utilization, substance abuse, criminality, unemployment, divorce, suicide, AIDS risk behaviors and premature mortality’ (Demontis et al., 2019). Such lists are not very cheerful, so it’s worth emphasising that all such data are probabilistic and never predict what is going to happen to an individual child growing up with ADHD. Fortunately, medical and other therapies are there to help, so the word ‘association’ should never be interpreted as ‘definitely going to happen’. But in our present context, these connections simply highlight the overlap between different human characteristics. Scientists like to try and break complex behaviours down into contributing components to make

https://doi.org/10.1017/9781108566520.008 Published online by Cambridge University Press

120

7 GENES, PERSONALITY AND PERSONALITY DISORDERS

them easier to study, but complex human beings so often frustrate their investigations. What about the molecular genetics of ADHD? The story here follows a by now familiar pattern, although it has to be said that the genetic data, at the time of writing at least, are trailing well behind many of the other traits that we have discussed so far. Given the high heritability and concordance between identical twins, this might seem surprising, but it’s less surprising when one considers the complex collection of behaviours that are incorporated into an ADHD diagnosis. Like so many other labels we use to describe human traits, ADHD is not a ‘thing’ but a collection of several different behavioural traits that, when they occur in the same person, are then given a label. The other familiar aspect of the ADHD genetic story is that the earlier work on candidate genes has now largely been left behind due to lack of replication, and the way has opened up to GWAS and other genomic approaches that make no prior assumptions about the biology of the trait, which is poorly understood anyway. Many GWAS were carried out with uncertain results before finally, in 2019, there was a reliable report published showing that twelve variant DNA regions reproducibly associate with ADHD, containing 304 candidate genetic variants in total (Demontis et al., 2019). This does not mean that all those variant genes are relevant to the development of ADHD, but it does indicate that some are. The secret of success was, as always, to increase the number of genotyped people with the trait under study – this time more than 55,000 individuals with ADHD were included. Once the number goes over a quarter of a million (as in other GWAS), the number of ‘hits’ will become much higher. But even the twelve variant DNA regions identified so far are providing some interesting information. For example, many of the variants involve regulatory regions of the DNA that are specific in their regulation to the cells found in the central nervous system. There was also genetic overlap – meaning having gene variants in common – with a long list of other disorders, including major depression, insomnia and anorexia. There was also a negative correlation with the genetic variants involved in educational enhancement and intelligence. Some specific genes were implicated in this study that are well known in neurological studies of the brain. For example, one of the variant DNA areas on chromosome 7 contains a gene known as FOXP2, which encodes a transcription factor known to be important in synapse formation and in the brain mechanisms involved in the development of language and learning. Other genetic variants are involved in the regulation of dopamine levels in the brain. Dopamine is a neurotransmitter, meaning that it is heavily

https://doi.org/10.1017/9781108566520.008 Published online by Cambridge University Press

7.4 PERSONALITY DISORDERS

121

involved in transmitting chemical signals between brain cells. Dopamine in the brain is associated with arousal and reward-motivated behaviour, among other actions. The drug methylphenidate (sold under the trade name Ritalin), commonly used in the treatment of ADHD, binds to proteins known to regulate dopamine levels in the brain. Linking genetics to the syndrome itself as well as to its treatment is the kind of connection always of great interest in the field of personality disorders. Before leaving the topic of ADHD, it is worth once again emphasising that this disorder is not a ‘disease’ just because its heritability is high and there are genetic variants associated with it. The same could be said of the normal range of human personality and character traits. We are all part of that range and genetic variation is involved in the whole range. Those interested in clarifying the definition of ADHD have pointed out that there are three correlated factors that can be interpreted in terms of hyperactivity/impulsivity, inattentiveness/dreaminess and nervous behaviour – and we are all on the spectrum with all three of these characteristics (Lubke et al., 2009). It’s just that ADHD is at the extreme of the spectrum for all three – that’s why it’s given a name.

7.4.2 Aggression and Antisocial Personality Disorder We can all think of people with rather aggressive personalities – and then of course we read about hyper-aggressive people every day in the media and just hope we don’t meet ‘someone like that’. Many people end up in prison as a result of aggression, often amplified by alcohol, whereas other crimes are cool and calculating. Those who end up in prison are often diagnosed as having antisocial personality disorder (ASPD; 40–70 per cent of prison populations have been so diagnosed), which is characterised by personalitytype criteria such as impulsivity, aggressiveness, disregard of safety for oneself or others, and lack of remorse for hurting others. Conduct disorder prior to the age of 15 is an essential diagnostic criterion for ASPD, and it markedly increases the risk for ASPD in adulthood (Washburn et al., 2007). But it’s also worth pointing out that the ‘diagnosis’ of ASPD remains controversial among clinicians as it comprises such a varied collection of behaviours – of which aggression is only one. What can genetics possibly have to do with aggression and ASPD? The answer is: quite a lot, although much depends on which aspects of aggression and ASPD are under investigation. ‘Aggression’, like the other traits being discussed in this chapter, is a complex mix of many different behaviours, as is ASPD. For both children and adults, there is a range of questionnaires that

https://doi.org/10.1017/9781108566520.008 Published online by Cambridge University Press

122

7 GENES, PERSONALITY AND PERSONALITY DISORDERS

are used to calculate ‘aggression scores’, given to parents when assessing their own children. One of these is called the ‘Child Behaviour Checklist’ and it is generally filled in by mothers, as in a study involving more than 10,000 twin pairs from cohorts from the Netherlands and the UK who were assessed at the ages of 7, 9/10 and 12 years (Porsch et al., 2016). Another parental rating assessed conduct disorder. From this study, the authors calculated heritability values for aggression levels in the range 50–80 per cent depending on which twin cohort was being assessed. The influence of shared environment was similar: in boys, shared environment explained around 20 per cent of the variation in aggression across all ages, while in girls, its influence was absent around age 7 and only came into play at later ages. It should be noted that this is quite different from ADHD, as mentioned above, where the environmental risks are non-shared with other siblings. As always, the precise heritability values should not be taken too seriously; the fact that they are positive is the main point – genetics has something to do with aggression and ASPD. But what? In the earlier literature, there were hundreds of papers describing candidate gene studies. Unfortunately, as for the other traits mentioned so far, none turned out to be reproducible. In one meta-analysis, 185 candidate gene studies were assessed, published in the period 1992–2011, covering thirty-one genes relating to anger, hostility, aggression, violence, irritability and criminality (Vassos et al., 2014). However, none of the reported genes was found to be associated with the trait in question in a way that could be replicated by others, the acid test in any scientific study of this kind. As the authors report: ‘Our study provides evidence that the candidate gene approach has not succeeded in identifying genes associated with these outcomes. This is consistent with recent observations in the field that candidate gene studies of human characteristics and complex diseases at large have failed to produce consistent and clinically useful findings’ (Vassos et al., 2014). One of these studies made quite a stir when it was published in 2010 in Nature, one of the top science journals in the world. It’s worth outlining the study just to illustrate how dangerous it is to jump to conclusions about links between genes and aggression without adequate support. The publication described a Finnish population in which a particular gene variant was found to correlate at a value greater than would be expected by chance in a population of prisoners characterised by severely increased impulsivity (Bevilacqua et al., 2010). Even the generally restrained Nature gave the news item reporting these results the headline ‘A gene for impulsivity’, a phrase carefully avoided by the authors of the study and indeed in the news report itself. It goes without saying that there is no gene ‘for’ impulsivity.

https://doi.org/10.1017/9781108566520.008 Published online by Cambridge University Press

7.4 PERSONALITY DISORDERS

123

Unfortunately, once inaccurate phrases are used in the scientific literature one cannot then really blame the media and the blogosphere for their own headlines on such reports, including ‘The violent gene: genetic mutation found only in Finnish men that makes them fight’,3 ‘Impulsive? It may be a genetic mutation?’,4 ‘Scientists discover gene that triggers violent gene’,5 ‘Drunken rage could be in your genes’6 and so forth. Note the deterministic language. Here is a single mutant gene as a powerful causative agent triggering violent behaviour in men over which apparently they have no control. The reality is much more mundane. The investigators discovered a novel mutation called Q20* in a gene encoding one of the fifteen serotonin receptors found in the human brain. Serotonin, like dopamine, is one of the brain’s important neurotransmitters and the mutation results in a very low level of the receptor protein in the brain – the protein that binds the serotonin and passes its signal along to the next brain cell in the chain. So it is a perfectly reasonable guess, supported by animal studies, that such a mutation might be linked to increased impulsivity. The novel mutation was discovered in 17 out of 228 cognitively normal albeit violent individuals displaying high impulsivity compared with its presence in 7 out of 295 normal controls. In other words, there was a more than twofold higher tendency for the mutant Q20* to be found in those scoring high on impulsivity than in those who were average with regard to this trait, although we note that the sample sizes were very small. So what are we to make of this correlation between a mutated gene and a violent subpopulation of offenders? We note that out of the total 5.3 million population of Finland it was estimated that about 106,000 individuals are carriers of the Q20* mutation, just 2 per cent of the population. But the number of Finns in prison around the time of the study was 3,500, so clearly the vast majority of people with the Q20* mutation are not violent offenders. Second, it’s quite possible that the seventeen violent offenders in this study also share thousands of other SNPs in common besides the Q20* mutation, which also identify them as a particular cohort, but none of these SNPs, including Q20*, is necessarily anything to do with impulsivity. Correlation does not necessarily entail causation. And in any event, the GWAS carried out since 2010 have not, so far at least, identified this particular genetic variant as being associated with either impulsivity or other measures of aggression, as we shall consider further below. But before going to more believable results, it is useful to flag up one of the other candidate gene studies on aggression that became rather famous in its day but again has now rather faded from view. This was all to do with a variant gene that encodes an enzyme called monoamine oxidase A (MAOA).

https://doi.org/10.1017/9781108566520.008 Published online by Cambridge University Press

124

7 GENES, PERSONALITY AND PERSONALITY DISORDERS

There are good biological reasons why MAOA might be thought to be involved in aggression. The gene encoding the enzyme is expressed in areas of the brain responsible for cognitive processing and breaks down the key neurotransmitters serotonin, dopamine and noradrenaline, thus reducing their actions in communicating messages from one brain cell to another. Numerous animal studies point to an important role for MAOA in modulating animal behaviour: for example, deletion of the MAOA gene from mice led to a colony displaying increased aggression (Cases et al., 1995). To cut a long story short,7 two variants of the MAOA gene were identified that supposedly led to higher or lower activity levels of the enzyme. When children experienced severe maltreatment, so the story went, those with the low activity MAOA variant were more likely to display aggressive antisocial behaviour later on in life, a classic G × E type of situation. Suffice it to say that, despite many further studies, these findings have never been clearly replicated with any assurance. One of the observations that made the overall story difficult to believe was that the low-activity MAOA gene variant is extremely common in all human populations, and is carried by 35–40 per cent of western white populations and 77 per cent of the Han Chinese population (Lu et al., 2002). Given the vast differences in aggression levels in such huge populations, all carriers of the low-activity variant, it is hard to know how to interpret some of the earlier claims made about the role of the MAOA variant gene. The many studies on this claim are, however, interesting – and indeed somewhat alarming – from a science communication perspective. The MAOA gene story is rather typical of the history of investigation of other candidate genes that were once thought to influence complex behavioural traits: the initial positive finding of an association was greeted with acclaim by both the academic community and, often in a distorted or exaggerated form, by the media; extensive attempts were then made to replicate the original finding, some positive, some negative in outcome; then, very often, the original finding eventually faded away through lack of positive replication. In 2004, the low-activity MAOA gene variant was already being dubbed the ‘warrior gene’, this in the respected journal Science, of all places, by a journalist who was commenting on the correlation between MAOA variants and an apparent increased level of aggression in a small group of macaque monkeys (Gibbons, 2004). The terminology unfortunately stuck, and for many years the ridiculous phrase was wheeled out in the media when referring to the MAOA genetic variant. The situation was not helped when, in 2006, two New Zealand geneticists from Wellington’s Victoria University

https://doi.org/10.1017/9781108566520.008 Published online by Cambridge University Press

7.4 PERSONALITY DISORDERS

125

reported during a talk at an international meeting that the low-activity variant of the MAOA gene was found in 60 per cent of Maori people, data based on a sample of just seventeen Maori individuals (Merriman and Cameron, 2007), an estimate later adjusted to 56 per cent (Lea and Chambers, 2007). Based on these data, they concluded that ‘positive selection of MAOA associated with risk taking and aggressive behaviour has occurred during the Polynesian migrations’ and termed the low-activity variant the ‘warrior allele’. In further publicity surrounding this announcement, the MAOA variant (allele) was then linked to antisocial behaviour in the Maori people. The Christchurch Press of 9 August 2006 reported that Maori men were genetically predisposed to ‘violence, criminal acts, and risky behavior’, while the Wellington-based Dominion Post opined that that the MAOA gene ‘goes a long way to explaining some of the problems Maori have. Obviously, this means they are going to be more aggressive and violent and more likely to get involved in risk-taking behaviour like gambling.’8 And all this based on the fact that Maori people have a somewhat higher proportion of individuals carrying the low-activity MAOA gene compared with some other populations. Remember that it was in 2006 that Bradley Waldroup, the murderer mentioned at the start of this chapter, had his sentence reduced to 32 years, rather than execution, as a result of his defence team’s genetic arguments. Their defence was based on the low-activity MAOA claims that were having such high media coverage during that year. This whole sorry story has been critically analysed by genetic bioethicists ever since (e.g. Perbal, 2013). Scientists need to take great care over the public presentation of their results, particularly seeking to avoid the pitfall of extrapolating wildly from scanty data to broad societal conclusions. Now having discarded the MAOA variant gene as being significant in the differences in aggression observed in different populations following childhood abuse, at least as far as present data are concerned, we might ask the question as to what would happen if there were a single gene associated with, let’s say, a fourteenfold higher relative risk of being an aggressive criminal? Could we then talk about a gene causing criminality, for example? Well as it happens, that fourteenfold risk gene exists and it’s called Sry. Out of 131 countries worldwide holding around 10.7 million prisoners, an average of 93 per cent of the prisoners are male,9 and the gene that identifies this population is the Sry gene found on the Y chromosome. So universal is the correlation between the Sry gene, without which males would not be male, and criminality that we can safely say that no other genetic correlation will ever be found between a particular gene and criminality that surpasses this

https://doi.org/10.1017/9781108566520.008 Published online by Cambridge University Press

7 GENES, PERSONALITY AND PERSONALITY DISORDERS

126

one. And yet we still hold nearly all males responsible for their criminal actions and put them in jail as soon as they are convicted. Furthermore, we note that most people who possess a Y chromosome go through life without committing a crime. So the Sry gene does not cause criminality, although clearly we cannot go to the opposite extreme and say that it has nothing to do with human behaviour at all (discuss!). The situation when we return to the MAOA gene and now consider its complete absence is quite different again. Here, we finally return to the type of family studies illustrated back in Figure 1.1 in Chapter 1 in which a single mutant gene is the direct cause, in the present case, of an aggressive behavioural condition. In 1978, there was a chance encounter between a geneticist called Han Brunner from the University Hospital of Nijmegen in The Netherlands and a lady who came to his office seeking help for the outbursts of aggression that characterised many but not all of the males in her extended family (Morell, 1993). Fifteen years later, Brunner described the family tree of this family (geneticists often need a lot of patience) and described the pattern of inheritance (Brunner et al., 1993b). Figure 7.2 shows the family tree in question.

C C

T

C

C CT

T

T T CC C T

CT

C

C C CT C C

C

Figure 7.2 A Dutch family in which a mutation in the monoamine oxidase A gene causes a complete lack of the monoamine oxidase protein. Squares are males and circles are females. A dot in the middle of the circles means that the female is a carrier of the mutant gene on one of her X chromosomes. Black boxes are males who have inherited the X chromosome from their mothers carrying the mutant gene and who display the aggressive syndrome. The C refers to the normal genetic letter at a specific position in the gene, whereas a T refers to the mutant letter that is causing all the trouble. From Brunner et al. (1993a).

https://doi.org/10.1017/9781108566520.008 Published online by Cambridge University Press

7.4 PERSONALITY DISORDERS

127

This was followed up soon after by a description of the mutation involved – a very rare mutation that causes complete deletion of the MAOA enzyme in affected males (Brunner et al., 1993a). This became known as Brunner syndrome. As expected, lack of the enzyme led to a major disruption in the brain levels of key neurotransmitters. The fourteen males that had the mutation, spread over four generations, were characterised by mild intellectual disability, together with outbursts of anger and aggression including arson, rape and exhibitionism. For example, one affected male raped his sister at the age of 23 years. Aggressive behaviour tended to cluster in periods of 1–3 days during which the affected male would sleep very little and experience frequent ‘night terrors’ (Brunner et al., 1993b). Violent outbursts were frequently triggered by bereavement or minor provocations. Female behaviours, meanwhile, were found to be within the normal range. This pattern of inheritance is typical of X-chromosome-linked disorders. The key point to remember is that males are XY whereas females are XX in terms of their chromosomes. So for the genes located on the X chromosome, females will be fine if only one copy of the gene is mutated, because she has the normal back-up copy. Thus, females become carriers of the mutation without displaying any phenotypic evidence of its presence. By contrast, the XY chromosomes of males mean that there is no ‘back-up’ gene and so the mutation of the gene on the single X chromosome causes complete deficiency of the MAOA enzyme. On average, 50 per cent of males will be characterised by the deficiency because there is a 50:50 chance of inheriting the X chromosome carrying the defective gene rather than the mother’s other X chromosome. In the present case, measurement of MAOA activity in cells from the females showed that they were within the normal range, whereas virtually no activity was detectable in the affected males (Brunner et al., 1993a). Unfortunately, little information is available from the study of this family concerning the relationship between early childhood experiences and the development of aggressive episodes. The abnormal behaviour was documented in males living in four different families in different parts of the country, so environmental differences must have been considerable. The aggression varied markedly in severity and over time, even within this single family. For 20 years, it was thought that this was the only extended family in the world with this particular mutation. But in 2014, males in three generations of a French family were found to display autism spectrum disorders, attention deficits and aggressive behaviour (Piton et al., 2014). Further analysis showed that there was a mutation in the MAOA gene in the affected males, different from the one in the Dutch family, but sufficient to reduce MAOA enzyme activity by around 80 per cent, a reduction associated with a much lower expression of MAOA protein than normal. Some amelioration of

https://doi.org/10.1017/9781108566520.008 Published online by Cambridge University Press

128

7 GENES, PERSONALITY AND PERSONALITY DISORDERS

symptoms was obtained by the use of psychotropic drugs. Affected males in their mid-30s had developmental ages of typical 2–5-year-olds. No symptoms were reported for the females in the family, again typical of X-linked inheritance. A further report has come from two Australian families who are also largely deficient in the MAOA enzyme due to different mutations in the MAOA gene (Palmer et al., 2016). In the French and Australian studies, some striking similarities between Brunner syndrome and autistic symptoms have been highlighted, particularly when these involve explosive reactions to frustrating situations. So a change in a single letter in a single gene can be the cause of aggressive behaviour in males in certain rare cases, a point that will require further discussion when we think about free will and determinism in Chapter 11. It is now time to return to the more normal situations in which not one but many genetic variants are involved in the wide range of different levels of aggression that one observes in normal populations. Hundreds of animalbreeding experiments, too many to describe here, have shown that multiple genetic variants are involved in distinguishing between tame and aggressive breeds of the same species. One famous experiment has been in progress in Russia since 1959. Red foxes were bred in such a way that more friendly and more aggressive breeds are now available for study. No less than 103 genetic variants have been detected that are different between these two breeds (Kukekova et al., 2018). But before some crazy person thinks of breeding less aggressive humans, it’s important to note that one of the variant genes highlighted in this study, SorCS1, has also been implicated in schizophrenia and autism in humans. So the extrapolation from animals to humans should always be treated with some caution. Coming back to humans, and in contrast to red foxes and to the other traits discussed so far, human GWAS on aggression and ASPD have really only just begun, most likely held up by the challenges involved in tackling such complex traits. GWAS of ASPD have made a start (Rautiainen et al., 2016), with a few possibly significant hits, but the authors are the first to admit that much larger samples are necessary to be sure. A GWAS on impulsivity has also reported a few hits, making the point that impulsivity can be separated genetically into several distinct components (SanchezRoige et al., 2019). Several significant hits were made in a GWAS of antisocial behaviour based on 16,400 individuals (Tielbeek et al., 2017), but by now it should be clear that much larger cohorts are really necessary to pick up a broader range of genetic variants. Work has also started on the epigenetic regulation of genes encoding proteins known to be involved in behavioural changes in animal studies

https://doi.org/10.1017/9781108566520.008 Published online by Cambridge University Press

7.4 PERSONALITY DISORDERS

129

(Palumbo et al., 2018). Epigenetic changes arising from childhood experiences, possibly leading to longer-term regulation of genes involved in aggression in adulthood, is a fascinating area of research, although one fraught with experimental challenges. Overall, there are hundreds of reports of particular gene variants in both animal and human studies that associate with various aspects of aggression, and indeed human criminality, but there is, at the time of writing, no well-established and reproducible data set of human genetic variants shown to associate with aggressive traits. Given the wellestablished heritability of having an aggressive personality, one assumes that many gene variants will one day be reliably established, but that day has not yet arrived.

https://doi.org/10.1017/9781108566520.008 Published online by Cambridge University Press

8

Genes, Food, Exercise and Weight

S

kinny genes the secret to staying slim proclaims the bbc website.1 ‘Scientists say they have discovered the secret behind why some people are skinny while others pile on the pounds easily’. Well, maybe, but such headlines can easily leave people with the wrong impression, or even the feeling that life has dealt them a deterministic set of genes leading to a pile of flab that no amount of exercise or dieting is going to shift. Looking at some changes in average weight in different populations can help in gaining some perspective on the question. The number of Americans with obesity has steadily increased over the past five decades until today more than one-third of the adult population is obese and over two-thirds are overweight (Yang and Colditz, 2015). Overall, 10–20 per cent of Europeans are now classified as obese. The lower people’s socio-economic status, the more likely they are to be obese. None of this has anything to do with a change in the genetics of these populations. The same could be said of the increased weight of UK children leaving primary school – nearly four in ten of 10–11-year-olds are predicted to be overweight or obese by 2024.2 A change in their genes has nothing to do with that either, but rather the overconsumption of unhealthy fattening foods coupled with a lack of exercise. Pre-school children raised in North Korea are 7 kg lighter than comparable children brought up in South Korea, and women also weigh up to 9 kg less on average than those in the south (Schwekendiek, 2009). The two populations are genetically no different – it is the differences in nutrition that make the difference. Similar conclusions come from studies on immigrant populations to the USA who generally have low levels of obesity upon arrival but who become similar to the US population within 10–15 years in terms of the proportions who are overweight or obese (Goel et al., 2004). Worldwide, nearly 40 per cent of adults are overweight and 10–15 per cent are obese (Goodarzi, 2018). The health implications of such figures are huge: 130

https://doi.org/10.1017/9781108566520.009 Published online by Cambridge University Press

8 GENES, FOOD, EXERCISE AND WEIGHT

131

obesity correlates with type 2 diabetes, increased risk of cardiovascular disease and certain types of cancer. In a study of 2.8 million people, even slightly overweight people were twice as likely to develop type 2 diabetes, and for those who were obese, the risk was nearly nine times higher.3 Just take a look at your parents and you might receive the impression that your size and shape is all dependent upon your inheritance. But genes are not all that you receive from your parents – you also inherit many of their social practices when it comes to the type of food you like, how much you are used to being served at each meal, and the nutritional value and calorific content of the food that your parents raised you to enjoy. Our friends also make a big difference. One study showed that a person’s chance of becoming obese increases 57 per cent if a friend becomes obese, 40 per cent if a sibling becomes obese and 37 per cent if a spouse becomes obese (Christakis and Fowler, 2007). One could go on, but the point is clear: certainly genetics makes a difference as we shall see in a moment, but it’s good to start with a reminder that, with rare exceptions, genes are not deterministic of how fat or thin we are, and there are many other factors involved. For example, blame your friends’ influence! Body mass index (BMI) is a better way of measuring body size and shape than simply measuring weight. BMI is your body weight in kilograms divided by the square of your height in metres (to give kilograms per metre squared, or kg/m2); BMI can also be measured using pounds and feet. There are plenty of online calculators if you want to check your BMI and see whether you are under- or overweight according to your national average for people of your age.4 Once you start thinking about your BMI, it’s fairly obvious that the genetics involved in its precise value is likely to be complex. First, there is the genetics of appetite. Then once you’ve eaten your food, there are all the chemical pathways involved in digestion and metabolism – meaning all the various ways in which your body uses your food to generate the energy and other chemical needs of the body. Some lucky people just seem to eat as much as they like, at least in their younger years, and their weight always seems to stay the same. Where does it all go? And then there are other aspects of your BMI where genetic variance makes a difference – as in how much fat is laid down and, most important for many people, where exactly that happens. Neuronal circuits in the brain, especially those involved in a brain area known as the hypothalamus, are also tightly integrated with all the various chemical signals involved in taste, hunger, appetite and satiety – and with the many other aspects of our enjoyment of food (van der Klaauw and Farooqi, 2015). So in what follows, we will go through the genetics of BMI

https://doi.org/10.1017/9781108566520.009 Published online by Cambridge University Press

132

8 GENES, FOOD, EXERCISE AND WEIGHT

layer by layer (if that’s the right phrase to use given the context) and see how ‘food and weight’ provides yet another example in which genetic variation and environmental variation are tightly integrated to produce the size of jeans that we can wear with comfort.

8.1 The Heritability of BMI It will be no surprise to hear that the heritability of BMI is rather high, varying from 31 to 90 per cent depending on the study and the variations in different populations around the world. The heritability values also depend on age. One large study pooled data from nearly 90,000 twin pairs from four different continents – Europe, East Asia, Australia and North America – and measured the heritability of BMI at different ages (Silventoinen et al., 2016). The investigators deliberately restricted their study to focus on children coming from affluent rather than poor populations so that huge differences in nutrition between their cohorts of twins might not make interpretations more difficult. Their results showed that heritability started out at 41 per cent for both boys and girls at the age of 4 and then increased to 75 per cent by 19 years of age. Why such a big difference with age? Remember that heritability in these kinds of studies is based on the comparative differences in the trait being studied (BMI in this case) between identical and non-identical twins. One can well imagine that for the twins aged 4, their parents will decide on their nutrition, so differences in nutritional intake will not differ that much between twin pairs, irrespective of whether they are identical or nonidentical (‘Eat up your beans!’). However, adolescents at the age of 19 will be far less dependent on their parents’ choices when it comes to food and it is quite likely that they will have left home to work or study, so the genetic variants become more significant in explaining the variation in a population. Now what happens when heritability values are estimated without taking into account the nutritional status of the twins – in other words, the twin cohorts are not selected based on coming only from affluent areas? A study from China provides some interesting clues in this respect. In this case, around 12,000 twin pairs from the Chinese National Twins Registry were studied, comprising twin pairs coming from all parts of China, rich and poor. Below 7 years of age, the BMI heritability was low – less than 20 per cent – but by the age of 17 had increased to around 60 per cent for boys but only 30 per cent for girls (Liu et al., 2015). So presumably in this case, the environmental factors, in particular differences in nutritional intake, were making a much bigger difference in the variation, particularly in the case of girls (maybe due to stronger social pressures to stay thin).

https://doi.org/10.1017/9781108566520.009 Published online by Cambridge University Press

8.1 THE HERITABILITY OF BMI

133

It will not be surprising to hear that the total amount and distribution of fat in our bodies is influenced by genetic variation. For example, DXA scanning5 has been used to estimate how fat is distributed in the body, and a Danish twin study showed that the total body fat, as well as the amount of fat distributed in the trunk and lower body, all showed high heritabilities in the range 83–86 per cent, irrespective of whether the twins being studied were in the age range 25–32 or 58–66 years old (Malis et al., 2005). The heritabilities of many behavioural traits relevant to body weight have also been assessed. Once again, the precise values are not that important, but the fact that the heritabilities are not zero – meaning genetic variation has something to do with the trait in question – is the most important point. We perceive the nature of our food through taste, texture and smell. ‘I’m not going to eat that,’ we say, ‘it smells horrid!’ From twin and family studies, heritabilities of 30–50 per cent have been measured for food’s pleasantness and its correlated consumption, together with a craving for sweet foods (Keskitalo et al., 2008). Other twin studies have found heritabilities of 53–62 per cent for the preference for, and intake of, foods that are high in fat and sugar. Satiety, the feeling of fullness that influences the timing to the next meal, is 63 per cent heritable (Carnell et al., 2008). Yet other studies have looked at traits such as ‘cognitive restraint’, meaning the ability to restrain from eating too much based on reason and will; ‘external eating’, the tendency to overeat in response to external stimuli, such as delicious foods (sounds as if that refers to all of us, but maybe my cognitive constraint is just weak); and ‘emotional eating’, which is the tendency to overeat in the presence of negative mood states such as depression, anxiety or loneliness. Such traits are measured using various questionnaires. In one large study of Korean twins, the heritability for cognitive restraint was found to be 31 per cent, and 25 per cent for both external and emotional eating (Sung et al., 2010). Comparable results have been found in a range of studies on Western populations. Other investigations have tried disentangling the various human feelings involved in hunger, estimating a heritability of 25 per cent for the degree to which reported hunger levels then lead to varying amounts of food intake (De Castro, 1999). Similar heritability levels have been reported for other relevant traits, such as the total daily calorie intake, meal frequency, meal size and the levels of fat metabolism (meaning how quickly fats are broken down once they enter the body). The basal metabolic rate – the number of calories required to keep your body going at rest – is 47 per cent heritable (Bouchard et al., 1990), and varies considerably among those who do, or do not, engage in regular exercise. Pretty much

https://doi.org/10.1017/9781108566520.009 Published online by Cambridge University Press

134

8 GENES, FOOD, EXERCISE AND WEIGHT

anything to do with our BMI has been found to be influenced by variant genes. In all this discussion of heritability, it’s once again worth emphasising that such values say nothing about the average values in a given population. In principle, these could all double, but the heritability values stay the same because heritability, as will be remembered, refers to the proportion of the total variation in a population that can be attributed to genetic differences among individuals. So everyone could become obese to varying degrees, but the differences among individuals might stay relatively the same.

8.2 The Search for Relevant Genes The extensive studies reporting positive heritabilities for the various mechanisms leading to differences in BMI provide strong justification for the search for relevant gene variants that contribute to these differences. The genetics of obesity is often the target for research, given its health implications. Obesity is defined as an increase in fat mass that is sufficient to adversely affect health. In practice, measurements of BMI are often used as an indication of obesity, although there are examples, some of them rather rare, showing that in some cases it can provide a rather poor indication (Speakman et al., 2018). For example, the actor Arnold Schwarzenegger, when he was a competing Mr Universe entrant, had a BMI above 30 but less than 10 per cent body fat. The WHO uses as its definition of obesity any adult with a BMI of more than 30 kg/m2, whereas those with a BMI of more than 25 kg/m2 are overweight. The vast majority of people with a BMI above 30 are just fat, and we can probably take Mr Schwarzenegger as a rare exception. Rare but revealing are the mutations in single genes that by themselves are both necessary and sufficient to lead to severe childhood obesity. There are at least eight different types of obesity of this type, and the Mendelian family histories can be drawn of the kind illustrated in Figure 1.1 back in Chapter 1. For example, there is a chemical messenger (known as a ‘hormone’) called leptin that is secreted by the body’s fat tissues and inhibits food intake. If the actions of leptin are inhibited in any way, then appetite can increase dramatically. In fact, many different chemicals have been described in our bodies that either increase or decrease our desire for food, so the regulation of appetite is a very finely tuned operation (Singh et al., 2017). In the case of leptin, children have been reported with an inherited mutation in the leptin gene so that they completely lack this chemical. In one report, a child aged 2 years weighed 29 kg whereas another child, aged 8 years, weighed 86 kg

https://doi.org/10.1017/9781108566520.009 Published online by Cambridge University Press

8.2 THE SEARCH FOR RELEVANT GENES

135

(Farooqi, 2005). The impact of lacking leptin can be directly measured using brain-scanning methods, which show a marked increase in the firing of brain cells in a specific area (known as the ‘ventral striatum’) of the brains of those unfortunate enough to lack leptin (Farooqi et al., 2007). The good news is that such patients can now be treated successfully with leptin made using genetic engineering. One of the effects of giving leptin is to restore the abnormal brain cell firing to normal within 7 days of treatment, even before any measurable difference in BMI has occurred, a further reminder of the way in which genes, the brain, hormones, taste, appetite, satiety and everything else, not to mention choice, are all integrated to produce our size and shape. Taken together, the rare cases of obesity caused by single-gene mutations make up only around 1 per cent of all cases of obesity, so here our emphasis will be on the 99 per cent of people where the genetic contributions are polygenic. Once again, GWAS come to the rescue and many different GWAS have identified more than 300 different SNPs that associate with obesity, waist-to-hip ratio or other measurements such as fat distribution, in both children and adults (Bradfield et al., 2012, Liu et al., 2018, Speakman et al., 2018, Goodarzi, 2018). When BMI is the variant being measured in a population, then more than 100 associated gene variants have been identified using GWAS (Goodarzi, 2018). All these numbers continue to rise as larger and larger cohorts are included in such investigations. It will be remembered that the SNPs simply identify DNA regions, acting as markers that flag up a nearby gene or regulatory region that contributes a small amount of the observed variation in a given population. Here, we will simply focus on the very first genetic variant that was found in this way as it provides an interesting example. The first variant gene to be identified by GWAS is known by the charming name of FTO – referring to the fat mass and obesity-associated gene (Frayling et al., 2007). Actually, when it was first shown to be associated with obesity in this way, FTO stood for ‘fused toes’ because mice lacking the gene had precisely that phenotype. But sensibly, the FTO terminology was engineered to bring it more in line with obesity. The variant FTO gene was originally linked to type 2 diabetes, and through that linkage shown also to be identified with obesity. It’s the link with obesity that counts – type 2 diabetes follows later once the obesity is established. Since then, FTO has routinely been picked up in many different GWAS on obesity in both children and adults. The discovery provides a great example of the power of GWAS as a fishing exercise. Nobody had guessed that FTO might be involved, and

https://doi.org/10.1017/9781108566520.009 Published online by Cambridge University Press

136

8 GENES, FOOD, EXERCISE AND WEIGHT

when we discuss next how FTO works, you will see why this was the case. It was an unknown that nobody knew was unknown. Not surprisingly, when the association of the FTO gene with obesity was first announced in 2007, the media loved it. ‘Secret of why we pile on the pounds’, said the Daily Express. Those who have a single copy of the FTO variant gene weigh an average 1.5 kg more than age-matched controls, whereas those who have two copies – one on each chromosome – weigh an average 3 kg more. In European populations, at least, around 16 per cent of people have double copies of the FTO variant ‘risk’ gene. So, just focusing on the double-copy people, this implies that the European population weighs at least 350 million kg more than it would have done without the variant FTO gene. The variant has been found to associate with increased hip circumference and waist-to-hip ratio, as well as a higher mass of fat. The association has been reported in multiple ethnic groups, but having the ‘risk’ version of the FTO gene varies widely among different populations. For example, around 40–60 per cent of European populations carry at least one copy of the risky version compared with only 12–20 per cent in East Asian populations (Loos and Yeo, 2014). Despite the variation in frequency, the modest effect size of an extra 3 kg for those having two copies of the variant remains consistent, regardless of ethnicity. The Avon Longitudinal Study of Parents and Children (ALSPAC) has already been mentioned as providing some very useful data from children as they grow up. When the FTO gene variant was examined in this population, it was found that the variant seemed to make no difference to birth weight, but children aged 7 years onwards did start becoming slightly heavier than controls as they grew older, so the variant gene doesn’t seem to be having any effect on fetal growth (Frayling et al., 2007). The gene makes no difference to height, but its effect on weight does seem to be wholly dependent on an increased mass of fat. Before we start thinking that the FTO gene variant explains all our recent weight gain, despite our best attempt at the latest diet, it is worth pointing out that the variant gene explains only about 1 per cent of the overall variation seen in a population with respect to the variation in BMI (Frayling et al., 2007). Given that FTO is one of the most influential genetic variants to make a difference to BMI discovered so far, this implies that there should be well over 100 variant genes contributing to the variation in BMI. But that number might need to be revised if it turns out that some of the gene variants are collaborating to make the difference, rather than each variant gene just making a difference all by itself.

https://doi.org/10.1017/9781108566520.009 Published online by Cambridge University Press

8.2 THE SEARCH FOR RELEVANT GENES

137

So what does FTO actually do? Giving a clear answer to this question is proving quite challenging. The gene encodes an enzyme that chemically modifies the genetic letters found in single-stranded DNA and RNA molecules such as the messenger RNA (mRNA) introduced in Chapter 2. ‘Singlestranded’ simply means that the enzyme can only gain access to DNA, for example, when it’s been unwound – in other words, it’s no longer present as a double helix as in Figure 2.1. By itself, this mechanism really provides no clues as to how such an action might lead to the regulation of BMI. So what happens if the FTO gene is defective so that FTO protein produced is completely lacking in its normal enzyme activity? There are, in fact, some rare cases where this happens. The story is not good. FTO deficiency leads to growth retardation, functional brain deficits and abnormalities in the face structure, with other problems described in some patients as well (Boissel et al., 2009). No patient with this deficiency has survived beyond the age of 30 months. Given the very broad actions of the FTO enzyme, this is perhaps not surprising. If you don’t have the enzyme at all, then lots of bad things happen. Colonies of mice have also been genetically engineered to completely lack the gene, as well as mice colonies in which the gene is only ‘knocked out’ (deleted) in the brain. What is especially interesting here is that the phenotype of the mice lacking FTO just in the brain is pretty much the same as the mice who lack it in every tissue of the body. This does point to FTO’s role in the brain as being of particular importance. Although FTO is expressed in all the tissues of the body, its expression is highest in the brain, especially in the hypothalamus, which, as already mentioned, plays a key role in food intake (Loos and Yeo, 2014). Its level of expression in the brain is influenced by diet – eat a fatty diet for 10 weeks and FTO expression increases. Maybe the most relevant question in the context of overweight and obesity might be this: given that so many people have at least one copy of the FTO gene variant associated with these phenotypes, what’s the difference at a functional level between the ‘risky’ gene and the ‘non-risky’ version? The simple answer is that we don’t know (Loos and Yeo, 2014). One possibility is that the variant ‘risky’ gene leads to slightly different levels of FTO enzyme being expressed in the hypothalamus, which in turn influences food intake. Certainly, possession of the FTO variant gene is associated with increased eating. People with a double dose of the risky gene eat more and have reduced satiety after eating, meaning that they still feel hungry even after eating plenty. They also prefer food containing more calories. Conversely, they do not display any reduced energy expenditure or physical activity, so the FTO influence seems to be on energy intake. In one study, a group of

https://doi.org/10.1017/9781108566520.009 Published online by Cambridge University Press

138

8 GENES, FOOD, EXERCISE AND WEIGHT

young children in the New York area, none of them obese, were recruited to measure how much they ate under controlled conditions (Ranzenhofer et al., 2019). On average, the children with one copy of the risky FTO gene ate around sixty-four more calories in a single meal than the children who had no risky copies at all, whereas children who had two copies of the risky FTO gene consumed a further sixty-four calories, making 128 calories more in all on average. That might not sound like a lot, but if such a behavioural difference is multiplied by three meals per day, week after week, then it’s easy to see how such increased food intake could end up making quite a difference. This kind of study is also important because the children involved were not obese and obesity itself makes a difference to hormone secretion, levels of food metabolism and the like, making it difficult to sort out cause and effect. In the New York study, there were behavioural differences in food consumption in the absence of any prior obesity, suggesting that the FTO risky gene exerts its earliest influence at the stage of eating more or less food. It should be emphasised that this brief overview of FTO represents just a summary of a big topic. Every year, more than 100 scientific papers are published with ‘FTO’ in their title. And we should not, of course, forget the many hundreds of other genetic variants that have been identified as being associated with variant BMI or the other aspects of being fat or thin already mentioned. But together, they only seem to explain less than 5 per cent of the overall genetic contribution to the variance in BMI in a given population (Speakman et al., 2018). So there still seems to be quite a way to go in terms of relevant gene discovery. In terms of possible mechanisms of action, many of the gene variants line up with the way things are looking for the main actions of FTO – meaning that many involve genes expressed mainly in the brain. So this is consistent with the idea that many of them might be involved in the hypothalamus – that key brain centre for appetite regulation. But other pathways are gradually being identified. One study used GWAS on more than 400,000 Europeans to investigate the variation in risk-taking in this particular population, finding that there was significant overlap between the variant genes that correlate with this trait and those that associate with BMI (Clifton et al., 2018). Some of the overlap pointed towards the involvement of a particular well-known brain pathway implicated in risk-taking. The take-home message is: if you’re more willing to engage in risky behaviours, then becoming overweight might just be one outcome. Another GWAS found a genetic correlation between a high BMI and increased susceptibility to loneliness and depression (Day et al., 2018). But when the variant trait being looked at

https://doi.org/10.1017/9781108566520.009 Published online by Cambridge University Press

8.2 THE SEARCH FOR RELEVANT GENES

139

was waist-to-hip ratio, many of the genes identified were found to be expressed in fat tissue, meaning that this facet of body size is more likely to be controlled within the fat tissue itself (Goodarzi, 2018). In fact, forty-four genetic variants have been found by GWAS to associate with variation in the waist-to-hip ratio, of which twenty-eight had larger effects in women, five had larger effects in men, and eleven had opposite effects in men and women (Goodarzi, 2018). Provided that we are referring to the 99 per cent of obesity that is polygenic, rather than to the 1 per cent that is caused by a single defective gene, then by now it should be clear that for most of us there is nothing deterministic about having a very high BMI and being obese. Remember that even the well-known FTO gene variant contributes only about 1 per cent of the variation in weight found in a European population (for example). So if you have two copies of the ‘risky’ gene, you will be likely to have a mild (and unconscious) predisposition to eat more but not in a way that cannot be countered by your own decision to eat less and take more exercise. Now, of course, if you had a whole load of the gene variants that contribute to a higher BMI, then you might feel that the odds are heavily stacked against you in terms of staying a healthy weight. But it’s worth bearing in mind that the relevant gene variants that have been looked at so far are nowhere near as common as the FTO gene variant already discussed. For example, one of the variant genes that GWAS have associated with increased BMI is known as ADCY3, but this variant is present in at most 3 per cent of the populations studied so far (Grarup et al., 2018). Furthermore, there is plenty of evidence showing that increased exercise counteracts the effects of having many of the ‘risky’ genetic variants (Goodarzi, 2018). For example, in a study of 109,000 British people who all had sixty-nine ‘risky SNPs’ associated with higher BMI, exercise was shown to counteract the risk rather successfully in terms of maintaining a healthy weight (Tyrrell et al., 2017). The type of exercise that helps best to counteract the effects of risky gene variants is also of interest. A study of 18,424 Taiwanese adults whose DNA had been sequenced revealed that, whereas swimming and cycling weren’t so great at mitigating the risk, jogging was the best, and fast walking and climbing mountains were also good (Lin et al., 2019). Some forms of ballroom dancing also worked quite well, including the foxtrot and the waltz. So there really is something here for everybody. Just remember that doing intensive exercise in rare bursts is no good (and can be dangerous) – what counts is regular exercise. From all this, we can conclude that there is nothing deterministic about having a large array of ‘risky’ gene variants in your genome, but it’s also the

https://doi.org/10.1017/9781108566520.009 Published online by Cambridge University Press

140

8 GENES, FOOD, EXERCISE AND WEIGHT

case that it will take more effort and discipline over diet and exercise to maintain a healthy BMI than for those fortunate enough to have less of a load of ‘risky’ variants. Having said that, it’s also of interest to note that genetic counselling based on known genetic risk of obesity typically resulted in reduced self-blame and increased motivation to make life-style changes but didn’t actually result in weight loss (Goodarzi, 2018). In fact, some people, perversely, actually put on more weight once they knew their genetic risk profile, maybe feeling that they were doomed by their genetics. Probably, the moral is not to worry too much about what ‘risky’ genes you may or may not have, but simply to eat a healthy diet and take plenty of exercise, watching out that your BMI doesn’t drift into the overweight category, and especially not into the obese category. For most people, life really should be that simple.

8.3 The Epigenetics of BMI Our BMI provides a classic example of the myriad ways in which genes and environments are tightly integrated to produce the shape we are, and, as already emphasised, our own choices play a central role in the outcome. A thorough survey of the huge literature on such gene–environment interactions is beyond our present scope and in any case is well described in many reviews for those who wish to dig further (e.g. Reddon et al., 2016, Goodarzi, 2018). If coping with more than 100 genetic variants that influence the variation in BMI seems a dizzying prospect, then once the epigenome starts being considered as well, a feeling of BMI-reducing indigestion might well be the result. The technical word ‘epigenome’ simply refers to the sum total of all the epigenetic marks on the DNA in a particular tissue at a particular moment in time. Biologists love adding ‘ome’ on to the ends of chemical words as it helps to establish a whole new field of enquiry. So now there are journals with titles such as Epigenomics and Proteomics, plus at least twenty journals with Genomics in their titles. Studies of the epigenetics of BMI variation are only really just beginning as the techniques for measuring epigenomes continue to be developed and the costs come down. As mentioned previously, one of the main epigenetic modifications of DNA involves what is called ‘methylation’. If this happens in a regulatory region of a gene, this creates a barrier that prevents transcription factors binding to that regulation site, causing a decrease in the expression of the gene. In one study, it was shown that, specifically in the genes found to be expressed in adipose (fat) tissue, no fewer than 2,825 genes were methylated differently between those with lower or higher BMI in a way that

https://doi.org/10.1017/9781108566520.009 Published online by Cambridge University Press

8.4 CONCLUSIONS

141

made a difference to the expression of those genes (Ronn et al., 2015). What does this mean? In such measurements of the epigenome, it’s impossible to distinguish between cause and effect. Very likely, in such cases it’s a mixture of both: what we eat, together with our daily exercise level, both influence the epigenetic status of multiple genes in our fat tissues, thereby regulating their metabolism, their size and their shape. Moreover, once an epigenetic difference is established in a particular gene, that difference can be replicated along with cell replication so that the difference has long-term effects. On the other hand, many of the epigenetic differences may be a consequence of the particular way in which the adipose tissue is functioning in those who have a higher BMI. To sort out which of the 2,825 genes do what, together with the physiological consequences of their different epigenetic status, is a huge enterprise, but in the first instance it’s the fact of methylation differences existing at all that is significant. Hearing about epigenetic differences leading to differential gene regulation between those of lower or higher BMI should in no way make us feel doomed to be overweight or obese. In fact, quite the opposite. Doing extra exercise has been shown to change the DNA methylation of 2,817 genes in skeletal muscle and 7,663 genes in adipose tissue – eighteen of which have previously been shown to be associated with obesity. Most of the genes in the former showed decreased methylation, implying that many genes were being increased in their expression, whereas in adipose tissue the reverse was the case – methylation was generally increased, thereby lowering gene expression (Reddon et al., 2016). A review of twenty-five different studies showed that exercise in each case made a marked difference to the epigenetic status of multiple genes (Voisin et al., 2015, Jacques et al., 2019). It’s also good to remember that, whereas many twins are epigenetically very similar at birth, during adult life their epigenomes can become remarkably different (Fraga et al., 2005). Our choices really do make a difference and our own DNA reflects those differences.

8.4 Conclusions Does a discussion about BMI and obesity belong in a book about behavioural genetics? Hopefully, this chapter has made it clear that the answer is ‘yes’. Our shape and size are to a large degree the outcome of a glorious mix of genetic variation, exercise practices, cultural backgrounds, food habits and personal choice. Some of the relevant variant genes involved, such as FTO, appear to exert their main effects on our brains by increasing appetite, and an increased desire for food certainly predisposes us to certain behaviours. On

https://doi.org/10.1017/9781108566520.009 Published online by Cambridge University Press

142

8 GENES, FOOD, EXERCISE AND WEIGHT

the other hand, other variant genes appear to exert their effects after the behaviours have been in action – such as how much fat is stored and where exactly it goes. And our own behaviour with respect to physical exercise – or lack of it – shapes the outcome in many different ways. So overall, ‘food, exercise and weight’ provides a classic example of the way in which genes, environments and human choice are woven together in a highly complex way to generate the person that we see in the mirror each day.

https://doi.org/10.1017/9781108566520.009 Published online by Cambridge University Press

9

Genes, Religiosity and Political Commitment

J

umping from the genetics of body size to the genetics of religion and politics might seem a bit of a leap – and in any event it can surely not be the case that genes have anything to do with such complex human behaviours. Or can they? As already emphasised in Chapter 4, behavioural geneticists continue to include within the scope of their enquiry pretty much any trait that differs among people. And certainly behaviours reflecting religious and political commitments, or the lack of them, vary among people. It may be remembered from Chapter 1 that the rather dramatic claim was made in a Nature article that ‘An increasing number of studies suggest that biology can exert a significant influence on political beliefs and behaviours’ (Buchen, 2012). Such claims require some critical attention, and the aim of this chapter is to do precisely that. As usual, we will consider the scientific data first and then proceed to think about what it means – or perhaps, more importantly, what it doesn’t mean.

9.1 Defining Religiosity If intelligence is tricky to define, then words like ‘religion’, ‘spirituality’ and ‘religiosity’ are ten times worse. Within the social sciences, religiosity is treated as a multi-dimensional construct, incorporating numerous elements, or domains. Although these are generally given new names with each new author, they include beliefs, values and attitudes, practices including participating in services, ceremonies or rituals, knowledge of the religion, religious experience, personal faith or devotion, and denominational or institutional loyalty. Religiosity thus covers a range of social, cultural, cognitive and behavioural elements and includes both how a person feels about or experiences their religion and how they act and behave religiously.

143

https://doi.org/10.1017/9781108566520.010 Published online by Cambridge University Press

144

9 GENES, RELIGIOSITY AND POLITICAL COMMITMENT

Domains do not necessarily align perfectly – a person can hold religious beliefs without engaging in religious practices. Others have come to the definitional questions more from the perspective of evolutionary psychology, defining religiosity as ‘the mental ability to be religious’ (Voland, 2009). This is a much narrower definition and redefines religiosity as a biological capacity, rather than as an actuality. Under Voland’s model, the phenotypic expression of religious behaviours, both thoughts and action, is called ‘religiousness’, defined as ‘the individually varying psychic and behavioural manifestation of religiosity’ (Voland, 2009). This terminology has been widely used in behavioural genetic studies, which generally use religiousness to refer to public or social domains such as church attendance and religious practices. Religiousness is also used to describe the importance of religion in a person’s life as assessed by their actions (known as religious ‘salience’). The words ‘religiousness’ and ‘religiosity’ have often been treated interchangeably. ‘Spirituality’ is also sometimes assessed in the behavioural genetics literature. ‘As the terms religiousness and spirituality have developed over time, they have acquired more specific connotations. Currently, religiousness is increasingly characterized as “narrow and institutional” and spirituality is increasingly characterized as “personal and subjective”’ (Zinnbauer et al., 1997). Overall, spirituality is generally described in terms of connection to, or a feeling of oneness with, the wider world, other persons, and the transcendent or supernatural. Its overlap with religious practice appears to be highly culture dependent.

9.2 The Heritability of Religiosity Behavioural genetic studies have tended to assess only one or a few religiosity domains, which are rarely the same from study to study. The most common domains assessed are religious practices, church attendance, and religious conservatism or fundamentalism, followed by personal devotion and religious attitudes. In eight different published studies investigating the heritability of religiosity in adolescents (aged 11–18) using the standard twin method, all were in agreement that the heritability is essentially zero or so small as to be unlikely to be of significance, especially in younger adolescents. This is reflected in the work of Polderman et al. (2015) who carried out a huge and very useful metaanalysis in which they summarised the heritability of multiple human traits based on 50 years of twin studies. As part of this study, they grouped sixtythree separate twin studies on ‘religion and spirituality’, reporting that for

https://doi.org/10.1017/9781108566520.010 Published online by Cambridge University Press

9.2 THE HERITABILITY OF RELIGIOSITY

145

the age range 12–17 years, the differences in religion and spirituality between identical twins was only very slightly higher than that between non-identical twins.1 The actual traits measured included religious beliefs, religious affiliation and practices such as church attendance, religious fundamentalism, a 9-point scale measuring ‘religiousness’, a 45-point scale measuring religious values and ‘spiritual involvement’, and religious salience. In marked contrast to these studies, measurements of religiosity in adults (over the age of 18) have consistently reported positive heritability values in twelve different studies,2 of which a representative sample are cited here (Truett et al., 1992, D’Onofrio et al., 1999, Vance et al., 2010). Most of the traits measured in the adolescent cohorts were also included in the adult cohorts and, in addition, some studies included extra traits, such as spirituality, personal devotion and ‘religious conservatism’. Heritabilities are, on average, between 30 and 50 per cent but for some domains are much lower (16 per cent) or much higher (64 per cent). Environmental effects on variance are as large as genetic effects in most domains. The data are inconsistent on whether church attendance and other external or public religious practices are heritable: Truett et al. (1992) found non-significant genetic effects in their study, but Bradshaw and Ellison (2008) estimated the heritability of church attendance at 32 per cent, which is broadly equivalent to other domains. No study found a positive heritability for affiliation with specific religions or denominations (D’Onofrio et al., 1999, Kendler et al., 1997). Spirituality does not appear to have greater heritability than religious practice or attitudes. Measured heritabilities for spirituality domains range from 23 to 65 per cent depending on which precise trait is being measured. One problem with several of these studies is small sample sizes – samples only rarely achieve more than 1,000 twin pairs, and twin studies need very large samples (over 10,000) to achieve really good statistical power; in consequence, confidence intervals tend to be very wide and often include zero, which renders some of the heritability values reported somewhat suspect on statistical grounds. In any case, how does one interpret the fact that heritability of these various religious traits in adolescence is zero or close to zero, whereas in adults the heritability values are at least generally positive, even if one does not take the precise values that seriously? The usual explanation provided is that shared environmental effects on religious practices and values should be considerably higher in childhood and adolescence than in adulthood, because childhood behaviour is under parental control. It would thus be expected that genetic effects would increase variation in early adulthood as adolescents gain more independence and are able to make their own

https://doi.org/10.1017/9781108566520.010 Published online by Cambridge University Press

146

9 GENES, RELIGIOSITY AND POLITICAL COMMITMENT

behavioural choices (known technically as ‘niche picking’). Well, maybe, although no coherent theories exist as to why genetic variation might relatively quickly start to exert measurable effects in adulthood on the variation in such complex behavioural traits, which, as already noted, are themselves difficult to define clearly. As the epigenetically regulated gene expression profile changes during development, perhaps this amplifies the role of genetic variation on the trait in adulthood? Variance in personality traits, as discussed in Chapter 7, are well established as displaying positive heritability values, so correlations between certain personality types and particular forms of spirituality or religious practice might perhaps go some way towards explaining the results (Kandler and Riemann, 2013). In science, it is usual to seek simpler explanations for data in place of more complicated or implausible explanations. In this context, we need to remember, as mentioned in Chapter 4, that in twin studies the contributions of the three different factors being looked at that cause the variation in the trait – the heritability, the shared environment and the non-shared environment – all have to add up to 1. So when the children leave home and go their separate ways such that the influence of the shared environment dramatically declines, ipso facto something has to change, and the correlation with genetic variation, the heritability, now begins to loom large. As it happens, there is a simple explanation for the heritability difference between adolescence and adulthood that need not involve genetic variation at all, at least not in any direct sense. It should be remembered that all that is necessary to generate a positive heritability value is for the similarities between pairs of identical twins to be more correlated than between pairs of non-identical twins with respect to a particular measurable trait (like church attendance). It is quite possible that the growing disparity in scores between the different twin types in adulthood is due to the closer links and emotional bonds that are known to be more of a characteristic of identical than of non-identical twins once twins leave home. No one is a more ‘significant other’ in the life of an identical twin than their co-twin. For example, an interesting study from Germany based on 133 monozygotic and sixty same-sex dizygotic older twin pairs reported data on the degree of closeness between the different twin pairs over time (Neyer, 2002). A couple of graphs from this paper are interesting to look at, as illustrated in Figure 9.1. There was much greater frequency of contact, emotional closeness and level of mutual support throughout adulthood reported by the identical compared with the non-identical twins. The dip for both types of twins in their frequency of contact and emotional closeness in adulthood compared with adolescence is explained simply by their leaving

https://doi.org/10.1017/9781108566520.010 Published online by Cambridge University Press

9.2 THE HERITABILITY OF RELIGIOSITY

(b)

6.0 5.5 5.0

MZ DZ

4.5 4.0

4.4 4.2 4.0

MZ

3.8

DZ

3.6 3.4

3.5

1

2

–7 63

2

–6 53

2

43

–4 34

–5

3

3

–3

–2

–7

Age intervals (years)

24

1

2 63

2

–6 53

2

43

–5

3

–4 34

–3

–2

3

3.2

24

17

Emotional closeness

4.6

17

Frequency of contact

(a)

147

Age intervals (years)

Figure 9.1 A comparison between identical (MZ = monozygotic) and non identical (DZ = dizygotic) twins in terms of (9.1a) their frequency of contact and (9.1b) their emotional closeness. Adapted from Figure 1 of Neyer (2002).

home and finding jobs in different places, whereas in retirement they tend to move back closer again. A more recent study on thousands of twins in the age range 20–41 reported that 24 per cent of the identical twins had ‘almost daily’ contact compared with only 10 per cent of the non-identical twins (McCaffery et al., 2011). It is not difficult to imagine how this ongoing interaction and greater emotional attachment throughout life might contribute to the measured greater correlation in traits of religious attitudes and practices between identical compared with non-identical twins in adulthood. So one possibility is that the heritability of the variation in religiosity, at whatever age, is actually zero, and genetic variation is irrelevant to differences in religiosity traits. Two observations, however, argue against this conclusion. The first is that when a small sample of adult twins ‘reared apart’ were measured for five different aspects of religious interests, attitudes and values, heritability values of around 50 per cent were determined (Waller et al., 1990), data consistent with other similar studies measuring various aspects of religiosity (Bouchard et al., 1999, Segal, 2012). On the face of it, given their separation, the ‘greater contact’ explanation could not apply in such contexts. Having said that, the numbers of each type of twin pair involved in such studies are small and the number of years of separation ranged from 0 to 69 (separation time was arbitrarily set at 0 years for twin pairs who were reared in different homes but who had periodic contact during childhood) (Waller et al., 1990). For such reasons, data obtained from such studies are not as easy to interpret as might first appear. The second observation that might count against the ‘greater contact’ argument is that there is no reported heritability for religious

https://doi.org/10.1017/9781108566520.010 Published online by Cambridge University Press

148

9 GENES, RELIGIOSITY AND POLITICAL COMMITMENT

affiliation in adulthood, whereas the greater social contact of identical twins in adulthood might be expected to generate more conformity in such commitments. So the jury is out on this one. But if you think about the much closer emotional attachments that, on average, identical twins appear to experience in adulthood compared with non-identical twins – and you then think how heritability values are based on a given trait being more similar between identical compared with non-identical twins – then my own view is that such a scenario is quite sufficient to explain the apparent heritability of religiosity in adults. Such a hypothesis can also be tested. All you need is a few thousand adult twin pairs, an estimate of their religiosity and how many times they see each other per year. If the heritability of religiosity is the same independent of how close they are, then my hypothesis will be falsified. As we shall see below, a similar discussion comes up again in the context of political commitments. A further curious fact usually ignored in reports on the heritability of religious traits is that if atheism is treated as merely the low end of the religiosity scale, then its heritability value will be identical. For example, a scale assessing church attendance is, at the same time, a scale assessing nonchurch attendance. If identical adult co-twins correlate very highly for church attendance (and they do), it follows that they also correlate very highly for non-church attendance. The statement ‘Church attendance is X per cent heritable’ could equally easily be written as ‘Non-church attendance is X per cent heritable.’ This is equally true of any of the religiosity scales used by the various twin studies. However, in the same way that the church attendance domain measures only a small facet of religiosity, so nonchurch attendance measures only a small facet of atheism. There could be numerous social reasons for not attending religious services, undertaking religious practices or observing religious holidays in addition to atheistic beliefs. But in any event, if the suggested non-genetic ‘simple explanation’ for the heritability of adult religiosity is correct, as it may well be, then any discussion of the heritability of religiosity or of atheism is likely to be irrelevant anyway. Somewhat related to the heritability of atheism are studies on the role of genetic variation in ‘apostasy’. By apostasy is generally meant a disengagement from religious belief, identity or practice. In highly religious societies, such as the United States of America, this has been linked to risk-taking behaviours, given that, in adolescence at least, leaving a religious community might be associated with certain negative consequences, such as loss of friends and support from that community (Zuckerman et al., 2016). The transition from

https://doi.org/10.1017/9781108566520.010 Published online by Cambridge University Press

9.3 HERITABILITY OF POLITICAL COMMITMENT

149

adolescence to adulthood is also associated with what are called ‘externalising’ behaviours, defined as explicit behaviours that have a negative effect on the environment. It is therefore of interest that a positive heritability has been reported for various aspects of apostasy (Freeman, 2019). Because apostasy is such a broad concept, three specific traits were investigated in this study – disengagement from religious institutions, cessation of prayer and religious disaffiliation – in the lives of 337 adolescent twin pairs followed up over 15 years. As the authors of this study highlight: ‘Each measure of apostasy captures disengagement from a different aspect of religion.’ In fact the numbers in their sample (twenty individuals) who displayed all three traits were too small to analyse, so each trait was assessed separately. The study reports heritabilities of 34 per cent for cessation of prayer and 75 per cent for religious disaffiliation, whereas the heritability value for disengagement from religious institutions was not significant. What does all this mean? As the author points out, and as every parent knows, puberty is associated with impulsive behaviour, and the role of genetic variation in such behaviour is well established. As the paper points out: ‘Declaring oneself free of a denomination or deciding to cease from daily prayers may be thrilling for a risk-seeking person, while they would be daunting to a risk-averse person.’ I’m not sure about ‘thrilling’, but we get the point. Religious disaffiliation may also be seen as a sign of rebellion against parents or religious authorities. So, as is often the case in genetic studies of religion, the results are not really about religion as such but rather reflect a whole package of trait differences that are related to personality and the contrasts between adolescent and adult behavioural characteristics. Believe it or not, there have been attempts to identify specific genetic variants that associate with religiosity, but as with other candidate gene studies, no replicable ‘hits’ have been identified, an unsurprising result given the discussion so far. In The God Gene: How Faith is Hardwired into our DNA, Dean Hamer claimed to have identified a variant version of a gene (VMAT2) involved in neurotransmitter levels in the brain that supposedly associates with the trait of ‘self-transcendence’ (Hamer, 2004). However, the study was never published in the peer-replicated literature, nor has it ever been taken seriously by others in the field. It is claims such as those made by The God Gene that often lead to hyped-up accounts in the media, but which, by the same token, unfortunately bring the field of behavioural genetics as a whole into disrepute.

9.3 Heritability of Political Commitment The literature on the heritability of political attitudes and commitments goes back more than 40 years, providing one includes broad social traits such as

https://doi.org/10.1017/9781108566520.010 Published online by Cambridge University Press

150

9 GENES, RELIGIOSITY AND POLITICAL COMMITMENT

‘conservatism’ and ‘radicalism’. More recent research has focused on the measurement of specific political traits using the standard twin methodology. As with ‘religiosity’, it is a challenge to know exactly what to measure, and different studies have homed in on different political traits (Hatemi et al., 2011, Hatemi and McDermott, 2012). Studies have included: identifying the heritability of voter turnout (Fowler et al., 2008); exploring common genetic influence on attitudes and voter choice (Hatemi et al., 2007); conducting longitudinal twin studies (from children to adulthood) on attitudes (Hatemi et al., 2009); and exploring the covariance among personality, partisanship and intensity of political comment (Verhulst et al., 2012). Twin studies have consistently reported positive heritability values in the range of 48–76 per cent for traits such as political participation (Fowler et al., 2008), ‘political sophistication’ (Arceneaux et al., 2012), ‘political interest’ (Klemmensen et al., 2012) and ‘foreign policy preferences’ (Cranmer and Dawes, 2012). By contrast, actual political party identification was reported to be due mostly to shared environmental influences (Hatemi et al., 2009). In this respect, the findings are rather similar to those for religiosity for which twin studies revealed no heritability for religious or denominational affiliation. How can one interpret such findings? No one believes that there are ‘genes for political preferences’ or ‘genes for conservatism’ or ‘genes for foreign policy preferences’, despite some media headlines to the contrary. The questionnaires used generally focus on twenty-first-century notions of American ‘conservatism’ or ‘liberalism’, including attitudes to ‘pyjama parties, nudist camps, computer music and horoscopes’ (Alford et al., 2005). But such ideas are highly culture and time dependent. Maintaining conservative attitudes does not even mean the same in the United Kingdom and the United States of America, let alone in Japan or North Korea or Saudi Arabia. The problem comes in treating a ‘political attitude’ as if it were something akin to height, body mass index or type 1 diabetes. It is not. The prevalence of type 1 diabetes is 45 per 100,000 in Finland and 1 per 100,000 in Venezuela. It is a trait with a clear universal definition and is highly dependent on differences in the polygenic and environmental compositions of different populations. A political attitude is simply nothing like that: it can mean something quite different depending on era and cultural context. This becomes even more problematic when trying to interpret the reports of positive heritabilities for mobile phone use (including amount of time spent texting) (Miller et al., 2012) and consumer preferences for soups and snacks, hybrid cars, science fiction movies and jazz (Simonson and Sela,

https://doi.org/10.1017/9781108566520.010 Published online by Cambridge University Press

9.3 HERITABILITY OF POLITICAL COMMITMENT

151

2011). An approach that ‘explains everything’ runs the risk of ending up explaining nothing. On a more positive note, clues about the interpretation of the heritability of political engagement traits arise from the fact that they correlate with many other facets of human personality, such as a prosocial personality and behaviour, ‘obedience to traditional authority’ (Ludeke et al., 2013), openness to experience and the ‘need to evaluate’ (Bizer et al., 2004). People who trust others more, a heritable trait (Sturgis et al., 2010), are more likely to join political organisations and civic associations (Uslaner and Brown, 2005). One study reported, with respect to the heritability of two key aspects of political orientation, that a ‘substantial proportion of this genetic variance can be accounted for by genetic variance in personality traits’ (Kandler et al., 2012). However, as in pretty much every aspect of the study of genetic variation in relation to politics, there are studies that come to different conclusions. Some attempts have been made to move beyond correlations to causation. One study suggests that, rather than personality traits causing people to develop different political attitudes, similarities between the two are largely due to common genetic influences (Verhulst et al., 2012). A further study measured personality and political traits 10 years apart in a large cohort of adult Australian twins, concluding that personality changes were not causal for differences in political attitudes over this time period (Hatemi and Verhulst, 2015). However, disentangling behavioural causes and effects in any biological organism is tricky, let alone when it comes to complex human behavioural traits, and it is not even clear what it really means to suggest that one complex trait is ‘causal’ for another: the fact that one trait changes, on average, before another trait in a population does not necessarily establish a causal relationship. For the moment, it is perhaps safer to accept the conclusions of the great bulk of the literature in this field, namely that the positive heritability values of political traits are most likely linked to the heritability of different personality traits (Friesen and Ksiazkiewicz, 2015). This is at least consistent with the data showing that whereas personality differences develop in early life, political attitudes and preferences reveal no heritability until adulthood after twins have left home. This last observation is again reminiscent of the data on religiosity. In one study, no heritability of political attitudes was found in adolescents, but measurements of heritability then increased sharply upon leaving home (Hatemi et al., 2009). During childhood and adolescence, individual differences in political attitudes were accounted for by a variety of environmental influences, with the influence of shared family environment increasing

https://doi.org/10.1017/9781108566520.010 Published online by Cambridge University Press

152

9 GENES, RELIGIOSITY AND POLITICAL COMMITMENT

markedly between the ages of 9 and 17. At the point of early adulthood (in the early 20s), for those who had left their parental home, there was evidence of a sizeable positive heritability value on political attitudes, which then remained stable throughout adult life. However, this was not observed in the population comprising those twins who continued to live at home from the age of 21–25. The authors’ conclusions were similar to the interpretation proffered for the comparable findings on religiosity: ‘genetic influences are only expressed in early adulthood and only when powerful social pressures such as the parental environment are removed’ (Hatemi et al., 2009). Again, it is tempting to speculate that it is the greater social contact between identical compared with non-identical twins in adulthood after leaving home that could most simply explain such results. The evidence for genetic influence starting at about the age of 21–25 arose mainly from a substantial drop in the similarities of political attitudes between non-identical twins, while the similarities between identical pairs remained stable. The political attitudes of non-identical twins with less contact after leaving home might well drift further apart when compared with identical twins. Counting against this suggestion, as with the similar suggestion made for religiosity, are measurements of traits such as ‘conservatism’ and ‘right-wing authoritarianism’ based on twins ‘raised apart’, studies that generate positive heritability values (Segal, 2012). It therefore remains an open question as to the extent to which decreased social contact among non-identical twins in adulthood compared with identical twins may explain the observed heritabilities of political attitudes that are positive in adults but zero in children and adolescents. Given the challenges facing GWAS on other behavioural traits as described earlier, it will come as no surprise that the use of GWAS in an attempt to correlate genetic variants with political ideological traits has failed to identify any SNPs at an acceptable level of significance (Hatemi et al., 2014). Earlier correlations reported between candidate genes and political traits are now generally accepted as being false positives (Charney and English, 2012). It is noteworthy that in the literature on the genetics of political attitudes and commitments, more than most, the language tends to drift imperceptibly from data and comments on variation in specified populations to inferences concerning the supposed ‘strength’ of different influences: for example, ‘To study the relative strength of genetic and environmental influences on individual traits and behaviours, a range of methods have been developed based on the comparison of family members with varying degrees of genetic similarity’ (Sturgis et al., 2010). Furthermore, the idea of

https://doi.org/10.1017/9781108566520.010 Published online by Cambridge University Press

9.4 WHAT DOES IT ALL MEAN?

153

heritability is often confusingly co-mingled with the idea of inheritance, which is something quite different, as in the title dealing with the heritability of political traits: ‘Are political orientations genetically transmitted?’ (Alford et al., 2005). The authors wish to reply in the affirmative, but their data concern heritability and have nothing to say about inheritance.

9.4 What Does It All Mean? It should be noted that the overall attitude of this chapter towards the findings of behavioural genetics has been one of general scepticism. Indeed, many in the social and political sciences treat the approaches described here as a waste of time, yet perhaps there is no need to adopt quite such a negative position. For example, let us imagine that the positive heritabilities of complex traits such as religiosity and various measurement of political commitment are indeed due to genetic variation. Such data would still be consistent with the intricate weaving together of genetic and environmental influences during early human development. Since the development of human traits is, according to this framework, 100 per cent genetic and 100 per cent environmental, any reductionist attempt to use an experimental approach to calve up the relative contributions to the variation found in populations with respect to a particular trait is bound to come up with various forms of ‘both/and’ answers. One reasonable assumption is that genetic variation does indeed contribute to personality differences that emerge during early childhood, as we have already noted in Chapter 7. Another reasonable assumption, with some data in support, is that people with certain types of personality find it easier to stick to certain commitments than others. With a bit of effort, the ‘others’ can achieve just as high levels of commitment as those who find it easy, but in a large population of twins under study, the ‘commitment is easier’ cohort numerically contribute to the positive heritability results described in adults. In such a scenario, the primary effects of genetic variation are on personality differences – and these in turn have an influence on someone’s commitments to pretty much anything they have an interest in, be it sport, religion, politics or stamp collecting. So the data described in this chapter say nothing about which religious or political beliefs are, or are not, well justified. The heritability of which specific religious or political beliefs people adhere to is zero. When it comes to personal beliefs, genes don’t exert a ‘pull’ on believing one thing rather than another – that’s up to environmental influences and personal choice – but genetic variation might be involved, via personality differences, in the

https://doi.org/10.1017/9781108566520.010 Published online by Cambridge University Press

154

9 GENES, RELIGIOSITY AND POLITICAL COMMITMENT

degree of personal effort required to practise those beliefs with equivalent enthusiasm. What behavioural genetics seeks to do is to take a unified cake and cut it into proportional slices using the approach of population genetics. As an experimental approach, this is a perfectly valid methodology, the kind of methodological reductionist approach that science uses all the time. But it has to be remembered that it is the functioning of the organism as an integrated whole that, in the end, is of most interest to biologists – and indeed to human beings in general – and that questions about genetic determinism relate to the individual and not to the population. As already emphasised, the question of the ‘strength’ of the genetic influence on a given trait is not something that can be inferred from population studies. A higher heritability value does not entail a ‘greater influence’ of genetic variation on a trait than a lower value. A variant gene or a small collection of variant genes, might, in principle, contribute a large amount of variation to a population with respect to a particular trait, and at the same time the genome of each individual might be 100 per cent necessary for the normal development of the 99 per cent of that trait that is invariant between individuals. The field of behavioural genetics at present is somewhat fragmented into the ‘true believers’ who will defend traditional methodologies at all costs, without really engaging in the critiques of the field, and those who adopt a more self-critical stance, recognising that there is a great need for new ideas and new techniques if the field is to make progress. Closer interactions between biologists and philosophers could be very beneficial for all parties, and likewise more interdisciplinary research between animal and human behavioural geneticists might help to move the field forward. Investigating traits using many different experimental approaches in parallel can be informative. In the chapter that follows, an attempt is made to give an interdisciplinary overview of the causes of a complex human behavioural trait – same-sex attraction – a composite story in which genetic variation represents but one possible contributory factor among many.

https://doi.org/10.1017/9781108566520.010 Published online by Cambridge University Press

10

Gay Genes? Genetics and Sexual Orientation

I

n previous chapters, the emphasis has been on the role of genetic variation in potentially explaining differences in traits in human populations. The focus here broadens to provide a general review of the current academic literature on the causes of same-sex attraction (SSA), a review that includes genetic data but which also aims to survey a wide range of personal, environmental and biological causes that have been proposed to explain this trait.1 It is perhaps worth emphasising at the outset of this discussion that the focus of this chapter simply concerns the present scientific data involved. Some people feel uncomfortable with even the attempt to delve into the roots of such traits. But the focus throughout this book is simply to ask the question: ‘What role does genetic variation play in the variation in this trait in a given population?’ And the motivation to ask that question comes very much from the need to counter some of the exaggerated accounts on the subject that appear in both the scientific literature and in the media. So it doesn’t matter if the trait is intelligence, personality differences, differences in size and shape, or any of the other hundreds of traits that have been studied – our aim is just to see where the scientific data are currently pointing, and the need to interpret those data in a careful and, if necessary, sceptical manner. It is a simple fact that most people experience attraction to the opposite sex, whereas a minority experience attraction to the same sex, and some more or less to both. So the question is: why? Many hypotheses about the causes of SSA have been proposed, which can be divided into three broad types of cause: environmental, biological and personal choice. Following some comments on the definition and measurement of SSA and related terms, the personal choice option is briefly discussed, then the environmental explanations, followed by the biological explanations in greater detail. However, in separating in this way the 155

https://doi.org/10.1017/9781108566520.011 Published online by Cambridge University Press

156

10 GAY GENES? GENETICS AND SEXUAL ORIENTATION

proposed causes, it should not be assumed that a single cause or causal chain is responsible for such a complex phenomenon as SSA. In reality, as will be discussed, it is very likely that many different causes are operating in tandem, and that causes are operating across the cohort of SSA individuals in different ways that are likely to be gender and culture specific. Although considered under separate headings, all influences are in reality completely integrated within the life of a developing individual, meaning that no one should expect to find ‘the’ cause of an individual’s SSA.

10.1 Defining and Measuring Same-sex Attraction Before reviewing the causes of SSA, it is important to clarify what is meant by SSA, and to acknowledge the complexities of definition and meaning in the field of sexual behaviour studies (Savin-Williams, 2006, Gates, 2011, Bailey et al., 2016).2 Sexual attraction refers to erotic desire experienced towards other individuals. Attraction is not a discrete variable, and exists along a continuum, from attraction exclusively towards the opposite sex (oppositesex attraction, OSA) to attraction exclusively towards the same sex, with attraction to both sexes equally in the middle. An individual’s sexual attraction status is frequently measured with a Kinsey scale, which usually defines seven points ranging from 0 (exclusively OSA) to 6 (exclusively SSA). Although sexual attraction is continuous, in practice almost all the studies reviewed in the following paragraphs that use the Kinsey scale collapse the continuum into discrete categories to increase statistical power: OSA (0–1), bisexual (2–4) and SSA (5–6), or OSA (0–1) and SSA (2–6), which should be remembered when interpreting empirical data. Other scales besides the Kinsey scale have been proposed and are in active use, such as the Klein Sexual Orientation Grid, the Sexual–Romantic scale and the Gendered Sexuality scale (Galupo et al., 2018). However, as many of the data reported here depend on the Kinsey scale, it should be assumed that this is the scale that has been used, unless otherwise stated. Accurately measuring SSA, either demographically or in experimental populations, is not straightforward. First, attraction is often conflated with, or extrapolated from, measures of related but distinct concepts including sexual behaviour, sexual fantasy and self-identity. All these facets together contribute to an individual’s sexual orientation. These facets do not always, or even frequently, exactly correlate but interact in multiple ways in different individuals. For example, a person who experiences SSA may not engage in same-sex behaviour, and someone who self-identifies as heterosexual may well experience some degree of SSA. On average, nearly

https://doi.org/10.1017/9781108566520.011 Published online by Cambridge University Press

10.1 DEFINING AND MEASURING SAME SEX ATTRACTION

157

three times as many people report some degree of SSA than the number of people who self-identify as gay, lesbian or bisexual. Consequently, studies that measure concepts other than attraction should be used to make inferences about SSA with caution. Having said that, in many of the studies reviewed below, it is a reasonable assumption that same-sex behaviour and/or a homosexual identity accompany SSA (although the converse is not necessarily true). Second, even where attraction is directly assessed, measuring instruments (usually self-report questionnaires or surveys) can differ enormously. Variables include how questions are phrased, the degree or frequency of attraction deemed sufficient to categorise as SSA, the number of possible options offered and the degree of anonymity. Different instruments can thus produce significantly differing results (Gates, 2011). The time period assessed (lifetime SSA versus current or recent SSA) is critical, as sexual attraction can be a dynamic trait. Long-running longitudinal surveys in the Uniited States of America and New Zealand have found that although the majority of individuals, between 80 and 90 per cent, have a stable sexual attraction or sexual self-identity across their lifespan, a sizeable minority experience change in their sexual attraction status over time (Ott et al., 2011, Mock and Eibach, 2012). Changes occur in both directions (i.e. SSA to OSA and vice versa) across the lifespan (Figure 10.1), so it is incorrect to assume that individuals who report current OSA never have or never will experience SSA. In absolute terms, more individuals move from exclusive OSA to degrees of SSA than move in the other direction, although in percentage terms, the reverse is true. There is also a significant sex difference, with female sexual attraction status much more labile than in men (Diamond, 2008). These data are important for interpreting the causes of SSA – the variation in when same-sex attraction develops in different people suggests that no one cause is sufficient to explain all forms of SSA. Because of the difficulties of measurement, estimates of the prevalence of SSA, in either national or global populations, vary widely.3 Table 10.1 lists some recent prevalence estimates for the national UK population for SSA and same-sex behaviour, aggregated from multiple data sources. Across surveys, some trends can be identified. Consistently, more women report experiencing some degree of SSA than men. Experiencing some SSA is more prevalent in the population than undertaking same-sex behaviour, which in turn is generally more prevalent than having a nonheterosexual self-identify (although measures of identity are highly variable across surveys). The prevalence of having exclusive SSA or same-sex behaviour is much lower than the prevalence of having any degree of SSA or same-sex behaviour.

https://doi.org/10.1017/9781108566520.011 Published online by Cambridge University Press

10 GAY GENES? GENETICS AND SEXUAL ORIENTATION

158

How old were you when you first thought that you were something other than straight or heterosexual?

Number of gay, lesbian and bisexual respondents

160 12 Median age (all)

140 120 100

10 Median age for gay men

13 Median age for lesbians and bisexuals

80 60 40 20 0 0

10

20

30 40 Age (years)

50

60

Figure 10.1 The development of same sex attraction (SSA) in a US demographic sample. The majority of SSA individuals develop stable SSA in early puberty, with a median age of 10 years for men and 13 years for women. A minority first identify their SSA much later in life. Adapted, with permission, from Pew Research Center (2013) A survey of LGBT Americans: Attitudes, Experiences and Values in Changing Times. Washington, DC: Pew Research Center.

10.2 The Question of Choice Before reviewing in detail putative biological and environmental explanations for SSA, it is necessary to briefly consider a third category, that of personal choice. Many would argue that an experience of sexual attraction, by definition, cannot be consciously chosen or willed, as attraction is a fundamental mental state that is not controlled by the conscious mind. Within this model, sexual attraction is a trait to be discovered within oneself, not created by personal choice. This ‘standard model’ (Wilkerson, 2009), so named because it is the model most commonly held by biologists, sociologists and the general public, underlies many survey instruments of sexual attraction (Savin-Williams, 2009) and popular writings about

https://doi.org/10.1017/9781108566520.011 Published online by Cambridge University Press

10.2 THE QUESTION OF CHOICE

159

Table 10.1 Summary prevalence rates in the United Kingdom for same-sex attraction and behaviour

Any degree of same sex attraction across the lifespan Exclusive same sex attraction Any same sex behaviour across the lifespan Exclusive same sex behaviour

Men (%)

Women (%)

6 8

9 10

2