The Microbiology of Safe Food 9781119405016, 9781119405252, 9781119405535, 1119405017

Citation preview

The Microbiology of Safe Food

The Microbiology of Safe Food Third edition

Stephen J. Forsythe Nottingham Trent University Nottingham, UK

This edition first published 2020 © 2020 John Wiley & Sons Ltd Edition History John Wiley and Sons (1e, 2000), John Wiley and Sons (2e, 2010) All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, electronic, mechanical, photocopying, recording or otherwise, except as permitted by law. Advice on how to obtain permission to reuse material from this title is available at http://www.wiley.com/go/permissions. The right of Stephen J. Forsythe to be identified as the author of this work has been asserted in accordance with law. Registered Offices John Wiley & Sons, Inc., 111 River Street, Hoboken, NJ 07030, USA John Wiley & Sons Ltd, The Atrium, Southern Gate, Chichester, West Sussex, PO19 8SQ, UK Editorial Office The Atrium, Southern Gate, Chichester, West Sussex, PO19 8SQ, UK For details of our global editorial offices, customer services, and more information about Wiley products visit us at www.wiley.com. Wiley also publishes its books in a variety of electronic formats and by print‐on‐demand. Some content that appears in standard print versions of this book may not be available in other formats. Limit of Liability/Disclaimer of Warranty The contents of this work are intended to further general scientific research, understanding, and discussion only and are not intended and should not be relied upon as recommending or promoting scientific method, diagnosis, or treatment by physicians for any particular patient. In view of ongoing research, equipment modifications, changes in governmental regulations, and the constant flow of information relating to the use of medicines, equipment, and devices, the reader is urged to review and evaluate the information provided in the package insert or instructions for each medicine, equipment, or device for, among other things, any changes in the instructions or indication of usage and for added warnings and precautions. While the publisher and authors have used their best efforts in preparing this work, they make no representations or warranties with respect to the accuracy or completeness of the contents of this work and specifically disclaim all warranties, including without limitation any implied warranties of merchantability or fitness for a particular purpose. No warranty may be created or extended by sales representatives, written sales materials or promotional statements for this work. The fact that an organization, website, or product is referred to in this work as a citation and/or potential source of further information does not mean that the publisher and authors endorse the information or services the organization, website, or product may provide or recommendations it may make. This work is sold with the understanding that the publisher is not engaged in rendering professional services. The advice and strategies contained herein may not be suitable for your situation. You should consult with a specialist where appropriate. Further, readers should be aware that websites listed in this work may have changed or disappeared between when this work was written and when it is read. Neither the publisher nor authors shall be liable for any loss of profit or any other commercial damages, including but not limited to special, incidental, consequential, or other damages. Library of Congress Cataloging‐in‐Publication Data Names: Forsythe, S. J. (Steve J.), author. Title: The microbiology of safe food / Stephen J. Forsythe. Description: Third edition. | Hoboken, NJ : Wiley-Blackwell, [2019] | Includes bibliographical references and index. Identifiers: LCCN 2019024541 (print) | LCCN 2019024542 (ebook) | ISBN 9781119405016 (paperback) | ISBN 9781119405252 (adobe pdf) | ISBN 9781119405535 (epub) Subjects: LCSH: Food–Microbiology. Classification: LCC QR115 .F675 2009 (print) | LCC QR115 (ebook) | DDC 579/.16–dc23 LC record available at https://lccn.loc.gov/2019024541 LC ebook record available at https://lccn.loc.gov/2019024542 Cover Design: Wiley Cover Image: © Kateryna Kon/Shutterstock Set in 10/12pts Minion Pro by SPi Global, Pondicherry, India Copyright for Plate 20 belongs to CDC 10 9 8 7 6 5 4 3 2 1

Contents

Preface to third edition Preface to second edition Preface to first edition 1  Foodborne infections

1.1 1.2 1.3 1.4 1.5 1.6

1.7 1.8 1.9 1.10 1.11 1.12

1.13 1.14

The microbial world and its relationship to food Origins of safe food production Overview of foodborne illness Public perception of safe food Causes of foodborne illness Food poisoning due to common food commodities 1.6.1 Milk and milk products 1.6.2 Meat products 1.6.3 Fresh produce 1.6.4 Low‐water activity (aw) and low‐moisture foods Host‐related issues Hygiene hypothesis Chronic sequelae following foodborne illness The size of the foodborne illness problem The cost of foodborne diseases Changes in antimicrobial resistance of foodborne pathogens 1.12.1 Bacterial antibiotic resistance in agriculture and aquaculture 1.12.2 Antibiotics of concern and resistance mechanisms 1.12.3 Polymyxin and plasmid‐encoded colistin resistance 1.12.4 Livestock‐associated methicillin‐resistant Staphylococcus aureus (LA‐MRSA) Food safety following natural disasters, and conflict Food microbiology, foodborne diseases and climate change

2  Basic aspects

2.1 The human intestinal tract 2.2 The normal human intestinal flora 2.3 Host resistance to foodborne infections

xvii xix xxi 1

2 6 7 14 17 20 20 20 20 21 22 23 23 24 36 38 38 40 41 42 42 43 45

45 46 51

v

vi

Contents

2.4 Bacterial cell structure 2.4.1 Morphology 2.4.2 Cell membrane structure and the Gram stain 2.4.3 Lipopolysaccharide (LPS, O antigen) 2.4.4 Flagella (H antigen) 2.4.5 Capsule (K and Vi antigen) 2.5 Bacterial toxins and other virulence determinants 2.5.1 Bacterial endotoxins and exotoxins 2.5.2 Pathogenicity islands 2.5.3 Bacterial toxins encoded in bacteriophages 2.6 Microbial growth cycle 2.7 Death kinetics 2.7.1 Expressions 2.7.2 Decimal reduction times (D values) and z values 2.8 Factors affecting microbial growth 2.8.1 Intrinsic and extrinsic factors affecting microbial growth 2.8.2 Water activity 2.8.3 pH 2.8.4 Temperature 2.8.5 Interplay of factors affecting microbial growth in foods 2.9 Microbial response to stress 2.9.1 General stress response (GSR) 2.9.2 pH stress 2.9.3 Heat shock 2.9.4 Cold shock 2.9.5 Osmotic shock 2.10 Predictive modelling 2.10.1 Predicting modelling development 2.10.2 Primary models and the Gompertz and Baranyi equations 2.10.3 Secondary models 2.10.4 Tertiary models 2.10.5 Application of predictive microbial modelling 3  Food preservation and spoilage organisms

3.1 Spoilage micro‐organisms 3.1.1 Spoilage micro‐organisms 3.1.2 Spoilage of dairy products 3.1.3 Spoilage of meat products 3.1.4 Fish spoilage 3.1.5 Egg spoilage 3.1.6 Cereals and grain 3.2 Shelf life indicators 3.2.1 Glucose 3.2.2 Gluconic and 2‐oxogluconic acid 3.2.3 L‐ and D‐lactic acids, acetic acid and ethanol 3.2.4 Biologically active amines 3.2.5 Volatile compounds

52 52 52 54 55 55 55 56 60 62 63 63 63 64 68 68 69 70 71 73 73 75 75 76 77 77 78 78 79 80 81 82 85

85 87 89 89 90 90 91 91 92 92 93 93 93

Contents

vii

3.2.6 Storage trials 93 3.2.7 Challenge tests 93 3.2.8 Predictive modelling 93 3.3 Methods of preservation and shelf life extension 93 3.4 Preservatives 95 3.4.1 Organic acids 96 3.4.2 Hydrogen peroxide and lactoperoxidase system 97 3.4.3 Chelators 97 3.4.4 Non‐acidic preservatives 98 3.4.5 Preservation due to weak acids and low pH 99 3.4.6 Biopreservatives 99 3.5 Physical methods of preservation 100 3.5.1 Preservation by heat treatment 101 3.5.2 High‐pressure treatment 102 3.5.3 Ohmic heating and radio frequency 103 3.5.4 Pulsed electric fields 103 3.5.5 Ultrasound 104 3.5.6 Intense light pulse 104 3.5.7 Food irradiation 104 3.5.8 Cold plasma and low‐energy electrons for food surface decontamination 108 3.6 Packaging 109 3.6.1 Reduced oxygen packaging, modified atmosphere packaging and active packaging 109 3.6.2 Antimicrobial packaging and nanotechnology 110 3.7 Fermented food products 111 3.7.1 Fermented milk products 113 3.7.2 Fermented meat products 117 3.7.3 Fermented vegetables 118 3.7.4 Fermented protein foods: shoyu and miso 118 3.8 Organisms involved in the production of fermented foods 118 3.8.1 Lactic acid bacteria 120 3.8.2 Bifidobacterium species 126 3.8.3 Other organisms 127 3.9 Functional foods: probiotics and gut modulation 128 3.9.1 Qualified Presumption of Safety (QPS) and Generally Regarded As Safe (GRAS) 128 3.9.2 Functional foods and probiotics 129 3.9.3 Probiotic studies 131 3.9.4 Novel organisms – modulation of gut microbiota 132 4  Bacterial foodborne pathogens

135

4.1 Indicator organisms 137 4.1.1 Coliforms 137 4.1.2 Enterobacteriaceae 138 4.1.3 Enterococci 138 4.1.4 Bacteriophages 138

viii

Contents

4.2 Campylobacter jejuni, C. coli and C. lari 139 4.2.1 General description 139 4.2.2 Campylobacter infections 140 4.2.3 Campylobacter jejuni typing 141 4.2.4 Virulence factors 142 4.2.5 Whole‐genome sequence analysis 145 4.2.6 Sources and control of Campylobacter jejuni 147 4.3 Salmonella serovars 148 4.3.1 General description 148 4.3.2 Salmonella serotypes 150 4.3.3 Infections caused by Salmonella serovars 152 4.3.4 Virulence factors of Salmonella serovars 154 4.3.5 Whole‐genome analysis 155 4.3.6 Sources and control of Salmonella serovars 156 4.3.7 Salmonella serovar outbreaks 157 4.4 Pathogenic E. coli 160 4.4.1 General description 160 4.4.2 E. coli pathovars 161 4.4.3 Infections caused by E. coli pathovars 163 4.4.4 Virulence factors 168 4.4.5 Whole‐genome analysis 169 4.4.6 Sources and control of E. coli pathovars 170 4.4.7 Outbreaks caused by E. coli pathovars 172 4.5 Sh. dysenteriae and Sh. sonnei 176 4.5.1 General description 176 4.5.2 Shigellosis 177 4.5.3 Virulence factors 177 4.5.4 Sh. sonnei outbreak 178 4.6 Cronobacter species 178 4.6.1 General description 178 4.6.2 Infections due to Cronobacter species 179 4.6.3 Identification and typing methods for Cronobacter species 180 4.6.4 Virulence factors 181 4.6.5 Sources and control of Cronobacter species 182 4.6.6 Cronobacter outbreaks 183 4.7 Vibrio cholerae, V. parahaemolyticus and V. vulnificus 184 4.7.1 General description 184 4.7.2 Infections due to Vibrio species 184 4.7.3 Virulence factors 186 4.7.4 Sources and control 187 4.8 Brucella melitensis, Br. abortus and Br. suis 188 4.8.1 General description 188 4.8.2 Brucellosis 188 4.9 Yersinia enterocolitica 189 4.9.1 General description 189 4.9.2 Yersiniosis 189

Contents

4.10

4.11

4.12

4.13

4.14

4.15

4.16

4.17

4.18

ix

4.9.3 Sources and control 190 4.9.4 Outbreaks due to Y. enterocolitica 190 Aeromonas hydrophila, A. caviae and A. sobria 191 4.10.1 General description 191 4.10.2 A. hydrophila gastroenteritis 192 4.10.3 Sources and control 192 Plesiomonas shigelloides 193 4.11.1 General description 193 4.11.2 Plesiomonas infections 193 4.11.3 Sources and control 193 Listeria monocytogenes 194 4.12.1 General description 194 4.12.2 Listeriosis 195 4.12.3 Lineages and typing L. monocytogenes 196 4.12.4 Virulence factors 198 4.12.5 Whole‐genome analysis of L. monocytogenes 200 4.12.6 Sources and control of L. monocytogenes 201 4.12.7 L. monocytogenes outbreaks 203 Staphylococcus aureus 207 4.13.1 General description 207 4.13.2 Infections associated with St. aureus 207 4.13.3 Virulence factors 208 4.13.4 Sources and control 209 Clostridium perfringens 210 4.14.1 General description 210 4.14.2 Cl. perfringens infections 210 4.14.3 Sources and control 211 Clostridium botulinum 211 4.15.1 General description 211 4.15.2 Cl. botulinum intoxication 212 4.15.3 Sources and control 212 B. cereus group 213 4.16.1 General description 213 4.16.2 B. cereus foodborne infections 214 4.16.3 Virulence traits 215 4.16.4 Sources and control 216 Enterococcus and Streptococcus species 217 4.17.1 General description 217 4.17.2 E. faecalis and E. faecium 217 4.17.3 Streptococcus pyogenes, group A streptococci 218 4.17.4 Virulence traits 218 Emerging and uncommon foodborne pathogens 219 4.18.1 Arcobacter genus 221 4.18.2 Campylobacter concisus 223 4.18.3 EAEC, E. coli O55 and E. coli O26 223 4.18.4 Escherichia albertii 226

x

Contents

4.18.5 Providencia alcalifaciens 228 4.18.6 Clostridium difficile 228 4.18.7 Mycobacterium paratuberculosis 230 4.18.8 Acinetobacter species 231 4.18.9 Nanobacteria 231 5  Foodborne pathogens: viruses, toxins, parasites and prions

233

5.1 Foodborne viruses 233 5.1.1 Norovirus 235 5.1.2 Hepatitis A 239 5.1.3 Hepatitis E 241 5.1.4 Rotaviruses 242 5.1.5 Small round viruses, astroviruses, sapporo‐like viruses, adenoviruses and parvoviruses243 5.1.6 Human enteroviruses 244 5.2 Seafood and shellfish poisoning 244 5.2.1 Ciguatera poisoning 245 5.2.2 Scombroid poisoning 246 5.2.3 Paralytic shellfish poisoning 246 5.2.4 Diarrhoeic shellfish poisoning 246 5.2.5 Neurotoxic shellfish poisoning 246 5.2.6 Amnesic shellfish poisoning 247 5.3 Foodborne parasites: eucaryotes 248 5.3.1 Toxoplasma gondii 249 5.3.2 Taenia saginata and T. solium 249 5.3.3 Echinococcus multilocularis and E. granulosus 250 5.3.4 Cyclospora cayetanensis 250 5.3.5 Cryptosporidium parvum 251 5.3.6 Anisakis simplex 252 5.3.7 Trichinella spiralis 252 5.4 Mycotoxins 253 5.4.1 Aflatoxins 255 5.4.2 Ochratoxins 256 5.4.3 Fumonisins 256 5.4.4 Zearalenone 256 5.4.5 Trichothecenes 256 5.4.6 Prions and transmissible spongiform encephalopathies 257 6  Methods of detection and characterisation

6.1 Prologue 6.2 Conventional methods 6.2.1 Culture media 6.2.2 Sublethally injured cells 6.2.3 Viable but non‐culturable bacteria (VBNC) 6.3 Rapid sampling methods 6.3.1 Sample preparation

259

259 264 265 267 268 269 269

Contents

xi

6.3.2 Separation and concentration of target 269 6.4 Rapid end‐detection methods 273 6.4.1 ELISA and antibody‐based detection systems 273 6.4.2 Reversed passive latex agglutination 274 6.4.3 ATP bioluminescence techniques and hygiene monitoring 275 6.4.4 Protein detection 276 6.4.5 Flow cytometry 276 6.4.6 Biosensors 276 6.4.7 Impedance (Conductance) microbiology 278 6.5 DNA‐based molecular typing and proteomic methods 279 6.5.1 Polymerase chain reaction (PCR) 280 6.5.2 Microarrays 282 6.5.3 Loop‐mediated isothermal amplification (LAMP) technique 283 6.5.4 Pulsed‐field gel electrophoresis (PFGE) 284 6.5.5 Restriction fragment length polymorphism (RFLP) 285 6.5.6 Amplified fragment length polymorphism (AFLP) 285 6.5.7 Random amplification of polymorphic DNA (RAPD) 285 6.5.8 Repetitive‐element polymerase chain reaction (Rep‐PCR) 285 6.5.9 Nucleic acid sequence‐based amplification (NASBA) 286 6.5.10 Multiple‐locus variable number tandem (VNTR) repeat analysis 286 6.5.11 PCR‐probe based serotyping 287 6.5.12 Ribotyping 287 6.5.13 Matrix‐associated laser desorption ionisation – time of flight (MALDI‐TOF) 287 6.6 Identification and typing methods based on high‐throughput DNA sequencing288 6.6.1 Conventional seven‐loci MLST 288 6.6.2 Genome sequence‐based MLST 289 6.6.3 CRISPR‐cas array typing 290 6.6.4 Single nucleotide polymorphism (SNP)‐based analysis 291 6.7 Specific detection procedures and accreditation 292 6.7.1 Aerobic plate count (APC) 292 6.7.2 Salmonella serovars 292 6.7.3 Campylobacter species 297 6.7.4 Enterobacteriaceae and E. coli 299 6.7.5 Pathogenic E. coli, including E. coli O157:H7 300 6.7.6 Shigella species 301 6.7.7 Cronobacter genus 302 6.7.8 Aeromonas species 304 6.7.9 Arcobacter species 304 6.7.10 Listeria monocytogenes 305 6.7.11 Staphylococcus aureus 308 6.7.12 Clostridium perfringens 308 6.7.13 B. cereus, B. subtilis and B. licheniformis 309 6.7.14 Mycotoxins 310 6.7.15 Viruses 310

xii

Contents

7  Microbiological criteria

313

7.1 Background to microbiological criteria and end‐product testing 313 7.2 International commission on microbiological specifications for foods (ICMSF)313 7.3 Codex Alimentarius principles for the establishment and application of microbiological criteria 314 7.4 Sampling plans 316 7.5 Variables plans 318 7.6 Attributes sampling plan 321 7.6.1 Two‐class plan 322 7.6.2 Three‐class plan 322 7.7 Principles 322 7.7.1 Defining a ‘lot’ of food 322 7.7.2 Sample unit number 323 7.7.3 Operating characteristic curve 323 7.7.4 Producer risk and consumer risk 324 7.7.5 Stringency of two‐ and three‐class plans, setting n and c 324 7.7.6 Setting the values for m and M 327 7.8 Microbiological limits 329 7.8.1 Definitions 329 7.8.2 Limitations of microbiological testing 329 7.8.3 Examples of sampling plans 329 7.9 Implemented microbiological criteria 333 7.9.1 Microbiological criteria in the European Union 333 7.9.2 EU Directives specifying microbiological standards for foods 333 7.10 UK guidelines for ready‐to‐eat foods 333 8  Hygienic production practices

8.1 Contribution of food handlers to foodborne illness 8.2 Personnel hygiene and training 8.3 Cleaning 8.4 Detergents and disinfectants 8.5 Microbial biofilms 8.5.1 Microbial biofilm formation 8.5.2 Bacterial biofilm induction 8.5.3 Biofilm removal and control 8.6 Assessment of cleaning and disinfection efficiency 9  Food safety management tools

9.1 9.2 9.3 9.4 9.5 9.6

The manufacture of hygienic food Microbiological safety of food in world trade Consumer pressure effect on food processing The management of hazards in food in international trade Hazard analysis critical control point (HACCP) Prerequisite programme

337

337 337 340 343 343 343 346 346 348 351

351 357 358 359 359 360

Contents

9.7 Outline of HACCP 9.7.1 Food hazards 9.7.2 Preparation for HACCP 9.7.3 Principle 1: hazard analysis (HA) 9.7.4 Principle 2: critical control points (CCPs) 9.7.5 Principle 3: critical limits 9.7.6 Principle 4: CCP monitoring 9.7.7 Principle 5: corrective actions 9.7.8 Principle 6: verification 9.7.9 Principle 7: record keeping 9.8 Microbiological criteria and HACCP 9.9 Microbiological hazards and their control 9.9.1 Sources of microbiological hazards 9.9.2 Temperature control of microbiological hazards 9.9.3 Non‐temperature control of microbiological hazards 9.10 HACCP plans 9.10.1 Production of pasteurised milk 9.10.2 Swine slaughter in the abattoir 9.10.3 Chilled food manufacture 9.10.4 Generic models 9.11 GMP and GHP 9.12 Quality systems 9.13 Total quality management 10  Microbiological risk assessment

10.1 Risk analysis and microbiological risk assessment 10.2 Origin of MRA 10.3 MRA – an overview 10.4 MRA – structure 10.4.1 Risk assessment 10.4.2 Risk management 10.4.3 Risk communication 10.5 Risk assessment 10.5.1 Statement of purpose 10.5.2 Hazard identification 10.5.3 Exposure assessment 10.5.4 Hazard characterisation 10.5.5 Dose–response assessment 10.5.6 Dose–response models 10.5.7 Dose and infection 10.5.8 Risk characterisation 10.5.9 Production of a formal report 10.5.10 Triangular distributions and Monte Carlo simulation 10.6 Risk management 10.6.1 Risk assessment policy 10.6.2 Risk profiling

xiii

363 363 363 364 364 366 366 366 366 367 367 369 369 370 371 371 371 372 373 376 382 382 382 385

385 387 389 392 393 394 395 395 396 396 396 401 403 405 409 413 414 414 415 419 419

xiv

Contents

10.7 Food safety objectives (FSO) 10.8 Risk communication 10.9 Future developments in MRA 10.9.1 International methodology and guidelines 10.9.2 Risk assessment database 10.9.3 Training courses and use of resources 11  Application of microbiological risk assessment

419 421 422 422 423 424 425

11.1 Salmonella serovars 425 11.1.1 Salmonella enteritidis in shell eggs and egg products 425 11.1.2 Hazard identification and hazard characterisation of Salmonella in broilers and eggs 428 11.1.3 Exposure assessment of Salmonella serovars in broilers 430 11.1.4 Salmonella serovars in cooked chicken 432 11.1.5 Salmonella serovars in cooked patty 433 11.1.6 Poultry FARM 434 11.1.7 Domestic and sporadic human salmonellosis 435 11.2 Campylobacter 435 11.2.1 C. jejuni risk from fresh chicken 435 11.2.2 Risk profile for pathogenic species of Campylobacter in Denmark 437 11.2.3 Risk assessment of C. jejuni in broilers 438 11.2.4 Campylobacter fluoroquinolone resistance 438 11.3 L. monocytogenes 442 11.3.1 L. monocytogenes hazard identification and hazard characterisation in ready‐to‐eat foods 442 11.3.2 L. monocytogenes exposure assessment in RTE foods 444 11.3.3 Relative risk of L. monocytogenes in selected RTE foods 446 11.3.4 L. monocytogenes in European Union trade 446 11.3.5 L. monocytogenes in meat balls 447 11.3.6 Listeriosis from RTE meat products 449 11.4 E. coli O157 449 11.4.1 E. coli O157:H7 in ground beef 449 11.5 Bacillus cereus 451 11.5.1 B. cereus risk assessment 451 11.6 Vibrio parahaemolyticus 453 11.6.1 Public health impact of V. parahaemolyticus in raw molluscan shellfish 453 11.7 Cronobacter species and Salmonella in powdered infant formula (PIF) 455 11.8 Viral risk assessments 457 11.8.1 Viral contamination of shellfish and coastal waters 457 12 International control of microbiological hazards in foods: regulations and authorities 459

12.1 Control of foodborne pathogens 459 12.1.1 Control of Salmonella serovars in poultry 459 12.1.2 Control of Escherichia coli pathovars and Salmonella serovars in fresh produce 461

Contents

12.1.3 Control of pathogens in low‐moisture foods (LMFs) 12.2 World Health Organisation (WHO), global food security from accidental and deliberate contamination 12.3 Regulations in international trade of food 12.4 Codex Alimentarius Commission (CAC) 12.5 SPS measures, Technical Barriers to Trade (TBT) and the WHO 12.6 EU legislation 12.7 International food safety agencies 12.7.1 European Food Safety Authority (EFSA) 12.7.2 Food authorities in the United States 13  Surveillance and foodborne outbreak investigation

13.1 Surveillance programmes 13.1.1 International Food Safety Authorities Network (IFSAN) 13.1.2 Surveillance systems in the United States 13.1.3 PulseNet international 13.1.4 European Centre for Disease Prevention and Control (ECDC) and European surveillance for salmonellosis and shiga toxin‐producing E. coli (STEC) 13.1.5 European Food‐Borne Viruses in Europe network (FBVE) 13.1.6 Rapid Alert System for Food and Feed (RASFF) 13.1.7 Global Salm‐Surv (GSS) 13.1.8 Surveillance of ready‐to‐eat foods in the United Kingdom 13.2 Outbreak investigations 13.2.1 Preliminary outbreak investigation 13.2.2 Case definition and data collection 13.2.3 Data collation and interpretation 13.3 Social media, crowd sourcing and reporting food poisoning cases 13.4 Mobile phones and food safety 13.5 Food terrorism and biocrimes

xv

462 464 467 468 469 470 471 472 472 475

475 476 476 478 479 480 480 483 483 483 486 486 487 492 493 493

14 Whole‐genome sequencing, microbiomes and genomic epidemiology

14.1 High‐throughput DNA sequencing 14.2 Microbiome analysis 14.3 Genomic epidemiology 14.3.1 Whole‐genome sequencing for microbial source tracking 14.3.2 Genome Trakr network (US) 14.3.3 NCBI pathogen detection site 14.3.4 Center for Genomic Epidemiology (Denmark) 14.4 Key outbreaks investigated using genomic epidemiology 14.4.1 Ready‐to‐eat meat products L. monocytogenes outbreak, Canada, 2008 14.4.2 E. coli O104:H4 outbreak, Germany, 2011 14.4.3 C. jejuni outbreak investigations 14.4.4 Salmonella enteritidis in eggs, European outbreak, 2014 14.4.5 Multinational outbreak of Salmonella Agona through infant formula contamination, 2017

499

499 501 503 503 505 505 505 505 505 506 508 509 510

xvi

Contents

14.4.6 Retrospective Cronobacter sakazakii neonatal intensive care unit outbreak, France, 1994 14.4.7 L. monocytogenes ST6, polony sausages, South Africa, 2017–2018

512 512

Glossary of terms 515 List of abbreviations 521 Food safety resources on the world wide web 525 Plates and credits 531 References533 Index563

Preface to third edition

I can hardly believe the last (second) edition of MoSF was back in 2009. It seems so long ago now. But what has changed in the meantime? Well, we have the recognition of various emergent foodborne pathogens as well as new and improved methods of detection, identification and profiling, not forgetting the increasingly affordability and application of next‐generation DNA sequencing. But has the incidence and burden of foodborne disease decreased? Apparently not. Essentially, we are probably more aware and better informed, and through social media make a complaint quicker. So what warranted another version of this book? Primarily the area of high‐throughput sequencing was high on my list of ‘must cover’. It is redefining microbiology as a whole, especially clinical and environmental microbiology, and to an extent food microbiology. Not just bacterial genome determination of individual laboratory isolates, but microbiome analysis of the human body flora and the flora of food, leading to an appreciation of interactions between the ‘normal’ gut flora and human well‐being. High‐throughput sequencing is now being applied to outbreak investigations. Between the second edition and this third edition there was a particular Escherichia coli outbreak in Germany that hospitalised over 1000, and killed over 50 people. This was the first outbreak to be investigated using whole‐genome sequencing. In fact it became a race between competing genome sequencing platforms for which one could sequence and publicise their results before their competitors. Related to outbreak investigations, in order to distinguish between strains to a greater degree than ‘conventional’ methods such as pulsed‐field gel electrophoresis (PFGE), which has repetitively (ad nauseum) been cited as the ‘gold standard’ for such analysis, DNA sequencing‐based methods are now being applied, such as multi‐locus variable analysis (MLVA), clustered regularly interspaced short palindromic repeats (CRISPR)‐cas array profiling and single nucleotide polymorphism (SNP) analysis. My own research area is being similarly swayed, with genomic analysis being used to inform laboratory‐based research, and vice versa with laboratory observations guiding areas to investigate bacterial populations at the genomic level. Subsequently, some useful detection methods (even early molecular ones) are infrequently used now and so have been removed from this edition – for example, denaturing gradient gel electrophoresis (DGGE). However, I have reluctantly retained microarrays, despite more cost‐effective equivalent methods involving rapid DNA sequencing being available, as they continue to be referred to in the literature. A large number of outbreak investigations have been included in this new edition. This is  because they contain a considerable amount of information and data that can illustrate

xvii

xviii

Preface to third edition

important points in food safety, in particular exampling the consequences of failure to produce safe food. Another reason for the update has been the expansion of social media and Internet resources. Social networks can be used to assist and alert possible outbreaks, though the potential for  misinformation and misattribution is considerable causing unwarranted cost to food ­producers. Currently social media, high‐throughput sequencing and food poisoning outbreak investigations are being combined, and these aspects were not covered in the previous two editions. Obviously, a single‐author book will have certain biases due to personal experience and expertise. As such I have tried to bring to this edition an emphasis reflective of my role as advisor to national and international regulatory and non‐regulatory authorities. Also I have drawn where appropriate on my own publications, but will leave my research students to ‘self‐surf ’ their own papers in the References section. Will this third edition become dated? Inevitably ‘yes’ – however the major organisms causing food poisoning have not changed since the first edition of MoSF in 2000 and probably will not change in the near future. Hence food safety advisors have a job for life! Please enjoy reading this latest edition, and please contact me via email ([email protected]) or various social sites if you want further information or help. Finally, my latest enterprise, this third edition, is again dedicated to my ever‐supportive wife Debbie, my now grown up children James and Rachel, and my parents – without whom none of this would have been possible. Steve Forsythe Nottingham, UK March 2019

Preface to second edition

Although I was pleased with the first edition of this book (MoSF), I nevertheless felt that it was not complete. This new edition tries to address this by including new sections on bioinformatics, biothreats and personnel, as well as updating many other sections. Since 2000, the topic of microbiological risk assessment has increased and subsequently I have incorporated parts of my other Blackwell’s book Microbiological Risk Assessment of Food (2002) into Chapter 10 as it was a substantial improvement on the first MoSF edition’s few pages. My appreciation is due to Simon Illingworth (LabM, Bury, UK) for reviewing Chapter 5 on detection methods for me. A major change is the complementing websites at www.foodmicrobe.com. This was available with the first edition, but was an afterthought and so unfortunately was not fully utilised. In fact it was one of the first web‐based supported books by Blackwells, and the listing of URLs in the Appendix was considered a ‘novelty’! How much has changed since 2000. I am using the web for two main purposes. First, to keep some chapters up to date, and second to offer various data exercises that are not in keeping with the book format. One aspect that I have been wanting to expand and encourage ‘younger’ readers to explore is the application of genomics, post‐genomics and bioinformatics to food microbiology. Again the first edition included microarrays but not the tools for one to investigate microbial genomes for oneself. In fact, 2000, the year MOSF was published, was also the year that the first version of the Campylobacter jejuni genome was released, and since the MOSF text was written in 1999, the whole topic of microbial genomics was not even on the radar. The fact that genomes are sequenced faster than they can be fully annotated means that one can quickly discover something that no‐one else has even known before, and I hope the bioinformatics aspects will enable and encourage readers to try in silico research. One topic that was gaining increasing public attention in 1999– 2000 was bovine spongiform encephalopathy‐variant Creutzveldt–Jakob disease (BSE‐vCJD). It appears in the intervening years that we have possibly passed the peak incidence, fortunately. However, over the same period, the spectre of bioterrorism has arisen and so this issue is addressed in a new section of this edition. One thing that has not changed between these two editions is the unacceptably high incidence of foodborne disease. Even more alarming is that we still are only aware of the ‘tip of the iceberg’ with regard to its true incidence. When one considers that it has been estimated that in the United States 3400 deaths are due to unknown foodborne agents (Frenzen 2004), then there is evidently a considerable amount of research and investment still to be undertaken. Confession time: it was my full intention to complete this new edition for publication in 2005. However, our intensive research into Cronobacter spp. (Enterobacter sakazakii) and related organisms has taken up more of my time than the hours in the day can permit. This emergent pathogen, sadly, can infect neonates causing severe illness, and even death. In order not to unbalance this book by excessive reference to this organism of my own personal interest, xix

xx

Preface to second edition

readers should consult Enterobacter sakazakii (ASM Press 2007), edited by myself and Jeff Farber (Health Canada) as well as my homepage www.foodmicrobe.com. As always my thanks and appreciation go to Nigel Balmforth, David McDade and especially Katie Loftus at Blackwells for their patience as the deadlines made a whooshing sound as they went by (frequently). Finally a special thanks to my forever supportive wife Debbie, my children James and Rachel, and my parents – without whom none of this would have been possible. Steve Forsythe Professor of Microbiology Nottingham Trent University July 2009

Preface to first edition

Throughout the world, food production has become more complex. Frequently raw materials are sourced globally and the food is processed through an increasing variety of techniques. No longer does the local farm serve the local community through a local shop, nowadays there are international corporations adhering to national and international regimes. Therefore approaches to safe food production are being assessed on an expanding platform from national and European to trans‐Atlantic and beyond. Against this backdrop there have been numerous highly publicised food safety issues such as bovine spongiform encephalopathy (BSE) and Escherichia coli O157:H7, which have caused the general public to become more cautious and vociferous concerning food issues. The controversy in Europe over genetically modified foods is perceived by the general public within the context of ‘food poisoning’. This book aims to review the production of food and the level of micro‐organisms that humans ingest. Certain circumstances require zero tolerance for pathogens, whereas more frequently there are acceptable limits set, albeit with statistical accuracy or inaccuracy depending upon whether you subsequently suffer from food poisoning. Microbes are traditionally ingested in fermented foods and this has developed into the subject of pre‐ and probiotics with reputed health benefits. Whether engineered ‘functional foods’ will be able to attain consumer acceptance remains to be seen. Food microbiology covers both food pathogens and food spoilage organisms. This book aims to cover the wide range of micro‐organisms occurring in food, both as contaminants and deliberate inoculation. Due to the heightened public awareness over food poisoning it is important that all companies in the food chain maintain high hygienic standards and assure the public of the safety of the produce. Obviously over time there are technological changes in production methods and in methods of microbiological analysis. Therefore the food microbiologist needs to know the effect of processing changes (pH, temperature, etc.) on the microbial load. To this end this book reviews the dominant foodborne micro‐organisms, the means of their detection, microbiological criteria as the numerical means of interpreting end‐product testing, predictive microbiology as a tool to understanding the consequences of processing changes, the role of ‘Hazard Analysis Critical Control Point’ (HACCP) and the objectives of microbial risk assessment (MRA) and the setting of Food Safety Objectives, which have recently become a focus of attention. In recent years the web has become an invaluable source of information and to reflect this a range of useful food safety resource sites are given in the back to encourage the reader to boldly go and surf. Although primarily aimed at undergraduate and postgraduate courses I hope the book will also be of use to those working in industry.

xxi

xxii

Preface to first edition

The majority of this book was written during the last months of 1999, a time when France was being taken to the European Court over its refusal to sell British beef due to BSE/new variant Creutzveldt–Jakob disease and there had been riots in Seattle concerning the World Trade Organization. Whilst large organisations were wondering about the impact of the Millennium bug, in the UK the public were waiting to see the impact of the BSE ‘bug’ (a few hundred or a few thousand cases?). As usual no book can be achieved without assistance and special thanks are due to Phil Vosey concerning MRA; Ming Lo for considerable help with the computer packages; Alison at Oxoid Ltd for the invaluable information on microbiological testing procedures around the world; Pete Silley and Andrew Pridmore at Don Whitley Scientific Ltd for the RABIT diagrams; and Garth Lang at Biotrace Ltd for the ATP bioluminescence data. Not forgetting of course Debbie and Cathy for reading through the draft copy – nevertheless all mistakes are the author’s fault. This book is especially dedicated to Debbie, James and Rachel, Mum and Dad for their patience whilst I have been burning the midnight oil. Dr Steve Forsythe 6 January, 2000

1 Foodborne infections

This may seem a rather self‐defeating opening sentence, but there is no universally accepted definition of ‘safe food’. The reason is that it is a relative term, which is linked to determining the acceptable level of risk to a mixed population or maybe a specific subgroup. Our food is very diverse, and to ensure it is safe requires a systematic, proactive approach of minimising contamination from ‘farm to fork’. However, our food supply involves international movement of ingredients and processed products, and therefore the ‘farm’ is much removed from the ‘fork’. Some procedures of food preservation are well known to the general public, such as refrigeration and canning. There’s also the implementation of ‘hazard analysis and critical control point’ (HACCP), in which the producer anticipates the likely hazards in the final product and ensures the processing reduces or eliminates them to an acceptable level. Foodborne illness can be defined as diseases commonly transmitted through food, and comprise a broad group of illnesses caused by microbial pathogens, parasites, chemical contaminants and biotoxins. An alternative phrase ‘food poisoning’ has often been used but nowadays is regarded as being too restrictive. Unfortunately, illness due to foodborne contamination is still a major cause of morbidity and mortality. Dealing with food safety problems is challenging, in part because they change over time. We have changes in our economy, and therefore lifestyle, eating habits (ranges of food, eating at home or eating out), and an ageing population who are more prone to infection and are slower to recover. The causative agents of foodborne illnesses are also changing, with the emergence of previously unrecognised emergent pathogens. Food producers both industrially and domestically need be aware of these changes in order to improve the safety of our food. This first chapter will consider the magnitude of the foodborne illness, the diversity of sources and diseases, along with its economic consequences. These key topics will be covered in greater depth later in specific chapters. Definition of terms will be found in the glossary at the end of the book, where there is also a listing of useful core hypertext links. Food microbiology is a multidisciplinary topic, and there are rapid advances being made in a number of areas. In order to keep this book as up‐to‐date as reasonably possible the reader should also refer to the supporting websites at www.foodmicrobe.com/info.htm where the reader will find additional information for specific chapters.

The Microbiology of Safe Food, Third Edition. Stephen J. Forsythe. © 2020 John Wiley & Sons Ltd. Published 2020 by John Wiley & Sons Ltd.

1

2

The microbiology of safe food

1.1  The microbial world and its relationship to food

The world of microbiology covers a diverse range of life, albeit very small in size. Simply using the loose definition that microbiologists study life forms which are not clearly visible to the naked eye means that it includes complex organisms such as protozoa and fungi, as well as the simpler bacteria and viruses. Nevertheless, the major micro‐organisms studied to date have been the bacteria. This is because of their medical importance, and because they are easier to cultivate than other micro‐organisms such as viruses, let alone prions, which are uncultivable. Despite this predominance of bacteria in our understanding of microscopic life, there are numerous important organisms in the other microbial categories. The cell structure reveals whether an organism is ‘eucaryotic’ (also spelt ‘eukaryotic’) or ‘procaryotic’ (also spelt ‘prokaryotic’). Eucaryotes have cellular organisation with mitochondria, endoplasmic reticulum and a defined nucleus, whereas procaryotes have no obvious organelle differentiation and are in fact similar in size to the organelles of eucaryotes. Analysis of the genetic information encoding for part of the ribosome (16S rDNA gene) has revealed a plausible relationship and evolution of life from procaryotes to eucaryotes through intracellular symbiotic relationships. Initially, the classification of micro‐organisms has been based on morphology, and phenotypical (biochemical) properties. Morphological analysis considers whether the bacterium is spherical (coccoid), filamentous, curved or rod‐shaped, and if the bacterium has a Gram‐ negative or Gram‐positive type cell wall (Section 2.4.2). The possession of certain enzymes, i.e. α‐galactosidase for lactose fermentation, has helped in identifying and defining a general group of bacteria associated with faecal contamination that would have come from the colon, the origin of the old term ‘coliforms’ (Section 4.1.1). In general, these biochemical and enzymic activities enabled cultivable organisms to be classified. Inevitably defining species according to biochemical (phenotyping) traits will be problematic, as such functions represent only a small portion of the total coding of an organism. DNA sequencing has enabled a fuller understanding of the genetic capabilities of micro‐ organisms (see Chapter 14). Certain genes, termed ‘housekeeping genes’, have undergone limited variation during evolution and can be used as a means of constructing a family tree of relationships between organisms. This is called ‘phylogenetics’. The 16S rDNA encoding gene was initially used for this purpose. It is a housekeeping gene as it encodes for one of the rRNA molecules that is essential for the ribosome to function for protein synthesis. This molecule is called the 16S rRNA gene in procaryotes, and the 18S rRNA gene in eucaryotes due to the respective difference in the size of their ribosomal particles (Section 14.2). A phylogenetic tree using these genes for organisms that cause foodborne disease is given in Figure 1.1. Viruses do not possess this gene and therefore do not appear on this ‘Tree of life’. As DNA sequencing became more accessible, other housekeeping genes of longer lengths, such as gyrB, have been used as their longer lengths enable greater discrimination. The combined sequence of several housekeeping genes is the basis of multi‐locus sequence typing (Section  6.6.1). Such phylogenetic analysis is now undertaken in silico with core genome sequences. Being able to sequence genes in organisms of interest, and compare the order of DNA nucleotides, is just one example of ‘bioinformatics’, which is enabling us to better understand the microbial world (see Chapter 14). Although the ‘species’ concept is conventionally applied to bacteria and fungi, there are difficulties. In higher organisms, a species can be defined as an interbreeding population that is reproductively isolated. The genome of each individual is derived most equally from its two

Foodborne infections Acetobacter pasteurianus Brucella melitensis Arcobacter butzleri Campylobacter jejuni Bacillus cereus St aureus Brochothrix thermosphacta Listeria monocytogenes Lactobacillus plantarum Pediococcus pentosaceus Lactobacillus acidophilus Streptococcus thermophilus Clostridium botulinum Desulphotomaculum nigrificans Bifidobacterium longum Brevibacterium linens Mycobacterium tuberculosis Pyrococcus furiosus

Alcaligenes faecalis Coxiella burnetii Aeromonas hydrophila Shewanella putrefaciens Escherichia coli Shigella dysenteriae Salmonella enterica LT2 Yersinia enterocolitica Vibrio cholerae Alteromonas stellpolaris Acinetobacter calcoaceticus Moraxella Pseudomonas fluorescens

3

Firmicutes; low GC% Gram-positive organisms

Actinobacteria; high GC% Gram-positive organisms Taenia solium Saccharomyces cerevisiae Penicillium chrysogenum Aspergillus niger Rhizopus oryzae Cyclospora cayetanensis

Eucaryotes

Enterobacteriaceae

Figure 1.1  Phylogenetic tree of organisms associated with food microbiology and the archaea organism Pyrococcus furiosus, based upon the DNA sequence ribosomal small subunit, constructed using ClustalW (freeware). Details of sequence accession numbers and tree construction can be obtained from www. foodmicrobe.com/info.

parents and as species diverge, the hybrids become less viable. In contrast, bacteria multiply by binary fission. The acquisition of additional genes by bacteria is generally limited to the transfer of genes between related strains, and less often between different species (horizontal gene transfer or HGT). Whole‐genome sequence analysis can reveal regions of DNA that differ from the majority of the genome due to their G:C ratio or codon bias. These regions can be from bacteriophages, and less closely related bacteria. In multi‐locus sequence analysis (Section 6.6.1), housekeeping genes are sequenced and form a core of genes that help to define the species. Therefore, although there are differences between the ‘species’ concept in bacteria and higher organisms, in both cases individual ‘species’ can be recognised by reference to a gene pool. Inevitably as our knowledge of organisms at the genome level improves, so previous species may be redefined. The relatedness of organisms used to be determined using laboratory‐based DNA–DNA hybridisation analysis. The DNA was extracted from two organisms, heated to separate the two DNA strands and then they were allowed to anneal to each other and the extent of hybridisation measured. The common standard was that 95% ± 0.5% has been commonly used to indicate that two strains are the same species.

4

The microbiology of safe food

A very brief survey of food poisoning micro‐organisms can start with eucaryotic ­ rganisms such as helminths (See Table 1.1 and Figure 1.1; cf. Section 5.3). These include the o cestode worms that are responsible for taeniasis; Taenia solium the pork tapeworm, and T. saginata the beef tapeworm. Both have worldwide distribution. Infection results from the ingestion of undercooked or raw meats containing the cysts. The mature worms can infect the eye, heart, liver, lungs and brain. A third tapeworm is Diphyllobothrium latum, which is found in a variety of freshwater fish including trout, perch and pike. Contaminated water supplies can carry infectious organisms including pathogenic protozoa such as Cyclospora and Cryptosporidium.

Table 1.1  Diversity of hazards associated with food. Biological

Chemical

Physical

Macrobiological Microbiological

Veterinary residues: antibiotics, growth stimulants Plasticisers and packaging migration: vinyl chloride, bisphenol A

Glass Metal

Viruses Hepatitis A Norovirus Rotavirus

Chemical residues: pesticides (DDT), cleaning fluids

Stones

Pathogenic bacteria Spore‐forming Bacillus cereus Clostridium perfringens Clostridium botulinum Non‐spore‐forming Campylobacter jejuni Pathogenic strains of Escherichia coli Listeria monocytogenes Salmonella serovars Bacterial toxins Staphylococcus aureus B. cereus Shellfish toxins: domoic acid, okadaic acid NSP, PSP Parasites and protozoa Cryptosporidium parvum Entamoeba histolytica Giardia lamblia Toxoplasma gondii Fasciola hepatica Taenia solium Anisakis species Trichinella spiralis Mycotoxins: ochratoxin, aflatoxins, fumonisins, patulin

Allergens Toxic metals: lead, cadmium, arsenic, tin, mercury

Wood Plastic

Food chemicals: preservatives, processing aids

Parts of pests

Radiochemicals: 131I, 127Cs

Insulation material

Dioxins, polychlorinated biphenyls (PCBs)

Bone

Prohibited substances Printing inks

Fruit pits

NSP = neurotoxic shellfish poison, PSP = paralytic shellfish poisoning. Source: Adapted from Snyder (1995) and Forsythe (2000).

Foodborne infections

5

Fungi (a collective term that includes yeast and moulds) are also eucaryotic. Well‐known examples of moulds are Penicillium chrysogenum and Aspergillus niger. As shown in Figure 1.1, they form a unique branch of life and, despite their superficial appearance, they are not plants. They do not differentiate to give the root, stem and leaf systems associated with plants. They also do not produce the green photosynthetic pigment chlorophyll. Their morphology can be a branching mycelium with differentiating cells producing hyphae to disperse spores or they can exist as single‐cell forms, commonly referred to as yeast. Yeasts such as Saccharomyces cerevisiae are very important in food microbiology. The brewing and bakery industries are dependent upon yeast metabolism of sugars to generate ethanol (and other alcohols) for beer and wine production, and carbon dioxide for bread manufacture. Mycotoxicoses are caused by the ingestion of poisonous metabolites (mycotoxins), which are produced by fungi growing in food (cf. Section 5.4). Aflatoxins are produced by the fungi Aspergillus flavus and A. parasiticus. There are four main aflatoxins designated B1, B2, G1 and G2 according to whether they show blue (B) or green (G) fluorescence when viewed under a UV lamp. Ochratoxins are produced by A. ochraceus and Penicillium viridicatum. Ochratoxin A is the most potent of these toxins. The bacteria are procaryotes and are divided into the ‘eubacteria’ (true bacteria) and archaea organisms according to 16S rDNA analysis, detailed cell composition and metabolism studies. Curiously, there are no archaea organisms of clinical relevance or of importance in the food industry. Pyrococcus furiosus (found in hydrothermal vents, growing at 100°C) has been included in Figure  1.1 only to illustrate the separation of the archaea from eubacteria and eucaryotes. As a gross generalisation the size of a rod‐shaped bacterium is about 2 μm × 1 μm × 1 μm. Nevertheless, despite their size, even 500 cells of Listeria monocytogenes can cause an infection in a pregnant woman resulting in a stillbirth. Familiar foodborne pathogens such as Salmonella serovars, Escherichia coli and Campylobacter jejuni are eubacteria that are able to grow at body temperature (37°C) and in situ produce various toxins that damage human cells resulting in ‘food poisoning’ symptoms such as diarrhoea and vomiting. See Chapter 4 for more information on (eu)bacteria causing food poisoning. Viruses, such as noroviruses, are very much smaller than bacteria. The large ones, such as the cowpox virus are about 0.3 μm; the smaller ones, such as the foot and mouth disease virus, are about 0.1 μm. See Chapter 5 for more information on foodborne viral pathogens. Because of their small size, viruses pass through bacteriological filters and are invisible under a light microscope. Bacterial viruses are termed bacteriophages (or ‘phages’). They can be used to ‘fingerprint’ bacterial isolates, which is useful in epidemiological studies; this is termed ‘phage typing’. Bacteriophages are the most abundant micro‐organisms on Earth, and have had a major role in the evolution of bacteria due to their ability to integrate with host chromosomes, transferring remnant DNA from their previous host. Bacteria have evolved immune systems against them, called clustered regularly interspaced short palindromic repeat (CRISPR)‐cas arrays (Section 6.6.3). Prions (short for ‘proteinaceous infectious particles’) have a very long incubation period (months, or even years) and resistance to high temperature, formaldehyde and UV irradiation (see Section 5.4.6 for more details). With regard to sheep and cattle, the isomer of the normal cellular protein PrPC, termed PrPSC, accumulates in the brain causing holes or plaques. This leads to the symptoms of scrapie in sheep and bovine spongiform encephalopathy (BSE) in cattle. The equivalent disease in humans, variant Creutzfeldt–Jakob disease (vCJD), is probably due to ingestion of the infectious agent from cattle.

6

The microbiology of safe food

1.2  Origins of safe food production

The necessity of eating safe food must go back to early humans. It presumably developed with the hunter–gatherer lifestyle, where the domestication of animals and cultivation of crops required cooking and storage activities. For example, barley production flourished in the Egyptian Nile Valley about 18 000 years ago. This necessitated the preservation of the grain by keeping it dry to prevent fungal spoilage. Hence reducing the spoilage of more perishable foods by drying could easily have co‐developed at this time as well. Preservation by the addition of honey and olive oil was another early form of food preparation. Once salt had been found to have a preservative capability, it became a major commodity. In fact, the word ‘salary’ originally meant ‘soldier’s allowance for the purchase of salt’. Over time humans learnt to select edible animals and plants. We also learnt how to cultivate and farm, to harvest and organise our food resources according to the seasons, and habitat. Undoubtedly there was considerable trial and error, but gradually the beneficial habits were learnt and passed down through the generations. Numerous religious and cultural practices related to food have sound scientific basis. This may include not eating pork, which we now know carries the tapeworm Trichinella spiralis. The use of running water to bathe is more hygienic than using standing water. Various common foods require microbial activity. Beer, wine and bread require the yeast S. cerevisiae to ferment sugars to produce ethanol and carbon dioxide. Coffee and chocolate production involve the fermentation of naturally occurring bacteria and yeast to weaken the outer covering as well as produce desirable odours (Ozturk and Young 2017). There is a considerable industry centred on the metabolism of lactic acid bacteria and milk (Section 3.7). The beginning of a more scientific approach to food preparation occurred with the development of food preservation by heat treatment. In 1795, the French government realised the strategic usefulness of preserved food for its troops and offered a large reward for anyone who could develop a new method of preserving food. The prize was won by Nicolas Appert, a Parisian confectioner by trade. His preservation method was putting the food into wide‐mouth glass bottles, which were then sealed with a cork and put in boiling water for 6 h. The use of tins cans instead of glass bottles was the idea of Durand in 1810 and is the basis of the canning industry today. The thermal processing worked, but the rationale behind the procedures was not known until the work of Louis Pasteur and Robert Koch. Although previous workers such as Antonie van Leeuwenhoek in 1677 had discovered ‘little heat‐sensitive animalcules’, it was Louis Pasteur who started the science of microbiology. Due to his studies, between 1854 and 1864, he demonstrated that bacteria were the causative agents of food spoilage and disease. As a consequence the French wine industry adopted the process of heating the wine to kill the spoilage organisms, before inoculating with favourable micro‐organisms for the fermentation process. Due to its success, the ‘pasteurisation’ process was then applied to other foods such as milk, although this later application is principally for the control of pathogenic micro‐organisms. Another founding figure in microbiology was the German Robert Koch, who first developed a method of growing pure cultures of micro‐organisms. In 1884 he was the first to isolate the bacterium Vibrio cholerae (Section  4.7). From then on, the isolation and study of pure cultures has been a major activity for microbiologists. From those early days, medical, veterinary, environmental and food microbiology have each become established disciplines in their own right. Food microbiology itself encompasses a number of topics such as the detection of unwanted micro‐organisms and their products, as

Foodborne infections

7

well as the desirable use of microbial activity in the production of fermented foods such as beer, wine, cheese and bread. Or, as more simply put in the cliché, ‘the good, the bad and the ugly’, referring to food production, food poisoning and food spoilage, respectively. Advances continue to be made and the development of HACCP (Section 9.5) was driven by the need for safe food for the US manned space programme in the 1960s, and was supported by the US army as well, a modern‐day echo of Nicolas Appert’s contribution to safe food for Napoleon’s troops. 1.3  Overview of foodborne illness

Problems with food quality and safety have existed for many centuries; for example, the adulteration of milk, beer, wine, tea leaves and olive oil. Contaminated food causes one of the major health problems in the world, and leads to reduced economic productivity. Table 1.1 lists the pathogens that may be transmitted through contaminated food. Although foodborne illness is often presumed to mean illness due to eating food containing bacterial pathogens, as Table 1.2 shows, a wide range of organisms, toxins and chemicals can cause foodborne illness. Some compounds and organisms are external food contaminants, whereas others are intrinsic; for example, oxalic acid in rhubarb and the alkaloid solanine in potatoes. Microbial food poisoning is caused by a variety of micro‐organisms with various incubation periods and duration of symptoms (Tables  1.3 and 1.4). Organisms such as Salmonella and E. coli O157:H7 are well known by the general public, but there are also viruses and fungal toxins that have been relatively poorly studied and in the future we may more fully recognise their contribution to the general incidence of food poisoning. Micro‐organisms causing food poisoning are found in a diverse range of foods including milk, meat and eggs. They have a wide range of virulence factors that can cause a range of adverse responses, which may be acute, chronic or intermittent. Some bacterial pathogens, such as Salmonella spp., are invasive and can get through the intestinal wall into the bloodstream and cause generalised systemic infections. Other pathogens produce toxins in the food prior to ingestion, or during infection, and cause severe damage in susceptible organs, such as the kidney as in the case of E. coli O157:H7 infection. Complications can occur due to immune‐mediated reactions (i.e. Campylobacter infection leading to reactive arthritis and Guillain–Barré syndrome) where the host immune response to the pathogen is unfortunately also directed against the host’s tissues. Therefore, foodborne illness can be much more severe than a short period of gastroenteritis, and instead may result in hospitalisation. The severity can be such there may be residual (chronic) symptoms and even the risk of death, especially in the elderly and severely immunocompromised. Hence there is a considerable public health burden attributable to foodborne infections, which is being increasing recognised as causing considerable economic loss. Estimates of costs of foodborne illness have been made and will be considered in detail later (Section 1.11). Because the consumer is unaware that there is a potential problem with the food, a significant amount of contaminated food is ingested and hence they become ill. Consequently it is hard to trace which food was the original cause of food poisoning since the consumer will not recall noticing anything relevant in their recent meals. They are likely to recall food that smelt ‘off ’ or looked ‘discoloured’. However these attributes are related to food spoilage and not food poisoning. Food poisoning micro‐organisms are normally divided into two groups: • Infections; for example, Salmonella serovars, C. jejuni and pathogenic E. coli. • Intoxications; for example, Bacillus cereus, Staphylococcus aureus and Clostridium botulinum.

Table 1.2  Sources of foodborne pathogens. Food

Pathogen

Meat, poultry and eggs

Arcobacter species Campylobacter jejuni Salmonella serovars Staphylococcus aureusa

Fruit and vegetables

Clostridium perfringensb Clostridium botulinum Escherichia coli O157:H7 Bacillus cereusb Listeria monocytogenes Yersinia enterocolitica Hepatitis A virus Trichinella spiralis Tapeworms C. jejuni Salmonella serovars St. aureusa L. monocytogenes

Milk and dairy products

Powdered infant formula Shellfish and fin fish

Shigella spp. E. coli O157:H7 Y. enterocolitica Aeromonas hydrophila Hepatitis A virus Norovirus Giardia lamblia Cryptosporidium species Cl. botulinum B. cereusb Mycotoxins Salmonella serovars Y. enterocolitica L. monocytogenes E. coli C. jejuni Shigella species Hepatitis A virus Norovirus St. aureusa Cl. perfringensb B. cereusb Mycotoxins Salmonella serovars, Cronobacter species Salmonella serovars Vibrio species Shigella species Y. enterocolitica B. cereus E. coli

Incidence (%)

Raw chicken and turkey (45–64) Raw poultry (40–100), pork (3–20), eggs (0.1%) and shellfish (16) Raw chicken (73), pork (13–33) and beef (16) Raw pork and chicken (39–45) Raw beef, pork and poultry Raw ground beef (43–63), cooked meat (22) Red meat (75), ground beef (95) Raw pork (48–49)

Mushrooms (2) Artichoke (12), cabbage (17), fennel (72), spinach (5) Lettuce (14), parsley (8), radish (37) Potatoes (27), radishes (37), bean sprouts (85), cabbage (2), cucumber (80) Celery (18) and coriander (20) Vegetables (46) Broccoli (31)

Milk (48–49) Soft cheese and pâté (4–5)

Pasteurised milk (2–35), milk powder (15–75) cream (5–11), ice cream (20–35)

Raw seafood (33–46)

Fish products (4–9) (Continued )

Foodborne infections

9

Table 1.2  (Continued) Food

Pathogen

Incidence (%)

Cl. botulinum Hepatitis A virus Norovirus Giardia lamblia Cryptosporidium spp. Metabolic byproducts Algal toxins Salmonella serovars L. monocytogenes Shigella species E. coli St. aureusa Cl. botulinumb B. cereusb b

Cereals, grains, legumes and nuts

Spices

Water

Mycotoxinsa Salmonella serovars St. aureusa Cl. perfringensb Cl. botulinumb B. cereusb G. lamblia Vibrio cholerae Cronobacter species

Raw barley (62–100), boiled rice (10–93), fried rice (12–86)

Herbs and spices (10–75) Water (30)

 Toxin not destroyed by pasteurisation.  Spore‐forming organism. Not killed by pasteurisation. Source: Various including Synder (Hospitality Institute of Technology and Management, web address in the Internet directory) and ICMSF (1998a). a

b

The first group are bacteria that multiply in the human intestinal tract, whereas the second group are bacteria that produce toxins either in the food or during passage in the intestinal tract. As a generalisation, bacterial infections cause gastroenteritis, whereas the ingestion of a toxin causes vomiting. This division is also very useful to help recognise the routes of food poisoning, since bacterial infection will be due to something ingested in the previous 18–24 h, whereas vomiting is likely to be due to preformed toxin ingested in the past few hours. Gastroenteritis with a fever could be due to a Gram‐negative infectious organism, as the host’s immune system responds to the bacterial outer surface, which will be composed of lipopolysaccharide. This compound can be very pyrogenic in some Gram‐negative bacteria, therefore causing the symptom of a fever. Viral infections cause both vomiting and gastroenteritis. Vegetative organisms are killed by heat treatment, whereas spores (produced by B. cereus and Clostridium perfringens) may survive and hence germinate if the food is not kept sufficiently hot or cold. An alternative grouping is according to severity of illness. This approach is useful in setting microbiological criteria (sampling plans) and risk analysis. The International Commission on  Microbiological Specifications for Foods (ICMSF 1974 and 2002) divided the common

10

The microbiology of safe food

Table 1.3  Incubation period and duration of common food poisoning micro‐organisms. Micro‐organism

Incubation period

Duration of illness

Aeromonas species Bacillus cereus emetic diarrhoeal Campylobacter jejuni Clostridium botulinum Clostridium perfringens Escherichia coli ETEC EPEC EIEC EHEC Hepatitis A Listeria monocytogenes Norovirus Rotavirus Salmonella serovars Shigellae Staphylococcus aureus Vibrio cholerae Vibrio parahaemolyticus Yersinia enterocolitica

Unknown 0.5–6 hours 8–24 hours 3–5 days 12–36 hours 8–12 hours

1–7 days 1 day 1 day 2–10 days 2 h–14 days 1–14 days

16–72 hours 16–48 hours 16–48 hours 72–120 hours 3–60 days 3–70 days 24–48 hours 24–72 hours 16–72 hours 16–72 hours Few hours 6 hours – 3 days 4–96 hours 3–7 days

3–5 days 2–7 days 2–7 days 2–12 days 2–4 weeks Variable 1–3 days 4–6 days 2–7 days 2–7 days 2–3 days 1–6 days 3 days 1–3 weeks

foodborne pathogens into such groups in order to aid in the decision making of sampling plans. These ICMSF groupings are given later in Chapter 7. Detailed descriptions of certain ­foodborne pathogens are given in Chapters 4 and 5, and extensive details can be found in the numerous ICMSF publications listed in the References section. Despite an increasing awareness and understanding of food and waterborne micro‐ organisms, these diseases remain a significant problem and are an important cause of reduced economic productivity. While everyone is susceptible to foodborne diseases, there is a growing number of people within the general population who are more likely to experience such diseases, often with more severe consequences. These people include infants and young children, pregnant women, those who are immunocompromised through medication or illness and the elderly. There is evidence that the microbial causes of gastroenteritis vary with age and that viral agents are probably the major infectious agent in children under 4 years (Figure 1.2). There is also a difference between the sexes (Figure 1.3), which is possibly due to differences in personal hygiene, i.e. males have a lower tendency to wash their hands after going to the toilet. The production of food has increased by 145% since the 1960s. Of particular importance is the increase in developing countries such as Africa (140%), Latin America (200%) and Asia (280%). Food production has doubled in the USA, and by 68% in Western Europe. Yet hunger is still a worldwide problem. In this century there are over 800 million people suffering from malnutrition. Over the same time period, the world’s population has increased from 3 to 6 billion, and is expected to attain 9 billion by 2050. Subsequently there will be ever‐increasing demands in food production and food security. In developing countries, foodborne diseases are a major (up to 70%) cause of diarrhoea in children under 5 years of age. They may suffer two or three episodes of diarrhoea per year, and for some this may even be as many as 10. Pathogenic bacteria can contaminate weaning foods

Foodborne infections

11

Table 1.4  Usual incubation and onset period for foodborne illness.

Incubation period Agent

Symptoms

Chemical allergen Heavy metal: copper, tin, lead, zinc

Nausea, vomiting

Fish toxins: PSP, ciguatera and soon

Gastrointestinal and neurological symptoms

Monosodium glutamate

Burning sensation on body, tingling, dizziness, headache, nausea

Food allergens: nuts, eggs, milk, wheat

Anaphylactic shocks. Respiratory failure, rashes, nausea, vomiting

Bacillus cereus, emetic

Nausea, vomiting, abdominal pain Nausea, vomiting, abdominal pain Abdominal pain, watery diarrhoea Abdominal pain, watery diarrhoea Abdominal pain, diarrhoea, chills, fever, nausea, vomiting, loss of appetite Vertigo, blurred vision, difficulty in speaking, progressive nervous system failure and paralysis Sore throat, fever, nausea, vomiting, rhinorrhoea, tonsillitis, may be rash Profuse watery diarrhoea, feer, chills, headache

S. aureus Bacillus cereus, diarrhoeal Clostridium perfringens Salmonella serovars

Clostridium botulinum

Streptococcus Group A

Vibrio parahaemolyticus

106/g) Vibrio spp. Yersinia spp. Protozoa Cryptosporidium parvum Giardia intestinalis Viruses Adenovirus group F Astrovirus Calicivirus Rotavirus group A Rotavirus group C Norovirus No organism identified Total

Rate/1000 person years

General practice Number of cases

Rate/1000 person years

46 0 32 6

12.4 0 8.7 1.6

165 4 354 17

1.88 0.05 4.14 0.20

9

2.4

114

1.30

0

0

3

0.03

Number of community cases/GP cases

6.7 – 2.1 8.0 1.9 –

20

5.4

119

1.32

4.1

23 18 0 1 10 3

6.2 4.9 0 0.27 2.7 0.82

103 141 0 4 52 6

1.18 1.62 0 0.05 0.59 0.06

5.3 3.0

8 1 1

2.2 0.27 0.27

146 23 10

1.57 0.27 0.11

1.4 1.0 2.5

0 25

0 6.8

1 51

0.01 0.58

– 11.7

3 2

0.81 0.54

39 28

0.43 0.28

1.9 1.9

11 14 8 26 2 46 432 781

3.0 3.8 2.2 7.1 0.54 12.5 117.3 194

0.88 0.86 0.43 2.30 0.06 1.99 14.82 33.1

3.4 4.4 5.1 3.1 8.9 6.3 7.9 5.8

81 77 40 208 6 169 1305 8770b

– 5.4 4.6 13.4

 Excluding cases where individual follow‐up was not known.  Total cases are greater than the sum of individual organisms due to cases for which a stool sample was not sent for testing. The general practice total includes cases from the enumeration arm, for which full stool testing was not carried out. Source: Reproduced from Wheeler et al. (1999), with permission from BMJ Publishing Group Ltd. a

b

Foodborne infections

17

to conventional produce, which is not necessarily true. Cow manure can contain Salmonella and E. coli O157:H7, and there have been outbreaks linked with organically grown produce. Industry and national regulators strive for production and processing systems which ensure that all food is ‘safe and wholesome’, although complete freedom from risks is an unattainable goal. Safety and wholesomeness are related to a level of risk that society regards as reasonable in the context, and in comparison with other risks in everyday life. Putting food poisoning into context is not an easy task due to the high level of publicity that it receives in some countries. One can see in Table 1.5 that the risk of death due to food poisoning is nearly equivalent to the risk of being a pedestrian who is struck by a car. However, such tragic incidences affect individuals and generally only reach the local newspaper headlines, whereas food poisoning outbreaks involving a large number of people are more ‘significant’ to the media. A reported increase in the number of ‘food poisoning cases’ is commonly cited by the media. However, these trends must be regarded with caution since they are the number of gastroenteritis cases that have been investigated and the causative organism identified. Not all gastroenteritis cases may be due to food vectors, and an increase in public awareness can increase the number of members of the general public seeking medical help. Sometimes referred to as the ‘worried well’, this can put a strain on the medical profession and distract from investigating the real cases. In addition improved detection methods could ‘increase’ the apparent number of identified cases over time. In fact, across Europe and the United States the number of reported gastroenteritis cases has been decreasing. In 2005, the Centers for Disease Control and Prevention (CDC) reported that the most common diseases associated with foodborne pathogens, parasites and other bacteria were on the decline, or at least not increasing in incidence. Escherichia coli O157:H7 infections decreased by 36% in just 1 year. Similarly, the number of Campylobacter, Salmonella, Yersinia and Cryptosporidium infections declined by 28, 17, 51 and 49%, respectively, in the 8‐year surveillance period. However, this trend did not significantly change in the following 3 years (FoodNet 2009), but did decline thereafter (Imanishi et al. 2014). In 2009, the number of reported foodborne disease outbreaks declined 32% compared with the mean of 2004–2008, and was below the pre‐2009 average during 2010–2012. The decline was mainly in norovirus‐related outbreaks. Figure 1.4 shows the 20‐year trends in Campylobacter and Salmonella cases in England and Wales; note the decline in Salmonella cases since 1997. As given in Section 1.3, sentinel studies indicate a more informed incidence of gastroenteritis may be ~20%, though its mild nature means the illness is unnoticed and hence unreported. Possibly due to changes in eating habits there is a marked seasonality in food poisoning incidences with pathogens such as Salmonella and Campylobacter. Figure 1.5 show the peak incidence over the summer months. Typically, Campylobacter infections peak a month or two before Salmonella serovars do. 1.5  Causes of foodborne illness

There are a number of factors that contribute to food being unsafe and causing illness (Table 1.9). The principal causes can be summarised as: • inadequate control of temperature during cooking, cooling and storage; • inadequate personal hygiene; • cross‐contamination of raw and processed products; • inadequate monitoring of processes.

18

The microbiology of safe food 60 000

Number of reports

50 000

40 000

30 000

20 000

10 000

0 1980

1985

Campylobacter

1990

1995

S. Typhimurium

2000

S. Enteritidis

2005

Other Salmonella

Figure 1.4  Twenty‐year trend in Campylobacter and Salmonella reported cases for England and Wales. Source: Data source hpa website.

5000 4500

Number of reports

4000 3500 3000 2500 2000 1500 1000 500 0

2

4 2006

6

8

10

12

2

4

6

Four-week period Campylobacter

8

10

12

2007

Salmonella

Figure 1.5  Example of seasonal trends in reported cases of Salmonella and Campylobacter enteric infections in England and Wales. Source: Data source hpa site.

Foodborne infections

19

These contributing factors can be considerably reduced by adequate training of staff, implementation of HACCP combined with risk assessment (Chapters 9 and 10). As Chapter 9 will explain, it is now widely accepted that it is inadequate to rely on testing the final end product for the presence of micro‐organisms and as a means of controlling the hygienic status of the process. The World Health Organisation (WHO) summarises safe food in the ‘Five keys to safer food’: 1 keep clean; 2 separate raw and cooked; 3 cook thoroughly; 4 keep food at safe temperatures; 5 use safe water and raw materials. The key to the production of safe food is producing food that is microbiologically stable. In other words, any intrinsic microbes are unable to multiply to an infectious dose, ideally they die off and their toxins are absent. In the case of foodborne viruses, they do not multiply in foods and therefore it is essential to avoid initial contamination from sources such as water and infected food handlers. Essentially the cooking and cooling temperature profile should be aimed at: 1 Reducing the number of infectious organisms by 6 log orders (i.e. 106 cells/g to 1 cells/g). 2 Not providing suitable conditions for the outgrowth of microbial spores that survive cooking. 3 Avoiding conditions that enable bacteria such as St. aureus to produce heat‐stable toxins. By definition these toxins are resistant to 100°C for 30 minutes and hence are not destroyed by cooking before eating the food. Cross‐contamination causes post‐processing (i.e. after the cooking step) contamination of the food. This can be avoided through: 1 Careful design of the factory layout. 2 Control of personnel movement. 3 Good personal hygiene habits. Foods that do not undergo a cooking process are normally acidified (i.e. fermented foods) and stored under chilled conditions. These practices rely upon the pH and temperature of the food stopping microbial growth. The growth range of the major food poisoning organisms has been documented (see ICMSF 1996a for details). Hence the pH and storage temperature of the food that will restrict the growth of foodborne pathogens can be predicted. Although contamination of food can occur at any point from farm to table, restaurant food workers are a common source of foodborne illness. Angelo et al. (2017) analysed the characteristics of restaurant‐associated foodborne disease outbreaks and, in particular, the role of food workers by analysing outbreaks associated with restaurants from 1998 to 2013, as reported to the CDC. In total there were 9788 restaurant‐associated outbreaks, and most outbreaks (79%) were reported at sit‐down establishments. The median annual number of outbreaks was 620. Food workers contributed to 2415 (25%) outbreaks. The most commonly reported contributing factors were those related to food handling and preparation practices in the restaurant (2955 outbreaks, 61%). Norovirus caused the largest number of outbreaks (1425/3072, 46%) for which a single confirmed aetiology was reported. Fish (254 outbreaks, 34%) was most commonly implicated when the outbreak was attributed to a single food and a confirmed aetiology. These outbreaks were commonly caused by scombroid toxin (219 outbreaks, 86% of fish outbreaks).

20

The microbiology of safe food

1.6  Food poisoning due to common food commodities 1.6.1  Milk and milk products

Milk and milk products used to be a major cause of morbidity and mortality before the introduction of the pasteurisation process in the 1930s. Those who drank raw milk before then were at risk of typhoid fever, scarlet fever and tuberculosis. Nevertheless milk and milk products can still be the cause of foodborne disease, albeit a small proportion (1–2%) of the total number of cases. Zoonoses, which have been transmitted to humans through unpasteurised milk, include ­brucellosis, infections by pathogenic strains of E. coli (including E. coli O157:H7), leptospirosis, listeriosis, paratyphoid fever, salmonellosis, staphylococcal enterotoxic gastroenteritis, streptococcal infections, tuberculosis, foot and mouth disease, tickborne encephalitis, Q‐fever and toxoplasmosis. The sources of pathogenic bacteria can be the cows, environmental sources and those involved with the milk handling operations from the farm to food service facilities. The cows can transmit pathogens through their udders, skin and through faecal contamination. Many of these organisms are killed by pasteurisation, but post‐pasteurisation contamination has been reported for Salmonella serovars, L. monocytogenes, Cl. botulinum, Cl. perfringens and St. aureus, as well as viruses (adenoviruses, enteroviruses, hepatitis A) and protozoa (amoebiasis, giardiasis). Any subsequent temperature abuse enables the bacteria and protozoa to multiply. There are various heat treatments used to reduce the microbial load (especially pathogens) to a level at which they should not be a significant health hazard. The commonly used ones are thermisation, pasteurisation, ultra‐high temperature treatment (UHT) and sterilisation. Pasteurisation is intended to inactivate pathogenic vegetative cells. The destruction of spoilage organisms is often another benefit of pasteurisation, but is not the focus of milk safety programmes. 1.6.2  Meat products

Meat (beef, pork, poultry, lamb) consumption worldwide is in the order of 41 kg per person. Poultry consumption in particular is increasing and is ~14 kg per person per year. The largest consumer groups of poultry are in the developed Western countries, with the US being the largest consumers at ~50 kg per person per year. The importance of microbial safety of meat products is therefore considerable. Despite advances in animal husbandry, food processing and awareness of good hygienic practices, meat products are still a major source of foodborne infections. Foodborne bacterial pathogens on meat products may be able to grow if stored at temperatures >8°C. The pathogens may originate from the intestinal tract of the animal during slaughter, from the processing environment, as well as from human contact prior to ingestion. The most prolific causes of food poisoning, Salmonella, pathogenic E. coli (i.e. E. coli O157:H7) and Campylobacter, are most frequently through the ingestion of ­contaminated meat products. 1.6.3  Fresh produce

Eating fresh fruit and vegetables is encouraged as part of a healthy lifestyle. Unfortunately since a large proportion of them are consumed raw, there have been a significant number of foodborne outbreaks (Lynch et al. 2009; Olaimat and Holley 2012). In the US between 1996 and 2016, there were 46 reported food poisoning outbreaks linked to sprouts, with a total of 2474 illnesses, 187 hospitalisations and three deaths. The large 2011 E. coli O104:H4 outbreak

Foodborne infections

21

from sprouted fenugreek seeds in Germany and France illustrates the relevance of the consumption of these products with respect to food safety issues (Section 14.4.2). The globalisation of the food trade can further increase the risk, if produce comes from countries with lower safety standards. A study of reported foodborne outbreaks over 8 years (2004–2012) associated with fresh fruit and vegetable consumption in the United States and European Union (EU) showed norovirus and Salmonella serovars were responsible for ~75% of outbreaks; 294/377 in the USA and 148/197 in the EU. Norovirus was responsible for 59 and 55% of outbreaks in the USA and the EU, respectively, whereas the figures for Salmonella serovars were 19 and 20%, respectively. Norovirus was mainly linked with salad‐related outbreaks in the USA and with berries in the EU, while Salmonella serovars were the causative agent for the majority of sprout‐associated outbreaks. 1.6.4  Low‐water activity (aw) and low‐moisture foods

Low‐water activity (aw) foods and low‐moisture foods are not exactly the same, the latter being generally described as ‘dry food’. Some foods are not low in moisture (i.e. water content) but have a low aw due to water‐binding solutes (such as sodium chloride) that bind the available water, subsequently lowering the aw (see Section 2.8.2). Common examples of water‐binding solutes in food products are sodium chloride (salt) and sucrose (sugar), which has a water‐ binding ability 6× greater than sodium chloride. Low‐water activity and low‐moisture foods have aw values of 1700 children infected with norovirus through school meal Contamination of chocolate in the UK with S. Montevideo. £20 million loss by company

2008 2008

USA Canada

>1000 S. St. Paul infections from jalapeño peppers and tomatoes 28 listeriosis cases, with 1 death, of pregnant mothers and premature births linked to various cheeses 12 deaths, out of >38 cases linked to meat products

2008 2017–2018

China South Africa

Melamine in milk products; >50 000 illness, >4 deaths Listeriosis >1000 cases, 247 deaths, linked with polony sausages

Foodborne infections

29

Table 1.9  Factors contributing to outbreaks of foodborne disease (various sources). Contributing factors Factors relating to microbial growth Storage at ambient (room) temperature Improper cooling Preparation too far in advance of serving Improper warm holding Use of leftovers Improper thawing and subsequent storage Extra large quantities prepared Factors relating to microbial survival Improper reheating Inadequate cooking Factors relating to contamination Food workers Contaminated processed non‐canned foods Contaminated raw foods Cross‐contamination Inadequate cleaning of equipment Unsafe source Contaminated canned foods

Percentagea

43 32 41 12 5 4 22 17 13 12 19 7 11 7 5 2

 Percentages exceed a total of 100 since multiple factors often contribute to foodborne illness.

a

Table 1.10  Chronic sequelae following foodborne infection (based on Lindsay 1997 and Mossel 1988). Disease

Associated complication

Brucellosis Campylobacteriosis

Aortitis, orchitis, meningitis, pericarditis, spondylitis Arthritis, carditis, cholecystitis, colitis, endocarditis, erythema, nodosum, Guillain–Barré syndrome, haemolytic uraemic syndrome, meningitis, pancreatitis, septicaemia, reactive arthritis, irritable bowel syndrome

Cronobacter species E. coli (EPEC and EHEC types) infections

Meningitis Erythema nodosum, haemolytic uraemic syndrome, seronegative arthopathy

Listeriosis Salmonellosis

Meningitis, endocarditis, osteomyelitis, abortion and stillbirth Aortitis, cholecystitis, colitis, endocarditis, orchitis, meningitis, myocarditis, osteomyelitis, pancreatitis, Reiter’s syndrome, rheumatoid syndromes, septicaemia, splenic abscess, thyroiditis, irritable bowel syndrome

Shigellosis

Erythema nodosum, haemolytic uraemic syndrome, peripheral neuropathy, pneumonia, Reiter’s syndrome, septicaemia, splenic abscess, synovitis

Taeniasis Toxoplasmosis Yersiniosis

Arthritis, epilepsy Foetus malformation, congenital blindness Arthritis, cholangitis, erythema nodosum, liver and splenic abscesses, lymphadenitis, pneumonia, pyomositis, Reiter’s syndrome, septicaemia, spondylitis, Still’s disease

Table 1.11  Under‐reporting of foodborne pathogens in the USA.

Organism Bacterial pathogens Campylobacter species Salmonella, non‐typhoid serovars Clostridium perfringens Staphylococcus aureus Escherichia coli O157:H7 STEC other than O157 Shigella species Yersinia enterocolitica E. coli enterotoxigenic (ETEC) Streptococcus Group A Bacillus cereus E. coli, other diarrhoegenic Vibrio species other than those above Listeria monocytogenes Brucella species S. Typhi Vibrio cholerae O1 or O139 Vibrio vulnificus Clostridium botulinum Parasitic pathogens Giardia lamblia Toxoplasma gondii Cryptosporidium parvum Cyclospora cayetanensis Trichinella spiralis Viral pathogens Norovirus Rotavirus Astrovirus Hepatitis A

Under‐reporting factor

Estimated infections (1997)

38 38 38 38 20 Half as common as E. coli O157:H7 cases 20 38 10 38 38 Assumed to be as common as ETEC 20

1 963 000 1 342 000 249 000 185 000 92 000 Combined 90 000 87 000 56 000 51 000 27 000 23 000 5000

2 14 2 2 2 2

2000 777 659 49 47 56

20 7 45 38 2

200 000 112 000 30 000 14 000 52

11% of all acute primary gastroenteritis Not given (number of cases taken as equal to birth cohort) Not given (number of cases taken as equal to birth cohort) 3

9 200 000 39 000 39 000 4000

Source: Mead et al. (1999), Tauxe (2002). Table 1.12  Medical costs and productivity losses estimated for selected human pathogens.

Pathogen Bacillus cereus Campylobacter jejuni Clostridium botulinum Clostridium perfringens Cronobacter sakazakii Escherichia coli pathovars Listeria monocytogenes Salmonella serovars Staphylococcus aureus Vibrio cholerae Norovirus Toxoplasma gondii

Total cost (US$b)a

QALYsa

1.9

13 300

2.8 3.7

94 000 17 000

2.3 3.3

5000 11 000

QALY = Weighted quality‐adjusted life years.  Hoffmann et al. (2012, 2015). b  Minor et al. (2015). a

Weighted quality‐adjusted life daysb

Monetary loss per case (US$)b

0.18 2.09 29.41 0.18 4023 7.29 23.29 1.62 0.28 0.94 0.28 5.3

208 4903 1 525 594 218 7 013 777 11 936 1 468 384 5915 459 1429 363 41 652

Foodborne infections

31

Table 1.13  Antibiotic resistance changes in Salmonella Typhimurium DT104. Antibiotic

1990

1991

1992

1993

1994

1995

1996

Ampicillin Chloramphenicol Streptomycin Sulphonamides Tetracyclines Trimethoprim Ciprofloxacin

37a 32 38 37 36 0.4 0

50 49 52 53 50 3 0

72 60 75 76 74 3 0.2

85 83 85 86 83 2 0

88 87 92 93 88 13 1

90 89 97 90 90 30 7

95 94 97 97 97 24 14

 Percentage resistant.

a

These US studies can be compared with the two infectious intestinal disease survey sentinel studies in England (Table 1.6; Sethi et al. 1999; Tompkins et al. 1999; Wheeler et al. 1999; Tam et al. 2012). These studies were undertaken to determine a more accurate estimate of the incidence of food poisoning in England. The two UK sentinel studies estimated the overall extent of under‐reporting and found that for every case detected by laboratory surveillance, there were 136 in the community (Figure 1.6). Hence the scale of infectious intestinal disease in England was estimated at 9.4 million annual cases, of which 1.5 million cases are presented to general practitioners. Under‐reporting for individual organisms varied, most likely due to the severity of the illness. The reporting of Salmonella (3.2:1), Campylobacter (7.6:1) was higher than rotavirus (35.1:1) and norovirus (approximately 1562:1). These under‐reporting values differ from those of the USA (Table 1.11). The second sentinel study (Tam et al. 2012) was based on a population size of 800 000 people served by 88 general practitioners (GPs). It showed that in the UK, there are up to 17 million cases and 1 million GP consultations due to intestinal infectious diseases every year. Although the incidence of disease appeared to have increased since the 1990s, the number of consultations with GPs has halved. Norovirus was the most common recognised cause of infectious disease (3 million cases, 130 000 GP consultations per year). The total burden of foodborne diseases is that 20% of people in the general population of the UK are infected. This proportion is similar to the estimate of 28% in the USA. In England the organisms most commonly detected in patients with infectious intestinal disease are Campylobacter species (12.2% of stools tested), rotavirus group A (7.7%) and small round structured virus (6.5%). No pathogen or toxin was detected in 45.1–63.1% of cases. Surprisingly Aeromonas species, Yersinia species and some enterovirulent groups of E. coli were detected as frequently from controls as from cases. It is evident that causes of gastroenteritis vary with age (Figure 1.2). Noroviruses, caliciviruses and rotaviruses probably cause the majority of cases of gastroenteritis in children under 4, whereas bacteria (Campylobacter and Salmonella serovars) are the major cause of gastroenteritis in other age groups. Figure 1.3 indicates that males suffer from gastroenteritis more than females, except for one age group (>74 years, probably because of the lower ratio of men to women in this age group). A possible reason for part of this difference is that fewer men than women wash their hands after using the lavatory (33% compared with 60%, taking an average of 47 seconds, compared with 79 seconds). Helms et  al. (2003) compared the mortality rates of 48 857 cases of gastrointestinal ­infection with 487 138 controls from the general population. In their study they considered mortality rates over longer periods of time than normal studies (30 days of infection. The Public Health Agency of Canada estimates that, based on data collected for 2000–2010, the country experiences 4 million cases of foodborne illness each year in Canada, and this total is due to 30 known pathogens and unspecified organisms (Thomas et al. 2015). Their analysis accounted for under‐reporting and under‐diagnosis, and estimates of the proportion that were foodborne and those which were travel‐related were incorporated for each pathogen. Monte Carlo simulations were performed to account for uncertainty and these generated mean estimates and 90% probability intervals. Their calculations gave total estimated values of 11 600 (range 9250–14 150) hospitalisations and 238 (range 155–323) deaths associated with domestically acquired foodborne illness in Canada. This could be divided between those related to 30 known foodborne pathogens; 4000 hospitalisations (range 3200–4800) and 105 (range 75–139) deaths, compared with an estimated 7600 (range 5900–9650) hospitalisations and 133 (range 77–192) deaths associated with an unspecified organism. The prominent pathogens associated with these hospitalisations or deaths were norovirus, non‐typhoidal Salmonella serovars, Campylobacter species, E. coli O157 and L. monocytogenes. These data illustrate the considerable burden of foodborne illness in Canada. Even in a country such as Denmark, which has a significant health surveillance system, the number of people who get food poisoning each year is an under‐estimate and hence the cost to society has been unknown. Now researchers have calculated the real burden of disease from infections caused by the major foodborne pathogens Campylobacter species, Salmonella serovars and Shiga toxin‐producing Escherichia coli (STEC) using data for 2012. These calculations made allowances for under‐reporting and under‐diagnosis, thereby giving a higher number for how many people actually become ill owing to the three bacteria. The burden of disease is reported in DALYs. Campylobacter was determined to cause the highest disease burden. This gives authorities and other decision makers the scientific basis with which to prioritise initiatives aimed at increasing food safety and reducing the health consequences of infection by this organism. The amount of under‐reporting for the three organisms in Denmark was (i) Salmonella serovars – for every reported case, seven cases of disease were not reported; (ii) Campylobacter species infections – one out of 12 cases was reported; and (iii) for STEC infections –one out of 31 cases. The total burden of disease was highest for Campylobacter species with 1593 DALYs, followed by Salmonella serovars (389 DALYs) and STEC (113 DALYs). For both Campylobacter species and Salmonella serovar infections the disease that contributed the most to the total burden of disease was irritable bowel syndrome, while for STEC it was renal failure. For Campylobacter infections, 38% of the total number of DALYs was associated with foreign travel, while the major source of the burden of disease in Denmark was broilers. This was either from direct consumption of chicken meat or from environmental contamination. Therefore, reducing broiler contamination by Campylobacter should reduce the number of days of productivity lost through sickness. Belgium has estimated and forecast the number of cases due to Salmonella serovars, Campylobacter species and L. monocytogenes for 2012 through to 2020, and calculated the corresponding number of DALYs. These were based on the number of cases per month that were available: from January 2001 to December 2012 for salmonellosis, from January 1993 to December 2013 for campylobacteriosis and from January 2011 to December 2013 for

Foodborne infections

33

listeriosis. The proportion of salmonellosis and campylobacteriosis cases attributable to food is ~76%. The average monthly number of cases of salmonellosis was 264 in 2012 and is predicted to decrease slightly to 212 in 2020, while the numbers of Campylobacter cases were 633 and 1081, respectively. The numbers of L. monocytogenes infections are much smaller at 5 in 2012 and 6 in 2014. These values can be used to estimate DALYs for the three organisms. Salmonella serovars would be 102 in 2012 and were predicted to be 82 in 2020. Campylobacter cases would cause 1019 and 1736 DALYS, while listeriosis DALYs were higher than Salmonella ­serovars – 208 in 2012 and 252 in 2014. These figures indicate that Belgium needs to consider new action to reduce the risk of foodborne infection in particular from Campylobacter species, since its incidence is predicted to double by 2020. In England and Wales, Salmonella enterica serovar Enteritidis has caused the largest and most persistent epidemic of foodborne infection attributable to a single subtype of any pathogen since a systematic national microbiological surveillance programme was established in the 1940s (Lane et al. 2014). The epidemic was associated with the consumption of contaminated chicken meat and eggs. It has been estimated that during the epidemic >525 000 people became ill, causing 6 750 000 days of illness, 27 000 hospitalisations and 2000 deaths. There was only a decline in the number of infections when vaccination and other control measures for Salmonella were introduced for the production and distribution of chicken meat and eggs. Consequently, there has been a significant reduction in foodborne infections in England and Wales. It should be noted that up to 2009 across Europe, the numbers of food poisoning cases have been decreasing in recent years for all foodborne pathogens except C. jejuni (FoodNet 2009). The European Food Safety Authority (EFSA) reports (EFSA and ECDC 2007; EFSA 2017) that in 2016, Campylobacter continued to be the most commonly reported gastrointestinal bacterial pathogen in humans in the EU. The number of reported confirmed cases of human campylobacteriosis was 246 307, giving a notification rate of 66.3 per 100 000 population. Despite the high number of human campylobacteriosis cases, their severity in terms of reported case fatality rate remains low (0.03%), even though this was the third most common cause of mortality amongst the pathogens considered. The second most common cause was Salmonella serovars, with 94 530 confirmed cases, which was at the same notification rate as in the previous 5 years. There has been a statistically significant decreasing trend of salmonellosis between 2008 and 2016, however during 2012–2016 the trend has not shown any statistically significant increase or decrease. Seven EU Member States reported an increasing trend and four MS a decreasing trend over the period 2012–2016. The top five most commonly reported serovars during 2016 were, in decreasing order: S. Enteritidis, S. Typhimurium, monophasic S. Typhimurium, S. Infantis and S. Derby. The proportion of human salmonellosis illnesses due to S. Enteritidis has continued to increase in 2016. In Denmark there were few outbreaks due to Salmonella serovars in 2014 and nearly none in 2015. The total number of Salmonella infection cases was 925 in 2015 – the first time it has been lower than 1000, and more than half of these cases were acquired abroad. Salmonella Enteritidis is now rare in Denmark and has been nearly eradicated from Danish egg and chicken production. In contrast, S. Typhimurium and its monophasic variant are still recovered frequently in Danish food products and more than 4300 cases of Campylobacter infections were recorded, which was an increase on previous years. This increase could be due to improved surveillance coordination and methods of detection. As in other European countries, Campylobacter causes more intestinal infections than any other bacteria, with about a third being associated with travel abroad. It is still not fully established how many sources of

34

The microbiology of safe food

infection there are, though the probable primary source of infection is chicken meat either through consumption or through cross‐contamination in the kitchen. It should be noted that these values for clinical cases are where the agent has been isolated and identified. There are a significantly large number of cases for which the aetiological agent is unidentified. Frenzen (2004) estimates that such foodborne organisms cause 3400 deaths annually in the United States. Salmonella infections are believed to be mostly foodborne. The estimated numbers of Salmonella infection are 1.4 million and 80 million in the United States and worldwide per annum, respectively (Lynch et al. 2009). These numbers equate to 27 000 cases of salmonellosis/week in the United States. Despite major strides in foodborne surveillance in the United States in the past 16 years, detection of the 3800 cases of Salmonella/day in the United States is now occurring. In the USA in 2015, there were 902 foodborne disease outbreaks reported, resulting in 15 202 illnesses, 950 hospitalisations, 15 deaths and 20 food product recalls. This did not include outbreaks that had started before 2015, i.e. L. monocytogenes in caramelised apples (Section  4.12.7). The most common cause of confirmed, single‐aetiology outbreaks was norovirus, which accounted for 164 (37%) outbreaks and 3893 (39%) illnesses. This is similar to that already stated for Denmark. Salmonella serovars were the next most common cause, causing 149 (34%) outbreaks and 3944 (39%) illnesses. Salmonella Enteritidis was the most common serovar (51 outbreaks, 35%), followed by serovar 1, 4,[5],12:i:‐ (15, 10%), S. Newport (8, 6%) and S. Braenderup (7, 5%). This was followed by STEC, which caused 27 (6%) confirmed single‐aetiology outbreaks and 302 (3%) illnesses. Seventeen outbreaks (63%) were caused by serogroup O157, 3 (11%) by O26, 2 (7%) by O103, 1 (4%) by O45, 1 (4%) by O111, 1 (4%) by O121, 1 (4%) by O145 and 1 (4%) by multiple serogroups. Outbreaks caused by Cl. botulinum resulted in the highest proportion of ill persons hospitalised (97%), followed by L. monocytogenes (90%) and hepatitis A virus (38%). Amongst the 15 deaths reported, 14 (93%) were attributed to bacterial pathogens: Salmonella (9), Cl. botulinum (2), Cl. perfringens (1), L. monocytogenes (1) and Vibrio vulnificus (1). One death was attributed to norovirus. Fish (34 outbreaks), chicken (22) and pork (19) were the most common single food categories implicated. For the period 2009–2015, the CDC recorded 100 939 confirmed food poisoning cases, of whom 5699 were admitted to hospital and 145 died (Dewey‐Mattia et al. 2018). Chicken was responsible for the most illnesses with 3114 cases (12% of total). Pork was linked to 2670 cases (10% of total), and seeded vegetables caused a similar number of infections (2572 cases, 10% of total). There were 2953 outbreaks with a single confirmed aetiology, for which the four most prominent pathogens were: 1 Norovirus: 1130 (38%) outbreaks, and 27 623 (41%) of the cases. 2 Salmonella serovars: 896 (30%) outbreaks, and 23 662 (35%) of the cases. 3 Listeria monocytogenes, Salmonella serovars and STEC were responsible for 82% of hospitalisations and 82% of reported deaths. Amongst 1281 outbreaks in which the food reported could be classified into a single food ­category, the top three were: 1 Fish: 222 outbreaks, 17%. 2 Dairy: 136 outbreaks, 11%. 3 Chicken: 123 outbreaks, 10%. Multi‐state outbreaks were only 3% of all outbreaks reported, but led to 11% of illnesses, 34% of hospitalisations and 54% of deaths.

Foodborne infections

35

It is obvious from the plethora of differing detection procedures adopted by different c­ ountries that food poisoning statistics cannot be directly compared between countries due to the differing methods of analysis applied. Nevertheless, for several years the WHO has been encouraging Member States to quantify the national burden and causes of foodborne disease. Although there are a number of foodborne disease burden estimates, they are mainly from developed countries. In large parts of the world the data required to underpin such estimates are completely lacking. Additionally our understanding of foodborne infections is changing with the recognition of previously unrecognised pathogens and previously well‐ recognised pathogens that have increased in their prevalence or become associated with new food vehicles. In November 2007, the Foodborne Disease Burden Epidemiology Reference Group (FERG) was established by the WHO (Kuchenmüller et al. 2009). This group is composed of internationally renowned experts in a broad range of disciplines relevant to global foodborne disease epidemiology. Its objective was to determine the global burden of foodborne disease. This is the incidence and prevalence of morbidity, disability and mortality associated with acute and chronic infection. In order to achieve this, FERG needs to collaborate with the Food and Agriculture Organisation of the United Nations (FAO) and others to undertake regional consultations to discuss the regional specific profiles of foodborne syndromes and aetiological agents for future burden estimation. FERG’s remit is to consider microbial, parasitic, zoonotic and chemical contamination of food. Assessing the chemical and parasitic causes of foodborne disease burden is of particular interest since little work has been done on these topics to date. See Box 1.1 for a list of action points. FERG have reported their first estimates of the global incidence, mortality and disease burden (Havelaar et al. 2015; WHO 2015). According to their estimates, in 2010 there were ~600 million cases of foodborne illnesses, which led to >400 000 deaths. These infections could be attributed to 31 foodborne hazards. The WHO estimates that up to 30% of people in developed countries are infected by food and water every year (WHO Food safety and foodborne illness. Fact sheet 237, www.who.int/mediacentre/factsheets/fs237/en/print.html). This number is supported by US, Canadian and Australian authorities (Mead et al. 1999; Majowicz et al. 2004; OzFoodNet Working Group 2003).

Box 1.1  FERG Action Points Acute infectious diseases Examine pathogen specific global burden in children. Develop detailed analysis of cause of death data on WHO mortality database. Conduct or commission relevant burden work. Develop cause attribution models and estimate percentage foodborne. Recommend and/or conduct intervention studies to increase data availability. Chronic infectious diseases Evaluate the known microbial and chemical causes of foodborne disease. Develop cause attribution models and estimate percentage foodborne. Acute and chronic chemicals Identify major causes, particularly for developing countries, and commission relevant burden work. Develop cause attribution models and estimate percentage foodborne.

36

The microbiology of safe food

1.11  The cost of foodborne diseases

Several countries have estimated the economic consequences of foodborne illness. These costs include: • loss of income by the affected individual; • cost of health care; • loss of productivity due to absenteeism; • costs of investigation of an outbreak; • loss of income due to closure of businesses; • loss of sales when consumers avoid particular products. It is clear that foodborne illness has considerable implications for a country’s economy. The estimates, reviewed below, are primarily for developed countries. Unfortunately, there are fewer reliable estimates for developing countries, where the problem of diarrhoeal disease is far greater, and therefore the economic burden of foodborne illness must be even more severe. Early economic estimations for Canada were Can$1.3 billion loss due to foodborne pathogens (Todd 1989a). Back in 1991, Sockett estimated that in England and Wales, the 23 000 cases of salmonellosis were estimated to have resulted in an overall cost of £40–50 million (US$53m–$66m). Five years later Roberts (1996) estimated that the medical cost and value of lives lost due to just five foodborne infections in England and Wales was £300–700 million per year (US$400m–$900m). In Australia the cost of an estimated 11 500 food poisoning cases equated to AU$2.6 billion per year (ANZFA 1999). The Swedish Salmonella‐free poultry programme costs about $8 million per year, but saves an estimated $28 million per year in medical costs. The cost of Campylobacter infections in the USA has been estimated at $1.5–8.0 billion (Table 1.12). The total economic impact of foodborne diseases is a loss of $5–17 billion according to the US Food and Drug Administration. The Economic Research Service (ERS) is part of the United States Department of Agriculture, and estimates the ‘cost of foodborne illness’ for seven major foodborne pathogens (Plate 19). These are based on estimated medical costs and productivity losses over the infected individual’s lifetime. The medical costs include both acute illness and long‐term chronic complications. Therefore, the amount of medical attention required is determined according to the severity of illness. This ranges from those who do not visit the doctor, those who develop chronic complications, through to those who are hospitalised, and die prematurely. For each of these groupings the medical costs are estimated for the number of days of treatment, average cost per treatment or service and the number of patients (Buzby and Roberts 1997a). Productivity losses include changes in income and fringe benefits. High and low estimates of economic losses are calculated for cases where the person is unable to resume their normal job. This can be due to disability or death. The low estimate is based on loss of lifetime earnings and household production, whereas the high estimate is based on the ‘risk premiums’ in labour markets. The FDA uses the high estimate for productivity losses in its evaluation of food safety programmes, and adjusts this according to the person’s age. The cost of foodborne infections in the USA in 2011 has been estimated as $14.0 billion. This is based on an estimated total number of 9.4 million cases of infection attributable to 31 known organisms. The vast majority (95%) of these cases were due to 14 pathogens, and 98% of the deaths. In turn, 90% of the cases were attributable to non‐typhoidal Salmonella serovars, Campylobacter species, L. monocytogenes, Toxoplasma gondii and norovirus. However, there were an estimated 38.4 million further cases that were not attributable to an organism (Scallan et  al. 2011). These unattributed cases resulted in approximately 71 878 hospitalisations and

Foodborne infections

37

1686 deaths. It is plausible that the unattributed cases are due to organisms for which detection methods are currently not sufficiently robust, toxins and as‐yet‐unrecognised emergent pathogens. In the US, each year ~48 million people suffer from foodborne illness, yet only in 20% of these cases can a specific pathogen be attributed to the infection (Hoffmann et al. 2015). Of these, 90% are due to 15 pathogens and cost $15.5 billion in economic burden annually (84% due to deaths). In turn, 90% of the economic burden is due to only 5 of these 15 pathogens: Salmonella serovars ($3.7b, 17 000 QALYs), T. gondii ($3.3b, 11 000 QALYs), L. monocytogenes ($2.8b, 94 000 QALYs) norovirus ($2.3b, 5000 QALYs) and Campylobacter species ($1.9b, 133 000, QALYs) (Hoffmann et al. 2012 and 2015). According to similar calculations by Minor et al. (2015) the annual economic burden in the US is ~$36 billion. They also reported that Cronobacter was associated with the highest monetary loss due to the high incidence of infant mortality (Section 4.6). The Centers of Disease Control and Prevention (USA) estimates that 95% of Salmonella infections are foodborne in origin. Consequently the FoodNet estimate of 1.4 million annual salmonellosis cases means that 1.3 million cases were due to consumption of foods contaminated by Salmonella. The annual cost of campylobacteriosis in the USA is approximately $0.8–$5.6 billion (Buzby and Roberts 1997b) and estimated total costs of Campylobacter‐associated Guillain– Barré syndrome (GBS, an autoimmune reaction) are $0.2–$1.8 billion (Section  4.2). Hence reducing the prevalence of Campylobacter in food could prevent up to $5.6 billion in costs annually in the US. The economic burden of campylobacteriosis has been estimated to be €2.4 billion annually in Europe, and €82 million in the Netherlands in 2011. In the UK, Campylobacter infections cause more than 100 deaths per year, with considerably more people experiencing disease symptoms and decreased quality of life. As a result, this pathogen costs the UK economy ~£900 million (~$1.2b) per year. Therefore, it is not surprising that national regulatory authorities have set targets for the reduction of infections due to major foodborne pathogens. The estimated productivity costs due to foodborne infections are likely to increase in the future as the calculation methods improve. One omission is that the current methods do not include estimates for how much consumers are willing to pay for improved food safety. Additionally, it is highly probable that more foodborne pathogens will be identified in the future along with increased recognition of chronic complications. Currently chronic sequelae are estimated to occur in 2–3% of cases, and therefore the long‐term consequences to the individual and our economy may be more detrimental than the initial acute disease. Given these economic costs there is a considerable need, even in developed countries, for more systematic and aggressive steps to be taken to significantly reduce the risk of microbiological foodborne diseases. Since the cost of ‘gastroenteritis’ in developed countries is determined to be in the billions of US$, how much more is the human cost of food and water disease in developing countries where WHO estimates that worldwide almost 2 million children die every year from diarrhoea. A significant portion of deaths are caused by microbiologically contaminated food and water. Hence this is a major challenge for the future, which requires a global response. Therefore, this book frequently refers to the activity of international organisations activities such as Codex Alimentarius Commission and the WHO. The microbial contamination of food has a major impact on food production. Worldwide it is estimated that at least 10% of grain and legume production is lost, annually and this may even be as high as 50% for vegetables and fruits. This contamination affects trade in two ways. First, the contaminated food may not be accepted by the importing countries. Second, the loss in a country’s reputation for food safety may result in loss of further trade.

38

The microbiology of safe food

The cost benefits in preventing food poisoning through the assured food safety system HACCP (Section 9.5) have been estimated. Due to the range of economic models used the estimates varied considerably from $1.9 to $171.8 billion. Regardless of the exact figure it can be predicted that implementing HACCP will on balance cost less than the current medical costs due to foodborne infections, as well as decreasing productivity losses. In 2007, following a well‐publicised outbreak of S. Montevideo with ~40 victims, a major UK manufacturer of chocolate was fined $2 million. The company pleaded guilty to violating food and hygiene regulations, and was also ordered to pay $309 000 in legal costs. The cost of the product recall in 2006 was at least $60 million, and the company is reported to have spent $40 million on improvements, including changes to quality control procedures. Eventually the company was sold to a major food manufacturer based in another country. The problem came to light after three people, including two young children, were taken to a hospital in a food poisoning alert linked to the manufacturer’s chocolate. The victims’ age ranged from babies to adults of 52, and most were children under 4 years. Sales of the company’s chocolate fell by 14% during the first 6 months following the outbreak. The crux of the loss of Salmonella control by the manufacturers was the change to using a Most Probable Number (MPN) approach to detect the organism, which effectively permitted chocolate to be released onto the market despite positive results being obtained. 1.12  Changes in antimicrobial resistance of foodborne pathogens 1.12.1  Bacterial antibiotic resistance in agriculture and aquaculture

Antibiotic resistance, or antimicrobial resistance (AMR), is a major area of concern as the effectiveness of antibiotics to treat infections in humans and animals is falling. This results in a significant economic burden due to resulting higher costs of treatments and reduced productivity caused by absence. In 2009, the European Centre for Disease Prevention and Control (ECDC) and the European Medicines Agency (EMA) estimated that each year about 25 000 patients die in the EU from an infection caused by multidrug‐resistant bacteria. The resultant loss (health care costs and productivity losses) is in the order of €1.5 billion per year (~$1.8b/ year). In the US, antibiotic resistance adds $20 billion in excess direct health care costs, with additional costs to society for lost productivity as high as $35 billion a year. The European Commission has published two Action Plans (2011 and 2017) to reduce the increasing incidence of AMR. There are seven priority areas, which include the monitoring of AMR in bacteria from food‐producing animals and food. In order to evaluate the level of AMR, it is necessary to harmonise the method of determining antibiotic susceptibility and definition of resistance. The microdilution method needs to be used and interpreted according to the European Committee on Antimicrobial Susceptibility testing (EUCAST) epidemiological cut‐off (ECOFF) values for the interpretation of ‘microbiological resistance’. ECOFF values are preferred for surveillance purposes as they enable the identification of small changes in bacterial susceptibility, which may indicate emerging resistance and allow appropriate control measures to be considered at an early stage. The concentration ranges used in the testing ensure that both the ECOFF and the clinical breakpoints are included, making comparability of results with human data possible. Many bacterial species can develop resistance to the antibiotics they are exposed to in clinical and veterinary settings. An example can be the veterinary use of enrofloxacin causing cross‐resistance due to selection of DNA gyrase mutants (gyrA) in Salmonella serovars. The dissemination of antibiotic resistance (regardless of origin) within the microbial gene pool

Foodborne infections

39

means governments are now taking a ‘One Health’ approach for responsible use of antibiotics. The ‘One Health’ approach recognises that human and animal health are interconnected, with diseases being transmitted from humans to animals and vice versa. Resistance can occur by: • Random mutation with subsequent positive selection of mutants resistant to the antibiotic, i.e. fluoroquinolone resistance in C. jejuni. • Mobilisation, including HGT, of resistance genes, i.e. colistin resistance in E. coli. • Dissemination of strains with previously developed resistance, i.e. S. Typhimurium DT104. A microbiological risk assessment of fluoroquinolone‐resistant C. jejuni is reviewed in Section 11.2.4. Isolation of S. Typhimurium DT104 resistant to ampicillin, chloramphenicol, streptomycin, sulphonamides and tetracycline (R‐type ACSSuT) was first reported in 1984, and appears to be highly clonal. It is likely that the genes were acquired through HGT, and afterwards the strain was widely disseminated due to the veterinary use of antibiotics. It has been proposed that the selective pressure was in part due to the use of antibiotics in aquaculture (Section 1.12.1). This organism has been isolated from cattle, poultry, sheep, pigs and horses. Antimicrobial therapy is used extensively to combat S. Typhimurium infection in animals, and the evolution of a strain resistant to the commonly used antibiotics has made infections with S. Typhimurium in food animals difficult to control. The primary route by which humans acquire infection is by the consumption of a large range of contaminated foods of animal origin. Concern over the increase in AMR is not new. Table 1.13 shows the significant increase in antibiotic resistance in S. Typhimurium DT104 isolates between 1990 and 1996 in England and Wales. In addition to the antibiotic‐resistant type ACSSuT, many isolates are also resistant to trimethoprim and ciprofloxacin. Ciprofloxacin resistance has been most notable in S. Hadar (39.6%) compared with other Salmonella serovars. A number of international surveillance networks monitoring antibiotic resistance are covered later, such as FoodNet and Global Salm‐Surv (Sections 13.1.2 and 13.1.7). In Europe there have been a number of concerted efforts to control microbial antibiotic resistance including a ban in 2006 on using antibiotics as growth promoters (Eurosurveillance 2008). It has been proposed that the reduction in prophylactic use of antibiotics has led to a reduction in animal health leading to infection by E. coli and Lawsonia intracellularis in early post‐weaning pigs and clostridial infection in broilers. Consequently there may have been an increased use of therapeutic antibiotics in animals that are also used in human treatments (Casewell et  al. 2003). Nevertheless good husbandry reduces animal infections and subsequently reduces the use of antibiotics. While the use of antibiotics in agriculture has received much attention, there is considerable use of antibiotics in aquaculture (Done et  al. 2015). China is the largest single aquaculture producer and exporter with ~62% of the global aquaculture production volume (2012 figures). A considerable amount of antibiotics are used in fish cultivation as disease control agents and growth promoter in aquaculture. This raises concerns as this extensive use of antibiotics may result in water and sediment contamination and the development of antibiotic resistance genes (Mo et al. 2017). There are ~50 antibiotics commonly used in both aquaculture and agriculture, of which ~39 are also of importance in human medicine. Six classes of antibiotics commonly used in both agriculture and aquaculture are also included on the WHO list of critically important/highly important/important antimicrobials. The emergence of antibiotic resistance amongst fish bacterial pathogens increases the risk not only of these antibiotic‐resistant bacteria causing an

40

The microbiology of safe food

infection, but also the transfer of antibiotic resistance determinants to other bacteria of concern in animal and human health. For example, tetracycline resistance, encoded by Tn1721, has been disseminated between Aeromonas salmonicida, a fish pathogen, and the human pathogens A. hydrophila, A. caviae and E. coli. Similarly, A. salmonicida plasmids resistant to trimethoprim, sulphonamide and streptomycin can transfer with high frequency to E. coli and Salmonella. It has been argued that molecular and epidemiological evidence has demonstrated that multiple antibiotic resistance determinants in S. Typhimurium DT104 (a major cause of salmonellosis in humans and animals in Europe and the USA) possibly originated in aquaculture (Section 4.3). The antibiotic resistance determinants of S. Typhimurium DT104 are encoded by a chromosomal transmissible genetic element that contains a resistance gene (floR) for florfenicol. This is an antibiotic that is extensively used in aquaculture in the Far East. Additionally, florR was first reported in the fish pathogen Vibrio damsel (Bolton et al. 1999; Angulo et al. 2004). The increasing generation of whole‐genome sequences for bacterial isolates means the organisms can be screened for antimicrobial factors. In the US, the National Antimicrobial Resistance Monitoring System for enteric bacteria (NAMRS) monitors for changes in the antimicrobial susceptibility of certain enteric bacteria (Salmonella serovars, E. coli, Shigella, Vibrio and Campylobacter species) isolated from humans, retail meats and food animals. They have also compared E. coli whole‐genome sequences to a database containing over 2500 resistant genes and gene variants. They were then able to predict antimicrobial resistant profiles that correlated with laboratory‐generated profiles (Tyson et  al. 2015). Similarly, the Center for Genomic Epidemiology (Denmark) (www.genomicepidemiology.org) predicts the antibiotic resistance traits in user’s uploaded genomes. This is accessible to researchers, educators and non‐governmental investigators unlike NARMS. At present in silico‐predicted antibiotic resistance profiles cannot be reliably used to replace laboratory‐based antibiotic resistance determination (ECOFFS, clinical breakpoints, etc.). Nevertheless, using whole‐genome sequence data in the future will enhance antimicrobial monitoring systems and improve the tracking of antimicrobial resistant isolates within the food supply, as well as supporting epidemiological investigations. 1.12.2  Antibiotics of concern and resistance mechanisms

Antibiotics are currently our most successful means of treating infectious diseases. However, resistance to commonly used antibiotics is increasing through various mechanisms and selection pressures. The acquisition of genes from another strain or species can provide resistance to an entire class of antimicrobials. These genes are frequently associated with large transferable extrachromosomal DNA elements, plasmids, on which there may be other mobile DNA elements such as transposons and integrons. Plasmids in the incompatibility group IncA/C are of considerable interest with respect to multiple antibiotic resistance in enteric pathogens of humans and animals. These large, low‐ copy number plasmids are of increasing relevance in the spread of antibiotic resistance. They have at least three hotspots for the integration of mobile genetic elements. They were first identified from multidrug‐resistant fish pathogens A. hydrophila and Vibrio species, and coincided with the use of antibiotics in aquaculture (Section 1.12.1). IncA/C plasmids and their multiresistance phenotype emerged in clinical isolates of Salmonella Newport after 1980. The reservoir appears to be environmental bacteria, followed by the acquisition of antibiotic gene modules due to the selective pressure of antibiotic exposure. IncA/C plasmids have acquired the NDM‐1 gene encoding the New Delhi metallo‐β‐lactamase (NDM) and are found in clinical strains of E. coli and Klebsiella pneumoniae.

Foodborne infections

41

In Enterobacteriaceae, resistance to β‐lactam antibiotics is mediated by production of β‐ lactamase enzymes, which inactivate the drugs by hydrolysing the β‐lactam ring. Some of the clinically most important inactivating enzymes are: 1 Extended‐spectrum β‐lactamases (ESBLs). This includes the SHV, TEM and CTX‐M type β‐lactamases. The prevalence of CTX‐M is increasing in Enterobacteriaceae and predominates as a cause of extended‐spectrum cephalosporin resistance. 2 Carbapenemases. This includes class A (KPC types), class B metallo‐β‐lactamases (MBLs) and class D oxacillinases. The carbapenem‐resistant Enterobacteriaceae have been increasingly reported worldwide. The carbapenemases include NDMs, KPCs and OXA‐48. KPCs are currently the most common cause of carbapenem resistance worldwide. The emergence of New Delhi metallo‐ β‐lactamase (NDM‐1) and its variants has raised a major public health concern. NDM‐1can hydrolyse a wide range of β‐lactam antibiotics, including carbapenems. 3 AmpC cephalosporinases. Another issue with increasing antibiotic resistance is the co‐resistance to microbiocides, such as triclosan and quaternary compounds. Exposure of Salmonella serovars and E. coli O157 to sublethal concentrations of antibacterial agents can lead to adaptive resistance to both biocides and antibiotics. For example, benzalkonium chloride‐resistant Salmonella Virchow showed elevated resistance to chlorhexidine. Escherichia coli O157 with adapted resistance to triclosan also had increased resistance to chloramphenicol, erythromycin, imipenem, tetracycline and trimethoprim. Increased co‐resistance to sanitary products and antibiotics can also occur due to the co‐location of heavy metal resistance and antibiotic resistance genes on the same plasmid, or the same genes encoding for resistance to both metals and antibiotics. 1.12.3  Polymyxin and plasmid‐encoded colistin resistance

Polymyxins are active against most members of the Enterobacteriaceae family. There is, ­however, some natural resistance amongst Proteus, Brucella, Legionella, Campylobacter and Vibrio species. Acquired resistance to polymyxins in Enterobacteriaceae may be due to genes encoding lipopolysaccharide‐modifying enzymes, regulators of the PmrAB and PhoPQ two‐ component systems, and the intrinsic regulator RamA (Poirel et al. 2017). Colistin, also known as polymyxin E, is an antibiotic of last resort for the treatment of ­multiple antibiotic‐resistant Gram‐negative infections. Chromosomal‐encoded resistance had been known for some time. However, it was not until November 2015 that the plasmid‐borne colistin resistance gene mcr‐1 was first identified in E. coli isolates from Chinese pigs, pork products and patients (Liu et al. 2016). It has since been reported in food‐producing animals and people in more than 30 countries across five continents. This caused an international alarm as plasmid‐borne genes are far more transmissible than if they were chromosomal. Additionally, mcr‐1 encoding plasmids may also carry other antibiotic resistance genes: carbapenemases and extended‐spectrum β‐lactamases. The mcr‐1 gene (and variants mcr‐2 to ‐5) has also been identified in other members of the Enterobacteriaceae from South America, Asia, Europe and Africa, demonstrating that it is widespread though at a low level such that it was not found in bacterial genome databases (Olaitan et al. 2016). It is now thought that the mcr‐1 gene emerged about 10 years before it was first identified, and most likely originated in Chinese livestock (Wang et al. 2018). Analysis of ~450 whole‐ genome sequenced isolates from 31 countries was carried out, with China (212 isolates) and Vietnam (58) having the most MCR‐1–positive samples. Nearly 90% of the isolates (411) were E. coli, but isolates carrying the gene were found in seven other bacterial species: S. enterica,

42

The microbiology of safe food

K. pneumoniae, Escherichia fergusonii, Kluyvera ascorbata, Citrobacter braakii and Klebsiella aerogenes. Between 2002 and 2008, mcr‐1 first became mobilised as part of a composite transposon. The MCR‐1 transposon then jumped to different plasmids resulting in a dramatic demographic diversification and expansion. The highest proportion of isolates carrying transposon sequences similar to the ancestor are environmental and agricultural isolates from China, where colistin was used widely in food‐producing animals until 2016. Dissemination globally was probably primarily driven by the trade of food animals and meat, although movement of colonised humans may have contributed as well. 1.12.4  Livestock‐associated methicillin‐resistant Staphylococcus aureus (LA‐MRSA)

Methicillin‐resistant Staphylococcus aureus (MRSA) can cause a severe invasive disease in humans, as well as infections in animals – mastitis in cattle, exudative dermatitis in swine and mild skin infections. Unfortunately, it can be a commensal of humans and other mammals. MRSA started to circulate in hospitals in the 1960s. Since then MRSA has been isolated outside the hospital setting (community‐acquired MRSA, CA‐MRSA) and from livestock (livestock‐ associated MRSA, LA‐MRSA). The prototypical LA‐MRSA isolate was characterised by multi‐locus sequence typing (MLST; Section 6.6.1) and defined as sequence type 398 (ST398). The identification and prevalence of this strain in livestock species indicate that livestock are potentially the largest reservoir for MRSA outside the hospital setting. This lineage has been identified in a variety of livestock species and is considered to be adapted for the colonisation of non‐human mammals, although colonisation and infection have been noted in humans. In Europe and Asia, LA‐ MRSA isolates are primarily MLST sequence type (ST) 398 and ST9, respectively, while in the United States it also contains ST5. These latter strains are of concern because, unlike ST398 and ST9, which are considered livestock‐adapted, ST5 isolates are a widespread and successful hospital‐acquired MRSA lineage. The pathogenicity of ST5 isolates may be associated with the capacity and ease with which they acquire mobile genetic elements. 1.13  Food safety following natural disasters, and conflict

Following natural disasters, it is essential that safe water and food are available. The victims may be suffering from injuries, shock, disorientation and fear. The risk of further illness will be high due to poor hygiene, inability to cook and the lack of toilet facilities. Hence food and water can quickly become contaminated with microbial pathogens causing hepatitis A, typhoid fever, cholera and dysentery. To help governments in their planning and response to natural disasters, the WHO has developed the guide Ensuring Food Safety in the Aftermath of Natural Disasters (see www.who.int/foodsafety/foodborne_disease/emergency/en/). It offers specific advice to those involved in food storage, handling and preparation during disaster situations. In summary: 1 Preventive food safety measures in the aftermath of natural disasters. All water should be treated as contaminated with surface water unless it is boiled or otherwise made safe. Only after treatment can it be consumed or used in preparing food. However, the food may still contain chemical hazards. Agricultural areas should be assessed for where food can still be harvested or where food has been safely stored after harvesting. 2 Inspecting and salvaging food. All food stocks should be inspected and assessed for their safety. Canned foods with broken seams, serious dents or leaks, and jars with cracks should be discarded. Food deemed as safe should be segregated from other foods. In areas of flooding,

Foodborne infections

43

food should be stored in a dry area. Mould growth (and possible toxin production) is more likely to occur on vegetables, fruits and cereals that are stored under moist conditions, or have become wet. Refrigerated food, in particular meat, fish, poultry and milk, that can no longer be kept cold should be consumed before it has been in the ‘danger zone’ (5–60°C) for more than 2 hours. Other foods, normally refrigerated, can be kept for longer than 2 hours, but should be discarded if showing signs of spoilage. 3 Provision of food after a natural disaster. When cooking facilities become re‐established it is usual to distribute dried food. Therefore, instructions on food preparation, especially the crucial need for safe water not only for reconstitution, but also washing hands and utensils, will be needed. 1.14  Food microbiology, foodborne diseases and climate change

There is now widespread consensus that the Earth’s climate is changing, especially in regions that will affect humans. With respect to food production global climate change may have various effects, some of which may include the microbial safety of food (Tirado et al. 2010; Lake and Bakker 2018). The geographic distribution and production of crops used for our staple diet and the phyllosphere microflora of those crops may be strongly affected by climate change. A few examples are given below. Aspergillus flavus is able to grow under high temperatures and drought conditions and can produce highly toxic mycotoxins called aflatoxins (Section 5.4.1). It is able to persist under extreme heat and dry conditions and is an increasing problem in the Mediterranean and other temperate regions (Cotty and Jamie‐Garcia 2007; Perrone et al. 2014). For example, the impacts of climate change have been observed in Serbia, where no contamination occurred previously, but prolonged hot and dry weather in the year 2012 resulted in 69% of maize plants being contaminated with A. flavus. A similar case was also found in Hungary, where the increase in A. flavus contamination may be due to climate change conditions. The world’s largest agri‐food exporters include countries such as Brazil and Argentina and parts of Asia including China and India. These are also areas predicted to be impacted by climate change. Therefore, a greater understanding of the effect of climate change on fungal contamination and mycotoxin production is needed for food security. Warming seas will have an effect on the distribution of bacterial pathogens. Vibrio vulnificus is rarely isolated from waters with salinities greater than 2%, consequently it is not found in regions such as the Mediterranean Sea, and hence no reported infections. It also rarely occurs when water temperatures are less than 13°C, so the vast majority of cases occur in the warmer summer months of May to October. Vibrio vulnificus is known to enter a viable but non‐ culturable state (VBNC; Section  6.2.3) at temperatures below 13°C. Within these limits, this pathogen occurs in estuarine environments worldwide, but with an increasing incidence and geographic distribution due to global climate change. Similarly, the rise in sea water ­temperature from 1982 to 2010 in the Baltic Sea coincides with the increase in Vibrio infections in the Baltic region and western North Sea, especially in 2006 when there was a heatwave (Baker‐Austin et al. 2013).

2 Basic aspects

2.1  The human intestinal tract

The human intestinal tract is approximately 30 ft. (~9 m) long, and is divided into a number of regions: oesophagus, stomach, small intestine, large intestine and anus (Figure  2.1). In the mouth food is chewed, which breaks it into smaller pieces and mixes it with saliva. On swallowing, the food passes through the pharynx and oesophagus into the stomach. The stomach produces gastric enzymes (endopeptidases, gelatinase and lipase) to start the breakdown of food components, and has a low acidic pH (~pH 2). Most digestion and absorption occurs in the small intestine, which is about 20 ft. (~6 m) long. The structure of the small intestine maximises the area available for absorption. The surface of the mucosa is convoluted and folded. The surface is covered with finger‐like projections called villi, which are, in turn, covered with absorptive cells. The effective surface area of the mucosal cells is further increased by the microvilli that occur on the luminal membrane of the enterocyte (Figure 2.2). The brush border of the enterocytes contains various enzymes including many disaccharides such as maltase, isomaltase, sucrase and lactase. These are involved in both the digestion and absorption of carbohydrates. Pancreatic enzymes, which include trypsin, chymotrypsin, carboxypeptidase, amylase, lipases, ribonuclease, deoxyribonuclease, collagenase and elastase, assist in digestion. Also bile salts secreted from the liver aid the absorption of fats. The tissue underneath the epithelial cells contains blood capillaries that absorb monosaccharides and amino acids, and lymph capillaries to absorb fatty acids and glycerol. Mucosal enzyme levels are affected by bacterial activity and lactose deficiency is a sensitive indicator of the colonisation of the small intestine by pathogenic bacteria. Interactions between the cell membranes and the luminal contents are facilitated by the glycocalyx, a complex mucus layer overlaying the enterocytes. The small intestine is the site where major absorption (>80%) occurs, but is also perturbed by many intestinal infections, leading to acute diarrhoea and dehydration. For example, the cholera toxin inhibits sodium uptake, and stimulates chloride secretion (Figure  2.3). This results in fluid and electrolyte loss. Undigested food enters the colon, which has a neutral pH and a slow transit time (up to 60 h).

The Microbiology of Safe Food, Third Edition. Stephen J. Forsythe. © 2020 John Wiley & Sons Ltd. Published 2020 by John Wiley & Sons Ltd.

45

46

The microbiology of safe food Mouth

Stomach

Duodenum Jejunum

Small intestine

Ileum

Colon and appendix

Large intestine

Anus

Figure 2.1  The human intestinal tract.

2.2  The normal human intestinal flora

In the womb, the intestinal tract of the human foetus is microbiologically sterile. However, during birth the neonate acquires a microbial flora from the vagina, contact with the environment and through feeding. Consequently a dense, complex bacterial community becomes established in the intestinal tract (O’Hara and Shanahan 2006). The maturation of the immune system requires continual stimulation from the developing gut flora. The lack of immunity development has been linked with an increase in prevalence of atopic diseases: the ‘hygiene hypothesis’ (Section 1.8). The intestinal flora changes in relationship to diet, age and disease. It is a complex ecosystem and may have a number of effects on the health of the host. Microbiome studies of the infant and adult intestines based on high‐throughput sequencing (Section 14.1) have significantly expanded the study of the human intestinal flora. This is primarily because it is culture‐independent and is therefore not dependent upon the use of numerous isolation agars, and incubation conditions. This enables larger and more in‐depth studies to be undertaken, including the interaction between the intestinal flora, immune status and cognitive behaviour (Fung et al. 2016; Feng et al. 2018). The role of the intestinal microbiome in multiple diseases identifies it as a potential therapeutic target through the ingestion of live probiotic bacteria or their modulation by dietary prebiotics.

Microvili

Lamina propria

Villus crypt Figure 2.2  Gut mucosal structure.

Intestinal lumen

Vibrio cholerae K+ Na+ HCO3– Cl– H2O Type II secretion system Neuramidase

A1–A2:B5 toxin

A1–A2:B5 toxin GM1 ganglioside CFTR Cystic fibrosis transmembrane A1–A2 A2 A1

cAMP-dependent protein kinase A

G protein

Adenylate cyclase

cAMP GTPase inhibited

Figure 2.3  Osmotic balance; NaCl flux across gut mucosa.

Mucosal cell

48

The microbiology of safe food

Bacterial colonisation of neonates starts soon after birth and facultative anaerobic bacteria can initially be detected in faecal samples indicating their presence in the large intestines. These organisms remove oxygen and subsequently lower the redox potential, which enables strict anaerobes to grow. In vaginally delivered neonates the first colonising bacteria are of maternal origin. In Caesarean section‐delivered neonates, the environment and hospital staff are a major source of the colonising bacteria. After this initial colonisation, the diversity of the subsequent bacterial flora community is influenced by the diet. Human breast milk is not necessarily sterile, and can have low numbers of streptococci, micrococci, lactobacilli, staphylococci, diphteroids and bifidobacteria. Breast‐fed infants are often colonised by staphylococci due to increased contact with the mother’s skin during feeding. In breast‐fed full‐term infants, Gram‐positive bifidobacteria, lactobacilli and streptococci dominate. The intestinal microbial flora of formula‐fed infants is initially more diverse than that of breast‐fed infants. Instead, the predominant flora is a mixture of Enterobacteriaceae (Escherichia coli and Klebsiella species), Staphylococcus, Clostridium, Bifidobacterium, Enterococcus and Bacteroides species. Despite this common distinction between breast‐ and formula‐fed infants there is a significant issue if the neonate is born prematurely (low birth weight). Premature neonates do not have the teat/suckling reaction and need to be fed either the mother’s milk or reconstituted formula through nasogastric tubes. These tubes rapidly become colonised with a large variety of bacteria, yeast and fungi. For example, single nucleotide polymorphism (SNP) analysis has shown the colonisation of the infant intestines by antibiotic resistance‐encoding bacteria (Enterococcus faecalis and Enterobacter hormaechei), which were 150 kDa) single‐chain toxins that are proteolytically cleaved during activation and cellular entry. The botulinum toxin blocks the action of peripheral nerves (Figure 2.9). See Section 4.3.9 for more detail.

Basic aspects

59

150 kDa toxin activated by protease activity, either gastric or clostridial Toxin is a zinc-requiring endoprotease Nicked AB toxin: A, 50 kDa (light chain) B, 100 kDa (heavy chain) B subunit binds to sialic acid containing glycoprotein on peripheral neurons Toxin internalised into neuron Toxin prevents release of the neurotransmitter acetylcholine Consequently, nerve pulse transmission stops causing flaccid paralysis Figure 2.8  Structure and activation of Cl. botulinum toxin.

C

S S N

Protease cleavage Heavy chain (100 kDa) C

N S S N

C Light chain (50 kDa)

Figure 2.9  Cl. botulinum toxin mode of action.

2 AB5 where five B subunits (pentamer) form a doughnut‐like ring structure and possibly the A subunit enters the cell through the central hole. Examples include cholera toxin, E. coli heat‐ labile toxin and the Shiga toxins (Figure 2.10). Cholera and Shiga toxins bind to glycolipid ganglioside receptors on the host cell. The Shiga toxins attack 28S rRNA to depurinate adenine

60

The microbiology of safe food

A subunit 32 kDA

B subunit 7.7 kDA

Figure 2.10  AB5 structure of Shiga toxin.

4324, which is involved in elongation factor 1‐mediated binding of tRNA to the ribosomal complex and subsequently inhibits protein synthesis. The A subunit of Shiga toxin has sequence and structural homology with the ricin family of plant toxins, which have an identical mode of action. The A subunit is activated by proteolytic cleavage on cell entry to yield two fragments joined by a disulphide bridge, which is subsequently reduced: cholera toxin and E. coli heat‐labile (LT) toxin ADP‐ribosylate G3, a heterotrimeric G protein involved in the stimulation of adenylate cyclase (Figure 2.11). The A subunit of Gs is modified at Arg201, which inhibits the GTPase activity. This keeps the Gs protein in the ‘on’ position, leading to permanent activation of adenylate cyclase. The resultant high concentration of cAMP in the gut epithelial cells causes massive fluid accumulation in the gut lumen and watery diarrhoea, which can be fatal. Cholera toxin is chromosomally encoded on a phage, whereas E. coli LT and ST toxins are plasmid‐encoded and Shiga toxin is chromosomally encoded. Transported Toxins

Toxins in this group do not display toxic activity when purified from bacteria. They are transported directly from the bacterial cytoplasm to the eucaryotic cytoplasm by a complex array of proteins that bridge the membranes of the two cell types. An example is the Type III secretion system of Salmonella and E. coli O157. This secretion system is often encoded on pathogenicity islands (PAIs), described in the next section. Bacteria known to produce such toxins include Salmonella, E. coli, Shigella and Yersinia species. 2.5.2  Pathogenicity islands

PAIs are large (>30 kb) distinct chromosomal elements encoding virulence‐associated genes (Table 2.2) (Schmidt and Hensel 2004; Bertelli and Brinkman 2018). Because the %GC content of PAIs are often distinct from the rest of the bacterial DNA %GC content, it has been ­proposed that they may have been acquired in the past by horizontal gene transfer from other bacterial species. They are often found at tRNA loci. In addition to PAIs, there are shorter sequences of DNA, termed ‘pathogenicity islets’, which are transferred between bacterial pathogens. PAIs constitute major routes in the evolution of bacterial pathogens since in one acquisition they

Basic aspects

61

Vibrio cholerae ingested

Virulence genes expressed in stomach

Adherence to small intestinal mucosa

Colonisation of small intestinal mucosa

Production of cholera toxin which activates adenylate cyclase

Extensive fluid loss due to osmotic effect caused by sodium loss into gut lumen. Production of rice water stools Figure 2.11  Mode of action of cholera toxin.

can transform a benign organism into a virulent one; for example, the PAI called locus of enterocyte effacement (LEE) in E. coli pathovars. Both enterohaemorrhagic E. coli and enteropathogenic E. coli (EHEC and EPEC) contain a 35‐kb PAI called the LEE island (Section 4.4). This encodes for a Type III secretion system and other virulence factors that are essential for disease. It induces the attaching and effacing lesions on enterocytes and encodes the secretion system for toxin transfer from the E. coli cell to the host cell. This results in cytoskeletal rearrangements and the formation of a pedestal on which the E. coli cell is located. Uropathogenic E. coli cause urinary tract infections. These have a completely different PAI inserted in exactly the same site as EPEC strains. The PAI encodes

62

The microbiology of safe food

Table 2.2  Pathogenicity islands of three important foodborne pathogens.

Pathogen

PAI designation

Size (kb)

G + C ratio (PAI/host)

Escherichia coli

PAI I PAI II LEE (PAI III)

70 190 35

40/51 40/51 39/51

Salmonella Typhimurium

SPI‐1

40

42/52

Vibrio cholerae

SPI‐2 SPI‐3 VPI

40 17 39.5

45/52 Unknown 35/46

Listeria monocytogenes Listeria ivanovii

LIPI‐1 LIPI‐2

9 22

Phenotype Haemolysin production Haemolysin, P fimbriae production Induction of attaching and effacing lesions on enterocytes Invasion of non‐phagocytic cells Survival in macrophages Survival in macrophages Colonisation, expression of phage CTXɸ receptor Intracellular survival and spread Cell and host trophism

Source: Adapted from Henderson et al. (1999). PAI = Pathogenicity island.

for P fimbriae (an adhesin) and haemolysin (a toxin), which together are virulence factors required for urinary tract colonisation. PAIs are covered in more detail with the specific pathogens E. coli (Section 4.4) and Salmonella (Section 4.3). Although Yersinia and Shigella species have Type III systems on their plasmids, they have other virulence attributes that are on the chromosome and not localised in islands. Vibrio cholerae produces cholera toxin encoded by the ctxA and ctxB genes that are encoded by the filamentous phage CTX. The bacterial receptor for phage infection is the toxin co‐regulated pilus (TCP), which is also an important adherence determinant. TCP is found on the 39.5‐kb PAI called VPI in V. cholerae. It is believed that the acquisition of VPI enables aquatic V. cholerae strains to colonise the human intestine and the subsequent generation of epidemic and pandemic V. cholerae strains. Pathogenic strains of L. monocytogenes contain LIPI‐1. The 9‐kb element encodes for six genes required for listeriolysin O (LLO), a thiol‐activated haemolysin (hly), actA and plcB (responsible for intra‐ and intercellular movement). A second island LIPI‐2 (22 kb) is found in L. ivanovii. This encodes for several secreted internalin genes and the sphingomyelinase gene (smlC), which may be important in the organism’s tropism for ruminants. 2.5.3  Bacterial toxins encoded in bacteriophages

Bacteriophages are involved in the transfer of virulence factors between pathogens. For example, the Shiga toxin of Sh. dysenteriae is encoded on a bacteriophage that is integrated in the chromosome. Escherichia coli O157:H7 causes haemorrhagic colitis and haemolytic uraemic syndrome and contains the Shiga toxin genes. It is plausible that a bacteriophage encoding the Shiga toxin from Sh. dysenteriae was transferred to an EPEC strain and this led to the generation of a new pathogen, EHEC. Under certain circumstances, the phage encoding the Shiga toxin becomes lytic, multiples and subsequently releases the toxin. Another example of bacteriophage‐ encoded toxin is in V. cholerae, which produces cholera toxin encoded by the ctxA and ctxB genes. These are encoded by the filamentous phage CTX. The bacterial receptor for phage infection is the TCP, which is also an important adherence determinant. Consequently, the phage only infects bacteria that already possess an essential adhesin, thus ensuring virulence.

Basic aspects

63

2.6  Microbial growth cycle

Microbial growth cycle is composed of six phases (See Figure 2.12): • Lag phase: cells are not multiplying, but are synthesising enzymes appropriate for the environment. Lag times are more variable than growth rate possibly due to the effects of the physiological history of the cell, and the environment. • Acceleration phase: an increasing proportion of the cells are multiplying. • Exponential (or log) phase: the cell population are multiplying by binary fission (1‐2‐4‐8‐16‐32‐64, etc.). The cell numbers are increasing at such a rate that to graphically represent them it is best to use exponential values (logarithms). This results in a straight line, the slope of which represents the μmax (rate of maximum growth) and the doubling time ‘td’ (time required for the cell mass to increase twofold) can be determined. • Deceleration phase: an increasing proportion of cells are no longer multiplying. • Stationary phase: the rate of growth equals the rate of death, resulting in equal numbers of cells at any given time. Death is due to the exhaustion of nutrients, the accumulation of toxic end products and/or other changes in the environment, such as pH changes. The length of the stationary phase is dependent upon a number of factors such as the organism and environmental conditions (temperature, etc.). Spore‐forming organisms will develop spores due to the stress conditions. • Death phase: the number of cells dying is greater than the number of cells growing. Cells that form spores will survive longer than non‐spore‐formers. The length of each phase is dependent upon the organism and the growth environment, ­temperature, pH, water activity, etc. The growth cycle can be modelled using sophisticated computer programs and leads to the area of microbial modelling and predictive microbiology (see Section 2.10). 2.7  Death kinetics 2.7.1 Expressions

There are a number of expressions used to describe microbial death: • D value: decimal reduction time. Defined as the time at any given temperature for a 90% reduction (=1 log value) in viability to be effected.

Viable count (log10 cfu/ml)

10 9

Stationary

8 Exponential

7 6 5 4

Lag

3 2 Time Figure 2.12  The microbial growth curve.

Death

64

The microbiology of safe food

• z value: temperature increase required to increase the death rate tenfold; or in other words reduce the D value tenfold. • P value: time at 70 °C. A cook of 2 minutes at 70 °C will kill almost all vegetative bacteria. For a shelf life of 3 months the P value should be 30–60 minutes according to other risk factors. • F value: this value is the equivalent time, in minutes at 250°F (121°C), of all heat considered, with respect to its capacity to destroy spores or vegetative cells of a particular organism. Since these values are mathematically derived they can be used in predictive microbiology (Section 2.10) and microbiological risk assessment (Chapter 10). 2.7.2  Decimal reduction times (D values) and z values

To design an effective temperature treatment regime, it is imperative to have an understanding of the effects of heat on micro‐organisms. The thermal destruction of micro‐organisms (death kinetics of vegetative cells and spores) can be expressed logarithmically. In other words for any specific organism, in a specific substrate and at a specific temperature, there is a certain time required to destroy 90% (=1 log reduction) of the organism. This is the decimal reduction time (D value). Plotting the survival numbers (as log10 cfu/ml) for an organism against time g­ enerally gives a straight‐line relationship, which is more precisely known as a log‐linear relationship (Figure 2.13 ). The rate of death depends upon the organism, including the ability to form spores, and the environment (Table 2.3). Free (or planktonic) vegetative cells are more sensitive to detergents than fixed cells (i.e. biofilms). The heat sensitivity of an organism at any given temperature varies according to the suspending medium. For example, the presence of acids and nitrite will increase the death rate, whereas the presence of fat may decrease it. The D value is also dependent upon the inoculum preparation and the enumeration conditions. This has been demonstrated for E. coli O157:H7 and is summarised in Figure 2.14 (Stringer et al. 2000). Hence D values quoted in books and journals cannot be taken as fixed values and directly applied to processes. Although most frequently applied to thermal death rates, D values can also be used to express the rate of death due to other lethal affects such as acid and irradiation. Note, it is questionable whether the first order death kinetics (log‐linear) with temperature dependence is always appropriate for survival curve calculations. Although the log‐linear relationship has been the standard approach to thermal susceptibility determination, a ­ straight‐line relationship does not always occur when plotting experimental data. Complications arise because the microbial cell can become non‐viable for various reasons: disruption of the cell envelope, protein denaturation or nucleic acid damage. Shoulders and tails can be obtained when plotting log (N/N0) survival curves. This may be for reasons such as cell clumping, mixed populations with varying thermal sensitivities, changes in resistance during treatment or inactivation of a number of essential loci. Various mathematical models are available to address this issue, the derivation of which is outside the scope of this book. However a useful Excel™ add‐on called ‘GInaFIT’ is available as freeware (see www.ginafit.be) that enables the user to try nine different mathematical fits to their data from within Excel (Geeraerd et al. 2005). The models are (i) classical log‐linear curves; (ii) curves with a shoulder; (iii) curves showing tailing; (iv) survival curves with both shoulder and tailing; (v) concave curves; (vi) convex curves; (vii) convex/concave curves followed by tailing; (viii) biphasic inactivation kinetics; and (ix) biphasic inactivation kinetics preceded by a shoulder. Plotting the D value against temperature can be used to determine the change in temperature required to obtain a tenfold increase (or decrease) in the D value. This co‐efficient is called the z value (Figure 2.15). The integrated lethal value of heat received by all points in a

Basic aspects

65

Viable count (log10 cfu/g)

8 6 4

1 log D value

2

2.9 minutes 0 0

2

4

6 Time (minutes)

8

10

(a) Viable count (log10 cfu/g)

8 6 1 log

4

D value 5.2 minutes

2 0 0

5

10 Time (minutes)

D value (log10)

(b) 3 2.5 2 1.5 1 0.5 0 –0.5 –1 –1.5 –2

Liquid whole egg Egg yolk Meat

50

(c)

15

55

60

65

Temperature (°C)

Figure 2.13  (a) Death rate of Escherichia coli O157:H7 in beef at 60°C. (b) Death rate of Staphylococcus aureus in liquid egg at 60°C. (c) D value for Salmonella Enteritidis in eggs and meat. Source: FSIS (1998); Fazil et al. (2000a).

container during processing is designated Fs or Fo. This represents a measure of the capacity of a heat process to reduce the number of spores or vegetative cells of a given organism per container. When we assume instant heating and cooling throughout the container of spores, vegetative cells or food, Fo may be derived as follows:

Fo

Dr log a log b

66

The microbiology of safe food

Table 2.3  Variation in microbial heat resistance of micro‐organisms according to test conditions.

Organism

Medium

pH

Aeromonas hydrophila Brucella species Bacillus cereus (spores) (toxin destruction; diarrhoeal/emetic) Bacillus coagulans

Saline – –

– –

Buffer Red pepper Buffer “

4.5 “ 7.0 4.0

– Buffer Beef – –

– 7.0 – – –

–

– –

Bacillus licheniformis Bacillus stearothermophilus Bacillus subtilis Campylobacter jejuni Clostridium butyricum Clostridium perfringens (spores) (toxin) Clostridium botulinum type A and B proteolytic strains spores Type E and non‐proteolytic types B & F (toxin destruction) Clostridium thermosaccharolyticum Desulphotomaculum nigrificans Escherichia coli O157:H7

Listeria monocytogenes Salmonella Enteritidis Salmonella Senftenberg Staphylococcus aureus (toxin destruction) Streptococcus Group D Yersinia enterocolitica Saccharomyces cerevisiae Zygosaccharomyces bailii

– – Beef Apple juice Growth, 23°C Growth, 37°C Beef Liquid whole egg Beef –

Cured meat Saline “ “ “

3.6 4.5

–

4.5 “ “

Temperature (°C)a

D value (minutes)

51.0 65.5 100 100 56.1/121 110 “ 110 “ 120 100 50 50 100 100 98.9 90 121.1

8.08–122.8 0.1–0.2 5.0 2.7–3.1 5/Stable 0.064–1.46 5.5 0.27 0.12 4.0–5.0 11.0 0.88–1.63 5.9–6.3 0.1–0.5 0.3–20.0 26–31 minutes 4 0.2

82 85 120 120 62.8 58 58 58 58 62.0 62.8 62.0 65.5 98.9

0.49–0.74 2.0 3–4 2–3 0.47 1.0 2.5 1.6 5.0 2.9–4.2 0.06 2.65 0.2–2.0 >2 hours

70 60.0 60

2.95 0.4–0.51 22.5b 0.4c 14.2b

“ “

z value 5.22–7.69 – 6.9 6.1 – – – – 10 6.0–6.4

3.8 7.2 10 5.6–10.7 4.0–6.2 7.2–10 4.65 4.8 4.8 4.8 4.8 5.98 3.30 5.91 4.8–5.4 (approx. 27.8) 10 4.0–5.2 5.5 3.9 –

 To convert to °F use the equation °F = (9/5)°C + 32. As a guidance: 0 °C = 32 °F, 4.4 °C = 40 °F, 60 °C = 140 °F.  Ascospores. c  Vegetative cells. a

b

Where a  =  number of cells in the initial population, and b  =  number of cells in the final population. See Table 2.4 for an example of the effect of cooking temperature on the survival of E. coli O157:H7 in beef using D and z values.

Basic aspects Inoculum grown 25°C D 60 = 0.74 min 37°C D 60 = 1.02 min

67

Pre-treatment heat shock 45°C, 5 minutes D 60 = 1.1 minutes

Control Inoculum grown 30°C, 24 hours Suspended in peptone salt dilution fluid Enumerated on TSA, aerobically, 30°C, 7 days D 60 = 0.79 min

Enumeration: Incubation 4 days D 60 = 0.80 minutes Incubation 37°C D 60 = 0.62 minutes

Enumeration: MPN-cooked meat broth D 60 = 1.24 minutes MPN TSB D 60 = 1.22 minutes

Figure 2.14  Changes in Escherichia coli O157:H7 D value at 60°C with inoculum and recovery conditions. Source: Adapted from Stringer et al. (2000).

1 0.5

Log10 D value

0 –0.5 1 log

–1 –1.5 –2

Z value

–2.5 8°C

–3 –3.5 50

60 70 Temperature (°C)

80

Figure 2.15  Campylobacter jejuni z value in lamb cubes.

12‐D Concept

The 12‐D concept refers to the process lethality requirement, which has for a long time been used in the canning industry. It implies that the minimum heat process should reduce the probability of survival of the most resistant Cl. botulinum spores to 10−12. Since Cl. botulinum spores do not germinate and produce toxin below pH 4.6, this concept is observed only for foods above this pH value.

Fo

Dr log a log b



Fo

0.21 log 1 log 10

12



68

The microbiology of safe food Table 2.4  Effect of cooking temperature on the survival of Escherichia coli O157:H7.

Temperature (°C)

Time for viable count to be reduced by 6 log cyclesa (minutes)

58.2 62.8 67.5

28.2  2.8  0.28

Source: Data derived from D62.8 = 0.4, z = 4.65. a  For example, 107 cfu/g to 10 cfu/g.



Fo

0.21 12 2.52

Processing for 2.52 minutes at 250 °F (121 °C) should reduce the Cl. botulinum spores to one spore in 1 million containers (1012). When it is considered that some flat‐sour spores have Dr values of about 4.0 and some canned foods receive Fo treatments of 6.0–8.0, the potential number of Cl. botulinum spores is reduced even more (See Table 2.3). Application of D Value Concept in Pasteurisation

Pasteurisation treatment aims to reduce the number of pathogenic spoilage organisms by a set amount (frequently a 6 log reduction) and to ensure that the product formulation and storage conditions inhibit the growth of any surviving cells during the intended shelf life of the product. For more detail, see Section 3.2. For example, if the D value of a target organism is 0.5 minutes at 70°C and the z value was 5°C, a process of 3 minutes at 70°C would give a 6 log reduction. At 75°C the organism would die 10 times faster hence 0.3 minutes (20 seconds) at 75°C is an equivalent treatment giving a 6 log reduction. Since the determination of actual z values for each organism and product would require considerable work, assumed appropriate values are often used. Although this may be satisfactory for general quality assurance purposes, it is important to establish accurate values for the most appropriate micro‐organism present in the product. 2.8  Factors affecting microbial growth

Traditional ways to control microbial spoilage and safety hazards in foods are summarised in Table 2.5. How these are used in methods of food preservation is considered in more detail in Sections 3.4 and 3.5. In order to design adequate treatment processes an understanding of the factors affecting microbial growth is necessary. To complement this, Table 2.6 lists the ­common faults during processing that result in the growth of foodborne pathogens or toxin production. 2.8.1  Intrinsic and extrinsic factors affecting microbial growth

Food is a chemically complex matrix, and predicting whether or how fast micro‐organisms will grow in any given food is difficult. Most foods contain sufficient nutrients to support microbial growth. Several factors encourage, prevent or limit the growth of micro‐organisms in foods, the most important are aw, pH and temperature.

Basic aspects

69

Table 2.5  Methods of food preservation. Operation

Intended effect

Cleaning, washing Cold storage (below 8°C) Freezing (below −10°C) Pasteurising (60–80°C) Blanching (95–110°C) Canning (above 100°C) Drying Salting Syruping (sugars) Acidifying

Reduces microbial load Prevents the growth of most pathogenic bacteria; slows the growth of spoilage microbes Prevents growth of all microbes Kills most non‐sporing bacteria, yeast and moulds Kills surface vegetative bacteria, yeast and moulds ‘Commercially sterilises’ food; kills all pathogenic bacteria Stops growth of all microbes when aw $2.4 billion with 822.8 million tests, with molecular and immunological‐based methods accounting for about 35% of the total number of tests. Although they are more expensive per test, by definition they offer a faster throughput than conventional methods. The choice of test varies per organism. For example conventional methods predominate for Campylobacter species, whereas the opposite is true for E. coli O157. Around 68.5 million The Microbiology of Safe Food, Third Edition. Stephen J. Forsythe. © 2020 John Wiley & Sons Ltd. Published 2020 by John Wiley & Sons Ltd.

259

260

The microbiology of safe food

Salmonella tests are carried out annually, but the results may take several days. Hence there is still a need for even more rapid testing, which would benefit both food producers and consumers. This chapter reviews the various methods available to detect the major foodborne pathogens. Analysing food and environmental samples for the presence of food poisoning and food spoilage bacteria, fungi and toxins is standard practice for ensuring food safety and quality. Despite the considerable collection of detection methods that have been developed, the interpretation of results in food microbiology is far more difficult than is normally appreciated. It is not only the specificity and sensitivity of the method that needs to be appreciated, but also how representative was the sample that was analysed. This chapter considers the major groups of detection and characterisation methods available to the food microbiologist, and the issues of sampling plans and statistical representation of samples are covered in Chapter 7. The various reasons for caution in interpreting microbiological results include: • Micro‐organisms are in a dynamic environment in which multiplication and death of different species occur at differing rates. This means that the result of a test is only valid for the time of sampling. • Viable counts by plating out dilutions of food homogenate onto agar media can be misleading if no micro‐organisms are cultivated yet preformed toxins or viruses are present. For example, staphylococcal enterotoxin is very heat‐stable and will persist through the drying process in the manufacture of powdered milk. • Homogeneity of food is rare, especially with solid foods. Therefore, the results for one sample may not necessarily be representative of the whole batch. However, it is not possible to subject a whole batch of the food to such examination for micro‐organisms as there would be no product left to sell. • Colony counts are only valid within certain ranges and have confidence limits (Table 6.1). For of the reasons above, microbiological counts obtained through random sampling can only form part of the overall assessment of the product. There are a number of issues related to the recovery of micro‐organisms from food that must be addressed in any isolation procedure: 1 If solid food, then a liquidised homogenate is necessary for dilution purposes. 2 The target organism is normally in the minority of the microbial population. Table 6.1  Confidence limits associated with numbers of colonies on plates. 95% confidence intervals for the count Colony count

Lower

Upper

3 5 10 12 15 30 50 100 200 320

37% of all food processing plants, and over 30 million ATP monitoring tests are performed annually (Weschler, Strategic Consulting). One can only imagine the market for a real‐time test that covers at least the major foodborne pathogens. 6.2  Conventional methods

A number of steps are required in order to successfully isolate target organisms from food. • Choose representative samples to test the batch of ingredients/food. • Where practical, homogenise the food before sampling. Otherwise take representative samples from the different phases (liquid/solid). Volumes of food analysed are often 1 g when direct enumeration is used; or 25 g when a large sample size is required for presence/absence testing. Multiple numbers of samples may be required from each batch of food, depending on the appropriate microbiological criteria (Chapter 7). • Homogenise solid ingredients/food using a Stomacher™ or Pulsifier® machine. • A pre‐enrichment step may be required to allow any injured cells to repair their damaged membranes and metabolic pathways. Injury may have occurred during processing (cooking, desiccation, etc.). • Enrich the target organisms from within the mixed flora using a medium that encourages the growth of the target organisms and suppresses the growth of other micro‐organisms. Conventional methods are frequently plate counts obtained from homogenising the food sample, diluting and inoculating specific media to detect the target organism (Figure 6.1). The first step is normally preparing a 1:10 dilution of the food. The sample is usually homogenised in order to release attached micro‐organisms from the food surface. The methods are very sensitive and relatively inexpensive (compared with rapid methods), but require incubation periods of at least 18–24 hours for visible colony formation.

Methods of detection and characterisation

265

Pre-enrichment Recovery of injured cells 1–25 g of sample 1:10 dilution in pre-enrichment broth

Enrichment Suppress growth of non-target organisms

Subculture on selective agar Differential media to distinguish target from non-target isolates

Subculture on non-selective media (check isolate purity)

Biochemical and serological tests as appropriate Figure 6.1  General sequence of isolation of foodborne pathogens: presence/absence testing.

The target organism, however, is often in the minority of the food microbial flora and may be sublethally injured due to processing (cooking, etc.). Therefore, the above procedure is frequently modified to allow a recovery stage for sublethally injured cells, or to enrich for the target organism. For example, the recovery of Salmonella serovars from ready‐to‐eat foods occurs in the stages pre‐enrichment, enrichment, selection and detection, and uses a large sample size (25 g). The procedure is covered in more detail in Section  6.7.2. As referred to above, this approach is ‘bacteriological’ rather than ‘microbiological’ in that the presence of toxins, protozoa and viruses will not be revealed. Specific examples of methods for the detection of key target organisms are given in later sections. 6.2.1  Culture media

Conventional food microbiology requires the use of broths and agar plates to cultivate the ­target organism(s). These media must meet the organism’s nutritional and physiological requirements. Hence the media must be designed with sufficient protein, carbohydrates and minerals as well as a suitable pH, and incubated under favourable conditions of temperature and oxygen availability for an adequate length of time. In general terms, media whether as a broth or in a solid agar form may be (i) non‐selective and able to grow most organisms in a sample, (ii) selective to favour the growth of a target organism or (iii) semi‐selective and differential where the target organism is presumptively identified based on colony morphology including colour in the presence of other organisms. Such presumptive isolates require further confirmatory tests, which are frequently phenotypic (biochemical profiles). Selective media may be designed to suppress the growth of non‐target organisms while enabling the differential growth of the target organism. If the target organism outgrows the non‐target organism by >100‐fold in broth culture, there is a good chance that it will be isolated as a pure culture on plating out. Ideally selective media will be non‐inhibitory to the target organism. This is not always achieved, but preferably the media will recover more than 50% of the initial population.

266

The microbiology of safe food E. coli β-glucuronidase

β -D-glucuronide 4-methyl-umbelliferyl

Coliforms

4-methyl-umbelliferone (fluorescence, UV 365 nm)

β-galactosidase

β-D-galactopyranoside o-nitrophenyl

o-nitrophenol (bright yellow)

Enterococci β-glucosidase

β -D-glucoside-4-methyl-umbelliferyl

4-methyl-umbelliferone (fluorescence, UV 365 nm)

Figure 6.2  Fluorogenic substrates for specific detection of food pathogens.

Fermentation of carbohydrates is frequently used in differential agar. Violet red bile lactose agar (VRBA) contains the indicator neutral red causing lactose‐fermenting organisms (commonly referred to as ‘coliforms’) to acquire a pink‐red colour. This differentiates them from other bile‐resistant Gram‐negative organisms. Although pH indicators have a long history of use, they suffer in that the surrounding medium may change colour and the observer may be unable to select the target organism when large numbers of colonies have grown on the plate. The advantage of incorporating fluorogenic and chromogenic substrates into growth media is that they generate brightly coloured or fluorescent compounds after bacterial metabolism (Plate 5) (Manafi 2000). Therefore, colonies of the target organism can be spotted on a mixed flora plate, even when considerably outnumbered. The main fluorogenic enzymes substrates in  use are based on 4‐methylumbelliferone, such as 4‐methylumbelliferyl‐β‐D‐glucuronide (MUG) (Figure 6.2). Although these compounds are very specific for the distinguishing enzyme activity, they can diffuse into the medium making the target colony less noticeable in a mixed flora. In addition, they require the medium to be slightly alkaline, and require a UV light source for visualisation. Consequently, the application of MUG subtrates has been limited. Chromogenic substrates commonly used are indoxyl substrates, for example 5‐bromo‐4‐chloro‐3‐indolyl‐β‐D‐ glucuronic acid (BCIG), which in the presence of oxygen form coloured aglycones; examples are shown in Plate 4. In contrast to MUG substrates, these do not diffuse into the agar. TBX agar (Oxoid Thermo Fisher, Merck and LabM) is an example of a chromogenic agar for the detection

Methods of detection and characterisation

267

of E. coli, which is based on the splitting of BCIG by β‐D‐glucuronidase enzyme activity to form blue‐green colonies. Unlike the majority of E. coli strains, those in the serovar group O157:H7 do not ferment sorbitol or rhamnose in the presence of sorbitol. They are β‐D‐ glucuronidase negative and do not grow at 45.5°C. This has enabled the design of specific chromogenic agars such as CHROMagar O157 (CHROMagar) and Fluorocult E. coli O157:H7 (Merck), which can differentiate between the two types of E. coli based on sugar fermentation by colony colour. Although very popular with media manufacturers, the indoxyl substrates require the presence of oxygen or other oxidants for colour formation, and may produce toxic intermediates. Chromogens and fluorogens (ALDOL™) that do not require oxidation have been developed by Biosynth AG (Switzerland). These undergo intramolecular aldol condensation to form insoluble dyes. Consequently they can be incorporated into media for use under both aerobic and anaerobic incubation conditions, which was not feasible with indoxyl substrates. Media such as the modified semi‐solid Rappaport–Vassiliadis medium and Diagnostic Semi Solid Salmonella (DIASALM) agar have used bacterial motility as a means to enrich for the target organism. This principle has been applied to the improved detection of Salmonella serovars (as per above examples), Campylobacter and Arcobacter species. The semi‐solid Rappaport medium isolates motile Salmonella as they migrate through the medium ahead of competing organisms. This medium however will not isolate non‐motile Salmonella strains. Compact Dry and the Petrifilm® system (manufactured by 3M) are alternatives to the conventional agar plate. The Petrifilm system uses a dehydrated mixture of nutrients and gelling agent on a film. The addition of 1 ml of sample rehydrates the gel, which enables the colony formation of the target organism. Colony counts are performed as per the standard agar plate method. The throughput of samples is estimated to be double that of conventional agar plates. Petrifilm systems are available for various applications including APCs, yeast, coliforms and E. coli (Plate 6). 6.2.2  Sublethally injured cells

Sublethal injury implies damage to structures within the cells that causes some loss or alteration of cellular functions, the leakage of intracellular material and making them susceptible to selective agents. Changes in cell wall permeability can be demonstrated by the leakage of compounds from the cytoplasm (increase absorbance at 260 nm of culture supernatants) and the influx of compounds such as ethidium bromide and propidium iodide. Conditions that can generate sublethally injured cells include: • moderate heat: pasteurisation; • low temperature: refrigeration; • low water activity: dehydration; • radiation: gamma rays; • low pH: organic and inorganic acids; • preservatives: sorbate inclusion; • sanitisers: quaternary ammonium compounds; • pressure: high hydrostatic pressure; • nutrient deficiencies: clean surfaces. Cells in the exponential phase of growth are generally less resistant than cells in the stationary phase due to the synthesis of stress resistance proteins. ‘Metabolic’ injury is often taken as the inability to form colonies on minimal salt media while retaining colony‐forming ability on complex nutrient media, whereas ‘sublethal’ injury can be taken as the inability to proliferate or survive in media containing selective agents that

268

The microbiology of safe food

have no apparent inhibitory action upon non‐stressed cells. Injury is reversible by repair but only if the cells are exposed to favourable resuscitation conditions such as a non‐selective nutrient‐rich medium under optimal growth conditions. In practical analytical food microbiology the phenomenon of injury may present considerable problems as many of the physical treatments including heat, cold, drying, freezing, osmotic activity and chemicals (disinfectants, etc.) may generate injured cells causing variations in plate counts. The injured cells may remain undetected as selective media usually contain stressing ingredients such as increasing salt concentrations, deoxycholate lauryl sulphate, bile salts, detergents and antibiotics. The injured cells are ‘viable’ but are not metabolically active enough to achieve cell division. Subsequently microbiological examination for quality control can indicate low plate counts, when in fact the sample contains a high number of injured cells. An example of the difference between plate counts on selective and non‐selective agar can be seen in Figure 3.7 where food pathogens have been exposed to high pressure. In food and beverage products, once the stress‐causing injury is removed, these injured cells are often able to recover. The cells regain all of their normal capabilities, which include pathogenic and enterotoxin properties. Therefore, important food poisoning organisms may be undetected by analytical testing, but may cause a major food poisoning outbreak. For these reasons, substantial efforts need to be made to develop improved analytical procedures that will detect both injured and uninjured cells. In Salmonella detection (Section 6.7.2) the sample is incubated overnight in buffered peptone water (BPW) or lactose broth to allow injured Salmonella cells to recover and multiply to detectable levels. However, it is uncertain if BPW is the best recovery medium since other organisms can suppress the growth of low numbers of salmonellae and also there is the problem of ‘how do you know if injured Salmonella were present in the sample if you do not detect a colony on a plate?’. For other organisms that might be sublethally injured it has been recommended that food samples should be resuscitated in a non‐inhibitory medium for an hour or two, allowing injured cells to resuscitate yet prevent the population size increasing. This generalised approach is far from optimised and leaves plenty of opportunity for oversight in the detection of potentially pathogenic food poisoning organisms. Hence such techniques need to be appropriately validated. Extreme environments usually kill the majority of the bacterial population, and can result in the selection of resistant mutants. It is possible that stress induces hypermutability and leads to the greater chance of survivors. The cross‐protection effect in which exposure to one stress induces the resistance to another stress is of particular concern for food processing. For example, acid‐adapted E. coli O157:H7 is more heat tolerant. Similarly, heat‐shocked Listeria monocytogenes increases its resistance to ethanol and salt. 6.2.3  Viable but non‐culturable bacteria (VBNC)

It has been proposed that many foodborne bacterial pathogens are able enter a dormant state (Zhao et al. 2017). Such cells are not culturable by conventional methods, but remain viable (as demonstrated by substrate uptake) and virulent. Hence the term ‘viable but non‐culturable’ or VBNC was derived. This phenomenon has been shown in Salmonella serovars, Campylobacter jejuni, E. coli, Vibrio vulnificus and V. cholerae. For example, in the human intestine previously non‐culturable vibrios were shown to regain their ability to multiply (Colwell et  al. 1996). Therefore, VBNC bacterial pathogens pose a potential threat to health and are of considerable concern in food microbiology since a batch of food might be released due to the apparent negative presence of pathogens, yet contain infectious cells.

Methods of detection and characterisation

269

VBNC bacterial cells maintain their cell integrity but exhibit dwarfing and have coccoid morphology. The VBNC state may be induced due to a number of extrinsic factors such as temperature changes, low nutrient level, osmotic pressure, water activity and pH. Therefore, foodborne pathogens may enter the VBNC state during food processing techniques, such as high temperature, high pressure, disinfectant, preservation and low temperature storage. Consequently there will be a food safety risk as the organisms will not be detectable by the standard accredited detection methods. Hence current methods may not be recovering all the pathogens from foods and water. Therefore alternative end‐detection methods need to be further developed, such as those based on immunology and DNA sequences. The VBNC concept is not accepted by all microbiologists. Some argue that it is only a matter of time until we design the most appropriate recovery media and others that the cells have self‐destructed due to an oxidative burst causing DNA damage (Barer 1997; Barer et al. 1998; Bloomfield et al. 1998). 6.3  Rapid sampling methods

New pathogen identification technologies are faster than conventional methods, and increasingly automated. No single method is appropriate for all circumstances, so the selection of the most appropriate method is necessary. Conventional procedures are by nature labour‐intensive and time‐consuming. Therefore, a plethora of alternative, rapid methods have been developed to shorten the time between taking a food sample and obtaining results. These methods aim to either replace the conventional enrichment step with a concentration step (i.e. immunomagnetic separation [IMS]) or to replace the end‐detection method with one that requires a shorter time period (i.e. ATP bioluminescence). Major improvements have been in three areas: 1 sample preparation; 2 separation and concentration of target cell, toxins or viruses; 3 end‐detection. Sometimes a rapid technique will involve one or more of the above aspects, i.e. the hydrophobic grid membrane filter (HGMF) method both concentrates the organisms and enumerates on specific detection agar media. 6.3.1  Sample preparation

Agar slides containing selective or non‐selective agar can be pressed against the surface to be examined and directly incubated. This obviates the need for sampling and the errors inherent in releasing organisms from cotton wool swabs. Samples can be put onto agar surfaces in a spiral format allowing several effective dilutions to be countable on one plate (Plate 7). Another improvement in recent years in sample preparation is the automatic diluter, which simply enables the operator to take a food sample of ~10 g, and then an appropriate volume of diluent is added to give an accurate 1:10 dilution factor. 6.3.2  Separation and concentration of target

Separation and concentration of target organisms, toxin or viruses can shorten the detection time and improve specificity of a test procedure. Common methods include: • IMS; • Direct Epifluorescent Filter Technique (DEFT); • HGMF.

270

The microbiology of safe food

Immunomagnetic separation

IMS is increasingly being used as it significantly reduces the detection period due to the elimination of a culture‐enrichment step. It uses superparamagnetic particles (3–5 μm diameter) that contain γ‐Fe2O3 and are coated with antibodies against the target organism. Hence the target organism is ‘captured’ in the presence of a mixed population due to the antigen–antibody specificity. This has removed the need for an overnight enrichment broth incubation period in Salmonella isolation, and for E. coli O157 the enrichment step is only 6 hours (see Section  6.7.4). A generalised procedure is given in Figure  6.3, and Plate  8. Commercially available IMS kits target key food and water pathogens: Salmonella serovars, E. coli O157:H7, Mixed microbial culture, for example overnight pre-enrichment broth

Addition of antibody-coated paramagnetic particles

Incubate ca. 10 minutes Antibody beads bind to target organism

Placement of magnet temporarily immobilises target organism, facilitating removal of non-target cells, via pipetting, and so on

Resuspension and washing of antibody-bead bound target cells

Pipetted onto selective media, ELISA, DNA techniques, and so on Figure 6.3  Immunomagnetic separation (IMS) technique.

Methods of detection and characterisation

271

Table 6.3  Applications of immunomagnetic separations (IMS). Organism

Application

Escherichia coli O157 Salmonella serovars Listeria monocytogenes Staphylococcus aureus Cryptosporidium parvum Legionella species Yersinia pestis Chlamydia trachomatis HIV Erwinia chrysanthemi Erwinia carotovora Saccharomyces cerevisiae Mycobacterium species

Food and water microbiology

Clinical microbiology

Plant pathogen detection Biotechnology

Source: Adapted from Safarik et al. (1995).

L. monocytogenes and Cryptosporidium (Table  6.3). IMS can enrich for sublethally injured micro‐organisms that would otherwise be missed using the standard enrichment broth and plating procedures. These organisms might be killed in the enrichment broth due to changes in cell wall permeability. Dead cells can be detected using a combined IMS and PCR procedure. IMS can be combined with almost any end‐detection method: culture media, ELISA and DNA probe. The IMS Salmonella detection method is as efficient as the enrichment broth selection stage, which is the most efficient of the ISO procedures (Mansfield and Forsythe 1996, 2000a; Section 6.7.2). The selective enrichment step (overnight incubation) is replaced with the IMS (10 minutes). Hence the technique reduces the total time required for sampling and detection by 1 day. In addition IMS can have a greater recovery of stressed Salmonella cells than ISO protocols. Commonly the first step in isolating a pathogen is pre‐enrichment to aid the recovery of damaged cells, followed by enrichment to encourage the growth of the target organism and suppress non‐target cells. Yet at each stage only a small aliquot is taken, leaving the bulk of the microbial culture to be discarded. To increase sensitivity, a variation on the IMS technique is to circulate the total volume over the magnet during incubation. This approach has been demonstrated by Pathatrix (Matrix MicroScience) to be very effective and rapid for detecting the presence of Salmonella and E. coli O157 from a variety of food samples. Direct epifluorescent technique (DEFT) and hydrophobic grid membrane filters (HGMF)

Membrane filters can be used to shorten the overall detection time because: 1 They can concentration the target organism from a large volume to improve detection limits. 2 They remove growth inhibitors. 3 They transfer organisms to a different growth media without physical injury through centrifugation and resuspension. The membranes can be made from nitrocellulose, cellulose acetate esters, nylon, polyvinyl chloride and polyester. Because they are only 10 μm in thickness they can be directly mounted on a microscope and the cells visualised.

272

The microbiology of safe food

Milk, 2 ml

Add trypsin, 0.5 ml + Triton ×100, 2 ml

Incubate 50°C, 20 minutes

Filter through Nuclepore MF

Add acridine orange, 2 minutes

Wash + air dry

Observe under epifluorescence microscope

Dead cells fluoresce green

Living cells fluoresce red

Figure 6.4  Direct Epifluorescent Filter Technique (DEFT) for the detection of bacteria in milk.

The DEFT method concentrates cells on a membrane before staining with acridine orange (Figure  6.4). Acridine orange fluoresces red when interchelated with RNA and green with DNA. Subsequently viable cells fluoresce orange‐red, whereas dead cells fluoresce green. The DEFT count has gained acceptance as a rapid, sensitive method for enumerating viable bacteria in milk and milk products. The count is completed in 25–30 minutes and detects as few as 6 × 103 bacteria per ml in raw milk and other dairy products, which is three to four orders of magnitude better than direct microscopy. Because it is a microscopic technique one is able to distinguish whether the micro‐organisms present are yeasts, moulds or bacteria. HGMF is a filtration method that is applicable to a wide range of micro‐organisms (Entis and Lerner 1996, 2000). The pre‐filtered food sample (to remove particulate matter >5 μm) is filtered through a membrane filter, which traps micro‐organisms on a membrane in a grid of 1600 compartments, due to hydrophobicity effects. The membrane is then placed on an appropriate agar surface and the colony count determined after a suitable incubation period.

Methods of detection and characterisation

273

6.4  Rapid end‐detection methods

Improvements in end‐detection methods include: • immunoassays; ELISA, latex agglutination; • ATP bioluminescence; • gene probes linked to PCR. 6.4.1  ELISA and antibody‐based detection systems

ELISAs are widely used in food microbiology. ELISA is most commonly performed using mono‐ and polyclonal antibody‐coated microtitre trays to capture the target antigen (Figure 6.5; Plate 9). The captured antigen is then detected using a second antibody (usually monoclonal for specificity) that is conjugated to an enzyme. The addition of the enzyme substrate enables the presence of the target antigen to be visualised. ELISA methods offer considerable specificity and can be automated. Tray coated with monoclonal antibodies specific against target organism

Add sample

Target organism binds to antibodies

Washing procedure to remove nontarget organisms

Secondary antibody labelled with horse radish peroxidase or alkaline phosphatase to give colorimetric reaction upon addition of substrate

Figure 6.5  Enzyme‐linked immunosorbent assay (ELISA).

274

The microbiology of safe food

A wide range of ELISA methods are commercially available, especially for Campylobacter, Salmonella serovars and L. monocytogenes. The technique generally requires the target organism to be 106 cfu/ml, although a few tests report a sensitivity limit of 104. Hence the conventional pre‐enrichment and even selective enrichment might be required prior to testing. The VIDAS system (bioMérieux) has pre‐dispensed disposable reagent strips. The target organism is captured in a solid phase receptacle coated with primary antibodies and then transferred to the appropriate reagents (wash solution, conjugate and substrate) automatically. The end‐ detection method is fluorescence, which is measured using an optical scanner. The VIDAS system can be used to detect most major food poisoning organisms. The bioMérieux VIDAS® UP E. coli O157 uses bacteriophage‐binding sites for specific capture and detection of the target organism. 6.4.2  Reversed passive latex agglutination

Reversed passive latex agglutination (RPLA) is used for the detection of microbial toxins such as the Shiga toxins (from Shigella dysenteriae and enterohaemorrhagic E. coli), E. coli heat‐labile (LT) and heat‐stable (ST) toxins (Figure 6.6). Latex particles are coated with rabbit antiserum,

Latex particles coated with

V-shaped well microtitre plate well

antibodies against target antigen (e.g. bacterial toxin).

Absence of antigen

Particles settle to form dense dot

Presence of antigen

Particles remain dispersed in solution

Appearance viewed from above microtitre plate:

Figure 6.6  The principle of reversed passive latex agglutination (RPLA).

Methods of detection and characterisation

275

which is reactive towards the target antigen. Therefore, the particles will agglutinate in the presence of the antigen forming a lattice structure. This settles to the bottom of a V‐bottom microtitre well and has a diffuse appearance. If no antigen is present then a tight dot will appear. 6.4.3  ATP bioluminescence techniques and hygiene monitoring

The molecule ATP is found in all living cells (eucaryotic and procaryotic). Therefore, the presence of ATP indicates that living cells are present. The limit of detection is ~1 pg ATP, which is equivalent to approximately 1000 bacterial cells based on the assumption of 10−15 g ATP per cell. Since a sample is analysed in seconds–minutes it is considerably faster than conventional colony counts for the detection of bacteria, yeasts and fungi. Additionally food residues that act as the loci of microbial growth will also be detected rapidly. Hence ATP bioluminescence is primarily used as a hygiene monitoring method and not for the detection of bacteria per se. In fact, in a food factory, there will not necessarily be a correlation between plate counts and ATP values for identical samples, since the latter will additionally detect food residues (Plate 11). ATP is detected using the luciferase–luciferin reaction: ATP luciferin Mg 2

oxyluciferin ADP light 562 nm

The firefly (Photinus pyralis) is the source of the luciferase and the reagents are formulated such that a constant yellow‐green light (max. 562 nm) is emitted. ATP bioluminescence measurement requires a series of steps to sample an area (usually 10 cm2). Many instruments currently have the extractants and luciferase–luciferin reagents encased with the swab in a ‘single‐shot’ device. This saves the preparation of a series of reagents and the associated pipetting errors. ATP bioluminescence can be used as a means of monitoring the cleaning regime especially at a Critical Control Point of a HACCP procedure (Section 9.5). See Table 6.4 for a list of examples of ATP bioluminescence applications. There are three food production processes that are not amenable to ATP bioluminescence. These are milk powder production, flour mixes and sugar because the cleaning procedures do not remove all food residues. It has been noted that the luciferase–luciferin reaction can be affected by residues containing sanitisers (free chlorine), detergents, metal ions, acid and alkali pH, strong colours, many salts and alcohol (Calvert et al. 2000). Hence the commercially available ATP bioluminescence Table 6.4  Application of ATP bioluminescence in the food industry. 1 Hygiene monitoring 2 Dairy industry • Raw milk assessment • Pasteurised milk, shelf life prediction • Detection of antibiotics in milk • Detection of bacterial proteases in milk 3 Assessing microbial load • Poultry carcasses • Beef carcasses • Mince meat • Fish • Beer

276

The microbiology of safe food

kits may contain detergent neutralisers such as lecithin, Tween 80 and cyclodextrin. Enhancement and inhibition of the luciferase–luciferin reaction can lead to errors of decision and hence an ATP standard should be used to test the activity of the luciferase. An improvement has been the use of caged ATP as an internal ATP standard, whereby a known quantity of ATP is released into a solution upon exposing the swab to high intensity blue light (Calvert et al. 2000). 6.4.4  Protein detection

An alternative to ATP detection for hygiene monitoring is the detection of protein residues. The method uses the Biuret reaction (Figure 6.7). There are many simple kits available that are able to detect ~50 μg protein on a work surface within 10 minutes. The surface is sampled either by swabbing or a dipstick and reagents added. The development of a green colour indicates a clean, hygienic surface, grey is caution and purple is ‘dirty’. The technique is more rapid than conventional microbiology and less expensive than ATP bioluminescence since no capital equipment is required. It is however less sensitive than ATP bioluminescence. 6.4.5  Flow cytometry

Flow cytometry is based on light scattering by cells and fluorescent labels that discriminate the micro‐organisms from background material such as food debris (Figure 6.8). Fluorescence‐ labelled antibodies have been produced for the major food poisoning organisms such as Salmonella serovars, L. monocytogenes, C. jejuni and Bacillus cereus. The level of detection of bacteria is limited to approximately 104 cfu/ml due to interference and autofluorescence by food particles. Fluorescent labels include fluorescein isothiocyanate (FITC), rhodamine isothiocyanate and phycobiliproteins such as phycoerythrin and phycocyanin. These emit light at 530 nm, 615 nm, 590 nm and 630 nm, respectively. Viable counts are obtained using carboxyfluorescein diacetate, which intracellular enzymes will hydrolysis releasing a fluorochrome. Fluorescent‐labelled nucleic acid probes, designed from 16S rRNA gene sequences, enable a mixed population to be identified at genus, species or even strain level. However, as the organism might be non‐culturable, it is uncertain whether the organism was viable in the test sample and this throws into doubt whether its detection is of any significance. The method has been used for the detection of viruses in seawater (Marie et al. 1999). 6.4.6 Biosensors

Biosensors are analytical devices that integrate biological material for target recognition with a physicochemical transducer. They are currently the fastest growing pathogen detection technology and have the promise of multiple sample analysis, in real time with high specificity and Protein chain – N Protein + Cu2+ + nitrate

N – Protein chain Cu+

OH– Protein chain – N Purple colour Figure 6.7  The Biuret reaction.

N – Protein chain

Methods of detection and characterisation

277

Sample

Laser

Fluorescence detector

Deflection plates +

–

Left collection tube

Right collection tube

Waste Figure 6.8  Flow cytometry with cell sorting.

sensitivity (Vanegas et al. 2017). The advantages of small‐scale devices are reduced cost per unit, small sample volume requirement, shorter analysis times and possible multitarget analysis. Currently many biosensor systems use either DNA probes or specific antibodies to give the high specificity. A greater variety of end‐detection methods are used: optical, electrochemical and piezoelectric. Optical biosensors are very sensitive and specific, with methods based on fluorescence and the more recent technology of surface plasmon resonance (SPR) (refractive index measurement) being developed. Although optical systems may be more sensitive, electrochemical‐based biosensors detecting current or impedance changes may be preferable for turbid samples. Cell‐based biosensors based on mammalian and higher eucaryotic cells or cell components for the detection of bacterial pathogens and toxins are emerging (Banerjee and Bhunia 2009). Mammalian cells generate the initial response, which is converted via an electrochemical or optical system to a detectable signal. Examples include gangliosides to detect E. coli LT‐II toxin, E‐cadherin for the detection of L. monocytogenes and β1 integrins for Yersinia enterocolitica. Cell receptors can be the ligands for a number of pathogens and toxins and are therefore used as the detector. There is a demand by industry for affordable high‐throughput detection methods (Hyytiä‐ Trees et al. 2007). However, only a few biosensors have been developed for foodborne pathogens that demonstrate high specificity, quantitative sensitivity and rapid response time in complex matrices (Valderrama et al. 2016). Few, if any, have been validated by international accreditation organisations such as AOAC International. Early examples include a conductometric biosensor

278

The microbiology of safe food

to detect both E. coli O157:H7 and Salmonella serovars, and a disposable immunosensing amperometric strip for the quantitative detection of E. coli O157:H7 (Muhammad‐Tahir and Alocilja 2003; Lin et al. 2008). The most common techniques for the development of handheld quantitative biosensors for pathogens are SPR, quartz crystal microbalance (QCM), cantilever‐based sensors and electrochemical impedance spectroscopy (EIS). They are comparable in that the target binds to immobilised antibodies on the sensor surface and this causes a direct measurable signal. In SPR, when the target molecule binds to the gold sensor surface it causes a change in the angle of reflected light. A SPR sensor can simultaneously detect E. coli O157:H7, Salmonella Typhimurium, L. monocytogenes and C. jejuni (Taylor et al. 2006). QCM sensors are composed of a thin quartz disc with electrodes. The key aspect is that an oscillating electric field is applied to the disc, which induces an acoustic wave with a specific resonance frequency. Consequently, by coating the disc with a capture layer (antibodies, nucleic acids, etc.) it will then respond to the presence of the target by a change in the resonance frequency. These have been developed for S. Typhimurium, B. cereus and L. monocytogenes. Cantilever‐based sensors using antibodies are also being developed. There are essentially two modes of cantilever. In the static mode, the cantilever bending of the cantilever on binding the target is measured. Whereas the dynamic mode is similar to QCM sensors in that resonance frequency changes are monitored to determine when target molecules bind. There is considerable interest in using cantilevers for detection of pathogenic micro‐organisms. For example, Campbell and Mutharasan (2007) detected E. coli O157:H7 at 1 cell/ml without the conventional requirement of cultivation to increase target cell numbers. Detection times of 10 minutes for 10 E. coli O157:H7 cfu/ml have been given. It should be noted that these experiment are under ideal laboratory conditions in the absence of an interfering food matrix. Antibody microarrays can analyse many samples in parallel for foodborne pathogenic bacteria and biomolecules (Gehring et al. 2008; Karoonuthaisiri et al. 2009). However, with a few exceptions, most current methods do not have particularly low detection limits and may only equate to a conventional ELISA method: 105–107 cfu/ml of target organism. Hence, in general biosensors are not yet comparable with conventional methods for detection limit. One SRP sensor that has been commercialised is Spreeta™. These small (15 cm × 8 cm) sensors are very portable due to their low weight (600 g) and power source (9 V battery). They have been used for Campylobacter, E. coli and L. monocytogenes detection (Nanduri et al. 2007; Wei et al. 2007). Quantum dots are semiconductor nanomaterials, and therefore only nanometres in size. Specific antibodies can be conjugated to them to detect specific bacteria. Quantum dots characteristically have high fluorescence efficiency and sensitivity along with wide and continuous absorption spectra and narrow emission spectra. Highly fluorescent quantum dots (i.e. CdSe/ ZnS) have been used for the simultaneous detection of E. coli O157:H7 and S. Typhimurium. This is achievable as the two organisms have different emission wavelengths but the same excitation wavelength. They have also been used for the detection of mycotoxins. Such devices can be incorporated into a rapid lateral flow device and have a potential application in smart food packaging indicating the presence of specific microbial pathogens and toxins. 6.4.7  Impedance (Conductance) microbiology

Impedance microbiology is also known as conductance microbiology: impedance is the reciprocal of conductance and capacitance. It can rapidly detect the growth of micro‐organisms by two different methods (Silley and Forsythe 1996; Plate 10). 1 Directly due to the production of charged end‐products. 2 Indirectly from carbon dioxide liberation.

Methods of detection and characterisation

279

In the direct method, the production of ionic end‐products (organic acids and ammonium ions) in the growth medium causes changes in the conductivity of the medium. These changes are measured at regular intervals (usually every 6 minutes) and the time taken for the impedance value to change is referred to as the ‘time to detection’. The greater the number of organisms, the shorter the detection time. Hence a calibration curve is constructed and then the equipment can automatically determine the number of organisms in a sample. The indirect technique is a more versatile method in which a potassium hydroxide bridge (solidified in agar) is formed across the electrodes. The test sample is separated from the potassium hydroxide bridge by a headspace. During microbial growth carbon dioxide accumulates in the headspace and subsequently dissolves in the potassium hydroxide. The resultant potassium carbonate is less conductive and it is this decrease in conductance that is monitored. The indirect technique is applicable to a wide range of organisms including Staphylococcus aureus, L. monocytogenes, Enterococcus faecalis, Bacillus subtilis, E. coli, Pseudomonas aeruginosa, Aeromonas hydrophila and Salmonella serovars. Standard selective media or even an agar slant can be used for fungal cultures. The time taken for a conductance change to be detectable (‘time to detection’) is dependent upon the inoculum size. Essentially the equipment has algorithms that determine when the rate of conductance change is greater than the preset threshold. Initially the reference calibration curve is constructed using known numbers of the target organism. The microbial load of subsequent samples will be automatically determined. The limit of detection is a single viable cell since, by definition, the viable cell will multiply and eventually cause a detectable conductance change. Microbes frequently colonise an inert surface by forming a biofilm; for more detail see Section 8.5. Biofilms can be 10‐ to 100‐fold more resistant to disinfectants than suspended cultures and therefore the efficacy of disinfectants for their removal is very important. Impedance microbiology can be used to monitor microbial colonisation and efficacy of biocides. 6.5  DNA‐based molecular typing and proteomic methods

Typing micro‐organisms is an important tool in outbreak investigations and surveillance both at national and international levels. At any given time there may be several strains of a foodborne pathogen, plus non‐pathogenic variants circulating. Therefore, typing methods must have high discriminatory power in order to distinguish between related and unrelated isolates. Please note the term ‘related’ here is used in the context of relevance, and not in the phylogenetic context. A standardised method should be used that enables results to be comparable between national and international laboratories. Therefore, it can be used for international surveillance purposes. Until recently, determining the relatedness between bacteria relied on phenotyping and chemotyping methods. However, these are limited to well‐studied organisms, and growth conditions etc. must be carefully standardised to ensure reproducible and reliable results. Despite serotyping and phage typing being well‐recognised methods for Salmonella and E. coli, they have not been developed for many other organisms. Instead DNA‐ based methods of genotyping are more appropriate, being applicable across the microbial world. The major advantages of genotyping are: 1 DNA can be extracted from all organisms, including those that cannot be cultivated. 2 It is not dependent upon growth conditions. 3 Closely related strains can be distinguished. 4 Identical methods can be applied to different bacterial species. 5 DNA profiles can be digitised, and easily distributed. 6 DNA profiles can be used for comparative analysis.

280

The microbiology of safe food

The molecular typing techniques can be designed to target different areas of the genome. However, the stability of the target areas must be considered in case there is too much variability for typing purposes. The most common genotyping methods include: • pulsed‐field gel electrophoresis (PFGE); • DNA probe‐based hybridisations such as ribotyping; • PCR‐based methods, such as random‐amplified polymorphic DNA (RAPD); • sequence‐based methods, including multi‐locus sequence typing (MLST) and multiple‐ locus variable number tandem repeat analysis (MLVA). 6.5.1  Polymerase chain reaction (PCR)

PCR and qPCR (real‐time, quantitative PCR) offer the obvious advantages of a quick turnaround time to obtain a result, as well as high sensitivity and specificity of detection. However, there are some disadvantages that also need to be considered. These include the purchase and maintenance of specialised, expensive equipment; the need for trained personnel along with the use of extensive DNA clean‐up chemistries before addition to the PCR/qPCR reaction; and the need to avoid cross‐contamination. There is also still the need to pre‐incubate the samples as the target organism may be below the limit of detection (~100 cells) threshold. This will also reduce the issue of the inability of PCR to distinguish between live and dead cells, which is why it is not directly applied to the food. DNA probes for selected target organisms are commonly used in the food industry. However, the presence of a pathogen’s DNA in the food does not demonstrate the presence of a viable organism that is capable of multiplying to an infectious level. The PCR amplifies DNA to levels that can be detected (Figure 6.9). Specificity is obtained by the design of appropriate DNA probes to flanking regions unique to the target organism. The PCR technique uses a heat‐stable DNA polymerase, Taq or Pfu, in a repetitive cycle of heating and cooling to amplify the target DNA. The procedure is essentially: 1 The sample is mixed with the PCR buffer, Taq or Pfu polymerase, deoxyribonucleoside triphosphates and two primer DNA sequences (~20–30 nucleotides long). 2 The reaction mixture is heated to 94°C for 5 minutes to separate the double‐stranded target DNA. 3 The mixture is cooled to approximately 55°C for 30 seconds. During this time the primers anneal to the complementary sequence on the target DNA. 4 The reaction temperature is raised to 72°C for 2 minutes and the DNA polymerase extends the primers, using the complementary strand as a template. 5 The double‐stranded DNA is separated by reheating to 94°C. 6 The replicated target sites act as new templates for the next cycle of DNA copying. 7 The cycle of heating and cooling is repeated 30–40 times. The PCR will have amplified the target DNA to a theoretical maximum of 109 copies, though usually the true amount is less due to enzyme denaturation. The amount of amplified DNA is approximately 100 μg. 8 The DNA is stained with ethidium bromide or preferably for safety SYBR® Safe (In Vitrogen™) and visualised by agarose gel electrophoresis with UV transillumination at 312 nm.Negative control samples omitting DNA must be used in order to check for contamination of the PCR reaction by extraneous DNA. PCR is not usually directly performed on food samples for several reasons: • The technique would not distinguish between viable and non‐viable cells in the sample. • The reaction is inhibited by some food components. • The target cell number may be too low for detection.

Methods of detection and characterisation PCR cycle

281

Copies of target gene Target gene

0

1

5ʹ 3ʹ

3ʹ 5ʹ

DNA polymerase

Two primer sequences Heat 94°C, strand separation

Primer extension, 72°C

1

2

+

Heat 94°C, strand separation

Primer extension, 72°C 2

4

3

8

4

16

5

32 Repeat cycle of heating and primer extension for ca. 20 cycles Results in ca. 106 copies of the gene

Figure 6.9  The polymerase chain reaction (PCR).

Therefore, a major limitation of PCR is its inability to differentiate the DNA from dead and viable cells, hence the time‐consuming step of (pre‐)enrichment is used to recover viable organisms from food samples. One means of overcoming this limitation is using ethidium monoazide and propidium monoazide (PMA) to pretreat samples before DNA extraction to intercalate the DNA of dead cells in food samples, and then proceed with regular DNA preparation and qPCR.

282

The microbiology of safe food

By combining PMA treatment with qPCR (PMA‐qPCR), it is possible to directly detect viable cells of various bacterial pathogens in food samples. The basis of the method is that EMA and PMA dyes can only enter bacterial cells with damaged cell membranes. The internalised dyes will intercalate with nucleic acids, and also cross‐link to DNA after exposure to strong visible light. This intercalation to the DNA of dead bacterial cells will prevent subsequent DNA amplification by PCR, hence obtaining a positive PCR result can be interpreted as indicating the presence of viable target organisms. Apart from the principal method of amplifying a single target site, there are three other main PCR methods. In multiplex PCR different primers are used to amplify different DNA regions simultaneously. Whereas the previous methods give presence/absence for the target cell in a sample, real‐time PCR is quantitative. It uses a fluorescent tag on the amplicon. Consequently the increase in fluorescence is proportional to the number of target organisms in the sample; various methods have been developed including the molecular beacon and TaqMan™. Reverse transcriptase PCR (RT‐PCR) can be used to ensure only viable cells are detected. Certain genes are specifically expressed in the growth phases and these are the targets for the reverse transcriptase enzyme, which transcribes the mRNA to single‐stranded DNA, which can be amplified by PCR and subsequently detected. In common with other detection methods, results based on PCR techniques may not be comparable with results from other laboratories due to different protocols. Therefore, inter‐ laboratory proficiency tests, collaborative trials and standardised protocols are needed. 6.5.2 Microarrays

Microarrays were originally used to simultaneously measure multiple gene expression. However, oligonucleotide DNA microarrays can be used to detect foodborne pathogens. Microarrays consist of hundreds to thousands of specific oligonucleotides on a solid support. The oligonucleotides are short DNA sequences (25–80 bp) that have been generated by the PCR amplification of specific parts of genes within the selected genome (Plate 12; Schena et al. 1998; Graves 1999). In general, they are prepared by one of two methods: 1 Growing oligonucleotides on the surface, base by base. This is called a Genechip™. 2 Linking presynthesised oligonucleotides or PCR products to a surface. The DNA sequences are “spotted” onto the chips by a robot and act as the reference for comparison. The spots are positioned in a grid pattern, where each spot contains many identical copies of an individual gene. The positions of the DNA sequence are recorded by spot location, so that the appropriate gene sequence can be identified any time a probe hybridises with, or binds to, its complementary DNA strand on the chip. DNA array technology can also simultaneously detect different sequences in mixed DNA samples. Consequently, it is possible to detect and genotype different bacterial species in a single food sample. Multipathogen microarrays have been reported for food and biodefence analysis. The FDA‐1 microarray was designed for the simultaneous detection of (i) four Campylobacter species (C. jejuni, C. coli, C. lari and C. upsaliensis) using glyA and fur genes; (ii) six Listeria species were distinguishable by using the iap gene; (iii) 16 different St. aureus set genes encoding SEA‐SEE and SEG‐SEQ; and (iv) six Clostridium perfringens toxin genes (cpb1, cpb2, etxD, cpe, cpa, Iota). The reported sensitivity was 30–200 colony‐forming units. However since whole‐genome sequencing continues to become more affordable and accessible, the use of microarrays for bacterial detection and typing will diminish (Section  2.9). Nevertheless, having the whole genome sequence does not always inform the user regarding the conditions affecting the expression of key genes such as those involved in pathogenicity.

Methods of detection and characterisation

Genomic DNA

283

Genes amplified on 96-well microtitre plate

DNA sequences transferred to slide

Hybridisation

Comparison of gene expression or presence

Intrinsic effects (gene mutations) Bacterial cDNA

Extrinsic effects (different growth conditions)

Figure 6.10  Applications of bacterial microarrays.

Therefore, there is still a continuing application for microarrays in transcriptomics in which the mRNA expression is measured under the given conditions (Figure 6.10). Cells are grown under two different conditions: the experimental condition and the reference (control) condition. The mRNA from the bacteria grown under the two conditions are extracted (separately), and the enzyme reverse transcriptase is used to convert the mRNA into complementary DNA (cDNA). One set of cDNA will be labelled with a green fluorescent dye (Cy3) and the other a red fluorescent dye (Cy5). Therefore, cDNA from the two growth conditions can be distinguished by fluorescence. The two sets of cDNA are then incubated with the DNA array, during which time complementary regions with the array will bind together. The array is then scanned twice. Once to detect the spots containing cDNAs labelled with green dye, and then to detect the spots contain red‐labelled cDNAs. By merging the scan images, yellow spots correspond with DNA regions that bound both the red‐ and green‐labelled cDNAs. Therefore, genes (transcripts) that are expressed under both sets of growth conditions are identified. Similarly, based on the merged spot colours, genes which are up‐ or down‐regulated are also identified. 6.5.3  Loop‐mediated isothermal amplification (LAMP) technique

An alternative to PCR is the LAMP technique, which was invented in Japan in 2000. Being isothermal, the reactions are at one fixed temperature, which simplifies instrumentation. The key step is the production of stem‐loop DNA structures during initiation steps, which can then serve as the starting material for second‐stage LAMP cycling. LAMP uses a strand‐displacing Bst DNA polymerase large fragment, which allows autocycling amplification under a constant temperature (60–65°C). This means that thermal cycling, as used in PCR, is not necessary to denature DNA double strains in order to generate new initiation sites. LAMP uses four to six specially designed LAMP primers that target several (six to eight) regions of the template DNA. Amplification can generate 109 copies within an hour. The reaction appears to be less susceptible to interference from compounds in food that would normally inhibit the PCR reaction. The end‐product can be detected by various methods including colorimetry, fluorescence and bioluminescence. LAMP methods have been successfully used to detect a number of foodborne bacterial pathogens including Salmonella serovars (Yang et al. 2018).

284

The microbiology of safe food

6.5.4  Pulsed‐field gel electrophoresis (PFGE)

PFGE has been applied to a number of bacteria species, and has been one of the most widely used methods for microbial source tracking in epidemiology. It used to be the genotyping method of choice, or the ‘gold standard’ by which other genotyping methods were compared. However the method is dated, has many limitations and is being superseded. The CDC, like other health authorities, used standardised PFGE methods for specific pathogens. It was the key method used for establishing PulseNet for the active surveillance of foodborne infections due to Salmonella, E. coli O157, Shigella species, L. monocytogenes, C. jejuni and norovirus (see Section 1.12.3; Graves and Swaminathan 2001; Ribot et al. 2001). However, it should be noted that the PFGE network PulseNet is changing to using whole‐genome sequencing in place of PFGE due to its limitations and improvements in DNA sequencing techniques (Section 13.1.3). Genomic DNA is digested with a rare cutting restriction enzyme, such as XbaI for Enterobacteriaceae. The resulting DNA fragments will be too big to be separated by conventional gel electrophoresis. Instead the polarity of the electric field is briefly reversed periodically throughout the run. The field reversal causes the DNA molecules to be re‐orientated in order to move through the pores of the gel in the opposite direction. The longer the molecule is, the slower this process will be. Consequently, the net migration of the DNA fragments will be size‐dependent. It is important that the starting DNA is intact or at least very large fragments. To achieve this the bacterial cells are embedded in agarose before lysis and restriction digestion. Then the agarose plug is inserted into the gel for electrophoresis. The resultant banding patterns can be digitised and compared using appropriate cluster analysis software. Genomic variation occurs even between related bacterial strains resulting in changes in the length of DNA fragments between restriction enzyme digestion sites. Therefore there will be differences in the banding pattern of related strains. However there is no definitive threshold of difference between ‘indistinguishable’ and ‘unrelated’. A common guidance has been to initially assume the strains are indistinguishable if less than three bands different (Tenover et al. 1995). In practice the user will come to their own conclusion based on experience. Although PFGE has been of considerable value in outbreak investigate and control, it has a number of limitations (www.cdc.gov/pulsenet/whatis.htm#limitations). These include: • Time‐consuming procedure, requiring a high level of skill and is prone to operator variation. • Cannot optimise separation in every part of the gel at the same time. • More than one section of DNA could give the same band size, i.e. a band may be the result of two or more separate pieces of DNA. • Do not really know if bands of same size are the same regions of DNA. • Bands are not independent as the change in one restriction site can mean more than one band change. • ‘Relatedness’ of strains is not a true phylogenetic measure. Also, the word ‘related’ is ambiguous and can be misleading as it can refer to ‘relevant’ strains in the epidemiological investigation, or in the phylogenetic context of relatedness. • Frequently strains are untypable by PFGE due to the intrinsic presence of DNases and therefore no banding pattern is obtained. These can then be overlooked and not included in subsequent analysis using other methods. PFGE has limited usefulness with clonal organisms as by definition they are genetically homogeneous with little sequence variation. Examples are Salmonella Enteritidis, which has only five PFGE patterns, one of which (XbaI restriction pattern JEGX01.0004) is given by ~45% of all clinical isolates in the CDC database, and Cronobacter species (Allard et  al. 2013; den Bakker et al. 2014; Ogrodzki and Forsythe 2016).

Methods of detection and characterisation

285

6.5.5  Restriction fragment length polymorphism (RFLP)

The basis of the RFLP method is that if a DNA probe is added to a restricted genomic DNA preparation and it hybridises to a DNA sequence that is present only as a single copy in the genome, then only a single band will be detected. The size of the fragment will depend on the position of the restriction sites flanking the detected sequence. The band will be the same size between two strains if the structure and location of that gene is the same; whereas, if there is variation in the distance between the restriction sites, then there will be a difference in the size of the detected band. The variation can be caused by loss of a restriction site due to a point mutation (less frequent), and the insertion or deletion of a DNA region. This is called restriction fragment length polymorphism or RFLP. The most useful polymorphisms are due to duplication or transposition of repetitive sequences. These are short sequences that occur twice or more in succession in the genome. The technique is applicable to both bacteria and eucaryotes: Campylobacter (Messens et al. 2009) and Toxoplasma gondii (Velmurugan et al. 2009). 6.5.6  Amplified fragment length polymorphism (AFLP)

In the AFLP method, genomic DNA is cut with two restriction enzymes, and double‐stranded adaptors are specifically ligated to one of the sticky ends of the restriction fragments. The subsequent PCR step using primers complementary to the adaptor sequence will amplify the restriction fragments ending with the adaptor, the restriction site sequence and a number of additional nucleotides (usually one to three nucleotides) from the end of the unknown DNA template. At the start of the amplification, highly stringent conditions are used to ensure efficient binding of primers to fully complementary nucleotide sequences of the template. AFLP allows the specific co‐amplification of high numbers (typically between 50 and 100) of restriction fragments and is often carried out with fluorescent dye‐labelled PCR primers. The PCR amplicons can then be detected, after size separation, on an automated DNA sequencer. The genetic relatedness of bacterial isolates can then be determined by comparison of the banding patterns. AFLP can be as discriminatory as PFGE. 6.5.7  Random amplification of polymorphic DNA (RAPD)

Unlike the previous methods that used targeted DNA probes/primers, RAPD is based on the parallel amplification of a set of fragments by using short arbitrary sequences as primers (usually 10 bases) that target several unspecified genomic sequences. Subsequently, the PCR amplification step uses a low, non‐stringent annealing temperature, which allows the hybridisation of multiple mismatched sequences. When the distance between two arbitrary primer binding sites on both DNA strands is within the range of 0.1–3 kb, a PCR amplicon will be generated for the length between the two binding sites. RAPD amplicons can be analysed by agarose gel electrophoresis or DNA sequencing depending on the labelling of primers. The number and length of PCR amplicons will be unique to a particular bacterial strain and therefore can discriminate between non‐related, non‐clonal strains. The advantage of the technique is that it is simple and relatively inexpensive to perform for microbial source tracking. However, it may not be as discriminatory as PFGE and is not as reproducible. 6.5.8  Repetitive‐element polymerase chain reaction (Rep‐PCR)

Rep‐PCR uses primers that hybridise to non‐coding intergenic repetitive sequences that occur in bacterial genomes. PCR is then used to amplify the DNA sequence between adjacent repetitive elements. As this occurs across multiple sites so multiple amplicons will be produced according to the distribution of the repeat elements across the genome and they will differ between

286

The microbiology of safe food

unrelated, non‐clonal strains. The PCR amplicons can be separated according to size and the subsequent banding patterns used to assess the relatedness amongst the multiple isolates. Different groupings of repeat sequences have been used in the variants of rep‐PCR, such as ERIC‐PCR using enterobacterial repetitive intergenic consensus and BOX‐PCR sequences. Since rep‐PCR uses conventional PCR and gel electrophoresis results can be obtained relatively quickly with low cost and expertise. 6.5.9  Nucleic acid sequence‐based amplification (NASBA)

NASBA is a versatile technique that amplifies nucleic acids under isothermal conditions, unlike PCR, which is based on thermocycling. Essentially NASBA is used to amplify RNA to generate a single‐stranded RNA template. This can then be converted into cDNA using reverse transcriptase. The sequence of reactions is carried out at ~41°C, and requires two specific primers and three enzymes: avian myeloblastosis virus (AMV) reverse transcriptase, T7 RNA polymerase and RNaseH. The NASBA amplicons can be detected by conventional agarose gel electrophoresis, or in real time by using fluorescently labelled probes to detect the single‐ stranded RNA amplicons. Real‐time NASBA has been used for the detection of Salmonella serovars, V. cholerae, St. aureus and Campylobacter species. In addition, real‐time NASBA is able to directly detect viable target organisms in food samples (i.e. without a time‐consuming cultivation step) through mRNA amplification. The positive detection of RNA targets would indicate the presence of viable target organisms. NASBA offers high‐throughput analysis and it has been commercialised as kits. 6.5.10  Multiple‐locus variable number tandem (VNTR) repeat analysis

The MLVA method is based on the amplification of variable number of tandem repeat (VNTR) areas. In MLVA, primers are used to hybridise either side of the tandem repeat. Each strain can be given a number corresponding to the number of tandem repeats. The number of copies of each VNTR generates a profile that is then used for genotyping. MLVA schemes have been developed for Enterobacteriaceae: E. coli O157, S. Typhi and S. Typhimurium. It has been reported to be better than PFGE for both surveillance and outbreak investigations of the latter, probably as it is able to discriminate within clonal groups (Torpdahl et al. 2007). A limitation of MLVA is that the banding patterns generated by low‐resolution gel electrophoresis data cannot be compared directly between different laboratories. Subsequently the exact numbers of repeats cannot be determined and neither can the target site be identified. However, for a centralised typing system, akin to PulseNet, the exact number of repeat units in each MLVA locus must be determined. A better separation of the amplified DNA fragments by size is achievable using capillary electrophoresis. By labelling the primers with different fluorescently coloured dyes and combining capillary electrophoresis with DNA sequencing the mixture of MLVA amplicons can be analysed in one run and the strain described according to the number of repeat units at each MLVA locus. This allelic profile (e.g. 2‐9‐7‐3‐2) can be stored in a centralised reference database. Since specific primers have to be designed and tested for each target organism, this means that MLVA is not generally suitable to replace PFGE for epidemiological investigations in general. However, PFGE has limited application for highly clonal organisms such as S. Enteritidis (which only has approximately five PFGE profiles) as it cannot discriminate between unrelated strains. Similarly (in Belgium) one single MLVA profile of S. Enteritidis represented more than a quarter of 1498 S. Enteritidis strains isolated during 2007–2012. Nevertheless, the most common MLVA types can be further divided into

Methods of detection and characterisation

287

subgroups using phage typing and PFGE. Consequently, MLVA cannot be relied upon as a single typing method but can be used with parallel methods for prevalent MLVA types. In the EU and US the method has been standardised for S. Typhimurium, but is still regarded as problematic for S. Enteritidis due to its high level of clonality (Allard et al. 2013; CDC 2013; den Bakker et al. 2014; ECDC 2016; Peters et al. 2017). 6.5.11  PCR‐probe based serotyping

The O antigen, also known as lipopolysaccharide (LPS), is anchored in the outer membrane of Gram‐negative bacteria. Unlike the K antigen it elicits an immunological response. Consequently, LPS has been used for a typing scheme referred to as O‐serotyping  –  the Kauffmann–White Salmonella and E. coli serotyping schemes being well‐known examples (Section 4.3.2). Originally, bacterial extracts would have been injected into laboratory animals (rabbits or mice) and corresponding antigens produced for use in serotyping assays. The advent of molecular methods and a reduction in the use of animals have largely replaced this now historic approach. Alternative laboratory‐based serotyping approaches have used PCR probes designed for the galF and gnd genes as these flank the O antigen region, also known as the rfb locus. The method may use long‐range PCR to generate a large single PCR amplicon (size range 9.8–14.8 kb). This is then restricted using the restriction enzyme MboII, and the DNA fragments separated by standard gel electrophoresis. The RFLP banding patterns can then be further analysed. However, it is commonly overlooked that standard gel electrophoresis does not distinguish DNA fragments that are ±10% in size. Therefore, a single band may be composed of multiple bands (with differing DNA sequences) that happen to be similar in size. Also the O antigen does not necessarily follow phylogeny, resulting in the same antigen occurring in different species, even different genera; E. coli O29 and E. coli O103 and Cronobacter. There are also open access automated serotype prediction pipelines that can predict the serotype within minutes from the whole genome sequences (i.e. Center for Genomics Epidemiology; Section 14.3.4). 6.5.12 Ribotyping

Ribotyping is a molecular technique for bacterial identification and characterisation that is based on the DNA sequence encoding for the ribosomal rRNA (rDNA genes). Extracted DNA is cut with specific restriction enzymes, and the resulting fragments are separated by gel electrophoresis. These are then probed with labelled 16S or 23S rRNA sequences. The resulting hybridisation pattern can be digitised and compared with reference organisms in a database. The technique was popularised in the 1990s but is now largely redundant due to more discriminatory and cheaper genome sequence‐based methods. 6.5.13  Matrix‐associated laser desorption ionisation – time of flight (MALDI‐TOF)

Identification of bacteria by mass spectrometry (MS) has been an active research area for decades. Most MS‐based bacterial identification methods rely on the reproducible patterns generated from the measurement of masses of proteins and/or lipids from intact bacterial cells or cell extracts. Consequently, MS‐based methods can directly detect expressed bacterial proteins. Tens to hundreds of proteins are measured in a single experiment. The result is a reproducible mass fingerprint that will be unique to a given genus or species. The significant advantage of MS methods over phenotypic and molecular methods for bacterial identification is that no advanced knowledge of the microorganism is necessary in order to choose the method of analysis.

288

The microbiology of safe food

MALDI‐TOF MS is the most common technique used for bacterial analysis by MS. The method has been refined over the years such that it is easy to operate, the cost per sample of analysis is low and analysis is as fast as 10 minutes from colony selection to identification. As multiple laboratories began developing libraries of reference spectra representing species of interest, it became clear that lab‐to‐lab reproducibility was going to require standardisation of spectral libraries. Subsequently, in 2008, multiple commercial MALDI‐TOF MS instruments were released that included reference spectra libraries. The research use versions of these systems also include methods for creating user libraries. A thorough review of bacterial identification by MALDI‐TOF MS for clinical use can be found in Clark et al. (2013). Most commercially available MALDI‐TOF MS bacterial identification instruments contain reference databases that are primarily based on clinically relevant isolates. Consequently, MALDI‐TOF MS is an effective and rapid tool for genus‐ and some species‐level identifications. However, it is not able to identify bacteria directly within food samples, and cannot differentiate strains within mixed cultures. Therefore, a single colony or pure culture is generally required. However, the speed at which even a genus‐level identification can be made makes MALDI‐TOF MS a potentially important screening tool. 6.6 Identification and typing methods based on high‐throughput DNA sequencing

The epidemiological investigation of a foodborne outbreak and microbial sourcing such as in the food processing environment in order to curtail further infections depend on typing methods to differentiate relevant strains from non‐relevant ones. Historically methods such as serotyping and PFGE have been used. However, techniques based on high‐throughput sequencing of genomes offer greater resolution and discrimination than previously achievable. The microbial genomes will contain sequence variation between strains due to mutations and recombination events. DNA sequence‐based methods are more repeatable and reproducible than DNA gel electrophoresis‐based methods. There are a very large number of typing methods, such as core‐ genome MLST, single nucleotide polymorphism (SNP) and CRISPR‐cas array profiling, which are based on high‐throughput sequencing of whole genomes. These offer considerably better discrimination than methods based on PCR and PFGE (Carriço et al. 2013; Ronholm et al. 2016; Sekse et al. 2017). Additionally, high‐throughput sequencing enables the analysis of specific target sites, such as CRISPR‐cas array profiling, which is not easily achievable using conventional methods. Where large‐scale genome sequencing is not available due to limited equipment and expertise, the whole‐genome sequencing of a few representative strains can enable the design of primers etc. for a larger diversity analysis of laboratory strain collections using conventional methods. 6.6.1  Conventional seven‐loci MLST

MLST of bacteria is usually based on DNA sequence variation within alleles (short regions within a gene, also known as loci). Conventional, laboratory‐based MLST is based on the analysis of seven housekeeping genes, though sometimes this can be as low as five. The maximum length of the allele is usually between 450 and 500 bp, due to the original limitation of DNA sequencers. This limitation is no longer necessary as the allele sequences can be predicted in silico from whole genome sequences and includes ribosomal MLST (rMLST, 53 loci) (Figure 4.13), and core‐genome MLST (cg‐MLST, >1000 loci) (Section 6.6.2). Although overall

Methods of detection and characterisation

289

more expensive and limiting, due to accessibility and expertise, the laboratory‐based seven‐loci MLST continues in use. For each housekeeping gene analysed, each sequence variation is assigned a distinct allele number. If using conventional seven‐loci MLST, then the seven alleles will generate a seven‐ digit number. In turn, each unique seven‐digit number is assigned a number known as the sequence type (ST). For example, Cronobacter sakazakii allele profile atpD5, fusA1, glnS3, gltB3, gyrB5, infB5 and ppsA4 is simplified to ST 4 as it was the fourth seven‐digit profile in the database (Forsythe 2018a, b). Since by definition a single nucleotide difference generates a different allele number, this means two strains with one nucleotide difference would have different STs and not be easily recognised as related. Therefore, by convention STs that have five or more loci in common are referred to as a clonal complex, and are numbered after the central ST. A considerable advantage of MLST is that the sequence data are unambiguous and the allelic profiles can be compared with those in large online central databases that are curated and that have the standardised protocols. There are three such international open access MLST databases supporting an increasing range of organisms: http://pubMLST.org, http://bigsdb.pasteur. fr and www.mlst.net. For example, the Pasteur Institute E. coli MLST protocol uses PCR primers to specifically amplify seven genes: dinB (DNA polymerase), cdA (isocitrate dehydrogenase), pabB (p‐aminobenzoate synthase), polB (polymerase PolII), putP (proline permease), trpA (tryptophan synthase subunit A), trpB (tryptophan synthase subunit B) and uidA (β‐glucuronidase). Sequence data for each allele generate a seven‐digit profile that is designated a ST, and can be compared with other E. coli strains in the database (the URL is given in the ‘Food Safety Resources’ section). The MLST databases are curated, open access, public domain sites to which any laboratory can submit sequence data (alleleic or whole genome). In contrast, PulseNet is only accessible by member laboratories for reasons of privacy and confidentiality. MLST has been highly informative in the diversity of bacterial species and genera, and in identifying particular pathovars. Sopwith et  al. (2006) used MLST for the first continuous population‐based survey of campylobacteriosis. They demonstrated that MLST could identify variations in the epidemiology of campylobacteriosis between distinct populations over a 3‐ year period and described the distribution of major subtypes of interest. It was based on 493 cases, from which 93 distinct MLST STs of C. jejuni were obtained. The most common type was ST‐21 (102 cases), which was isolated three times more frequently than the next complex, ST‐45. The clonal complex ST‐21 has previously been reported from humans, cattle, chicken, milk, sand and water. Hence ST‐21 is associated with environmental and foodborne transmission of C. jejuni, whereas ST‐45 is principally from humans and chickens. Cronobacter sakazakii ST4 and ST12 are strongly associated with neonatal meningitis and necrotising enterocolitis, respectively (Forsythe 2018a, b; Joseph and Forsythe 2011) 6.6.2  Genome sequence‐based MLST

Whole‐genome sequencing of target organisms enables the further expansion of DNA sequence‐based typing methods, by substantially increasing the number of loci examined simultaneously. By combining MLST with whole‐genome sequencing the number of loci profiled can be considerably increased from the traditional seven. cg‐MLST schemes typically use a comprehensive set of loci (>1000) that are present in most members of the target bacterial group and therefore offer higher resolution than the conventional seven loci. For example, cg‐MLST for L. monocytogenes uses ~1700 loci, and that for Cronobacter species uses 1865 (Forsythe et al. 2014; Ruppitsch et al. 2015).

290

The microbiology of safe food

Genomic methods such as rMLST (Figure 4.13) and cg‐MLST are expansive methods and can be directly compared with the previous laboratory‐derived seven‐loci typing schemes. Given that rMLST and cg‐MLST use 53 and >1000 loci, respectively, it is self‐evident that these newer genomic methods provide considerable greater discrimination between strains. Similar to generating a genotype from a whole genome sequence as described above for rMLST, it is also possible to generate a genotype profile using a core‐genome approach. Forsythe et  al. (2014) demonstrated its application across the Cronobacter genus for 1865 loci to profile  all Cronobacter genomes in the database, and also generate a phylogenetic analysis across the genome. Additionally, as an example for studying the organism’s biology, they expanded on the cell wall division genes. This was achieved using the comparative genomics analysis facility within the Cronobacter PubMLST database. The cg‐MLST analysis used C. sakazakii ES15 as the reference genome since it is well annotated, but the user can choose their own genome for comparative analysis if they wish. The user can either profile all annotated genes, or choose their own interest area, for example, biochemical pathways, adhesins and virulence traits. The output is a self‐explanatory Excel spreadsheet, and no bioinformatic knowledge or specialist software is required; thus the approach is highly accessible with no usage costs. Loci that recurrently show variations between same‐patient isolates and epidemiologically linked isolates are commonly homopolymeric tracts in contingency genes. The cg‐MLST approach enables the clustering of isolates (which vary in only a few loci) that have been obtained from cases spread over a large geographical area, despite lacking an apparent epidemiological link between them. Investigators commonly define clusters of genetically indistinguishable isolates based on a five‐single nucleotide variant (SNV) cutoff value for allele differences, while excluding indels in homopolymeric tracts. A cg‐MLST scheme of 1748 loci for L. monocytogenes is more discriminatory than PFGE, and can separate unrelated isolates that share identical PFGE profiles and phylogenetically closely related isolates with distinct PFGE profiles. This helps to refine epidemiologic investigations by focusing on phylogenetically closely related isolates to improve source identification, and enables the recognition outbreaks at earlier stages. For C. jejuni and C. coli a set of 1343 loci has been collated as a human campylobacteriosis cg‐MLST scheme. The 1343 loci are present in ~95% of draft genomes of 2472 representative UK campylobacteriosis isolates, 89.3% C. jejuni isolates and 10.7% C. coli. In addition to the rapid high‐resolution analysis of large numbers of Campylobacter isolates, the cg‐MLST scheme enabled the efficient identification of very closely related isolates from a well‐defined single‐source campylobacteriosis outbreak (Cody et al. 2017). The scheme, as per other MLST schemes, is open access via the www.pubmlst.org site. 6.6.3 CRISPR‐cas array typing

Clustered regularly interspaced short palindromic repeats (CRISPRs) are chromosomal loci that are very useful for high‐resolution genotyping, even with highly clonal organisms. They were first identified in ~1990 as ambiguous repeats in the genome of E. coli, and are now known as CRISPR spacer arrays or CRISPR‐cas arrays. In brief, CRISPR‐cas systems may be composed of up to three sections: (i) a group of CRISPR‐associated (cas) genes; (ii) an AT‐rich leader sequence upstream of the array; and (iii) a CRISPR array, composed of short (~24–48‐nucleotide) direct repeat (DR) sequences separated by similarly sized, unique spacers (Plate  18). These spacers are usually derived from mobile genetic elements such as bacteriophages and plasmids. The CRISPR spacer array is

Methods of detection and characterisation

291

constitutively transcribed into a precursor CRISPR RNA that is cleaved by specific Cas proteins and then processed to generate small interfering crRNAs. These then lead to the degradation of invading bacteriophages or plasmids. New spacers are always added to the leader end of the array. The subject area is too large to consider in detail here, instead the reader should consult core texts for background information such as Koonin et al. (2017). Analysing CRISPR‐cas arrays has helped our understanding of bacterial lineages in E. coli, Yersinia, Salmonella enterica serovars Enteritidis and Heidelberg, as well as Cronobacter. CRISPR arrays may differ between closely related strains due to their different exposure histories to phages and plasmids, leading to differences in their spacer acquisitions. CRISPR loci between strains can also differ due to the loss of internal spacers and SNPs. Therefore, the CRISPR‐cas array loci can be used for molecular subtyping, offering greater discrimination between strains than MLST and PFGE. Since CRISPR‐cas arrays reflect the exposure of strains to phages and plasmids there should be differences between strains despite being related (Ogrodzki and Forsythe 2015; Zeng et al. 2017) (Plate  18). Therefore identifying epidemiologically associated isolates from unrelated strains which were very difficult to separate if not impossible with previous genotyping methods such as PFGE. CRISPR‐cas arrays have been analysed in Salmonella serovars. Salmonella have two CRISPR loci (CRISPR1 and CRISPR2) that are separated by ~16 kb. These share the same consensus DR sequence (29 nt), and the spacers are 32 nt in length. Analysis of >1000 Salmonella CRISPRs has shown that this organism is not rapidly acquiring new spacers. Consequently these two CRISPR loci appear no longer to be active. Instead CRISPR‐cas array variation is due to the multiplication and/or loss of internal spacers, as well as SNP variants of spacer or DR loci. Retrospective analysis of 783 Salmonella strains from 10 outbreaks that belonged to 130 serotypes has shown correlation between CRISPR‐cas arrays with serotyping and MLST. Furthermore, the spacer variants meant it was possible to discriminate between subtypes within the prevalent serotypes (Fabre et al. 2012). 6.6.4  Single nucleotide polymorphism (SNP)‐based analysis

SNPs occur in genes, non‐coding regions and mobile elements (i.e. plasmids). SNPs that occur within a gene are termed synonymous if they do not result in a change in the resultant amino acid sequence after transcription and translation, or non‐synonymous if there is a change in the resultant amino acid sequence. This is related to the wobble base and redundancy of codons. The term ‘diversifying population’ is used if there are more non‐synonymous than synonymous SNPs in a bacterial population. If the same SNP occurs in two or more isolates then it is termed ‘parsimonious’. To detect SNPs, genomes sequences need to be aligned. This can be computationally demanding, and often a reference genome is used to which the DNA sequencing reads are mapped. A ‘SNP caller’ program is then used to identify the SNPs that are different between the genomes. This method is very useful for determining the phylogenetic relationship between two or more genomes. However, it introduces an issue of reference bias and difficulties in standardising SNP analysis for outbreak investigations as the choice of reference strain is crucial. SNP analysis in the accessory genes is only possible if those regions have been sequenced in all genomes, and a plasmid, etc. has not been lost during subculturing. Alternatively SNP analysis can be achieved using de novo‐assembled genomes by annotating the contigs and then comparing the open‐reading frames. Examples of SNP analysis in outbreak investigations are given in Section 14.4.

292

The microbiology of safe food

6.7  Specific detection procedures and accreditation

Standardised protocols for the isolation of most foodborne pathogens have been defined by various regulatory and accreditation bodies such as the ISO, and the FDA. The methods also need to be validated by a national or international body. The AOAC International (Anon. 1999b) and ISO are the most widely accepted; others include UKAS (UK), EMMAS (European), AFNOR (French), DIN (Germany) and the European MICROVAL. Appropriate control organisms should always be run to ensure the media is conforming to its specification. There is however not always a single ideal method for each pathogen and countries may vary in their preferred technique. Therefore, only an overview of techniques will be given here, and specific instructions must be obtained from the relevant authoritative source before using the methods. The procedures for a number of regulatory authorities, such as the FDA and Health Canada, are available online and are listed in the ‘Food Safety Resources’ section at the end of the book. A consequence of the variation in methodology is the uncertainty about how confidently one can compare food poisoning statistics between countries. Because batches of media can vary in their composition, as a means of monitoring personnel proficiency, good laboratory management requires that standardised positive and negative control organisms are used to confirm the selectivity of the media. The control organisms originate from national and international culture collections such as the National Collection of Type Cultures (NCTC) and American Type Culture Collection (ATCC). This is indicated by the culture collection index numbering, i.e. St. aureus ATCC® 25923. These are well‐characterised strains available to all quality control laboratories and act as an international standard for referencing. 6.7.1  Aerobic plate count (APC)

The APC is commonly used to determine the general microbial load of the food and not specific organisms. It is a complex growth medium containing vitamins and hydrolysed proteins that enables the growth of non‐fastidious organisms. The agar plates may be inoculated by a variety of techniques (Miles‐Misra, spread plate, pour plate) that vary in the volume of sample (20 μl to 1 ml) applied. The plates are generally incubated at 30°C for 48 hours before enumerating the number of colonies. The accuracy of colony numbers is given in Table 6.1. The advantage of pour plates is that up to 1 ml of sample can be applied giving a lower limit of detection (10 cfu/g) than spread plate and Miles‐Misra (100 and 500 cfu/g, respectively). However, it is difficult to remove colonies from within the agar of a pour plate for confirmation of their identity and further characterisation. Consequently, most methods use the spread plate approach (0.1 ml/plate), which combines a lower detection limit with the generation of well‐separated colonies for further analysis, though requiring large volumes of agar preparation and disposal. 6.7.2  Salmonella serovars

Internationally standardised procedures are necessary for the isolation of Salmonella serovars from food in order to collate results together for microbial source tracing and epidemiological analysis. These methods commonly involve a mixture of conventional cultivation procedures followed by molecular analysis (Table  6.5). Nevertheless, the ISO approved method (ISO 6579‐1:2017) is considered primarily as it is an internationally approved method that accredited laboratories analysing food have to follow. Typically Salmonella cells are only present at very low levels in food, and are likely to be in a stressed condition following processing. Therefore, large volumes need to be sampled. Due to its severity, presence/absence testing is used instead of enumeration. Consequently, the

Methods of detection and characterisation

293

Table 6.5  Salmonella detection media.

Differential indicator

Medium

Common abbreviation

Brilliant green agar Bismuth sulphite agar Brilliance™ Salmonella agar Chromogenic Salmonella Esterase Desoxycholate citrate lactose agar Desoxycholate citrate agar Hektoen enteric agar Salmonella Shigella agar Xylose lysine deoxycholate agar

BGA BS BSA CSE DCLS DCA HE SS XLD

H2S positive

Lysine decarboxylase positive

Lactose negative

Sucrose negative

x

x

Xylose positive

Salicin negative

β‐glucosidase negative

Esterase positive

x

x x

x

x x x x

x

x x x x x x

x x x

x x

294

The microbiology of safe food Pre-enrichment Test portion, 25 g + buffered peptone water, 225 ml 16–20 hours, 37°C Selective enrichment Culture, 0.1 ml +

Culture, 10 ml +

Rappaport (RVS) broth, 10 ml

Selenite cystine broth, 100 ml

18–24 hours, 41–43°C

18–24 hours, 37°C

Selective diagnostic isolation Plate on xylose lysine deoxycholate agar and any other solid selective medium 24 hours, 35°C or 37°C (48 hours, if necessary) Pick five presumptive Salmonella colonies from each agar plate and inoculate on nutrient agar 18–24 hours, 35°C to 37°C Biochemical confirmation 4 hours, 37°C Serological confirmation Slide agglutinations – O, Vi, H antisera Figure 6.11  BSI/ISO Salmonella isolation procedure.

microbiological specification for ready‐to‐eat foods is the isolation of one Salmonella cell from 25 g of food. The isolation protocols have a number of steps to recover Salmonella cells, which are present in low initial numbers, and start with resuscitation of cells that may have been injured during processing (see Section 6.3.2 on IMS technique). It is possible to obtain an estimate of the number of Salmonella cells in the food by using the MPN approach whereby 10‐fold sample volumes are analysed. The flow chart for ISO 6579‐1:2017 is given in Figure 6.11, and FDA‐BAM in Figure 6.12. The general isolation procedure is: 1 Pre‐enrichment, to enable injured cells to resuscitate. Normally 25 g of food is homogenised and added to 225 ml BPW broth and incubated overnight (34–38°C). This step requires a nutritious non‐selective medium and therefore should result in the resuscitation of all bacteria present. The BPW broth may be modified if large numbers of Gram‐positive bacteria

Methods of detection and characterisation

295

Pre-enrichment Test portion, 25 g + pre-enrichment medium, 225 ml 16–20 hours, 37°C Selective enrichment Culture, 1 ml +

Culture, 1 ml + Selenite cystine

Tetrathionate broth, 10 ml

broth, 100 ml or Rappaport (RV) broth, 10 ml 24 h +/– 2h, 35°C

Selective diagnostic isolation Plate on Bismuth sulphite agar, Xylose lysine desoxycholate agar, Hektoen enteric agar 24 h +/– 2h, 35°C (48 h, if necessary) Pick two or more suspect Salmonella colonies from each agar plate for biochemical tests

Biochemical confirmation 24–48h, 35°C Serological confirmation Slide and tube agglutination – O, H antisera Figure 6.12  FDA‐AOAC BAM Salmonella isolation procedure.

are suspected to be present indicated by the addition of 0.002% brilliant green or 0.01% of malachite green. Since milk products are already highly nutritious, BPW broth can be distilled water plus 0.002% brilliant green instead. If the food is highly bacteriostatic (in the case of onion) this can be overcome by the addition of sodium thiosulphate or an increased dilution factor, for example 25 g in 2.25 l. 2 Selective enrichment, to suppress the growth of non‐Salmonella cells and enable Salmonella cells to multiply. This is achieved by the addition of inhibitors such as bile, tetrathionate, sodium biselenite (this compound is very toxic) and either brilliant green or malachite green dyes. Selectivity is enhanced by incubation at 41–43°C. Two selective broths are used. These are commonly Rappaport–Vassiliadis (RVS) broth and Mu(e)ller Kauffmann tetrathionate novobiocin (MKTTn) broth. More than one selective broth is used because the broths have different selectivities towards the 2500+ Salmonella serovars. Ideally this step will result in a 1000‐fold greater number of Salmonella compared with non‐Salmonella, and

296

The microbiology of safe food

therefore it will be easier to spot Salmonella colonies on the isolation agar plates used in the next step. 3 Selective differential isolation, to isolate Salmonella cells on an agar medium enabling single colonies to be isolated and identified. The media contain selective agents similar to the selective broths such as bile salts and brilliant green. They also contain substrates to enable Salmonella colonies to be differentiated from non‐Salmonella. Typically, Salmonella serovars do not ferment lactose, but do produce H2S. It should be noted however that atypical strains occur and are still virulent (Section  14.4.5). Commonly used selective agars are xylose lysine desoxycholate (XLD) agar, brilliant green agar (BGA), as well as Rambach™ agar and CHROMagar™ (Plate  13). As for the previous selective enrichment stage, more than one agar medium is used since the media differ in their selectivities. It should be noted that ISO approved selective agars are not used for the clinical isolation of Salmonella. The most commonly approved selective agar for the isolation of Salmonella from food is XLD. On this agar Salmonella colonies typically appear red with a black centre, or entirely black. This medium is designed for the isolation and differentiation of Salmonella (and Shigella) from other Enterobacteriaceae in the enrichment broth from the previous step. Almost all Enterobacteriaceae ferment xylose, except Shigella. The metabolism of xylose and lysine results in the pH of the medium becoming alkaline and blackening due to the H2S production. In contrast, acidification due to lactose and sucrose fermentation prevents non‐Salmonella lysine metabolising Enterobacteriaceae generating an alkaline pH value, and non‐pathogenic hydrogen sulphide producers do not decarboxylate lysine. The acid level also prevents blackening by these latter organisms until after the prescribed 18–24‐hour incubation period. XLD also contains sodium desoxycholate as an inhibitor of non‐enteric bacteria. Since Salmonella is so diverse, a second isolation agar must be used. This is commonly BGA in order to pick up the H2S‐negative Salmonella. On BGA, Salmonella colonies typically appear as red to pink‐white colonies surrounded by a red zone in the medium. Note this medium is not recommended for the isolation of Salmonella Typhi or Shigella species. According to the ISO 6579‐1:2017 procedure only one suspect colony needs to be confirmed. If negative, four more suspect colonies from different media combinations have to be tested. 1 Biochemical confirmation, to confirm the identity of presumptive Salmonella colonies. 2 Serological confirmation, to confirm the identity of presumptive Salmonella colonies and to identify the serotype of the Salmonella isolate, which is useful in epidemiology. Aside from the international standard ISO and FDA‐BAM methods for Salmonella detection, more rapid detection procedures have been developed based on molecular biology. These are based on PCR and real‐time, quantitative PCR (qPCR). One of the most frequently used target genes is invA. Genotyping Salmonella isolates is very important for epidemiological investigations. PFGE has been the common method used for most Salmonella serovars, but is not applicable for the highly clonal serovars S. Enteritidis and S. Heidelberg (den Bakker et  al. 2014; Peters et  al. 2017). More discriminatory methods have included genome‐based methods such as MLVA, SNP and CRISPR‐cas array profiling (Section 6.6). Examples of genotyping methods with S. Typhimurium and S. Enteritidis in outbreak investigations are given in Section 4.3.7 as well as in Section 14.4. As an example of the limitations of PFGE with clonal organisms, an investigation into S. Heidelberg is discussed. Whole‐genome sequence‐based typing methods were used to differentiate 145 S. Heidelberg strains, which were associated with four distinct outbreaks and sporadic

Methods of detection and characterisation

297

cases of salmonellosis in Québec (Canada) between 2007 and 2016. All isolates were indistinguishable by PFGE. The core‐genome SNV, and genome‐based MLST (i.e. cg‐MLST) were highly discriminatory. These methods separated the outbreak strains into four distinct phylogenetic clusters, which was in agreement with the epidemiological data. However, the CRISPR typing method was less discriminatory. 6.7.3  Campylobacter species

There are many methods, based on both cultivation and DNA amplification, for Campylobacter detection. It is likely, however, as is probably true for many other organisms, that no single method is ideal. Some of the major methods and protocols have been reviewed by Corry et al. (2002) and are summarised in Table 6.2 and Figure 6.13. Readers should consult the original articles for precise details. Campylobacter cells can be stressed during processing and hence a pre‐enrichment stage to enable injured cells to be resuscitated is commonly used prior to selection for the organism. Frequently selective agents are added as supplements and lower incubation temperatures are used. In order to aid the growth of the organism ferrous sulphate, sodium metabisulphite and sodium pyruvate are added to growth media to quench toxic radicals and increase the organisms’ aerotolerance. The organism is microaerophilic and is unable to grow in normal air levels of oxygen. The preferred atmosphere is 6% oxygen and 10% carbon dioxide. This is easily achieved in gas jars by using gas sachets that generate the required gases. PCR‐based detection methods include the detection of the genus Campylobacter using the highly conserved 16S rRNA gene sequence as a target for PCR, and the more specific C. jejuni probe targeting the hippurate gene. There are a number of typing methods for campylobacters. Traditional typing methods based on differences of key surface antigens such as LPS and flagella have been applied to campylobacters, but are problematic as their surface structures are variable even within an (a)

Pre-enrichment Dilute sample 1/10 in Bolton broth

Selective enrichment Incubate in a microaerobic atmosphere at 37°C for 4–6 hours, followed by 41.5°C for 44 ± 4 hours

Selective diagnostic isolation Isolate on mCCDA agar and a second medium as preferred Incubate in a microaerobic atmosphere at 41.5°C for 44 ± 4 hours

Confirmation Confirm presumptive Campylobacter colonies Figure 6.13  (a) Horizontal method for detection and enumeration of Campylobacter species (ISO, 2006). (b) FDA‐BAM method for detecting Campylobacter species in foods.

298

The microbiology of safe food

(b) Pre-enrichment 25g sample prepared in Hunt & Radle Campylobacter enrichment broth 1/10 dilution of sample preparation (optional)

Selective enrichment Pre-enrich for 4 or 5 h, 30°C and/or 37°C

Enrich at 42°C (or 37°C if C. fetusis being analysed)

Selective diagnostic isolation Plate out : (a) Undiluted enrichment culture (b) 1/100 dilution of enrichment culture

Karmali blood-free agar Incubate at 37°C (C. fetus) or 42°C

Campy-cefoperazone blood agar. Incubate at 42°C

Sub-culture to blood agar or heart infusion blood agar

Subculture to lysed horse blood agar for confirmation Figure 6.13  (Continued )

individual strain. The biotyping (biochemical activity profiling) scheme of Preston is more extensive than that of Lior. Two serotyping procedures are used: the Penner and Hennessy scheme for the heat‐stable antigens (lipopolysaccharide) and the Lior scheme for the heat‐ labile antigens (flagella). Phage typing can differentiate strains within a serotype and is a simple enough technique to apply to a large number of strains simultaneously. MLST (Section 6.6.1) of Campylobacter isolates has an advantage that it is unambiguous and more reliable than phenotyping. However, this is also a disadvantage as common clones will require further differentiation. Two genes involved in the synthesis of flagella proteins may also be included as they are more variable (Dingle et al. 2002). Different clonal groupings of C. jejuni have been identified in specific animals. However, some are also widespread, and can be found in humans. Real‐time PCR Taqman assays to detect SNPs specific for the six major

Methods of detection and characterisation

299

MLST clonal complexes have been developed (Best et  al. 2005). This can be applied to the rapid detection of C. jejuni and clonal allocation. Campylobacter typing methods have been reviewed by Wassenaar and Newell (2000) and the UK Advisory Committee for the Microbiological Safety of Food (ACMSF 2005). CRISPR‐cas array profiling (Section 6.6) has not been investigated as much in Campylobacter compared with Salmonella, probably because most Campylobacter infections are single cases, and not cluster outbreaks. The two most common molecular typing methods for Campylobacter isolates are laboratory‐based PFGE and seven‐loci MLST (Carrillo et al. 2012). However, these methods have been difficult to implement in the context of routine surveillance. The methods have limitations that lessen their robustness and reliability as the sole subtyping methods for assessing epidemiological links amongst C. jejuni isolates. Instead whole‐genome sequencing methods for typing are more promising especially with the declining costs of sequencing and increasing access to sequencing facilities. 6.7.4  Enterobacteriaceae and E. coli

Escherichia coli and other Enterobacteriaceae are often initially detected together in liquid media and then differentiated by secondary tests of indole production, lactose metabolism, gas production and growth at 44°C. Escherichia coli produces acid and gas at 44°C within 48 hours. MacConkey broth is a commonly used medium for the presumptive detection of lactose‐fermenting Enterobacteriaceae from water and milk. It selects for lactose‐fermenting, bile‐tolerant organisms, which in the past have been given the general term ‘coliforms’. Acid formation from lactose metabolism is shown by a yellow coloration of the broth (due to a pH indicator dye, neutral red or bromocresol purple) and gas formation is indicated by gas being trapped in an upturned Durham tube. Lauryl tryptose broth (also known as lauryl sulphate broth) can be used for the detection of lactose‐fermenting Enterobacteriaceae from food. Initially the inoculated medium is incubated at 35°C and afterwards presumptive positive tubes are used to inoculate duplicate tubes: one for incubation at 35°C and the other at 44°C. Both broths can be supplemented with 4‐methylumbelliferyl‐β‐D‐glucuronide (Section 5.2.1) to enhance E. coli detection. Enterobacteriaceae enrichment (EE) broth, also known as buffered glucose‐brilliant green bile broth, is an enrichment medium for Enterobacteriaceae from food. The broth is inoculated with samples that have been incubated at 25°C in aerated tryptone soya broth (1:10 dilution) for resuscitation of any injured cells. There are numerous solid media employed for the detection of E. coli‐coliforms and Enterobacteriaceae. For example, there are two types of violet red bile agars. First, VRBL for lactose‐fermenting Enterobacteriaceae in food and dairy products. Second, violet red glucose (VRBG) agar for the general detection of Enterobacteriaceae. These select for bile‐tolerant organisms, a predicted trait for intestinal bacteria. Other media include MacConkey agar, china blue lactose agar, desoxycholate agar and eosin methylene blue agar. These have different differentiation efficiencies and regulatory approval. A recent trend has been the inclusion of chromogenic and fluorogenic substrates in particular to detect β‐glucuronidase, which is produced by ~97% of E. coli strains (Section 6.2.1). The ISO 21528:2004 standardised methods for the detection and enumeration of Enterobacteriaceae include the MPN technique and colony counts. These involve several steps. • Pre‐enrichment. A portion of the food is added to BPW in the ratio 1:9, and incubated at 37°C for 18 hours to resuscitate cells. For the MPN procedure triplicate tubes are incubated of three different sample volumes equivalent to neat, 10−1 and 10−2 dilutions. • Enrichment. A total of 1 ml of incubated BPW and sample is pipetted into 10 ml of EE  broth, which is incubated at 37°C for 24 hours. This supresses the growth of non‐Enterobacteriaceae.

300

The microbiology of safe food

• Selective plating, by streaking the EE broth on VRBG agar, incubated at 37°C, for 24 hours. Presumptive Enterobacteriaceae colonies (pink to red or purple, with or without precipitation halo) are streaked on nutrient agar, incubated at 37°C, for 24 hours. • Confirmation of Enterobacteriaceae. Five presumptive colonies are selected and streaked for purity on nutrient agar plates before confirmation as Enterobacteriaceae according to negative oxidase reaction and glucose fermentation. For direct enumeration the pour plate method (ISO 21528‐2:2004) is used with 1 ml sample volume (in duplicate) in VRBGA, with an overlay of VRBGA to prevent spreading colonies and encourage anaerobic conditions during colony growth. Presumptive Enterobacteriaceae colonies, streaked on nutrient agar for purity checking and standardisation, are confirmed using the oxidase and glucose fermentation tests. E. coli can be detected using chromogenic agar (TBX) for the β‐glucuronidase reaction (i.e. ISO 16649‐2:2001). The medium contains the chromophore 5‐bromo‐4‐chloro‐3‐indolyl β‐D‐glucuronic acid (BCIG). A pour plate method is used, in which 1 ml test sample is pipetted into a sterile Petri dish (in duplicate) and molten TBX agar (at 44–47°C) is added. The plates are incubated at 44°C for 18–24 hours. Typical E. coli colonies are blue. In order to resuscitate cells, the sample is initially spread onto a cellulose membrane (pore size 0.45–1.2 μm, 85 mm diameter) on two minerals‐modified glutamate agar (MMGA) plates, which are then incubated for 4 hours at 37°C. Afterwards, the membranes are transferred to TBX agar plates for incubation at 44°C, for 18–24 hours. This method will not detect E. coli strains that do not grow at 44°C, and those that are β‐glucuronidase negative. Hence E. coli O157 will not be detected. 6.7.5 Pathogenic E. coli, including E. coli O157:H7

Since E. coli is a commensal organism in the human large intestine there is an inherent problem isolating and differentiating pathogenic strains such as serotype O157:H7 from the more numerous non‐pathogenic varieties. The key differentiation traits have been based on the observation that, unlike most non‐pathogenic E. coli strains, E. coli O157:H7 does not ferment sorbitol, does not possess β‐glucuronidase and does not grow above 42°C. Consequently, MacConkey agar was modified to include sorbitol in place of lactose as the fermentable carbohydrate (SMAC). This medium has been further modified by the inclusion of various other selective agents such as tellurite and cefixime. Pre‐enrichment in a modified BPW broth or modified tryptone soya broth is used to resuscitate injured cells before plating onto solid media. Because the cell surface antigenic determinant (O157:H7) is indicative of pathogenicity (though not 100%) the IMS technique (Section 6.3.2) greatly increases the recovery of E. coli O157:H7 (Chapman and Siddons 1996). The IMS technique is used worldwide and is one of the most sensitive methods for E. coli O157:H7. It has been approved for use in a number of countries. The toxins of E. coli O157:H7 can be detected using cultured Vero cells and RPLA, which are sensitive to 1–2 mg/ml culture filtrate (Figure  6.6). Polymyxin B is added to the culture to facilitate the release of the verocytotoxins/Shiga toxins. ELISA methods specific for pathogenic strains of E. coli have been developed. The ISO horizontal method for the detection of E. coli O157 (ISO 16654:2001) is applicable for a large variety of foods. It has four steps: 1 Enrichment in modified tryptone soya broth containing novobiocin. There is usually a 1:9 ratio of sample to broth. The mixture is incubated at 41.5°C for 6 hours before IMS, followed by a further 12–18 hours after which time the IMS can be repeated.

Methods of detection and characterisation

301

2 Separation and concentration of the target organism is achieved using IMS as previously described (Section 6.3.2). 3 Isolation by plating the IMS mixture onto cefixime tellurite sorbitol MacConkey agar (CT‐SMAC) and a second agar of the user’s choice. Escherichia coli O157 colonies are about 1 mm in diameter, transparent with a pale yellowish‐brown appearance. As stated in the ISO standard method, the enrichment broth may give heavy bacterial growth after overnight incubation, and this can give rise to problems recognising E. coli O157 colonies even on selective agars. Therefore, modified techniques such as plating out dilutions of the IMS mixture, or plating smaller volumes may be used, while also accepting this will decrease the sensitivity of the procedure. Five typical colonies are selected from each plate and streaked on nutrient agar, followed by incubation for 18–24 hours at 37°C. 4 Confirmation is by detection of indole formation and agglutination with E. coli O157 antiserum. 6.7.6  Shigella species

The differentiation between E. coli and Shigella is troublesome since the two organisms are genetically closely related. In particular, Shigella and enteroinvasive strains of E. coli are phenotypically very similar and they have a close antigenic relationship. Serologically Sh. dysenteriae

Peptone water dilutent 90 ml + sample 10 g

Method A Prepare serial ten-fold dilutions in peptone water diluent (10–1–10–6)

Spread-inoculate two plates of each dilution to sorbitol MacConkey agar

Incubate 35 °C, 18 h

Method B Tryptone soya broth + novobiocin 90ml

Incubate 35 °C, 20 h

HC agar

Incubation 35 °C, 20 h

Test non-sorbitol fermenting colonies serologically using 0157 and H7 antisera

Test sorbitol negative colonies serologically using O157 and H7 antisera

Figure 6.14  FDA‐BAM methods for isolation of enterohaemorrhagic Escherichia coli (EHEC).

302

The microbiology of safe food Table 6.6  Major differentiation characteristics between Escherichia coli and Shigella species. Property

Escherichia coli

Shigella species

Motility Lactose fermentation Indole fermentation Gas from glucose

+ + + +

− − − −

3 and E. coli 0124, Sh. boydii 8 and E. coli 0143, and Sh. dysenteriae 12 and E. coli 0152 appear identical. Distinguishing traits are given in Table 6.6. Other useful differential traits are: • Cultures that ferment mucate, utilise citrate or produce alkali on acetate agar are likely to be E. coli. • Cultures that decarboxylate ornithine are most likely to be Sh. sonnei. • Cultures that ferment sucrose are likely to be E. coli. Direct plating onto selective media is unlikely to be successful due to the close relationship of Shigella and E. coli. Hence the detection of Shigella is generally through use of the distinguishing biochemical differences of presumptive isolates (Table  6.6). For example, on MacConkey agar Shigella colonies initially appear as non‐lactose fermentors. The ISO method for Shigella species (ISO 21567:2004) has several stages: 1 Enrichment of sample in 9× volume of Shigella broth containing novobiocin, followed by anaerobic incubation at 41.5°C for 16–20 hours. 2 Selective plating on MacConkey agar, XLD agar and HE agar. These are incubated at 37°C for 20–24 hours. If no suspect colonies are visible then the plates are incubated for a further 18–24 hours. Five suspect colonies are picked and streaked on nutrient agar plates followed by incubation for 20–24 hours at 37°C. 3 Identification is by biochemical tests followed by serological analysis of positive isolates. Biochemical tests include growth on Triple Sugar Iron agar (TSI) for H2S and gas production, motility and a number of other phenotyping tests that can be achieved using commercial biochemical test kits (i.e. API 20E and Microbact). Polyvalent antisera are used with biochemically identified Shigella isolates to identify the species: Sh. flexneri, Sh. dysenteriae, Sh. boydii and Sh. sonnei. An abbreviated version of the FDA‐BAM method is given in Figure 6.15 for illustrative purposes. The FAD‐BAM method also uses anaerobic incubation as under these conditions Shigella cells can compete against other Enterobacteriaceae, and novobiocin is added to the media as a selective agent. Shiga toxins are virtually identical to the verocytotoxin produced by EHEC and therefore detection methods are applicable to both groups of foodborne pathogens. 6.7.7  Cronobacter genus

As previously described in Section  4.6, this organism is primarily associated with severe, albeit rare, neonatal infections. It does also cause infections in all other age groups. The link of some neonatal cases to contaminated reconstituted powdered infant formula has led to a revision in hygienic practices and microbiological criteria for this product. Similar to Salmonella isolation, the recovery of Cronobacter by conventional microbiology involves the three steps of pre‐enrichment, enrichment and plating on selective media. Just like Salmonella

Methods of detection and characterisation

303

225 ml Shigella broth + 25 g sample For Sh. sonnei add

For other Shigella spp., add

novobiocin (0.5 µg/ml)

novobiocin (3 µg/ml)

Adjust pH to 7.0 ± 0.2 Incubate anaerobically

Incubate anaerobically

44 °C, 20 hours

42 °C, 20 hours MacConkey agar 35 °C in air, 20 hours

Test for indole discard culture if positive Subculture presumptive colonies: Glucose broth, triple sugar iron agar, lysine decarboxylate broth, motility agar, tryptone water Incubate at 35°C, examine at 20 hours. Continue incubation to 48 hours

Discard cultures showing positive tests for the following: Motility, H2S, gas production, sucrose fermentation, lactose fermentation, indole

Confirm presumptive Shigella cultures as Gram-negative bacilli

Full biochemical screen

Serology Figure 6.15  FDA‐BAM method for enrichment culture of Shigella species in foods.

testing, this method is primarily a presence/absence screening and not enumeration. However, by using multiple different sample volumes, an MPN estimate of Cronobacter numbers can be made. Pre‐enrichment

Just like Salmonella detection, this uses BPW, but a 10 g sample volume is added to 90 ml BPW as the microbiological criteria are the absence of Cronobacter in 10 g quantities from a batch of infant formula.

304

The microbiology of safe food

Enrichment

The overnight BPW mixture pre‐enrichment is then enriched for any Cronobacter by adding to EE broth or a more specific Cronobacter selective broth. Selective plating

The enrichment broth is streaked for single colony isolation on Druggan–Forsythe–Iversen (DFI) agar or a similar chromogenic agar (Iversen et  al. 2004; Iversen and Forsythe 2007; Plate  5). Using chromogenic agars that differentiate Cronobacter from most other Enterobacteriaceae is necessary as the enrichment broths enrich for Enterobacteriaceae and not just Cronobacter. The DFI agar contains the chromogen 5‐bromo‐4‐chloro‐3‐indolyl‐α‐D‐ glucopryranoside (X‐αGlc) as Cronobacter have α‐glucosidase activity, which results in blue‐ green colony formation after 18 hours at 37°C. A number of other Enterobacteriaceae, including Salmonella and Proteus, have α‐glucosidase activity and are H2S producers. So the medium also contains sodium thiosulphate and ferric ammonium citrate, which together act as a H2S indicator distinguishing these organisms from Cronobacter by their black colonies. An added benefit is that Salmonella is also detected on this agar along with Cronobacter. Sodium desoxycholate is also an ingredient to suppress the growth of Gram‐positive organisms. 6.7.8  Aeromonas species

The detection of Aeromonas is not routinely undertaken by food microbiology laboratories; however, it is included here due to its association with contaminated water and food. A number of selective and differential isolation media can be used for the recovery of Aeromonas species from the environment, foods and clinical specimens. However, since no individual medium is optimal for the recovery of Aeromonas species, a number of combinations of approaches can be undertaken: direct plating, membrane filtration and most probable number. Approaches include: (i) starch ampicillin agar (SAA), bile salts inositol brilliant green agar (BIBG) with initial enrichment in alkaline peptone water (APW), or (ii) tryptose broth containing ampicillin (TSB‐30, ampicillin 30 mg/l) with commercially available media such as Aeromonas Medium (Ryan’s Medium). Starch glutamate ampicillin penicillin (SGAP‐10) medium has been used to detect aeromonads from foods. To facilitate recovery of aeromonads from highly contaminated samples such as faeces, an initial stage is incubation overnight in an enrichment broth such as APW and subculturing onto cefsulodin irgasan novobiocin (CIN) agar and blood ampicillin agar (10 mg/l ampicillin). The plates are then incubated aerobically at 35°C for 24–48 hours. Aeromonas species produce distinctive colonies, with or without haemolysis on blood agar. Colonies are screened by ­carrying out oxidase tests and identified phenotypically. 6.7.9  Arcobacter species

Arcobacter are not routinely tested for by food microbiology laboratories, nevertheless due to their association with contaminated food and water they have been included in this chapter. Methods of isolation are similar to those for Campylobacter species due to their physiological and genetic relatedness. In general Arcobacter species are more aerotolerant and have lower growth temperature limits than Campylobacter species. Their isolation requires selective media such as mCCDA and CAT. On blood‐based agars Arcobacter species produce round white, off‐white or greyish colonies, 2–4 mm in diameter after 3 days of incubation. The colonies are generally small, non‐pigmented and convex with entire edges. Swarming has been reported on fresh agar. Brain heart infusion agar supplemented with 0.6% (w/v) yeast extract and 10% (w/v) blood agar has been used recently for routine culturing.

Methods of detection and characterisation

305

Arcobacter isolates can be presumptively identified by their shape (small comma‐shaped or spiral rods) and motility (darting or corkscrew motility). They can be easily distinguished from Campylobacter and related genera by their ability to grow in air at 25°C. The main phenotypic traits used for Arcobacter species differentiation are catalase activity, nitrate reduction, cadmium chloride susceptibility, microaerophilic growth at 20°C, and growth on MacConkey agar and in the presence of 3.5% NaCl and 1% glycine. Organic acids and amino acids are used as carbon sources. Hydrogen is not required for growth. All Arcobacter isolates hydrolyse indoxyl acetate. A simple diagnostic characteristic useful for the presumptive identification of C. jejuni, C. coli and Arcobacter species is the cadmium chloride test. Isolates can be identified using 16S rRNA gene probes and genotyped. 6.7.10  Listeria monocytogenes

Unlike the isolation procedures for salmonellae, E. coli and Cronobacter, pre‐enrichment is not commonly used for the isolation of Listeria species. This is because other organisms present will outgrow Listeria cells. Instead various enrichment media have been developed and have regulatory approval. A common enrichment broth is Fraser broth (modified from UVM broth), which employs aesculin hydrolysis coupled with ferrous iron as an indicator of presumptive Listeria species. Enrichments are streaked onto agars such as ALOA, Oxford and PALCAM. Oxford agar is often incubated at 30°C, whereas PALCAM agar is incubated at 37°C under microaerophilic conditions; Plate  14. A large number of selective agents are used in Listeria media, such as acriflavine, cycloheximide, colistin and polymyxin B as L. monocytogenes can be quickly outgrown by competing flora. Typical L. monocytogenes colonies are surrounded by a black zone due to black iron phenolic compounds. On PALCAM agar the centre of the colony may have a sunken centre after 48 hours of incubation. Presumptive L. monocytogenes colonies are confirmed using biochemical and serological testing. Most non‐Listeria isolates can be eliminated using the motility test, catalase test and Gram staining. Listeria species are short Gram‐positive rods, catalase‐positive and are non‐ motile if incubated above 30°C. Motility of cultures grown at room temperature is characterised by a tumbling action. Listeria monocytogenes is β‐haemolytic on horse blood agar. The CAMP test (named after Christie, Atkins, Munch and Peterson) is used for species differentiation (Plate 15). The Listeria isolates are streaked on sheep blood agar and St. aureus NCTC 1803 and Rhodococcus equi NCTC 1621 are streaked in parallel close to the Listeria streaks (Figure 6.16). The phenomenon of enhanced zones of haemolysis is observed (Table 6.7). The ISO 11290 method for L. monocytogenes is in two parts: detection and enumeration methods (Figure 6.17). The detection method has four stages: 1 Primary enrichment is in a selective enrichment broth that has a reduced concentration of the selective agents acriflavine and nalidixic acid (i.e. half‐Fraser broth). Caution: this medium uses lithium chloride, which gives a strong exothermic reaction with water, and irritates the mucous membranes of the throat. Normally the ratio of test sample to broth is 1:9. This is incubated at 30°C for 24 hours before inoculating with the secondary enrichment broth and also plating onto selective agar. 2 Secondary enrichment in selective enrichment broth with full strength selective agents (i.e. Fraser broth), for 48 hours at 35 or 37°C. Typically 0.1 ml of the incubated primary enrichment mixture is added to 10 ml Fraser broth. After incubation the mixture is streaked onto selective agars. 3 Plating out and presumptive identification on two selective agars: ALOA (or equivalent formulation) and a second agar according to the user’s choice, such as Oxford agar or PALCAM. ALOA plates are incubated at 37°C for 24 hours, and up to 48 hours if necessary. As given

306

The microbiology of safe food Enrichment Test sample 1:

Test sample 2:

25 g + FDA enrichment broth 225 ml

25 g + FDA enrichment broth

After 24 + 48 hours, 30°C

225 ml, without selective agents + 0.1% sodium pyruvate Incubate 6 hours, 30°C Add selective agents. After 24 + 48 hours, 30°C

Selective diagnostic isolation

Oxford agar

LPM agar

Incubate 24–48 hours, 35°C

Incubate 24–48 hours, 30°C

After 24 and 48 hours

After 24 and 48 hours

Examine for brownish-black

Examine by Henry

colonies with brown halos

illumination

Subculture to tryptone soya agar + 0.6% yeast extract CAMP test motility test and biochemical tests Subculture to tryptone soya broth + 0.6% yeast extract Subculture to tryptose broth Serological tests Figure 6.16  FDA/BAM Listeria isolation procedure.

Table 6.7  CAMP reactions for Listeria species. Listeria species

Staphylococcus aureus

Rhodococcus equi

Listeria monocytogenes Listeria seeligeri Listeria ivanovii

+ + −

− − +

Methods of detection and characterisation

Zone of enhanced haemolysis

307

L.monocytogenes St. aureus

L. ivanovii

L. seeligeri

Rhodococcus equi

Figure 6.17  The CAMP test for haemolysis testing of L. monocytogenes.

above the selective agars are inoculated with primary and secondary enriched cultures. On ALOA typical Listeria colonies are green‐blue surrounded by an opaque halo. 4 Confirmation using morphological, physiological and biochemical tests. Five presumptive Listeria colonies are picked from each plate, and streaked on tryptone soya yeast extract agar (TSYEA), and incubated at 35 or 37°C for 18–24 hours. Confirmation tests include Gram stain, catalase production and motility of culture grown at 25°C. Listeria are small Gram‐positive rods, which are catalase‐positive and are motile with a tumbling motion. Listeria monocytogenes isolates are confirmed by positive β‐haemolytic reaction on sheep blood agar plates, utilisation of rhamnose but not xylose sugars and enhanced β‐haemolysis by St. aureus and not Rh. equi in the CAMP test (Plate  15). When observed by the Henry illumination tests, Listeria species colonies have a bluish colour and a granular surface. This test involves observing the colonies that are illuminated underneath by reflected white light. The ISO enumeration method has several stages: 1 Sample preparation in the ratio 1:9 in BPW or half‐Fraser broth (without selective agents). 2 Resuscitation for 1 hour at 20°C before inoculation of selective agar given below. After which the selective agents (lithium chloride, acriflavine and nalidixic acid) can be added and the sample incubated as given above for the detection method. 3 Inoculation of PALCAM agar; usually 0.1 ml or 1 ml if low numbers of L. monocytogenes are expected. Incubate plates at 35 or 37°C for 24 hours, or incubate for a further 24 hours if only slight growth or no growth is obtained. 4 Identification and enumeration by counting colonies with typical Listeria morphology. 5 Confirmation of Listeria species, and L. monocytogenes by selecting five presumptive colonies from each plate. Same tests are applied as previously described in the detection method.

308

The microbiology of safe food

6.7.11  Staphylococcus aureus

Large numbers of St. aureus cells are required to produce sufficient amounts of heat‐stable toxin. Therefore small numbers of St. aureus in food are of little significance, and so an enrichment step is not used for the organism’s isolation. Consequently tests for viable cells are applicable for samples before heat treatment, and tests for the enterotoxin and heat‐stable thermonuclease are for heat‐treated samples. The Baird–Parker agar is the most widely accepted selective agar for St. aureus (Plate 16). This medium includes sodium pyruvate to aid the resuscitation of injured cells. The selectivity is due to the presence of tellurite, lithium chloride and glycine. Staphylococcus aureus forms black colonies due to tellurite reduction and clearance of egg yolk due to lipase activity. Glycine acts as a growth stimulant and is an essential component of the staphylococcal cell wall. An alternative medium, mannitol salt agar, has better recovery efficiency of St. aureus from cheese. The selective agent is salt (7.5%) and mannitol fermentation is indicated by the pH indicator phenol red (reddish‐purple zones surrounding St. aureus colonies). The coagulase test (clotting of diluted mammalian blood plasma) is a reliable test for pathogenic St. aureus. DNase production correlates with the coagulase test and is therefore also indicative of pathogenicity. Testing for DNase activity used to involve growing the organism on agar containing DNA and then flooding with HCl to visualise zones of DNA degradation: a clear zone being due to the lack of precipitated DNA. However, the acid kills the organism resulting in a non‐viable culture. Therefore, current alternatives use indicator dyes toluidine blue and methyl green, which are included in the agar medium. The dyes form coloured complexes with the DNA and so presumptive pathogenic St. aureus colonies will show colour changes on DNA hydrolysis: toluidine blue produces pink zones, whereas methyl green goes almost colourless. Staphylococcal enterotoxins can be detected using RPLA (Section 6.4). The limit of sensitivity is about 0.5 ng of enterotoxin per g food. A number of enzyme immunoassays are available for staphylococcal enterotoxin detection. ELISA kits are also available that have a detection limit of >0.5 μg toxin per 100 g food and require 7 hours to obtain the result. Staphylococcus aureus causes a variety of diseases including boils and carbuncles, but in food‐related illness it is principally the production of heat‐stable toxins, termed enterotoxins, that causes vomiting (Section 4.13.3). Initially the enterotoxins were divided serologically into five major types – SEA to SEE – of which SEC is further divided into three variants. More recently a number of other SEs have been identified: SEG to SEQ. In addition, SEs are members of the pyrogenic toxin (PT) superantigens. 6.7.12  Clostridium perfringens

Clostridium perfringens is a strict anaerobe which produces spores that can survive heating processes. Therefore enrichment is required to detect low numbers of Clostridium cells, which may be outnumbered by other organisms (Figure 6.18). Numerous media include sulphite and iron, which result in a characteristic blackening of Cl. perfringens colonies (Table 6.2). However, this blackening reaction is not limited to Cl. perfringens and hence the term ‘sulphite reducers’ is often used instead of specifying the species Cl. perfringens. The lecithinase (phospholipase c) activity of Cl. perfringens is also a common test in diagnostic media resulting in opaque zones surrounding the colonies. Selectivity is by the inclusion of cycloserine or neomycin. All media are incubated under anaerobic conditions, which are either generated using an anaerobe jar or an anaerobic cabinet. Tests to distinguish Cl. perfringens from other anaerobic sulphite reducers are the Gram stain reaction, metronidazole sensitivity, the Nagler reaction and the reversed CAMP test. The

Methods of detection and characterisation Anaerobic plate count

309

Optional enrichment if cell numbers are very low

Blend sample 1/10 in 0.1%

Suspend sample (0.2 g) in 2 ml of liver

peptone water

broth or peptone–glucose–yeast extract broth

Decimal dilute to 10–7 in 0.1%

Incubate 18–24 hours, 35–37°C

peptone water

Plate out (0.1 ml in duplicate)

Plate out tubes showing gas on TSC agar with

on TCS with egg yolk or 1 ml on

gas production on TSC agar with egg yolk

TSC without egg yolk

Overlay 5–10 ml of egg yolk

Incubate anaerobically 18–24 hours, 35–37°C

free TSC agar

Incubate 18–24 hours, 35–37°C

Confirmatory tests on egg yolk positive black colonies

Confirmatory tests on egg yolk positive black colonies on TSC with egg yolk and black colonies on TSC without egg yolk Figure 6.18  Procedure for the isolation and quantification of Clostridium perfringens.

Nagler test uses Cl. perfringens type A antitoxin to neutralise lecithinase activity. The reversed CAMP test involves streaking sheep blood agar with Streptococcus agalactiae and the test isolate at right angles, without touching. After anaerobic incubation at 37°C for 24 hours, a positive result is indicated by arrow‐shaped areas of synergistic enhanced haemolysis at the junction of the two streaks (Figure 6.19). Production of ‘stormy clot’ in litmus milk and the detection of acid phosphatase are useful confirmatory tests. A number of biological methods are available including the rabbit ligated ileal loop tests, which though very effective and widely used, do require live animal testing. To date, few commercially produced kits are available for the detection of the extracellular toxins produced by Cl. perfringens. RPLA (Section 6.4.2) is available for Cl. perfringens enterotoxin. 6.7.13  B. cereus, B. subtilis and B. licheniformis

Enrichment is not normally used for the detection of B. cereus since low numbers of the organism are not regarded as being of significance (Figure 6.20). Direct plating onto selective media containing the antibiotic polymyxin B is often used. The two key distinguishing features incorporated into media design are the demonstration of phospholipase c activity and the inability to produce acid from the sugar mannitol as shown in Plate 17. If it is necessary to only count

310

The microbiology of safe food Area of enhanced haemolysis at junction of α-toxin and CAMP factor

Area of Cl. perfringens α-toxin

Horizontal streak of Cl. perfringens

Vertical streak of St. agalactiae

Area of CAMP factor β -haemolysis Figure 6.19  Reversed CAMP test for Clostridium perfringens haemolysis.

spores, then the vegetative cells must be killed by heat treatment (1:10 dilution, 15 minutes, 70°C) or alcohol treatment (1:1 dilution in 95% ethyl alcohol, 30  minutes at room temperature). Bacillus subtilis and B. licheniformis can be isolated easily using routine non‐selective media. They have a similar appearance on PEMBA medium, but are distinguishable from B. cereus. ELISA and RPLA tests are commercially available for Bacillus diarrhoeal enterotoxin. They have a sensitivity limit of 1 ng toxin/ml of material and take approximately 4 hours to obtain a result. However, no test has been developed for the emetic toxin due to problems of purification. 6.7.14 Mycotoxins

The toxins are visualised under UV light since the four naturally produced aflatoxins, B1, B2, G1 and G2, are named after the blue and green fluorescence they produce. Trichothecenes, fumonisins and moniliformin (from Fusarium species) can be detected using high‐performance liquid chromatography. ELISA and latex agglutination assays are commercially available for the detection of aflatoxins. ELISA assays can be used for the quantitative detection of moulds in foods. Compared with ELISA tests, which require 5–10 hours to complete, the latex agglutination method is much faster taking only 10–20 minutes. Test kits for mycotoxins have been developed that are rapid and simple to use. They are designed to be used in the field, country elevators, feed mills and processing plants. 6.7.15 Viruses

In most studies of food and waterborne viruses, samples have been screened for viruses by electron microscopy, DNA probe, ELISA or tissue culture. But not all these techniques are feasible with all viruses, i.e. norovirus viruses cannot be grown in tissue culture. Electron

Methods of detection and characterisation

311

Vegetative cells Make an initial dilution (1/10) of the sample in an appropriate diluent (e.g. phosphate buffer saline, maximum recovery diluent, peptone water) Decimal dilute to 10–6

Incubate duplicate plates of B. cereus selective medium (PEMBA) with 0.1 ml of 10–3 –10–6 dilutions

Spread the inoculum over the entire surface

Incubate aerobically for 24–48 hours, 35–37°C

Examine for peacock-blue colonies with blue egg yolk precipitate zone

Confirm with rapid screening procedure

If necessary, verify with biochemical tests

Spores If necessary to count spores, then first treat with heat or alcohol to destroy the vegetative cells Heat treatment: Heat the initial 1/10 suspension for 15 minutes 70°C. Proceed as for detection of vegetative cells Alcohol treatment: Dilute the initial suspension 1:1 in 95% ethyl alcohol. Leave for 30 minutes, room temperature. Proceed as for detection of vegetative cells, adjusting the dilutions to account for the 1:1 dilution of the sample suspension in alcohol Figure 6.20  Typical procedure for detection of Bacillus cereus.

microscopy has a limit of sensitivity of 105–106 particles per ml of faecal suspension. ELISA‐ based assays have been developed for group A rotavirus and adenovirus in clinical samples only. Although the presence of bacterial indicators (Enterobacteriaceae, E. coli) does not correlate with the presence of viral pathogens, there are no established virus detection methods for foods other than shellfish. Consequently foodborne viral gastroenteritis is not usually diagnosed and detection methods for viruses and viral genomic material have not been adopted by routine food analysis laboratories.

312

The microbiology of safe food

For epidemiological studies, large volumes (25–100 g) of food need to be analysed for enteric viruses since the level of food contamination is assumed to be low. If tissue culture is to be used to detect enteric viruses, then the viruses need to be separated from the bulk of the food. The viruses may be enumerated using plaque assays with dilution endpoint assays. Cell culture technique, however, is not applicable for noroviruses (formerly Norwalk‐like virus and small round structured viruses) due to the lack of host cell line, and is only moderately successful with hepatitis A virus (HAV). The sequencing of norovirus and HAV genomes has enabled the reverse transcriptase polymerase chain reaction (RT‐PCR) to be developed for detection and characterisation of the viruses in faecal, vomit and shellfish samples. The norovirus can be characterised by sequencing the amplified products of the test (amplicons), and this has enabled outbreaks in different locations to be linked to a single source. Electron microscopic analysis for norovirus is only of use with samples during the first 2 days of symptoms, whereas RT‐PCR is usable within 4 days of symptoms appearing (Atmar and Estes 2001). HAV genome is detectable by RT‐PCR in artificial sterile seawater for 232 days, compared with only 35 days for cell culture. Therefore, RT‐PCR is not a reliable indicator of infectious HAV.

7 Microbiological criteria

7.1  Background to microbiological criteria and end‐product testing

Testing food at the end of production for micro‐organisms (‘end‐product testing’) has been standard practice in the food industry for decades. However, a statistical appreciation of its usefulness has been largely overlooked. In 1974 the International Commission on Microbiological Specifications for Foods (ICMSF) wrote an excellent text regarding the setting of microbiological criteria. The book was written at a time of increasing global food transportation and was primarily aimed at being applied to food entering a country (port of entry) with no known history. Nevertheless these criteria have been applied within industry for their own products despite in‐house knowledge of the product. It is very important nowadays to recognise that microbiological testing of foods should be carried out under the umbrella of Hazard Analysis Critical Control Point (HACCP) as part of the verification principle (Section 9.7). In other words, end‐product testing in itself does not guarantee a safe food product, but could support the HACCP plan implementation. Nevertheless microbiological criteria (levels of microbes acceptable in a particular food) are required by regulatory authorities and between companies in a supply chain. The specifics of these criteria are often historical, not necessarily the most appropriate. Also, as in the case of criteria between companies, they are often confidential. These criteria may eventually be replaced by ‘Food Safety Objectives’ (Section 10.7). 7.2 International commission on microbiological specifications for foods (ICMSF)

In the 1960s the role of micro‐organisms in foodborne disease had become well recognised, and in addition, international trade had significantly increased. However, microbiological testing of food was hampered due to the lack of standardised methods and the use of sampling plans that lacked statistical validity. Consequently in 1962 the ICMSF was formed by the International Committee on Food Microbiology and Hygiene. The ICMSF’s aims were to: 1 Assess data on the microbiological safety and quality (including spoilage) of foods. 2 Determine if the use of microbiological criteria would both improve and assure the ­microbiological safety of certain foods. 3 Recommend methods of sampling and examination.

The Microbiology of Safe Food, Third Edition. Stephen J. Forsythe. © 2020 John Wiley & Sons Ltd. Published 2020 by John Wiley & Sons Ltd.

313

314

The microbiology of safe food

These objectives have evolved as knowledge of foodborne micro‐organisms and their control has increased and is reflected in the series of ICMSF books and papers. The first ICMSF book (initially published in 1968 and revised in 1978, 1982 and 1988) was centred upon encouraging a comparison of worldwide testing methods in order to obtain agreed methods for use in international trade. In 1974 the ICMSF published Microorganisms in Foods 2. Sampling for Microbiological Analysis: Principles and Specific Applications. This recognised the need for scientifically based sampling plans for foods in international trade. It described two‐ and three‐class sampling plans and ‘choice of case’. The sampling plans were originally designed for application at port of entry; that is, when there is no prior knowledge on the history of the food. This pioneering work set forth the principles of sampling plans for the microbiological evaluation of foods and is also known as attributes and variables sampling depending on the extent of microbiological knowledge of the food. The second edition of the book in 1986 took note of the successful application of the acceptance sampling plans on a worldwide basis, not only at an international level but at national and local levels by both industry and regulatory agencies. The ICMSF book has subsequently been updated in 2002. The third (two‐volume) ICMSF book was on the Microbiological Ecology of Foods (Volume 1, ICMSF 1980a) and Factors Affecting Growth and Death of Microorganisms (Volume 2, ICMSF 1980b). In addition, this detailed the microbial load, spoilage patterns and microbial hazards of 14 food commodity groups. The use of HACCP as the most assured means of safety food production was the subject of the Commission’s fourth book (ICMSF 1988). In order to support HACCP and good hygienic practices (GHP), in 1996 the Commission produced their fifth book as a reference source on the growth and death response of foodborne pathogens (ICMSF 1996a). Due to increased knowledge of foodborne pathogens in 1998, the Commission updated its earlier 1980b publication on 16 commodity foods (ICMSF 1998c). This was again with regard to typical microbial loads, spoilage patterns and control measures. More recently (2002) the ICMSF has produced book seven (Microbiological Testing in Food Safety Management), which is a revision of the 1986 publication. It has a threefold purpose: 1 Supporting the use of statistically valid sampling plans at port of entry (i.e. where there is no prior knowledge concerning the processing conditions). 2 Demonstrating the application of HACCP and GHP for structured food safety management, along with end‐product testing (cf. Chapter 9). 3 Recommending the incorporation of ‘food safety objectives’ (Section 10.7) as a means of translating ‘risk’ into the principles of HACCP and GHP. 7.3 Codex Alimentarius principles for the establishment and application of microbiological criteria

The Codex Alimentarius Commission has become the reference for international food safety requirements (cf. Section 12.4 for more detail). The Codex Alimentarius Commission (1997b, p.1) definition of a microbiological criterion is: A microbiological criterion for food defines the acceptability of a product or a food lot, based on the absence or presence, or number of microorganisms including parasites, and/or quantity of their toxins/metabolites, per unit(s) of mass, volume, area or lot.

Microbiological criteria

315

Codex Alimentarius Commission (1997b, c) requires that a microbiological criterion consists of: • A statement of the micro‐organisms of concern and/or their toxins/metabolites and the reason for that concern. • The analytical methods for their detection and/or quantification. • A plan defining the number of field samples to be taken and the size of the analytical unit. • Microbiological limits considered appropriate to the food at the specified point(s) of the food chain. • The number of analytical units that should conform to these limits. A microbiological criterion should also state: • The food to which the criterion applies. • The point(s) in the food chain where the criterion applies. • Any actions to be taken when the criterion is not met. The value of microbiological testing as a control measure varies along the food chain. Therefore, a microbiological criterion should be established and applied only where there is a definite need and where its application is practical. Such need is demonstrated, for example, by epidemiological evidence that the food under consideration may represent a public health risk and that a criterion is meaningful for consumer protection, or as the result of a risk assessment. The criterion should be technically attainable by applying good manufacturing practices (GMPs) (Codes of Practice). Criteria should be reviewed periodically for relevance with respect to emerging pathogens (Section 4.18), changing technologies and new understandings of science. A sampling plan includes the sampling procedure and the decision criteria to be applied to a lot, based on examination of a prescribed number of sample units and subsequent analytical units of a stated size by defined methods. A well‐designed sampling plan defines the probability of detecting micro‐organisms in a lot, but it should be borne in mind that no sampling plan can ensure the absence of a particular organism. Sampling plans should be administratively and economically feasible (Codex Alimentarius Commission 1997b, c). In particular, the choice of sampling plans should take into account: 1 Risks to public health associated with the hazard. 2 The susceptibility of the target group of consumers. 3 The heterogeneity of distribution of micro‐organisms where variable sampling plans are employed. 4 The acceptable quality level (AQL) and the desired statistical probability of accepting a non‐ conforming lot. The AQL is the percentage of non‐conforming sample units in the entire lot for which the sampling plan will indicate lot acceptance for a prescribed probability (usually 95%). For many applications, two‐ or three‐class attribute plans may prove useful. Microbiological criteria should be based on scientific analysis and advice and, where sufficient data are available, on a risk analysis appropriate to the foodstuff and its use (CAC 1997b). These criteria may be relevant to the examination of foods, including raw materials and ingredients of unknown or uncertain origin, or when no other means of verifying the efficacy of HACCP‐based systems and GHP are available. Microbiological criteria may also be used to determine that processes are consistent with the general principles of food hygiene. Microbiological criteria are not normally suitable for monitoring Critical Limits as defined in the HACCP system.

316

The microbiology of safe food

The purpose of establishing microbiological criteria is to protect the public’s health by providing food that is safe, sound and wholesome, and to meet the requirements of fair trade practices. The presence of criteria, however, does not protect the consumer’s health since it is possible for a food lot to be accepted that contains defective units. Microbiological criteria may be applied at any point along the food chain and can be used to examine food at the port of entry and at the retail level. The statistical performance characteristics or operating characteristics curve should be provided in the sampling plan (Section 7.4). Performance characteristics provide specific information to estimate the probability of accepting a non‐conforming lot. The time between taking the field samples and analysis should be as short as reasonably possible, and during transport to the laboratory the conditions (e.g. temperature) should not allow increase or decrease of the numbers of the target organism, so that the results reflect – within the limitations given by the sampling plan – the microbiological conditions of the lot. 7.4  Sampling plans

In addition to the ICMSF 1986 and 2002 publications on sampling plans, Harrigan and Park (1991) wrote an excellent book on their practical mathematics. Just as it is impractical to test a sample for every possible food pathogen, so it is also impractical to destructively test 100% of an ingredient or end‐product. Whilst it is accepted that no sampling plan can guarantee the absence of a pathogen in a batch of food (food lot), there is a need to use sampling plans to appropriately test a batch of material and give a statistical basis for acceptance or rejection of a food lot. Microbiological sampling plans are frequently used in food production, import control and in contractual agreements with suppliers and customers. Sampling plans are used to check the microbiological status of a commodity, its compliance to safety requirements and adherence to GHP (Section 9.11) during or after manufacture. The results from single sample examinations may give valuable baseline data which can be used for trend analysis, particularly where samples form part of a specific survey. However, statistical principles should be observed when sampling particular food commodities many of which (usually end‐products) are heterogeneous, even when they have a similar formulation. In situations where a food inspector might be concerned about a particular food, a sample taken for microbiological analysis may provide evidence that food hygiene regulations have been contravened or may provide the basis for additional inspection and/or examination. The single sample concept is likely to retain a role in assessing food safety in small‐scale food production businesses, which will have fewer resources for implementation of HACCP, but will nevertheless have to take proper account of the risk to public health posed by each individual operation. There are two types of sampling plans: 1 Variables plans, when the microbial counts conform to a ‘log‐normal’ distribution (Section  7.5). These data would be known by a producer and are not applicable to an importer at the port of entry situation. These are less frequently used as primarily food is produced and widely distributed, rather than regionally. 2 Attribute plans, when no prior knowledge of the distribution of micro‐organisms in the food is known, i.e. at port of entry or the distribution of target organism is not log‐normal (Section 7.6). Attributes sampling plans can be either according to a two‐class plan or a three‐class plan. The two‐class plan is used almost exclusively for pathogens, whereas a three‐class plan is frequently

Microbiological criteria

317

used to examine for hygiene indicators. The main advantage of using sampling plans is that they are statistically based and provide a uniform basis for acceptance against defined criteria. The type of sampling plan required can be decided using Figure 7.1. Attributes plans also involve the concept of ‘choice of case’, based on microbiological risk. ‘Case’ is a classification of sampling plans ranging from 1 (least stringent) to 15 (most stringent). The choice of case, and therefore the sampling plan, depends on: • The relative severity of the hazard to food quality or consumer health on the basis of the micro‐organisms involved. See Chapters 4 and 5. • The expectation of their destruction, survival or multiplication during normal handling of the food. See Chapters 2 and 3. Table 7.1 and the decisions trees of Figures 7.1 and 7.2 should be referred to aid deciding on the appropriate sampling plan. For example, cases 1–3 refer to utility applications, such as shelf life, whereas cases 13, 14 and 15 refer to severely hazardous foodborne pathogens. The severity of the microbiological hazard has been covered in Chapters 4 and 5 and the foodborne pathogens grouped to assist in referring to Table 4.2.

Organism in question is to be measured by: Presence or absence test

Enumeration

Use a two-class plan

Use a three-class plan*

Is it permissible to accept

Choose the n and c values

presence of the target

(Table 7.6) for desired

organism in food?

probability

No

Yes

c=0

c>0

Choose value of n

Choose value of n and c

to give desired probability

to give desired probability

(Table 7.4)

(Table 7.5)

Figure 7.1  Decision tree for choosing a sampling plan Source: Adapted from ICMSF (1986). * A variable plan may be applicable if the organism is distributed in a log‐normal fashion.

318

The microbiology of safe food

Table 7.1  Sampling plans in relation to degree of health hazard and conditions of use. Conditions in which food is expected to be handled and consumed after sampling

Type of hazard No direct health hazard Utility, for example reduced shelf life and spoilage Health hazard Low, indirect (indicator) Moderate, direct, limited spread Moderate, direct, potentially extensive spread Severe, direct

Reduce degree of hazard

Cause no change in hazard

May increase hazard

Case l Three‐class, n = 5, c = 3 Case 4 Three‐class, n = 5, c = 3 Case 7 Three‐class, n = 5, c = 2 Case 10 Two‐class, n = 5, c = 0 Case 13 Two‐class, n = 15, c = 0

Case 2 Three‐class, n = 5, c = 2 Case 5 Three‐class, n = 5, c = 2 Case 8 Three‐class, n = 5, c = 1 Case 11 Two‐class, n = 10, c = 0 Case 14 Two‐class, n = 30, c = 0

Case 3 Three‐class, n = 5, c = 1 Case 6 Three‐class, n = 5, c = 1 Case 9 Three‐class, n = 10, c = 1 Case 12 Two‐class, n = 20, c = 0 Case 15 Two‐class, n = 60, c = 0

Source: ICMSF (1986). Reprinted with permission of the University of Toronto Press.

Sampling plans and recommended microbiological limits were published by ICMSF (1986) for the following foods: 1 raw meats, processed meats, poultry and poultry products; 2 pet foods; 3 dried milk and cheese; 4 pasteurised liquid, frozen and dried egg products; 5 seafoods; 6 vegetables, fruit, nuts and yeast; 7 cereals and cereal products; 8 peanut butter and other nut butters; 9 cocoa, chocolate and confectionery; 10 infant and certain categories of dietetic foods; 11 bottled water. 7.5  Variables plans

Variables plans can be applied when the micro‐organisms in the food are distributed ‘log‐normally’; that is the logarithms of the viable counts conform to a normal distribution (Figure 7.3; Kilsby et al. 1979). This applies to certain foods that have been analysed over a period of time by the producer and therefore does not apply at port of entry. If the micro‐organisms’ distribution within a lot is log‐normal then sampling plans can be used to develop acceptance sampling plans. The sample mean (x) and standard deviation (s) are determined from previous studies. The sample mean and standard deviation are used to decide whether a ‘lot’ of food should be accepted or rejected.

Microbiological criteria Is the criterion to be applied at the food production site?

Yes

Yes

No

Origin of food known?

Are GHP and HACCP expected to be applied and verified?

Is criterion to be used for official control? No

319

No

No

Producer to consider need for own criterion

Is there evidence of a risk to health for this food?

No criterion should be established

No Yes Will the application of a criterion benefit public health?

No

Yes Indicator Case 5

Pathogen Potential for unacceptable multiplication during storage, distribution, preparation or use?

Case 8, 11 or 14

Yes Case 6

Potential for acceptable reduction during storage, distribution, preparation or use?

Case 4

Killing before consumption assured?

Case 9, 12 or 15

Yes Case 7, 10 or 13

Yes No criterion should be established Figure 7.2  Decision tree for choice of criteria for microbiological pathogens and indicator organisms.

In addition: The proportion (pd) of units in a lot that can have a concentration above the limit value, V must be decided. The desired probability P can be chosen where P is the probability of rejecting a lot that ­contains at least a proportion pd above V. The lot of food is rejected if x + k1 s > V

320

The microbiology of safe food

Frequency of occurrence (%)

35 30 25 20 15 10 5 0 0

1

2

3

4

5

6

Log10 conc (cfu/g) Figure 7.3  Normal log distribution of a micro‐organism.

Table 7.2  Safety and quality specification (reject if x + k1 s > V). Number of sample units

Probability (P) of rejection

Proportion (pd) exceeding V

3

5

7

10

0.95

0.05 0.1 0.3 0.1 0.25

7.7 6.2 3.3 4.3 2.6

4.2 3.4 1.9 2.7 1.7

3.4 2.8 1.5 2.3 1.4

2.9 2.4 1.3 2.1 1.3

0.90

Source: adapted from ICMSF (1986) and reprinted with permission of the University of Toronto Press.

Where: k1 is obtained from reference tables (Table 7.2) according to the pd and P values. It is therefore dependent upon the stringency of the sampling plan and number of sample units, n, analysed. V is the microbial count as a log‐concentration that has been set as a safety limit. Deciding that a lot would be rejected if 10% (pd = 0.1) of samples exceeded the value V, with a probability of 0.95 and taking five sample units (n) gives k1s as 3.4. The more samples (n) are taken, the lower the chance of rejecting an acceptable lot of food. The value of V is set by the microbiologist from previous experience. It can be similar to M in the three‐class plan (Section 7.6.2). For example: The aerobic plate count for ice cream from the old milk products Directive 92 / 46 / EEC gives M 500000 cfu / g. 500000 log 5. ThereforeV

5.

Microbiological criteria

321

Table 7.3  Determining the good manufacturing practice limit (accept if x + k2