Detection and Epidemiology of Coxiella Burnetii Infection in Beef Cattle in Northern Australia and the Potential Risk to Public Health

Detection and epidemiology of Coxiella burnetii infection in beef cattle in northern Australia and the potential risk to public health

Caitlin Medhbh Wood BVSc (Hons1)

MANZCVS (Veterinary Epidemiology)

https://orcid.org/0000-0003-2716-0402

A thesis submitted for the degree of Doctor of Philosophy at

The University of Queensland in 2020

School of Veterinary Science Abstract

Q fever has a longstanding history in Queensland, Australia. Although it is not a new disease to Australia, Q fever continues to burden the country with some of the highest annual case notification rates globally (Tozer 2015). While there are undeniable links with Q fever cases and exposure to cattle, it appears that coxiellosis predominantly goes undiagnosed and unnoticed in cattle within Australia. Knowledge of the prevalence and distribution of coxiellosis in cattle populations across northern Australia is limited due to minimal surveillance and no standardisation of diagnostic test methods. The overall aim of this PhD project was to improve the understanding of the epidemiology of coxiellosis in beef cattle in northern Australia and gain insights into the potential risk to public health.

Descriptive analyses of 2,838 human Q fever notifications from Queensland between 2003 and 2017 were initially reported. Queensland accounted for 43% of the Australian national Q fever notifications for this period. From 2013–2017 the most common identifiable occupational group was agricultural/farming. For the same period, at-risk environmental exposures were identified in 82% (961/1,170) of notifications; at-risk animal-related exposures were identified in 52% (612/1,170) of notifications; abattoir exposure was identified in 7% of notifications. Improved surveillance since 2012 has highlighted the need for further education and heightened awareness of Q fever risk for all people living in Queensland, not just those in previously considered high-risk occupations.

A focused molecular survey was conducted at an abattoir in Queensland aimed to: 1) estimate the prevalence of C. burnetii infection in a population of beef cattle going to slaughter, and 2) enable identification of specific genotypes of C. burnetii infecting cattle in this population. Cattle originating from several Queensland regions and northern New South Wales had reproductive tissue and liver tissue collected for molecular testing. C. burnetii DNA was detected, although at a lower than expected frequency, from the liver of cattle originating from Darling Downs South West, Central Queensland and North Queensland regions. No evidence of C. burnetii infection was detected in placental tissue or amniotic fluid samples from pregnant cattle post-mortem during this study.

During the next study, a human indirect immunofluorescence assay (IFA) was modified and validated for the detection of IgG antibodies against phase I and/or phase II C. burnetii in bovine sera and determined an optimal screening dilution cut-off to be 1:160. Direct comparison of the modified IFA with the commercial IDEXX enzyme-linked immunosorbent assay (ELISA)

i kit (Q Fever Ab Test IDEXX Laboratories, United States of America) was performed by testing 458 serum samples from four distinct cattle populations across the east coast of Australia and New Zealand. Results were then analysed using Bayesian latent class modelling, to validate the tests in the absence of a gold standard reference test. This analysis indicated that the IFA had an estimated diagnostic sensitivity (DSe) of 73.6% (95% Credible Interval (CrI) 61.1–85.9) and diagnostic specificity of (DSp) 98.2% (95%CrI 95.1–99.7). The commercial IDEXX ELISA kit was found to have a higher DSe of 87.9% (95%CrI 73.9–96.4) and similar DSp of 97.7% (95%CrI 93.2–99.7).

The IFA was used to test 2,012 sera samples from beef cattle managed on commercial properties located in Queensland and the Northern Territory. Bayesian latent class models were then developed to estimate the true prevalence of exposure, adjusted for diagnostic test sensitivity and specificity and incorporating the hierarchical structure of the cattle within properties and regions. In this study, cattle in the Northern Territory had lower estimated true prevalence than cattle within most regions of Queensland with the exception of south-east Queensland. Results from this study have provided baseline true prevalence estimates and described the geographic distribution of C. burnetii exposure in a sample of extensively managed beef cattle located across the tropical grazing regions of northern Australia.

Finally, C. burnetii IFA results from the previous chapter and additional molecular testing of vaginal swab samples from the same sample of cattle were examined to investigate the relationship between coxiellosis and reduced reproductive performance in beef cattle. A large dataset investigating causes of reduced reproductive performance in beef cattle across northern Australia was analysed and results indicated that high levels of C. burnetii exposure on properties was associated with reduced pregnancy rates, although further research is needed to confirm this hypothesis.

This thesis has explored several aspects of Q fever in humans and coxiellosis in beef cattle from the northern Australian context. Laboratory and epidemiological research outcomes have provided tools for improved surveillance in cattle and has improved our understanding of potential public health risks. In Australia, current Q fever control practices are isolated to human vaccination without any surveillance, control programs nor licensed vaccines in animals. Further research in livestock and wildlife will further improve our ability to control Q fever and manage health and productivity within agricultural industries.

Declaration by author

This thesis is composed of my original work, and contains no material previously published or written by another person except where due reference has been made in the text. I have clearly stated the contribution by others to jointly-authored works that I have included in my thesis.

I have clearly stated the contribution of others to my thesis as a whole, including statistical assistance, survey design, data analysis, significant technical procedures, professional editorial advice, financial support and any other original research work used or reported in my thesis. The content of my thesis is the result of work I have carried out since the commencement of my higher degree by research candidature and does not include a substantial part of work that has been submitted to qualify for the award of any other degree or diploma in any university or other tertiary institution. I have clearly stated which parts of my thesis, if any, have been submitted to qualify for another award.

I acknowledge that an electronic copy of my thesis must be lodged with the University Library and, subject to the policy and procedures of The University of Queensland, the thesis be made available for research and study in accordance with the Copyright Act 1968 unless a period of embargo has been approved by the Dean of the Graduate School.

I acknowledge that copyright of all material contained in my thesis resides with the copyright holder(s) of that material. Where appropriate I have obtained copyright permission from the copyright holder to reproduce material in this thesis and have sought permission from co- authors for any jointly authored works included in the thesis.

iii

Publications included in this thesis

Tozer S, Wood C, Si D, Nissen M, Sloots T, Lambert S. (2020) The improving state of Q fever surveillance: A review of Queensland notifications, 2003–2017, Communicable Diseases Intelligence, Vol. 44; pp 1-22

Wood C, Tan T, Muleme M, Barnes T, Bosward K, Alawneh J, McGowan M, Stenos J, Gibson J, Perkins N, Firestone S, Tozer S. (2019) Validation of an indirect immunofluorescence assay for the detection of IgG antibodies against Coxiella burnetii in bovine serum, Preventive Veterinary Medicine, Vol. 169; article number 104698

Wood C, Perkins N, Tozer S.J, Johnson W, Barnes T, McGowan M, Gibson J, Alawneh J, Perkins N, Firestone S, Woldeyohannes S. (2021) Prevalence and spatial distribution of Coxiella burnetii seropositivity in northern Australian beef cattle adjusted for diagnostic test uncertainty, Preventive Veterinary Medicine, 189: 105282

Other publications during candidature

Clark N.J, Tozer S, Wood C, Firestone S.F, Stevenson M, Caraguel C, Chaber A-L, Heller J, Soares Magalhães R.J. (2020) Unravelling animal exposure profiles of human Q fever cases in Queensland, Australia using natural language processing, Transboundary and Emerging Diseases, Vol. 67, Issue 5; pp 2133 - 2145

Oral presentations

Wood C, Tozer S, Gibson J, Alawneh J, McGowan M, Perkins N, Prevalence and spatial distribution of Coxiella burnetii exposure in beef cattle across northern Australia, Meat and Livestock Australia Postgraduate Conference, Q Station, Sydney, November 2018

Wood C, Seroprevalence of Coxiella burnetii in north Australian beef cattle using an indirect immunofluorescence assay, Australian and New Zealand College of Veterinary Scientists (ANZCVS) Science Week 2018, Epidemiology Chapter Conference Proceedings, July 2018

Wood C, Seroprevalence of Coxiella burnetii in north Australian beef cattle using an indirect immunofluorescence assay, Australian and New Zealand College of Veterinary Scientists (ANZCVS) Science Week 2018, Cattle Chapter Conference Proceedings (invited speaker), July 2018

Wood C, Tan T, Muleme M, Barnes T, Bosward K, Alawneh J, McGowan M, Stenos J, Gibson J, Perkins N, Firestone S, Tozer S, Q fever: validation of serological tests for Coxiella burnetii exposure in Australian cattle, Proceedings of Australian Veterinary Association Annual Conference, Brisbane, May 2018

Wood C, Harriott L, Soares-Magalhães R J, Cobbold R, Perkins N, Tozer S, Detection of Coxiella burnetii in peri-urban wild dogs in Southeast Queensland, 9th International Tick & Tick-borne Pathogen Conference and the 1st Asia-Pacific Rickettsia Conference, Pullman International, Cairns, Australia, September 2017

Poster presentations

Wood C, Tozer S, Gibson J, Alawneh J, McGowan M, Perkins N, Seroprevalence and spatial distribution of Coxiella burnetii exposure in beef cattle herds of northern Australia, 9th International Tick & Tick-borne Pathogen Conference and the 1st Asia-Pacific Rickettsia Conference, Pullman International, Cairns, Australia 27th August-1st September 2017

Wood C, Tozer S, Gibson J, Perkins N, McGowan M, Alawneh J, Investigating the epidemiology of Coxiella burnetii in beef cattle in Northern Australia and the potential risk to public health, MLA Postgraduate Conference, Sydney, October 2016

Tozer S, Vincent G, Wood C, Gaydon J, Sloots T, Lambert S, Coxiella MLVA typing of Queensland samples, International Congress for Tropical Medicine and Malaria, Brisbane, September 2016

Tozer S, Vincent G, Wood C, Sloots T, Lambert S, Multiple-Locus Variable number tandem repeat Analysis of Samples from Queensland, Australia, International Union of Microbiological Societies, Singapore, 17 – 21 July 2017

Contributions by others to the thesis

I would like to acknowledge the contributions of others to this thesis. All of my PhD supervisors have contributed to some degree in the conception, design and interpretation of parts of this thesis. I would like to acknowledge assistance in editing and critically reviewing the work within this thesis by Nigel Perkins, Sarah Tozer, John Alawneh, Michael McGowan and Justine Gibson.

Chapter 2: This chapter is a descriptive literature review that was written by myself, with editorial input from all supervisors mentioned above.

Chapter 3: This chapter has been published in the Australian journal, Communicable Diseases Intelligence. I am co-first author on this publication with Sarah Tozer. We equally contributed to the conception, design and analyses for this chapter. I prepared the final manuscript and wrote this chapter with editorial input and review from Sarah Tozer, Nigel Perkins and Justine Gibson.

Chapter 4: All supervisors contributed to the conception and design of this chapter. Collection of all samples at the abattoir and processing of samples in the laboratory was conducted by myself with assistance and guidance from my supervisor Sarah Tozer. I performed the analysis and wrote this chapter with editorial input from all supervisors mentioned above.

Chapter 5: This chapter has been published in the international journal, Preventive Veterinary Medicine. I am first author of this publication. The idea for this validation study was conceptualised by myself in collaboration with Sarah Tozer, Simon Firestone, Michael Muleme and John Stenos. Laboratory test modification and optimisation processes described in this chapter were conducted by myself with assistance from Sarah Tozer. I performed all laboratory testing described in this chapter with some assistance from Tabita Tan, who performed the ELISA testing of the Victorian cattle samples. I performed the statistical modelling in this chapter with guidance from Simon Firestone and Tamsin Barnes. Nigel Perkins, Sarah Tozer, John Alawneh, Michael McGowan, Justine Gibson, Tamsin Barnes and Katrina Bosward all provided critical review and editorial input for the manuscript.

Chapter 6: This chapter has been submitted and accepted (pending revisions) for publication to the international journal, Preventive Veterinary Medicine. The idea for this study was conceptualised by myself in collaboration with my supervisory team and Simon Firestone, Nigel Perkins, Michael McGowan and Tamsin Barnes. I would like to acknowledge Brighid

O’Niel for assistance in labelling and re-organising the CashCow serum samples. All laboratory testing performed for this chapter was completed by myself with support from Sarah Tozer. Assistance with the development and interpretation of statistical modelling was provided from Solomon Woldeyohannes, Simon Firestone and Wes Johnson. Final statistical models were run and analysed by myself with assistance from Solomon and Simon. Nigel Perkins, Sarah Tozer, Tamsin Barnes, Michael McGowan and Justine Gibson all provided critical review and editorial input.

Chapter 7: I wrote this chapter with critical review and editorial input from Michael McGowan, Nigel Perkins and John Alawneh. I performed all serological and molecular laboratory testing reported in this chapter with the assistance of Sarah Tozer. I would like to acknowledge John Alawneh for his contributions with developing the statistical analysis carried out in Chapter 7.

Chapter 8: I wrote the general discussion with editorial input from my supervisory team.

I would like to acknowledge and thank Jane Gaydon for proofreading this dissertation document.

vii

Statement of parts of the thesis submitted to qualify for the award of another degree

“No works submitted towards another degree have been included in this thesis”.

viii

Research Involving Human or Animal Subjects

All research involving human or animal subjects requires prior ethical review and approval by an independent review committee. At UQ, the relevant committee for research involving human subjects is the Human Ethics Unit and the relevant committee for research involving animal subjects is the relevant Animal Ethics Committee. Please provide details of any ethics approvals obtained including the ethics approval number and name of approving committees.

1. Children’s Health Queensland Hospital and Health Service Human Research Ethics Committee (HREC) number: HREC/08/QRCH/66 amendment number HREC/08/QRCH/66/AM03 2. The University of Queensland Animal Ethics approval SVS/115/11/MLA (NF) 3. The University of Queensland Animal Ethics ANRFA/SVS/100/16

Acknowledgements

“Coming together is the beginning; keeping together is progress; working together is success.”

- Edward Everett Hale (1822–1909)

The completion of this thesis was definitely a team effort and I would like to acknowledge and thank my team of supervisors for their support throughout the years. Firstly, Nigel Perkins, I cannot thank you enough for your constant support. I would enter our meetings with a list of problems and leave with a list of directions and a smile on my face. You have a wonderful ability to make me feel calm and motivated, even if you do not have the answer I am looking for, right there and then! Never second, Sarah Tozer ….. my SSSSSSSSSSSSSSSSSSSSS supervisor! I could not have imagined gaining such a wonderful teacher, mentor and friend. How did I luck- in to find you in this mingled web of Q fever? To the wonderful Professor Michael McGowan, thank you for encouraging me to question everything that I thought I knew and for always having my back. Justine Gibson, thank you for your continual help and support and for always checking that I dotted every “i” and crossed every “t”. Finally, to John Alawneh, you helped me to get started when I really didn’t know if I could do this. I may never be as good as you using “R”, however, I will keep on trying!

To my unofficial supervisors - Simon and Tamsin - I hope that I always told you how much I have appreciated your mentorship and assistance over the years. You have not only provided support with thesis specific topics, however, you were always happy to extend our discussions, which really helped me to prepare for the ANZCVS Vet Epi membership examinations in 2019.

I would also like to acknowledge and thank Kathy Bachman, whom I am sure I have bombarded with a constant overload of emails. Thank you for being a fantastic middle-man and helping me to get Nigel’s attention on to the highly important items!!

A huge thank-you needs to go to the original QPID laboratory team! Sarah, Jane, Seweryn and Bec. I am forever grateful for the opportunity to join your team and learn so much about laboratory work. All of those Ramen and dumpling lunches certainly paid off!

Financial support

This research was supported by an Australian Government Research Training Program Scholarship and a top-up MLA Postgraduate Research Scholarship.

Keywords

Coxiella burnetii, Q fever, coxiellosis, epidemiology, northern Australia beef cattle

Australian and New Zealand Standard Research Classifications (ANZSRC)

Provide data that links your thesis to the disciplines and discipline clusters in the Federal Government’s Excellence in Research for Australia (ERA) initiative.

Fields of Research (FoR) Classification

ANZSRC code: 070704 Veterinary Epidemiology 70%

ANZSRC code: 070707 Veterinary Diagnosis and Diagnostics 20%

ANZSRC code: 111117 Public Health and Health Services 10%

xii

Dedications

I would like to dedicate this thesis to my parents, Lyn and Desmond McGuckin, whom instilled within me the love of learning from a very young age. Thank you, for taking the time to dissect bandicoots at the farm and for showing me how to examine plants under a microscope before I could even ride a bike.

To Matt, Ella and Jackson, I could not have completed this thesis without the constant support, love and laughter from you all - my wonderful family. I know that you have had to share my time with this PhD for some time now… well finally, it is finished!

To the long list of dogs whom have helped me along the way… Ted was by my side at the beginning, then Doc joined the family (even if only briefly) keeping my lap warm. To my Zambian dogs, Pica and Abbot, thank you for sitting in my office day in and day out making sure that I kept on writing… and finally Jessie back in Brisbane, for making sure that I finished what I started!!!

xiii

TABLE OF CONTENTS

List of Figures & Tables ...... 8

Abbreviations ...... 12

1. General introduction ...... 16

1.1. Aims and objectives ...... 18

1.2. Overview of thesis chapters: ...... 18

2. Review of the literature ...... 21

2.1. Introduction ...... 21

2.2. History and background of Q fever ...... 22

2.3. Coxiella burnetii bacteriology ...... 23

2.3.1. Classification and morphology ...... 23

2.3.2. Coxiella burnetii genome and virulence ...... 24

2.3.3. Host cell/tissue invasion ...... 24

2.4. Molecular epidemiology ...... 25

2.5. Q fever epidemiology and clinical features ...... 26

2.5.1. Transmission of infection ...... 26

2.5.2. Clinical presentation ...... 27

2.5.3. Animal reservoirs ...... 27

2.6. Australian Q fever notifications ...... 29

2.7. The Netherlands Q fever outbreak (2007–2011) ...... 32

2.8. Distribution of beef and dairy cattle in Australia ...... 33

2.9. Detection and diagnostic tests in ruminants ...... 36

2.9.1. Direct diagnosis ...... 36

2.9.1.1. Bacterial culture ...... 36

2.9.1.2. Polymerase chain reaction (PCR) ...... 36 1

2.9.1.3. Indirect diagnosis (Serology) ...... 37

2.9.1.3.1. Complement fixation test (CFT) ...... 37

2.9.1.3.2. Enzyme-linked immunosorbent assay (ELISA) ...... 37

2.9.1.3.3. Indirect immunofluorescent antibody test (IFA) ...... 38

2.9.1.3.4. Validation of serological tests ...... 38

2.10. Epidemiology of coxiellosis in cattle ...... 39

2.10.1. Experimental infection in cattle ...... 39

2.10.1. Transmission and pathogenesis in naturally infected cattle ...... 41

2.10.2. Excretion of Coxiella burnetii ...... 42

2.10.3. Prevalence of infection in Australian ruminants ...... 43

2.10.3.1. Australian cattle ...... 43

2.10.3.2. Australian sheep and goats ...... 44

2.10.4. Prevalence of infection in cattle globally ...... 48

2.10.5. Risk factors for coxiellosis in cattle ...... 50

2.10.6. Impact of coxiellosis on reproductive performance in cattle ...... 52

2.10.7. Prevention and control of coxiellosis in ruminants ...... 55

2.11. Conclusion ...... 56

3. Report of Q fever notifications in Queensland, Australia, 2003–2017 ...... 60

3.1. Introduction ...... 60

3.1.1. Chapter objectives ...... 61

3.2. Materials and Methods ...... 61

3.2.1. Ethics statement ...... 61

3.2.2. Q fever case definition ...... 61

3.2.3. Notification data ...... 61

3.2.4. Statistical analysis ...... 62

3.3. Results ...... 63

3.3.1. Q fever notification counts and rates by age and sex ...... 63

3.3.2. Distribution of Q fever notifications, by HHS and LGA ...... 65 2

3.3.3. Hospitalization and work absences, due to Q fever ...... 67

3.3.4. Occupational groups ...... 67

3.3.5. At-risk exposures within one month prior to onset of Q fever ...... 68

3.3.1. Additional animal contact surveillance data (2013–2017) ...... 73

3.3.2. Q fever vaccination and awareness of risk of contracting disease ...... 73

3.4. Discussion and conclusions ...... 75

3.5. Chapter three appendix ...... 79

4. Detection of Coxiella burnetii in beef cattle at an abattoir in Queensland, Australia ...... 85

4.1. Introduction ...... 85

4.1.1. Background and history ...... 85

4.1.2. Abattoir surveys for animal disease monitoring ...... 86

4.1.3. Clinical samples and molecular epidemiology ...... 87

4.1.4. Chapter objectives ...... 88

4.2. Material and Methods ...... 89

4.2.1. Sample strategy and sample collection ...... 89

4.2.1.1. Sample size estimates ...... 89

4.2.1.2. Study 1 ...... 90

4.2.1.2.1. Sampling strategy ...... 90

4.2.1.2.2. Sample collection and storage - placental fluid and tissue ...... 91

4.2.1.3. Study 2 ...... 92

4.2.1.3.1. Tissue optimisation ...... 92

4.2.1.3.2. Sampling strategy ...... 92

4.2.1.3.3. Sample collection and storage - liver tissue ...... 92

4.2.2. Laboratory methods ...... 94

4.2.2.1. Molecular extraction from different samples ...... 94

4.2.2.1.1. Amniotic fluid ...... 94

4.2.2.1.2. Placental tissue ...... 95

4.2.2.1.3. Liver and spleen tissue ...... 95 3

4.2.2.2. Real-time PCR methods ...... 97

4.2.2.2.1. Quality control of test samples ...... 97

4.2.2.2.2. Coxiella burnetii testing from test samples ...... 97

4.2.2.2.3. PCR testing for tissue optimisation ...... 98

4.2.2.3. Statistical analysis ...... 98

4.3. Results ...... 101

4.3.1. Summary of cattle sampled ...... 101

4.3.2. Results from Study 1 ...... 101

4.3.3. Results from Study 2 ...... 102

4.3.3.1. Tissue optimisation ...... 102

4.3.3.2. Liver collection ...... 103

4.4. Discussion and conclusions ...... 105

5. Optimisation and validation of an indirect immunofluorescence assay (IFA) for the detection of IgG antibodies against Coxiella burnetii in bovine serum ...... 112

5.1. Introduction ...... 112

5.1.1. Chapter objectives ...... 113

5.2. Materials and Methods ...... 114

5.2.1. Serum samples ...... 114

5.2.1.1. Sample collection, storage and animal ethics ...... 114

5.2.1.2. Control sera for IFA testing ...... 114

5.2.1.3. Samples for proficiency panel ...... 114

5.2.1.4. Samples for the diagnostic test comparison of IFA and ELISA ...... 115

5.2.2. Serological test methods ...... 115

5.2.2.1. Development of the IFA slides ...... 115

5.2.2.2. Brief overview of IFA method ...... 116

5.2.2.3. Screening of sera for the diagnostic test comparison ...... 117

5.2.2.4. Brief overview of ELISA method ...... 118

5.2.2.5. Analytical performance of the IFA ...... 118 4

5.2.2.6. Comparison of phase variation on IFA slides on a subset of samples ...... 118

5.2.3. Statistical analysis ...... 119

5.3. Results ...... 123

5.3.1. Optimisation of the IFA ...... 123

5.3.2. Analytical test performance of the IFA ...... 123

5.3.3. Comparison of phase variation on IFA slides for screening sera ...... 124

5.3.4. Comparison of the IFA and ELISA ...... 125

5.4. Discussion and conclusions ...... 129

5.5. Chapter five appendix ...... 133

6. Prevalence and spatial distribution of Coxiella burnetii exposure in northern Australian beef cattle adjusted for diagnostic test uncertainty ...... 138

6.1. Introduction ...... 138

6.1.1. Chapter objectives ...... 139

6.2. Materials and Methods ...... 139

6.2.1. Ethics statements ...... 139

6.2.2. Study design ...... 140

6.2.3. Property selection criteria and blood sampling protocol ...... 140

6.2.4. Serological methods ...... 142

6.2.5. Statistical analysis ...... 142

6.2.5.1. Apparent prevalence ...... 142

6.2.5.2. Bayesian latent class models...... 143

6.2.5.3. Spatial visualisation ...... 145

6.3. Results ...... 145

6.3.1. Laboratory IFA testing of bovine samples ...... 145

6.3.2. Bayesian latent class modelling ...... 147

6.4. Discussion and conclusion ...... 151

6.5. Chapter six appendix ...... 155

7. Is Coxiella burnetii exposure associated with reduced reproductive performance in beef cattle in northern Australia? ...... 161

7.1. Introduction ...... 161

7.2. Material and Methods ...... 162

7.2.1. Animal ethics statement ...... 162

7.2.2. Study design and study population ...... 162

7.2.2.1. Background and methodology from CashCow ...... 163

7.2.3. Data and sample collection ...... 163

7.2.3.1. Infectious disease monitoring ...... 163

7.2.3.1.1. Coxiella burnetii serology ...... 164

7.2.3.1.2. Coxiella burnetii testing from vaginal swabs ...... 164

7.2.3.2. Property and herd data ...... 167

7.2.3.1. Animal data ...... 167

7.2.3.2. Nutritional data ...... 168

7.2.4. Annual pregnancy status ...... 168

7.2.5. Preparation of the analysis dataset ...... 169

7.2.6. Statistical methods ...... 169

7.2.6.1. C. burnetii laboratory results ...... 169

7.2.6.2. Exploratory data analysis ...... 169

7.2.6.3. Logistic regression ...... 170

7.3. Results ...... 172

7.3.1. Serology - IFA ...... 172

7.3.2. Vaginal mucus PCR ...... 172

7.3.3. Population annual pregnancy status ...... 173

7.3.4. Factors associated with annual pregnancy status ...... 173

7.3.4.1. Multivariable analysis ...... 177

7.3.4.2. Estimated predicted probability ...... 179

7.4. Discussion and conclusions ...... 180 6

7.5. Chapter seven appendix ...... 184

8. General discussion ...... 190

9. List of References ...... 202

10. Appendix ...... 227

10.1. Ethics approval certificates ...... 227

List of Figures & Tables

Figures

Figure 2.1 Q fever case notification counts per year from 1991 to 2018. Data sourced from the National Notifiable Disease Surveillance Website (Australian Government 2019)...... 30 Figure 2.2 “Q fever notification rates for Australia (National), Queensland (Qld) and New South Wales (NSW) by year, 1991–2014, National Notifiable Diseases Surveillance System” (Sloan- Gardner et al. 2017); NQFMP; National Q fever Management Program ...... 32 Figure 2.3 Breakdown of the beef cattle numbers per state as a proportion of total number of beef cattle in Australia (Meat and Livestock Australia 2014) ...... 34 Figure 2.4 Distribution of beef cattle as of 30 June 2001 (‘Australian Bureau of Statistics, Australian Government’ 2019)...... 35 Figure 3.1 Q fever notifications and annual rates (per 100 000 population per year) for Queensland and Australia, by year, 2003 - 2017...... 63 Figure 3.2 Summary of Queensland Q fever notifications by sex and 5-year age groupings, 2003 - 2017 ...... 64 Figure 3.3 Cumulative Q fever notification rates per 100 000 population/ 5 years, aggregated by residential local government area (LGA), from 2003–2017...... 66 Figure 4.1 Map of Queensland showing major regions defined (https://www.business.qld.gov.au/invest/queenslands-regional-locations/map-of-queensland- regions) ...... 93 Figure 4.2 Physical layout of samples for tissue optimisation for liver and spleen samples prior to DNA extraction ...... 96 Figure 5.1 Depiction of the principle of how IFA works...... 117 Figure 5.2 Photograph of an IFA slide - fluorescence observed on C. burnetii combined phase I and phase II antigen staining with positive bovine IgG antibodies (under 100x lense with oil immersion)...... 117 Figure 5.3 Serial dilution of 47 known negative bovine sera screened for antibodies against C. burnetii IgG phase I (P1) and phase II (P2) indicated that background fluorescence was visible using the IFA at serum dilutions of 1/40 and 1/80...... 124 Figure 5.4 Estimates from the Bayesian latent class model for the diagnostic sensitivity and diagnostic specificity of the IFA at different cut-off titres of 1:80, 1:160, 1:320, 1:640, 1:1280. Solid black line represents diagnostic sensitivity, dashed line represents diagnostic specificity 8 and shaded areas represent 95% credible intervals. Red reference line at 160 indicates the cut-off with the highest Youden’s index (J)...... 125 Figure 5.5 Prior and posterior distributions for diagnostic sensitivity and diagnostic specificity of enzyme-linked immunosorbent assay (ELISA) and indirect immunofluorescence assay (IFA) for the detection of IgG antibodies against C. burnetii in bovine serum; dotted line represents the prior distribution, solid line represents the posterior distribution of each variable...... 126 Figure 6.1 Geographical distribution of participating cattle properties from northern Australia (n = 60) with serum samples tested for C. burnetii IgG exposure...... 141 Figure 6.2 Caterpillar plot of model estimates for property level true prevalence as a proportion; the point estimate is the median of the posterior distribution of predicted prevalence and the solid line indicates 95% Credible Intervals...... 149 Figure 6.3 Map of the Northern Territory and Queensland showing the distribution of within- property true prevalence estimates of exposure to C. burnetii (n=60). Point values are the median posterior true prevalence estimates...... 150 Figure 7.1 Map of enrolled properties across the Northern Territory and Queensland categorised by country type...... 168 Figure 7.2 This figure was generated post-hoc from the final multivariable model to explore the relationship between Coxiella burnetii (C. burnetii) proportion positive and the outcome (predicted probability of pregnancy) adjusted for the impacts of other explanatory factors included in the model. Shaded area represent 95% confidence intervals...... 179 Figure 8.1 Choropleth map displaying human Q fever notification rate/5-years, for the time- period of 2009–2013, aggregated by local government area; overlayed with within-property beef cattle true prevalence estimates as points...... 198

Tables

Table 2.1 Suggested classification of Australian states and territories according to prominent pattern of Q fever case notifications ...... 30 Table 2.2 Table of published diagnostic sensitivity and specificity of serological tests for use in cattle and goats ...... 39 Table 2.3: Reported C. burnetii prevalence studies from Australian cattle between 1954 and 2018 ...... 46

Table 2.4: Published studies of C. burnetii sheep and goats in Australia between 1949 and 2018 ...... 47 Table 2.5 Globally reported C. burnetii seroprevalence studies with beef cattle included ...... 49 Table 3.1 Trends in demographics of confirmed Q fever cases in Queensland between 2003 and 2017, summarised by 5-year periods and for entire period...... 65 Table 3.2 Notifications of Q fever by Hospital and Health Service (HHS) of residence, 2003– 2017, Queensland ...... 65 Table 3.3 Table of Q fever cases in Queensland, Australia (2013–2017) by occupational category ...... 67 Table 3.4 Summary of at-risk exposure groups for Q fever cases in Queensland, by 5-year period ...... 68 Table 3.5 Details of environmental exposures for Queensland Q fever cases during the one month prior to disease onset...... 70 Table 3.6 Details of at-risk animal-related exposures for Queensland Q fever cases during the one month prior to disease onset...... 71 Table 3.7 Details of abattoir exposures for Queensland Q fever cases during the one month prior to disease onset...... 72 Table 3.8 Table of direct animal or arachnid contact in the one month prior to onset of Q fever, 2013-2017 ...... 73 Table 3.9 Details of awareness of risk and vaccination from notified Q fever cases in Queensland, by 5-year periods ...... 74 Table 3.10 Personal awareness of risk of Q fever, in cases that responded yes to an “at-risk” exposure during the one month prior to disease onset, 2013–2017 ...... 74 Table 4.1 Guide used to estimate gestational age of foetus ...... 91 Table 4.2 List of oligonucleotide sequences for gene target primers and probe sequences for PCR assays ...... 100 Table 4.3 Regional origins of cattle sampled during abattoir surveillance ...... 101 Table 4.4 Results for the tissue optimisation of liver and spleen samples ...... 102 Table 4.5 Prevalence of C. burnetii from bovine liver samples according to geographical origin of cattle ...... 104 Table 4.6 Cycle threshold (Ct) values of C. burnetii PCR positive targets; limit of Ct value set that >45 is negative...... 104 Table 5.1 Table of prior distributions used in the Bayesian latent class model...... 121 Table 5.2 Table of less informed prior specifications used for sensitivity analysis ...... 122

Table 5.3 Indirect immunofluorescence assay (IFA) end point titre results for IgG antibodies against C. burnetii for a proficiency panel of 10 bovine sera across two laboratories...... 123 Table 5.4 Cross classified test results from the IFA (1:160 cut-off) and ELISA (IDEXX) across four cattle populations ...... 127 Table 5.5 Bayesian estimates of the diagnostic sensitivity and specificity for the IFA and ELISA and estimated true prevalence for the four cattle populations for the detection of IgG antibodies against C. burnetii in bovine serum ...... 127 Table 5.6 Results from sensitivity analysis: Bayesian estimates of diagnostic sensitivity and specificity for the IFA and ELISA for the detection of IgG antibodies against C. burnetii in bovine serum and true prevalence estimates for three cattle populations, using less informative priors...... 128 Table 6.1 Descriptive summary of crude indirect immunofluorescent assay (IFA) results for Coxiella burnetii exposure in bovine serum ...... 146 Table 6.2 Predicted regional true prevalence estimates and 95% Credible Intervals, derived from the Bayesian hierarchical latent class model ...... 148 Table 6.3 Posterior probabilities of estimated true prevalence being less than a specified design prevalence ...... 148 Table 7.1 Detailed oligonucleotide sequences used for Equine Herpes Virus PCR assay .... 165 Table 7.2 Oliglonucleotide sequences for gene target primers for house-keeping gene ...... 165 Table 7.3 Detailed oligonucleotide sequences for the two gene targets used for C. burnetii PCR assays ...... 166 Table 7.4 Descriptive summary of crude IFA results for Coxiella burnetii exposure in bovine test sera aggregated by country type ...... 172 Table 7.5 Results from univariable analysis for explanatory variables retained for the multivariable model building ...... 174 Table 7.6 Estimated regression coefficients for the final multivariable mixed-effects logistic regression model for reproductive outcome “annual pregnancy status” ...... 178

Abbreviations

ABS Australian Bureau of Statistics AP Apparent prevalence BCS Body condition score CCI Calving to conception interval CDNA Communicable Diseases Network Australia CFT Complement fixation test CI Confidence interval CrI Credible interval (Bayesian confidence interval) Ct Cyclic threshold DNA Deoxyribonucleic acid DSe Diagnostic sensitivity DSp Diagnostic specificity EHV Equine herpes virus ELISA Enzyme-linked immunosorbent assay ERP Estimated Residential Population GAPDH Glyceraldehyde-3-phosphate dehydrogenase GEE Generalised estimating equations HHS Hospital and Health Service HREC Human Research Ethics Committee IFA Indirect fluorescent antibody test IgG Immunoglobulin G IgM Immunoglobulin M LCM Latent class model LCV Large cell variant LGA Local government area LPS Lipopolysaccharide MCMC Markov Chain Monte Carlo MLVA Multiple locus variable number of tandem repeats analysis MST Multispacer sequence typing NLIS National Livestock Identification System NOCS Notifiable Conditions System NQFMP National Q Fever Management Program NSW New South Wales OD Optical density OIE World Organisation for Animal Health PBS Phosphate-buffered saline PC3 Physical Containment Level 3 laboratory PCR Polymerase chain reaction PHU Public Health Unit Qld Queensland Q-VAX® Q fever Vaccine RFID Radio frequency identification device. 12

RFM Retained foetal membranes SCV Small cell variant SNP Single nucleotide polymorphism typing SoNG Series of National Guidelines TMB Tetramethylbenzidine TP True Prevalence USA United States of America Vic Victoria WA Western Australia

Chapter One

General Introduction

“Research is formalised curiosity. It is poking and prying with a purpose.”

- Zora Neale Hurston (1881–1960)

1. General introduction

Infection with Coxiella burnetii has been identified in a diverse range of animals, including wildlife and domestic mammals. This infection is zoonotic, meaning that it exists naturally in animal species and can be transmitted to humans. The human disease is referred to as Q fever and the infection in animals as coxiellosis. However, many published texts refer to the infection in all species as Q fever. For precision and consistency throughout this thesis, animal infection will be referred to as coxiellosis and Q fever will be reserved for human disease.

Q fever has a longstanding history in Queensland, Australia. The first reported cases were clinically diagnosed in Brisbane and the aetiological agent, Coxiella burnetii, was then isolated and named in a collaborative effort between Australian and American scientists (Derrick 1937; Cooke 2008). Although it is not a new disease to Australia, Q fever continues to burden the country with some of the highest annual case notification rates globally (Tozer 2015). Historically, Q fever has been associated with residing in rural areas and occupations such as abattoir workers, farmers and veterinarians (Garner et al. 1997). It has been frequently reported that exposure to infected cattle, sheep and goats results in the highest risk for zoonotic transmission (Eldin et al. 2016; Garner et al. 1997; Porter et al. 2011).

While there are undeniable links with Q fever cases and exposure to cattle, it appears that coxiellosis predominantly goes undiagnosed and unnoticed in cattle within Australia. There are currently limited studies specifically focused on the regional prevalence of C. burnetii infection or possible production effects of coxiellosis in beef or dairy industries (Cooper et al. 2011; Tan 2018; Pitt 1997). Overall, it is evident that there are gaps in the current knowledge base regarding coxiellosis in cattle in Australia and the significance of this infection for animal and public health is uncertain. This thesis is a collection of research designed and implemented to improve our understanding of the epidemiology of coxiellosis in beef cattle in northern Australia and gain insights into the potential risk of C. burnetii infection in beef cattle to public health.

Coxiella burnetii is a highly infectious intracellular bacterium that has distinct morphological properties enabling survival in harsh environmental conditions and persistence of infection within multiple host species (Toman et al. 2012; McCaul & Williams 1981). Transmission occurs through inhalation of aerosolised bacteria and C. burnetii is categorised as a group B bioterrorism agent due to the potential for use as a biological 16 warfare agent. Q fever can occur as both acute and chronic forms of disease with a range of clinical symptoms and may initially be misdiagnosed as influenza due to non-specific malaise (Eastwood et al. 2018). Persistent focalised C. burnetii infections can result in chronic hepatitis, gestational complications, paediatric osteomyelitis and endocarditis (Parker, Barralet & Bell 2006; Eldin et al. 2016).

The Q fever epidemic in the Netherlands (2008–2011) was a substantial motivator at the start of this thesis. It highlighted the potential effect of a large-scale animal outbreak on public health (Roest et al. 2010). In this scenario, coxiellosis was initially noticed in commercial goat and sheep dairies and then after several months, human cases began to emerge. The extent of the epidemic included over 4,000 notified acute human Q fever cases and in response, more than 50,000 pregnant small ruminants were culled (Roest et al. 2010). This incident appeared to be the catalyst for new research focused on C. burnetii in ruminants in Europe. Then, in 2015, following a much smaller human Q fever outbreak at a Victorian dairy-goat farm, the Australian interest started to emerge (Bond et al. 2015). It was obvious that Q fever required research encompassing animal, human and environmental facets to inform policy and improve prevention and control.

Within Australia, individual state-government public health departments perform follow-up surveillance of Q fever cases to identify potential risk exposures. However, the information gathered from this follow-up does not always ensure the identification of a specific source of infection, as many cases have multiple co-exposures or no known exposures (Sloan-Gardner et al. 2017). In Queensland, where Q fever appears endemic, cases are not traced back with diagnostic confirmation of the putative animal or environmental source. Therefore, the exact source of C. burnetii is often only speculative. There are no current Australian policies to notify or include animal health departments in the follow-up of human cases, and coxiellosis is not notifiable in six out of the seven states of Australia (‘Notifiable Animal Diseases’ 2018).

The state of Queensland has approximately 5 million inhabitants (20% of the national population) and consistently reports 40–55% of the national annual Q fever case notifications (Barralet & Parker 2004). Queensland has a strong livestock-based agriculture sector, which may increase the general risk of Q fever transmission through direct and indirect exposures. Approximately 40% of the Australian national beef cattle herd are based in Queensland on 11,500 registered livestock businesses (‘Australian Bureau of Statistics, Australian Government’ 2019). Dairy farming (cattle) in Queensland is now less common, with approximately 620 registered businesses holding dairy cattle; the majority of the 17 national milk is produced in southern Australia. Mixed grazing of beef-cattle with small ruminants does occur in Queensland, with approximately 1,200 agricultural businesses holding sheep. Trends show reducing numbers of primary sheep producers in Queensland over the last 10 years with many destocking due to drought; however, there has been a shift towards the commercialisation and sale of feral goats for live export and meat. Abundant feral goats in Queensland are being marketed as “rangeland goat” to fill the demand. There has also been an increase in demand for non-cattle dairy products, hence the emergence of the Australian goat dairy industry (Foster 2014).

At the time of this thesis, there was minimal knowledge of the epidemiology of coxiellosis in beef cattle of northern Australia. The clinical and economical significance of this infection for beef cattle herds was unknown, and how the infection in cattle contributes to public health risks of Q fever were unsubstantiated.

1.1. Aims and objectives

The overall aim of this thesis was to improve our understanding of the epidemiology of coxiellosis in beef cattle in northern Australia and gain insights into the potential risk to public health. To thoroughly explore the overall aim, the thesis was written as individual chapters outlined below, with specific objectives to investigate molecular strain differentiation, laboratory diagnostics, prevalence and geographical distribution, and production effects of coxiellosis in beef cattle in Northern Australia.

1.2. Overview of thesis chapters:

Chapter two is a broad descriptive literature review that explores what has currently been published about coxiellosis in cattle globally and highlights what has been reported in Australia. It outlines the history, current knowledge and epidemiology of human Q fever and coxiellosis in cattle.

Chapter three is a descriptive analysis of Queensland human Q fever notification data from 2003 - 2017. Data collected from Queensland Public Health Units, including enhanced surveillance of at-risk exposure, is examined to investigate current surveillance and gain insights into the epidemiology of human disease within this state.

Chapter four describes two abattoir-based surveys with the aim to estimate the prevalence of bacterial infection and identify specific genetic strains of C. burnetii that were circulating in cattle going to slaughter in South-East Queensland.

Chapter five reports results from a laboratory validation study designed to develop, optimise and validate a human immunofluorescence assay (IFA) for the detection of IgG antibodies against C. burnetii in bovine sera. In this study, the analytical performance of the IFA and a commercial ELISA is assessed and validated to determine the diagnostic sensitivity and specificity of the two tests using Bayesian latent class analysis.

Chapter six describes the serological testing of a large sample of beef cattle across a broad geographic distribution of northern Australia. It incorporates the IFA diagnostic test parameters, estimated in the previous study, to estimate the true prevalence of exposure to C. burnetii with hierarchical Bayesian latent class modelling. Basic maps are presented to visualise the spatial distribution of C. burnetii exposure in this sample of beef cattle. The model developed for this analysis can be modified to assist future true prevalence estimates in populations with clustered data.

Chapter seven was designed to examine the putative relationship between C. burnetii infection and measures of reproductive performance in a large population of commercial beef breeding cattle. Molecular methods were used to test vaginal swabs from beef cattle for C. burnetii deoxyribonucleic acid (DNA). Serological results from IFA testing performed for chapter six were examined with reproductive performance data. The reproductive data analysed in this chapter were generated during a previous industry-driven epidemiological study investigating causes of poor reproductive performance in the north Australian beef industry: “The Northern Beef Fertility Project; CashCow” (McGowan et al. 2014). Linear mixed-effects logistical regression models were used to explore this relationship.

Chapter eight is the general discussion of this thesis and provides an overview of the research findings, discussing limitations and providing insights of how this work contributes to this field of research.

Chapter Two

Review of the literature

“Learning is not attained by chance,

it must be sought for with ardour

and attended to with diligence.”

- Abigail Adams (1744–1818)

2. Review of the literature

2.1. Introduction

Coxiella burnetii is a zoonotic intracellular bacterium that has been detected in a diverse range of animals and geographic regions. Coxiellosis, the infection in animals, is listed by the OIE (World Organisation for Animal Health) as a reportable disease for member countries (OIE 2018). Infection can easily spread from animals to humans and is reported to be a re-emerging zoonosis in many countries (Arricau-Bouvery & Rodolakis 2005). The human disease, Q fever, is notifiable in most developed nations.

Q fever manifests in humans as acute and chronic/persistent infections and has been recognised as the cause of a post-Q fever fatigue syndrome (Barralet & Parker 2004; Parker, Barralet & Bell 2006; Eldin et al. 2016). Acute Q fever may be subclinical in 60% of cases. A variety of clinical signs occur in the remaining 40% including persistent fever, nausea, severe headaches and fatigue (Marrie 1990). Chronic disease may occur months to years post exposure and can recrudesce as endocarditis, chronic pulmonary infections, chronic hepatitis, osteomyelitis and gestational infections (Parker, Barralet & Bell 2006). This zoonotic disease has been most historically diagnosed in people with animal industry occupations, including abattoir workers, farmers and veterinarians (Garner et al. 1997; Derrick 1961). It has been a notifiable illness in Australia since 1977 with high case notification rates relative to other countries (Tozer 2015). Within Australia, people living in Queensland have consistently accounted for between 40-55% of the total annual cases (Sloan-Gardner et al. 2017). Current Q fever case notifications may be an underestimation of true clinical cases, as the symptoms are non-specific, and many general practitioners may not routinely test for it.

Coxiella burnetii is currently categorised as a group B bioterrorism agent by the Centre for Disease Control (CDC), United States of America, and is a Risk Group 3a microorganism (Oyston & Davies 2011). The reported low infective dose of the bacteria and its ability to survive

a “Risk Group 3 (high individual risk, limited to moderate community risk) - a microorganism that usually causes serious human or animal disease and may present a significant risk to laboratory workers. It could present a limited to moderate risk if spread in the community or the environment, but there are usually effective preventive measures or treatment available” (Safety in laboratories. Part 3, Microbiological safety and containment. 2010) 21 and spread in a spore-like life-cycle has resulted in any laboratory or animal experimental work requiring a Physical Containment Level 3 laboratory (PC3; Roest et al., 2013).

Coxiellosis is rarely reported in the Australian cattle industry and there are currently no formal surveillance or control programs in place to assess disease prevalence in animals. Tasmania is the only state in Australia where coxiellosis is listed as a notifiable animal disease (Tasmanian Government, 2018). While infection is apparently subclinical in cattle in Australia, coxiellosis has been found to be associated with sporadic abortion, birth of weak offspring and reproductive failures in cattle from European countries (Hansen et al. 2011; Agerholm 2013; Plummer et al. 2018).

Although it is not a new disease entity, there are many unknowns regarding C. burnetii in beef cattle in Australia and the significance of coxiellosis to the beef cattle industry and public health has not been fully investigated. This literature review aims to provide a brief overview of Q fever in humans and focus on the pathogenesis, diagnosis and epidemiology of coxiellosis in cattle.

2.2. History and background of Q fever

The first record of human Q fever disease occurred in Brisbane in 1935; clusters of “fever of unknown origin” were observed periodically between 1933 and 1935 in local meat-workers and dairy farmers. At that time, the condition was brought to the attention of medical pathologist Edward Derrick for investigation (Derrick 1937). Initially, Derrick suspected the causative agent to be a virus, as he was unable to isolate or visualise the microorganism in his laboratory. He then recruited fellow Australian scientists, Frank MacFarlane Burnet and Mavis Freeman to help investigate this “Query fever virus” (Burnet & Freeman 1937). Burnet and Freeman were able to infect healthy guinea pigs from a sample Derrick provided and continued investigating this novel agent until they could prove it was not a virus. It was then classified as Rickettsia burnetii (Burnet & Freeman 1937).

At a similar time, in Nine Mile, Montana, United States of America (USA), a separate investigation was ongoing to describe an agent recovered from ticks collected in an ecological study (Davis et al. 1938). Scientists Cox and Davis had observed febrile illness in laboratory guinea pigs infected with the agent and had described the microbe as possessing “viral and bacterial properties” (Davis et al. 1938). When a febrile illness was then observed in a laboratory worker, similarities were recognised between this case and the Australian Q fever cases. It was realised that the Rickettsia diaspora discovered in the USA and Rickettsia burnetii in Australia 22 were most likely the same organism. Further studies of the microbe resulted in reclassification and renaming from Rickettsia burnetii to Coxiella burnetii in recognition of the involvement of Cox and Burnett in isolating and describing the microorganism (Maurin & Raoult 1999). The human disease today is still termed Q fever derived from the original name “Query fever” given by Derrick.

2.3. Coxiella burnetii bacteriology

2.3.1. Classification and morphology Coxiella burnetii was first classified as a Rickettsia in the class Alphaproteobacteria, due to its recovery from ticks and its inability to be grown axenically (Omsland et al. 2009). However, recent 16S rRNA sequence analysis has resulted in the Coxiella genus being reclassified into the class Gammaproteobacteria within the order Legionellales (van Schaik & Samuel 2012; Maurin & Raoult 1999). C. burnetii is the only species recognised in the Coxiella genus, although a second highly homologous species has been proposed, Coxiella cheraxi, that was isolated from Australian crayfish (Cooper et al. 2007).

Coxiella burnetii is a small pleomorphic, obligate intracellular bacterium. It has been described as coccoid, bacilli and granular in appearance, depending on the stage of its developmental cycle (McCaul & Williams 1981). While it is often reported to be gram negative, C. burnetii is not easily nor consistently stained using the Gram method (McCaul & Williams 1981). It does however, possess a cell membrane similar to gram negative bacteria and can also be visualised using the Gimenez stain (McCaul & Williams 1981; Samuel & Hendrix 2009).

Two morphologically distinct cell variations can be identified in the developmental cycle of C. burnetii; a large cell variant (LCV) and small cell variant (SCV; McCaul and Williams, 1981). The LCV is the metabolically active form which can replicate inside the host cell phagolysosome. The LCV form is able to change into the SCV form prior to excretion from the host. The SCV is spore-like and is resistant to changes in osmotic pressure, high ambient temperatures and ultra- violet radiation enabling its survival in relatively harsh environmental conditions, permitting transmission of viable bacteria in dust from the environment (Toman et al. 2012). When inhaled or possibly ingested, the SCV is able to enter host phagocytic cells and revert into the LCV to enable replication (Roest et al., 2013).

2.3.2. Coxiella burnetii genome and virulence The C. burnetii genome is circular and variable in size, ranging from 1.5 to 2.4 Mb (van Schaik and Samuel, 2012; Willems et al., 1998). Although this bacterium is often referred to as homogenous, molecular studies have identified and described six different genomic groups (I – VI; van Schaik and Samuel, 2012). The first whole genome sequence of C. burnetii (Nine Mile strain) was published in 2003 (Seshadri et al. 2003). The AuQ01 strain was the first to be completely sequenced in Australia after isolation from an acute Q fever patient (Walter et al. 2014). Since then, many C. burnetii strains have been successfully sequenced and researchers have been able to compare strains and draw correlations between genomic groups of C. burnetii and the development of varying human disease (van Schaik & Samuel 2012; Glazunova et al. 2005). However, there are still gaps in the knowledge of specific C. burnetii virulence factors and studies focused on identifying and describing strain virulence factors will no-doubt add significant knowledge to the field (Million & Raoult 2015).

There are two antigenic phase variations of C. burnetii, phase I and phase II as determined primarily by the composition of cell wall lipopolysaccharide (LPS; Hackstadt, 1990). The phase I antigen is phenotypically smooth and can be identified on C. burnetii isolated directly from animals and human patients; it is then converted to the rough, phase II antigen through multiple passage of embryonated eggs or an immunocompetent host (Narasaki & Toman 2012). The smooth phase I antigen appears to be the virulent phase, which can invade immunocompetent hosts and is protected from the host’s immune system. The phase II antigen has an incomplete LPS due to a genetic deletion and is therefore an avirulent variation that is unable to survive immunocompetent host responses (Williams & Thompson 1991; Hoover et al. 2002).

2.3.3. Host cell/tissue invasion It is well documented that host monocytes and macrophages are target cells for C. burnetii invasion in vertebrates (Amara, Bechah & Mege 2012). The bacterium has adapted to enable multiplication in the acidic vacuoles of eukaryotic cells without being destroyed, thus it is able to invade host phagocytes and survive and replicate inside phagolysosomes (Maurin & Raoult 1999; Flannagan, Cosío & Grinstein 2009; van Schaik & Samuel 2012). Once replication occurs in local lymph nodes, the intracellular bacteria may circulate to peripheral sites where optimal survival will ensue. A published case report from Australia found C. burnetii could persist in the bone marrow following “successful” treatment of acute Q fever (Harris et al. 2000). Adipose tissue and the placenta appear to be tissues where C. burnetii can also survive

24 concealed from the host’s immune response (Amara, Bechah & Mege 2012). Although the exact mechanisms are not completely understood, changes in the host’s immune response during pregnancy may allow the infection to persist with recrudescence of shedding at the time of parturition. The bacteria have been found in placental trophoblasts of animals and humans and within placental macrophages (Amara, Bechah & Mege 2012; Roest, Bossers & Rebel 2013).

2.4. Molecular epidemiology

Although C. burnetii is the only species in the genus Coxiella, strain variations have been identified from different human, animal and environmental sources globally (Ceglie et al. 2015; Jado et al. 2012). Conventional serological methods cannot discriminate between exposures to different bacterial strains. Characterisation of C. burnetii, using molecular genotyping methods, is advancing rapidly and appears to be an ideal tool to help unravel the complex epidemiology of infection across multiple host species and geographic regions (Massung, Cutler & Frangoulidis 2012; Roest et al. 2011). Studies in Spain, Portugal, Italy, France and the Netherlands have observed genotype clustering within animal species (livestock and wildlife) and within geographical locations (Astobiza et al. 2012; Pinero et al. 2015; González-Barrio et al. 2016). In particular, it has been found that cattle genotypes are highly species-specific; C. burnetii within these “cattle-clusters” have not commonly been associated with human Q fever cases (Ceglie et al. 2015; Joulié et al. 2017; Pinero et al. 2015). Genotypes of C. burnetii isolated from goats and sheep were more likely to be identified in human Q fever cases, however these small ruminant genotypes may occasionally be found to infect cattle (Astobiza et al. 2012).

During the Q fever outbreak in the Netherlands, molecular genotyping techniques were successfully utilised to trace the probable source of human infection (Roest et al. 2011). This enabled control measures to be implemented at the apparent origin (infected dairy goats) thus reducing human Q fever cases. Currently, no data are available on specific genotypes of C. burnetii found in cattle or other reservoir animals in northern Australia (Vincent et al. 2016). In Victoria, the C. burnetii genotype responsible for a human Q fever outbreak in 2015 was successfully isolated from goat and human samples (Bond et al. 2015). The identification of bacterial genotypes present in a range of Australian animals is crucial to ascertain the likely sources of human disease and to aid in potential outbreak investigations.

There have been recent advances in isolating novel genotypes from human Q fever cases in Australia (Vincent et al. 2016; Tozer 2015). In one study, 42 C. burnetii isolates from acute Q fever patients were found to be genetically distinct compared to more than 300

C. burnetii strains from patients of other countries (Vincent et al. 2016). These findings support the theory that Australian strains have evolved to produce a unique phylogenetic clade of C. burnetii most likely due to geographical isolation (Vincent et al. 2016). To extend the current knowledge of molecular epidemiology within Australia, it is necessary to compare isolates from diverse hosts and locations across Australia.

2.5. Q fever epidemiology and clinical features

Q fever has been diagnosed in humans globally, with a variety of epidemiological patterns seen in different countries and regions (Babudieri 1959; Maurin & Raoult 1999). Q fever can be endemic, appear sporadically as single cases or have common point-source outbreaks with multiple cases linked to one exposure. These patterns seem to vary between and within countries, likely due to the wide range of animal reservoirs, differences in C. burnetii strains and environmental factors that are not completely understood (Million & Raoult 2015). Often complex epidemiological investigations are necessary to identify the role of potential reservoirs as likely sources of human infection so that appropriate control measures can be implemented (Ladbury et al. 2015). This can be challenging due to the bacteria’s ability to survive outside of an animal host and withstand harsh weather. Environmental contamination via bacterial shedding and spread of viable bacteria in dust can lead to human infection without apparent animal contact (Maurin & Raoult 1999; Arricau-Bouvery & Rodolakis 2005).

2.5.1. Transmission of infection Human infection occurs mainly through inhalation of aerosolised bacteria via direct contact with infected animal body-fluid, tissue or dust (Garner et al. 1997; Gunaratnam et al. 2014). It has long been suggested that wind could play an important role in the aerosol transmission of Q fever (Babudieri 1959; Clark & Soares Magalhães 2018). An increase in human Q fever cases was associated with the mistral wind, correlated to seasonal lambing in an area of France (Tissot-Dupont et al. 2004). Less commonly, transmission has been reported through ingestion of unpasteurised milk from infected animals (Marrie 1990). Transdermal transmission directly through tick bites is possible, as shown by experimental infection (Marrie 1990); however, does not appear an important route for Q fever (Maurin & Raoult 1999). Human to human transmission is extremely rare, although infection has been reported in pathology staff following the care and subsequent post-mortem of a Q fever patient (Harman 1949). A single published report was found describing sexual transmission of Q fever from a man to his wife (Milazzo et al. 2001). In this case the semen from the man with acute Q fever was found to be

26 positive for C. burnetii DNA by polymerase chain reaction (PCR) for 15 months post disease onset (Milazzo et al. 2001).

2.5.2. Clinical presentation Acute Q fever can present with a variety of clinical signs including fever, nausea, headaches and fatigue, following an incubation period of 19–21 days (Gunaratnam et al. 2014). There are reports that suggest clinical illness in Australia appears unique from disease in other countries (Parker, Barralet & Bell 2006). While in Australia, Q fever patients commonly present with a classical “flu like illness” including fever, headache, night sweats and fatigue (Eastwood et al. 2018), hepatitis is more common in southern Spain, Ontario and France (Tissot Dupont et al. 1992); and pneumonia in areas of Switzerland and Crete (Maurin & Raoult 1999). It is common in Australia for acute Q fever cases to be misdiagnosed as influenza and not immediately recognised or notified (Eastwood et al. 2018). Of the acute Q fever cases, 2% require hospitalisation and the case fatality rate is approximately 1–2% (Parker, Barralet & Bell 2006; Raoult, Marrie & Mege 2005). Although rare, a severe life-threatening case of Q fever sepsis syndrome was reported in Brisbane in 2015 (Stevenson et al. 2015). A 28-year-old woman, who had been exposed to kangaroos, initially presented with flu-like illness which then developed into sepsis with multiple organ failure and progressive respiratory failure requiring intubation for 7 days.

Approximately 10–15% of acute Q fever cases develop a debilitating “Post-Q fever fatigue syndrome”. An additional 1–5% of acute cases progress to chronic C. burnetii infections, including vascular and osteo-articular infections and gestational complications (Gunaratnam et al. 2014; Million & Raoult 2015). A recent review by experts in the field have discouraged use of the term “chronic Q fever” as this oversimplifies the diagnosis and clinical implications; the term “persistent focalised C. burnetii infections” has been recommended (Eldin et al. 2016; Million & Raoult 2015). Several cases of paediatric osteomyelitis have also been reported in Australia as a result of persistent infection (Britton et al. 2015; Nourse et al. 2004). Without treatment, endocarditis from persistent focalised C. burnetii infection can be fatal. In Australia, even with treatment, C. burnetii endocarditis has a 10% fatality rate (Gunaratnam et al. 2014).

2.5.3. Animal reservoirs Coxiella burnetii has been identified worldwide (excluding New Zealand) in a wide range of vertebrate and invertebrate reservoirs, including mammals, some birds, Australian marsupials and ticks (Cooper et al. 2012; Babudieri 1959). Australia’s earliest clusters of Q fever cases were associated with working at an abattoir that processed a high number of pregnant dairy 27 cattle, an outbreak in sheep shearers and an outbreak involving a family with no obvious animal exposure, only a history of picking pineapples likely contaminated by bandicoot faeces (Derrick 1961; Derrick, Pope & Smith 1959; Derrick 1937).

Ruminants: It is commonly reported in literature that cattle, sheep and goats are the most frequent source of human Q fever, however this is not always confirmed with molecular epidemiology (Million & Raoult 2015; Porter et al. 2011; Maurin & Raoult 1999; Eldin et al. 2016). High numbers of bacteria are shed in milk and the placenta and associated fluids and therefore, zoonotic spillover is expected when humans are in close contact with infected parturient and lactating livestock (Guatteo et al. 2006). It is likely that between countries and regions, different ruminant species may be more commonly associated with human infection (Babudieri 1959). In the Netherlands, pregnant dairy-goats have been implicated as the most significant source of human infections (Roest et al., 2013). In the United States of America (USA), although high seroprevalence and PCR positive bulk tank milk samples have been reported in dairy-cattle, the predominant strain of C. burnetii identified in cattle has not been found in human disease (Pearson et al. 2014). Moreover, a study in Minnesota, USA, found a significant association between the number of sheep flocks in a region and incidence of human Q fever. No association was evident with cattle or goat properties (Alvarez et al. 2018).

In Australia, Q fever cases have been commonly associated with working at abattoirs that slaughter cattle, sheep and goats (Gidding et al. 2009). However, a recent non-abattoir outbreak occurred at a dairy-goat farm in Victoria (Bond et al. 2015). A significant outbreak (involving 25 notified cases) has also been associated with attending a sheep saleyard on a dry and windy day and another (involving 4 notified cases) was traced back to working at a cosmetics supply factory that used ovine-derived products (Wade et al. 2006; O’Connor, Tribe & Givney 2015).

Companion animals: It has long been known that cats and dogs have the potential to spread C. burnetii (Babudieri 1959; Kosatsky 1984; Pinsky et al. 1991). In Australia, two recent outbreaks have been associated with parturient cats (Kopecny et al. 2013; Malo et al. 2018). Seroprevalence studies have also identified exposure in several populations of domestic dogs, domestic cats and breeding cats (Shapiro et al. 2015, 2016). Horses are susceptible to infection as shown by positive serological and PCR testing, however it is not clear if they are a common source of human infection (Marenzoni et al. 2013; Tozer et al. 2014).

Wildlife: In Australia there are reports of high seroprevalence and PCR detection of C. burnetii in native marsupials, dingos, foxes and feral cats (Cooper et al. 2013; Shapiro et al. 2015). The exact role that native wildlife and feral animals may play in the spread of Q fever to 28 humans and in maintaining infection within livestock and companion animal populations is unknown. However, it seems likely that they are a potential source of C. burnetii. From a 10 year review of Q fever cases in a hyper-endemic region of North Queensland, 69.8% of cases reported exposure to macropods compared to only 23.8% which reported exposure to cattle (Sivabalan et al. 2017).

In Guyana, South America, repeated human Q fever outbreaks have been linked to the three toed sloth, a reservoir of a highly virulent strain of C. burnetii (Million & Raoult 2015). The European rabbit is a reservoir of C. burnetii in Spain and has been reported to be associated with human Q fever pneumonia cases (Gonzalez-Barrio et al. 2015; Marrie et al. 1986). This species of rabbit was introduced to Australia in 1879 and has since become one of Australia’s major feral-animal pests (Fenner 2010).

Ticks: Ticks are able to carry and transmit C. burnetii, however, they do not appear to have an essential role in the maintenance of infection within animal or human populations (Woldehiwet 2004). Recent genetic studies suggest that research relying on the PCR detection of C. burnetii from ticks may have high false positive results due to a “coxiella-like” endosymbiont (Duron 2015; Elsa et al. 2015). It has been suggested that the pathogenic C. burnetii bacteria may have evolved from such a tick endosymbiont, however for now this is simply a hypothesis (Duron et al. 2015).

2.6. Australian Q fever notifications

Q fever disease in Australia can be described as endemic, with national case notifications fluctuating between 317 to 868 cases per year between 1991 and 2018 (Figure 2.1). From available disease notification data, outbreak reports and published reviews on Q fever within Australia, it is evident that there are differences in the pattern of Q fever between states and territories. When comparing the 10 year average annual Q fever notification rates (cases per 100 000 population; 2009 – 2018), Queensland has an average rate of 4.3, New South Wales 2.5 and all other states have rates less than 1 case per 100 000 (Australian Government 2019).

Queensland and New South Wales combined, report approximately 80% of the national Q fever cases (Australian Government, 2019; Tozer, 2015). Although most other states report relatively low incidence, some states appear to have more outbreaks than others. South Australia and Victoria have reported several outbreaks over the years, whereas the Northern Territory, Tasmania and the Australian Capital Territory appear to have low incidence of Q fever cases with negligible outbreaks (O’Connor, Tribe & Givney 2015; Bond et al. 2015, 2018; Wade 29 et al. 2006). It could be suggested that Australian states and territories be described and categorised according to the prominent pattern of Q fever case notifications (Table 2.1).

1000

900

800

700

600 notifications

500 case

400 of

300

Counts 200

100

0 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 Year

Figure 2.1 Q fever case notification counts per year from 1991 to 2018. Data sourced from the National Notifiable Disease Surveillance Website (Australian Government 2019).

Table 2.1 Suggested classification of Australian states and territories according to prominent pattern of Q fever case notifications

Low incidence with Low incidence with sporadic Hyper-endemic occasional outbreaks cases

Queensland Victoria Northern Territory

New South Wales South Australia Tasmania

Western Australia Australian Capital Territory

A publication that analysed the Australian national Q fever notifications from 1991 to 1994, identified higher case notification rates in adult males of eastern Australia (southern Queensland and northern New South Wales), a definite association with the meat industry as well as spatial correlation between cases and the presence of livestock (Garner et al. 1997). There are no strong seasonal patterns seen in Q fever cases in Australia, unlike what is reported in Europe. Although there is a short report of an increase in acute Q fever cases three months after high rainfall in a small region of north Queensland. They hypothesised that during the wet season there was an increase in native wildlife in the area, and drier conditions immediately followed favouring transmission (Harris et al. 2013).

When interpreting Australian case notification data, it is important to note that a government funded National Q fever Management Program (NQFMP) took place from 2002 to 2006. Initially the vaccination program targeted abattoir workers and sheep shearers and later included farmers (sheep, beef and dairy cattle), their family and employees (Gidding et al. 2009). As shown in Figure 2.2, a 20% reduction in national cases was observed following this vaccination program.

Since the NQFMP, there have been reports of a “changing epidemiology” of Q fever within Australia (Sloan-Gardner et al. 2017). Queensland and NSW still consistently report more Q fever cases than other states and territories, however less are directly associated with occupational exposure and there is a trend of increasing notifications in females (Sloan-Gardner et al., 2017). From enhanced Q fever surveillance data reported in Queensland between 2013 and 2017, direct at-risk animal-related exposure to an animal and/or animal birth products was reported in 52% of cases (Tozer et al. 2020). However, some kind of at-risk “environmental exposure” was reported in 82% of all notified cases. This included a range of epidemiological factors from exposure to dust from paddocks and animal yards, to contact with untreated water and included consuming unpasteurised milk or milk products. In New South Wales, a recent outbreak was unable to identify the exact source of infection, however epidemiological analysis indicated exposure to kangaroos and contact with dogs were significantly associated to outbreak cases (Archer et al. 2017). There has previously been difficulties in comparing Q fever risk factors and exposures across Australia, due to discrepancies in data collection during case follow-up between public health hospital divisions (Sloan-Gardner et al. 2017). The recent introduction of a national guideline for the surveillance of Q fever will aid in the pursuit of better understanding the epidemiology of the disease within Australia (CDNA 2018).

Figure 2.2 “Q fever notification rates for Australia (National), Queensland (Qld) and New South Wales (NSW) by year, 1991–2014, National Notifiable Diseases Surveillance System” (Sloan-Gardner et al. 2017); NQFMP; National Q fever Management Program

Australia is the only country worldwide to have a licensed and accessible human vaccine (Q-VAX®, Seqirus). The vaccine appears to be effective in preventing Q fever in “high risk” occupations such as abattoir workers and veterinarians (Marmion et al. 1990). Q-VAX® is a formalin-inactivated whole-cell vaccine produced with the Henzerling phase I strain of C. burnetii. This vaccine is reported to cause adverse reactions and cannot be used in children under the age of 15 years, although a review has recently been published describing some paediatric use of the vaccine (Armstrong et al. 2019). Vaccination requires a pre-screening serologic and skin prick test to reduce the chance of local and systemic reactions associated with prior exposure. The necessity of pre-screening and occurrence of side-effects means that this vaccine is not suitable if fast delivery of vaccination is necessary in the face of a large outbreak. There is definite need for an effective human vaccine that has minimal side-effects and does not require pre-screening.

2.7. The Netherlands Q fever outbreak (2007–2011)

Between 2007 and 2011, the Netherlands experienced the largest recorded human Q fever outbreak to date (Delsing, Kullberg & Bleeker-Rovers 2010). Over 4 000 cases of acute Q fever were reported with many patients requiring hospitalisation. The outbreak was traced back 32 to dairy-goat and sheep farms infected with coxiellosis (Ladbury et al. 2015; Roest et al. 2011). The intensive, small-ruminant dairy industry had been steadily increasing and herds had experienced episodes of abortions due to C. burnetii for some years without known human cases (Roest et al. 2010). In 2007, the first stage of the human outbreak was noticed in a single village surrounding infected goat farms. From 2008 to 2009, over 2 300 cases were reported at the peak of the epidemic, with seasonal cases subsequently decreasing. Many cases had no direct contact to the infected animals, however, resided within a 5 km vicinity of infected farms (Roest et al. 2010).

In response to the outbreak, in June 2008, it was legislated in the Netherlands that coxiellosis become notifiable in small ruminants. This was the first of many animal control measures implemented, including mandatory vaccination of dairy sheep and goats and finally culling of all pregnant sheep and goats on Q fever positive farms (Roest et al. 2010). Approximately 50 000 pregnant dairy-goats were destroyed in accordance with legislation. This event highlights that the spread of C. burnetii can be a serious threat to public health, animal health and industry.

Since this epidemic, there has been an increase in published research on the epidemiology of coxiellosis in ruminants as well as the zoonotic spillover effects. These studies provide invaluable analysis and insights into the Netherlands’s outbreak and potential risks for surrounding areas. However, Australia has very different agricultural industries and ecosystems, therefore findings from studies in Europe may not be applicable to the Australian situation.

2.8. Distribution of beef and dairy cattle in Australia

In 1788, the first European settlers brought cattle to Australia for milk and meat production. Slow to start, the beef industry only began to prosper with increased commercialisation and the exportation of frozen beef to the United Kingdom in the 19th century. Since then, the national herd size has grown from 1 044 head in 1800, to a national head count estimated at 29.3 million in 2014 (Meat and Livestock Australia 2014). Over 60% of beef and veal production is currently exported to more than 100 foreign markets. The export value of beef and live cattle in 2014 was $AUD 7.5 billion (Australian Bureau of Statistics, 2018).

The northern regions of Australia are heavily focused on beef cattle production whereas the majority of dairy-cattle enterprises can be found in southern states. Approximately 60% of Australian beef cattle are produced in northern Australia, with Queensland contributing 44% of national production (Figure 2.3; MLA Fast Facts 2014; Australia’s beef industry, 2014). The 33 northern Australian beef industry is considered to include northern Western Australia, the Northern Territory and Queensland. The southern beef industry includes the remaining southern areas of Western Australia, South Australia, New South Wales, Victoria and Tasmania (Martin et al., 2013). These two broad regions are largely distinguished by weather and environmental conditions which in turn drive the selection of cattle genotypes and the differences in management approaches, production outputs and markets. Northern properties tend to be more extensive, and pasture-based with lower stocking density due to the severe northern environment with lower soil and pasture quality, high summer temperatures and distinct wet and dry seasons. Bos indicus cattle are common on northern properties as they have increased tick and heat resistance. Southern regions are characterised by smaller land areas, higher soil and pasture quality, improved pastures, higher stocking densities and a dominance of Bos taurus breeds.

44 OF

NUMBER

(%) 20

TOTAL

14 OF

CATTLE 8 7 4 3

QUEENSLAND NEW SOUTH VICTORIA THE WESTERN SOUTH TASMANIA

PROPORTION WALES NORTHERN AUSTRALIA AUSTRALIA TERRITORY

STATES AND TERRITORIES OF AUSTRALIA Figure 2.3 Breakdown of the beef cattle numbers per state as a proportion of total number of beef cattle in Australia (Meat and Livestock Australia 2014)

The dairy industry is ranked third in value of Australian rural productions behind beef and wheat (Halliday 2018). The national bovine dairy herd includes 1.56 million head of cows, although the majority of milk production occurs in the south-eastern states. Following deregulation of the dairy industry in 2000, the number of dairy farms in northern Australia decreased dramatically. Only 7% of all milk produced in Australia comes from Queensland on 393 registered dairy farms focused around south-east Queensland, central coastal Queensland and a focal region of North Queensland (Halliday 2018). There are no dairy-cattle properties in the Northern Territory and most of northern Australia is not suitable for dairy farming. Although the beef and dairy industries can be considered distinct with regards to breeds and production 34 outcomes, there is some overlap of these industries. Farms with beef as their main enterprises may still rear replacement cattle for dairy herds or provide agistment services for dairy cattle. There is also interaction of beef and dairy cattle at saleyards and slaughterhouses as beef and culled-dairy cattle are processed in the same meat-works.

Figure 2.4 Distribution of beef cattle as of 30 June 2001 (‘Australian Bureau of Statistics, Australian Government’ 2019).

Figure 2.4 shows a map of the estimated density of beef cattle distributed across Australia in 2001. The beef industry can be further broken down into 12 regions according to production intensity, climate and topography, and 6 main enterprise categories (AusVet 2006), however a detailed breakdown of the industry is beyond the scope of this review. According to the Australian Bureau of Statistics (2016), approximately 55% of farms within Australia contain beef cattle, making it the most common agricultural activity in the country.

2.9. Detection and diagnostic tests in ruminants

Q fever was originally diagnosed by inoculating mice or guinea pigs with the blood or urine of an infected patient and monitoring the animals to determine whether disease occurred or not (Derrick 1937). Since then, several methods have been developed and employed for the diagnosis of Q fever in humans and coxiellosis in animals. These include both direct detection of C. burnetii from blood, body discharge or infected tissue samples, and indirect serological methods for the detection of antibodies in serum or milk (Herremans et al. 2013; Vincent et al. 2015). Detection of infection in animals is often associated with human outbreak investigations or secondary to reproductive loss/abortion investigations in a herd. Q fever may be initially suspected from a clinical history and then confirmed with laboratory diagnostics (Eastwood et al. 2018). Coxiellosis, on the other hand, may be asymptomatic in cattle and therefore, often is underdiagnosed (El-Mahallawy et al. 2016).

2.9.1. Direct diagnosis

2.9.1.1. Bacterial culture Livestock that are investigated for abortive disease may have placenta, aborted foetal tissue or birthing/vaginal discharges sampled for isolation of C. burnetii (Omsland et al. 2009; Roest, Bossers & Rebel 2013). Specimens can be used to inoculate embryonated chicken yolk sacs or other conventional cell cultures, however culturing of C. burnetii can be inherently difficult (Omsland et al. 2009). The need for PC3 restricted laboratory requirements also makes culturing of C. burnetii expensive and often impractical. These complications have likely contributed to the lack of investigative studies using this method for detection. When designing a study to identify bacterial presence, detection that does not require culture, such as molecular methods, are often more practical and economical.

2.9.1.2. Polymerase chain reaction (PCR) Detection of C. burnetii DNA can be achieved using PCR assays to test animal, human and environmental samples (Roest, Bossers & Rebel 2013). Bacteraemia during the early phase of acute Q fever, allows whole blood and serum to be useful samples for PCR detection (Vincent et al. 2015). However, blood and serum have not been established as suitable samples in animals, as the early bacteraemia stage of disease is often missed. Instead, milk and tissue samples are commonly used for PCR detection in animals. Samples must undergo a DNA extraction process and then targets of the C. burnetii genome are used for PCR assays. Commonly used targets are the com1 and IS1111 elements of the C. burnetii genome as there

36 are multiple copies of the targets along the genome enabling a highly specific and sensitive detection method (Klee et al. 2006; Lockhart et al. 2011). Real-time PCR methods allows quantification of bacterial load (Stenos, Graves & Lockhart 2010).

2.9.1.3. Indirect diagnosis (Serology) Serological methods can be employed for animal and herd level investigations in livestock. Identification of specific C. burnetii antibodies in blood and/or milk provides evidence of past exposure or recent infection (Natale et al. 2012). There is currently no “gold standard” test for sero-diagnosis of coxiellosis in livestock; however, the most commonly used serological methods are: the complement fixation test (CFT), indirect immunofluorescent antibody test (IFA) and enzyme-linked immunosorbent assay (ELISA; Brom et al., 2015; Natale et al., 2012; Porter et al., 2011).

Due to the ability of C. burnetii to show phase variation, both phase I and phase II forms of the antigen can incite specific immune responses. Serological tests that distinguish between antibodies to phase I and phase II variations of C. burnetii would be preferable. In human studies, high phase I IgG and IgA antibodies have been associated with the chronic form of Q fever disease (Eastwood et al. 2018). In ruminants, no clear association has yet been identified between C. burnetii phase-specific antibody response and particular forms of clinical disease, although there are suggestions that phase-specific serology may be a useful tool to identify high- shedding cattle (Lucchese et al. 2015).

2.9.1.3.1. Complement fixation test (CFT) Complement fixation tests were commercially produced for the diagnosis of Q fever in human cases in 1941 (Hendrix & Chen 2012). This test does not distinguish between phase I and phase II antibodies. While it has been the standard serological test used in many diagnostic laboratories historically, it has now been shown to be significantly less sensitive than ELISA and IFA tests. One study investigating test sensitivities of CFT in cattle and small ruminants reported a sensitivity of 36.7% in cattle and 20.6% in goats (Horigan et al. 2011). CFT for the diagnosis of C. burnetii in animal investigations is considered to be declining in use (Emery, Ostlund & Schmitt 2012; Horigan et al. 2011; Pritchard et al. 2011).

2.9.1.3.2. Enzyme-linked immunosorbent assay (ELISA) Enzyme-linked immunosorbent assays can incorporate a combination of both phase variations of the antigen to detect total antibodies in the sample. Microplate wells coated with separate phase I and phase II antigen are also available, thereby providing phase specific

37 antibody detection. This process is relatively easy to conduct and allows large numbers of samples to be processed quickly. The OIE (World Organisation for Animal Health) reference guide suggests the use of serological ELISA for detection of herd coxiellosis status and ELISA in conjunction with PCR identification of bacteria for individual animal diagnosis (OIE 2018). While ELISA kits are reported to have high sensitivity and specificity, current literature highlights the fact that there have been conflicting results when comparing different ELISA kits for use in ruminants in different locations (Horigan et al. 2011; Kittelberger et al. 2009; Rousset et al. 2007).

2.9.1.3.3. Indirect immunofluorescent antibody test (IFA) The IFA test method is the preferred reference method for human Q fever diagnostics; however, it has not frequently been used in animal serology. The IFA method can be used to detect phase specific antibodies in serum and milk of animals. There are currently no commercially available IFA kits for ruminants in Australia. An in-house human IFA has recently been validated using Bayesian latent class analysis and found to have better sensitivity than an ELISA (IDEXX Q Fever Ab Test) and CFT for detecting IgG and IgM antibodies against C. burnetii in goat serum (Muleme et al. 2016).

2.9.1.3.4. Validation of serological tests For C. burnetii serological testing, there are challenges in the interpretation of results due to the relative discrepancies between tests and the lack of a gold standard reference for test evaluation (Table 2.2). Many publications will report raw test results from serology (apparent prevalence) without considering the diagnostic sensitivity or specificity of the test used. The diagnostic sensitivity of a serological test can be defined as the ability of a test to correctly identify positive, known seropositive animals. The diagnostic specificity is the ability of a test to correctly identify negative animals in a population of known non-exposed animals (Thrusfield 2007). In order to be able to accurately compare prevalence estimates between studies, diagnostic tests must be comparable and evaluated for biases and inaccuracies (Collins & Huynh 2014).

Traditional methods to estimate the sensitivity and specificity of a test are mainly based on comparison of test results with a gold-standard reference test, known to have 100% sensitivity and specificity. When this gold standard does not exist, there are alternative methods that can be used to estimate the accuracy of the test. Using the results of multiple imperfect tests and performing statistical latent class analysis is a valid way to produce estimated test parameters and is suggested by the OIE for diagnostic test validation in animal studies (OIE 38

2016a). When the diagnostic test parameters are known, a more accurate estimate of prevalence (true prevalence) can be calculated ensuring comparable results between studies.

Table 2.2 Table of published diagnostic sensitivity and specificity of serological tests for use in cattle and goats

Sensitivity Specificity Sample Diagnostic test Reference (%) (%) size (Muleme et CFT 1* 29.8 96.8 96 Complement al. 2016) Fixation Test (CFT) (Horigan et CFT 2¥ 36.7 100 246 al. 2011) (Horigan et ELISA 1¥ 51.9 100 246 al. 2011) (Horigan et ELISA 2¥ 100 92.9 246 al. 2011) Enzyme-linked (Horigan et immunosorbent ELISA 3¥ 82.6 100 246 assay (ELISA) al. 2011) (Paul et al. ELISA 4 84 99 800 2013) (Muleme et ELISA 5* 70.1 96.2 96 al. 2016)

Indirect immuno- (Muleme et fluorescent IFA 1* 94.8 92.5 96 al. 2016) antibody test (IFA)

¥ Diagnostic sensitivity and specificity results reported using TAGS (test accuracy in absence of a gold standard) software analysis. * Diagnostic test validation was performed with goat serum; all other validation studies were performed with cattle serum.

2.10. Epidemiology of coxiellosis in cattle

2.10.1. Experimental infection in cattle Early publications have described the experimental infection of cattle with C. burnetii (Bell, Parker & Stoenner 1949; Derrick, Smith & Brown 1942). Although sample sizes were small, this early research provided great insight into possible routes of infection and pathogenesis of coxiellosis. Derrick et al. (1942), experimentally infected calves via subcutaneous inoculum derived from infected guinea-pigs. Two out of four calves experienced a mild febrile condition 3 days after inoculation, bacteraemia was present by the fourth day and antibodies to C. burnetii were detected from 11 to 29 days post exposure. Results from these

39 experiments demonstrated that cattle are susceptible to infection with some experiencing asymptomatic acute infection.

In the USA, a controlled experimental study was performed in heifers, lactating cows and male calves (Bell, Parker & Stoenner 1949). Treatment groups were inoculated with C. burnetii infected yolk sac via teat canal, intranasal, intravenous, vaginal tract or alimentary tract through ingestion of milk. Results indicated that C. burnetii infection could be produced in cattle infected via teat canal, intranasal, intravenous, vaginal tract routes, although results were inconclusive regarding ingestion of infected milk, as it was not conclusively determined if calves became infected. Of the other treatment groups, bacteraemia was identified within the first 5 days and urine tested positive for up to 8 days post inoculation in some animals. C. burnetii antibodies were detected at high levels in the serum of experimentally infected cows (using complement- fixation test) for up to 191 days after inoculation. Lactating cows, inoculated via teat canal and cervical canal, developed systemic infections. Milk and blood became infective and an acute mastitis developed, following a noted systemic reaction including marked pyrexia, serous nasal discharge, severe depression, inappetence, decreased rumination, moderate tachycardia and moderate polypnea. The cows recovered spontaneously without antimicrobial treatment, however C. burnetii continued to be shed in the milk for over 200 days in some cases. Sacrificed infected cows were examined post mortem on days 5, 11, 22 and 63 post inoculation. Bacteria was isolated from many tissues including liver, spleen, lung, lymph nodes, intestinal tract and mammary glands, however, pathological changes were not obvious in these tissues with the exception of oedematous mastitis and local lymphadenopathy.

The above mentioned experiments did not include pregnant cows, therefore observations could not be made on the effect of experimental infection of C. burnetii on pregnancy outcomes in cattle. A controlled experimental study published in French was translated to English for this literature review (Plommet et al. 1973). This study aimed to investigate acute and chronic pathogenesis of C. burnetii infection in cattle, including pregnancy outcomes (Plommet et al. 1973). The treatment group consisted of 12 heifers that were inoculated with an intradermal suspension of C. burnetii, and a control group of 98 heifers of the same age and source. Eleven cattle from the treatment group and 54 from the control group were artificially inseminated 8 months following the inoculation date and observed longitudinally for their entire pregnancy or until abortion occurred.

Cattle became seropositive between 6 and 13 days after inoculation in all treatment group animals and titres progressively decreased, except at parturition or time of abortion when 40 serological titres increased markedly. All inoculated animals developed pyrexia, inappetence and acute respiratory signs within 24 – 36 hours. Acute symptoms began to self-resolve after 7 days. Chronic non-reproductive disease is suggested in this study, as one animal died of heart failure 6.5 months after inoculation. Four out of the 11 inseminated treatment heifers had a normal pregnancy and delivered live calves. Two were slaughtered during gestation and were found to have normal foetuses. Two heifers aborted dead foetuses and 3 heifers may have suffered early embryonic loss. Overall, this experimental study identified a fertility rate in the inoculated group of 73% compared to the control group of 93%; abortion rate respectively was 37% (treatment) compared to 1.7% (control) and full term delivery of a live healthy calf was observed in 55% of the treatment group compared to 81% in the control group. There were no statistical hypothesis testing performed or reported in this publication.

2.10.1. Transmission and pathogenesis in naturally infected cattle Inhalation and possibly ingestion (through the oropharynx) of the SCV form of C. burnetii, seems to be the natural route of transmission in animals (Porter et al. 2011; Woldehiwet 2004). The highly resilient, “spore-like” form of the bacteria can be aerosolised during shedding directly from infected animals (other livestock and wildlife) and can survive in the environment. There is a single report that suggests venereal transmission in cattle (Kruszewska & Tylewska- Wierzbanowska 1997); however, more research is required to substantiate this as a common route of infection. It has been shown through experimental studies that ticks can transmit infection between animals; however, it does not appear to be a necessary invertebrate host for infection to survive in livestock populations (Derrick, Smith & Brown 1942; Derrick 1961; Maurin & Raoult 1999). The SCV bacterium has an affinity towards host macrophages and monocytes and can survive phagocytosis, the LCV then replicates within phagolysosomes. Multiplication in regional lymph nodes occurs early in the infection with a short bacteraemia for 7- 21 days (Woldehiwet 2004). The placenta and mammary tissue are the primary target organs for infection and multiplication of C. burnetii in pregnant ruminants (Agerholm 2013; Brom et al. 2015; H. Roest et al. 2013). Although there is not much information on the pathogenesis of C. burnetii infection in non-pregnant animals, the bacteria has been identified in inflamed cardiac valves of cattle going to slaughter (Hansen et al. 2011).

There is little current research on “acute” infection and the intrauterine spread of C. burnetii in cattle. A published review on the clinical effect of coxiellosis in domestic ruminants has described the complexity of possible outcomes of an intrauterine C. burnetii infection in a pregnant animal (Agerholm 2013). After infection is established in the placenta, C. burnetii may

41 follow two main routes: a latent infection or an active infection. Likely outcome then depends on infection remaining localised to the placenta or spreading vertically to the foetus through transplacental or haematogenous spread. These two main routes can lead to two likely outcomes; firstly, normal offspring or secondly an “abortion, premature delivery, stillbirth and weak offspring (APSW Complex)” (Agerholm 2013). Agerholm (2013) proposed this model for natural infection in pregnant cows which may explain conflicting reports on the clinical effect of coxiellosis in pregnant cattle. Some research indicates mostly asymptomatic infection, others found coxiellosis to be associated with sporadic abortion, subfertility, placentitis, retained foetal membranes and metritis (Cabassi et al. 2006; Agerholm 2013; Bildfell et al. 2000; López-Gatius, Almeria & Garcia-Ispierto 2012). It is reported extensively in literature that cattle can then remain chronically infected with C. burnetii without showing overt signs of disease (Lang 1990; Guatteo et al. 2006; Agerholm 2013). Cattle are therefore more likely to have asymptomatic (latent) infections than small ruminants.

2.10.2. Excretion of Coxiella burnetii Bacterial shedding of C. burnetii by infected ruminants mainly occurs through placenta and placental fluids but may also occur inconsistently through urine, faeces, vaginal discharge, milk and semen (Guatteo et al. 2007). Experimental infection in goats revealed that C. burnetii has a strong affinity for the trophoblasts of the placenta (Roest 2013). C. burnetii was identified in trophoblasts by immunohistochemistry and bacterial deoxyribonucleic acid (DNA) was also recovered from allantoic and amniotic fluid and tissue from the base of the cotyledonary villi (Roest 2013). Bacterial shedding patterns specifically in cattle are not completely understood. Inconsistent shedding of C. burnetii has been reported from observational studies in naturally infected dairy cows and false negative results may be common, as not all infected cows will be shedding bacteria consistently in faeces, urine, milk and vaginal mucous (Guatteo et al. 2006, 2007; Guatteo, Joly & Beaudeau 2012). A recent longitudinal observational study, monitored serological patterns and bacterial shedding (from vaginal mucous and milk) in an endemically infected dairy-cattle herd in Germany (Freick et al. 2017). This study found that at 3 weeks prior to parturition, 0% of cows were positive for bacterial shedding via vaginal mucous or milk. The time when highest vaginal shedding was observed was at parturition and highest milk shedding was observed at 100 days in milk. These results are consistent with shedding patterns observed in an endemically infected dairy-goat herd (Muleme et al. 2017; Canevari et al. 2018). Overall, it seems likely that cattle can be latently infected with C. burnetii and excrete the highest numbers of bacteria at the time of parturition through birthing fluids, followed by an extended period of milk-shedding during lactation (Guatteo et al. 2007; Porter et al. 2011). 42

2.10.3. Prevalence of infection in Australian ruminants There are limited published studies assessing the immune response to exposure or active infection from C. burnetii in cattle, sheep and goats in Australia. In preparing this review, only eight Australian studies were found that reported C. burnetii seroprevalence in farmed cattle and sheep, two in farmed goats and two investigating feral goats at slaughterhouses. Australian investigations using serological diagnostic methods or direct PCR detection of C. burnetii from animal samples during prevalence surveys or outbreak investigations have been summarised into Table 2.3 for cattle and Table 2.4 for sheep and goats.

2.10.3.1. Australian cattle Only one publication could be found investigating C. burnetii seroprevalence in cattle from northern Australia. In 2009, a cross-sectional study of Queensland beef-cattle (n = 1 835) reported an animal-level seroprevalence of 16.8% (Cooper et al. 2011). When examining prevalence at the property-level, at least one positive animal could be detected on 78.5% of breeder properties (n = 46) and 40.0% of properties supplying cattle to one abattoir in Northern Queensland (n = 17; Cooper et al., 2011). At the time of this thesis, no studies could be found reporting C. burnetii in cattle from the Northern Territory. There are currently no reports of C. burnetii seroprevalence in dairy-cattle in Queensland. Global publications suggest that dairy cattle have a higher prevalence than beef cattle (Paul et al. 2014; Alvarez et al. 2012).

Table 2.3 shows that animal-level C. burnetii seroprevalence in cattle from southern Australian states have consistently been reported as less than 1% from 1954 to 2018 (Cronin 2015; Hore & Kovesdy 1972; Tan 2018; Banazis et al. 2010; Forbes, Wannan & Keast 1954). A study conducted in 1972 across dairy-cattle herds in a region of Victoria identified 12.0% of properties (n = 49) had at least one positive animal (Hore & Kovesdy 1972). A recent pilot study tested cattle on farms and abattoirs and found similar results. In the Gippsland region, herd level seroprevalence was 8.7% (2/23) and animal level seroprevalence 0.8% (2/247). In the Goulburn Valley no positive animals were identified, however, post-hoc sample size analysis revealed much larger numbers would be required for cattle in that region to be considered truly free of C. burnetii (Tan 2018). A New South Wales government case report identified that 9.1% (2/22) of properties had at least one positive animal, although only 0.7% (3/440) of total animals showed exposure (Cronin 2015).

Although there appears to be a difference in C. burnetii prevalence between cattle in northern and southern Australia, the reported results should be interpreted with caution, as the study design, sampling methods and diagnostic tests used, differ across most of the surveys. 43

While earlier studies used CFT method based on the Nine Mile strain antigen (Forbes, Wannan & Keast 1954; Hore & Kovesdy 1972), some studies did not specify the test used (Durham & Paine 1997) and the only study to report diagnostic sensitivity or specificity of the test used was Tan et al. (2018). All reported prevalence results in Table 2.3 represent crude estimates of number of test positive cattle divided by the total number cattle tested and have not been adjusted for diagnostic sensitivity and specificity of the test used.

2.10.3.2. Australian sheep and goats A critical review of the literature reporting global prevalence of C. burnetii in domestic ruminants has concluded that estimated prevalence in sheep and goats is slightly lower than in cattle; around 15.0% at the animal level and 25% at the herd level (Guatteo et al. 2011). It is difficult to directly compare global trends with Australian trends due to the scarcity of studies specifically designed to estimate C. burnetii prevalence in small ruminants.

Table 2.4 shows a summary of publications that have tested sheep and goat, however some of these are reports from outbreak investigations, and others have used convenience sampling. Serological tests methods also differ between the studies.

In 1959, Derrick et al. described a human outbreak of Q fever associated with sheep shearing in Queensland. In this study, the sera of 174 sheep were tested for C. burnetii antibodies. Results indicated 11.0% of animals were positive. Although at the time Derrick believed this to be quite low, it is consistent with globally reported seroprevalence in sheep (Derrick, Pope & Smith 1959). From what is available in the literature, it appears that in southern Australia the seroprevalence of C. burnetii in sheep and goats are higher than cattle, while in Queensland, cattle have been reported to have a higher seroprevalence than in sheep and goats.

Published studies investigating C. burnetii in Australian goat populations are still scarce, even though there is an increasing awareness of the potential role of farmed goats in the spread of Q fever. Two seroprevalence surveys from slaughtered feral goats in Queensland and South/East Australia were published more than 30 years ago and indicate a seroprevalence of 10.0% (n = 20) and 51.5% (n = 171), respectfully. A more recent epidemiological study investigating C. burnetii on a dairy-goat property was initiated after a localised Q fever outbreak in Victoria (Bond et al. 2015). Following the confirmation of Q fever in 17 farm employees and one family member over a 28 month period, a One Health approach was used to investigate and control the outbreak. As part of this investigation, the seroprevalence of a sample of non-

44 pregnant milking goats was estimated using serological CFT. Seroprevalence in the sample of goats was reported to be 15% (95% Confidence Interval (CI); 7, 27) (Bond et al. 2015).

Table 2.3: Reported C. burnetii prevalence studies from Australian cattle between 1954 and 2018

Number of samples Estimated prevalence (%) Publication Diagnostic Sample Study location Cattle type Reference date test tested cattle properties animal level property level

1954 Beef & dairy CFT serum 700 unknown 0 0 1 New South Wales 2013 Beef ELISA* serum 440 22 0.7 9.1 2

South Australia 1997 Beef unknown serum 617 10 0.16 10 3

1972 Dairy CFT serum 1 576 49 0.5 12.2 4

2018 Beef & dairy ELISA* serum 278 78 0 0 5 Victoria

2018 Beef & dairy ELISA* serum 247 23 0.8 8.7 5

2010 Beef ELISA** serum 329 unknown 0.6 unknown 6 Western faeces & Australia 2010 Beef PCR assays 329 unknown 7.9 unknown 6 urine

Queensland 2011 Beef ELISA *** serum 1 835 63 16.8 78.5 7

Key: 1 Forbes et al., 1954; 2 Cronin, 2015; 3 Durham and Paine, 1997; 4 Hore and Kovesdy, 1972; 5 Tan, 2018; 6 Banazis et al., 2010; 7 Cooper et al., 2011; CFT = Complement fixation test; ELISA = Enzyme-linked immunosorbent assay; PCR = Polymerase chain reaction;* Q Fever Ab Test (IDEXX Laboratories, Unites States of America);**CHEKiT Q Fever ELISA kit (IDEXX Laboratories Inc., Switzerland); *** In-house test kit (James Cook University, Townsville, Australia)

Table 2.4: Published studies of C. burnetii sheep and goats in Australia between 1949 and 2018

Number of Proportion positive Study location Species tested Diagnostic test Reference samples %

Feral goats CFT 20 10% McKenzie et al., 1979

Queensland

Sheep unknown 174 11% Derrick et al., 1959

New South Wales Feral goats CFT 42 57% Hein and Cargill, 1981

South Australia Feral goats CFT 129 49% Hein and Cargill, 1981

Non-pregnant dairy goat CFT 65 15% Bond et al., 2015

Victoria Farmed goat ELISA 500 4% Tan, 2018

Farmed sheep ELISA 590 1% Tan, 2018

Sheep ELISA 50 0% Banazis et al., 2010

Western Australia PCR assays Sheep 50 0% Banazis et al., 2010 (faeces)

Key: CFT = Complement fixation test; ELISA = Enzyme-linked immunosorbent assay; PCR = Polymerase chain reaction 47

2.10.4. Prevalence of infection in cattle globally Coxiellosis is likely present in cattle populations’ globally; New Zealand is the only country declared free by the OIE World Organisation for Animal Health (Woldehiwet 2004; Maurin & Raoult 1999; OIE 2018). Knowledge of the true prevalence of infection in cattle populations across different countries is increasing, however issues of study design and inconsistencies with diagnostic tests are not unique to Australian studies. It was highlighted in a critical review on the prevalence of C. burnetii in domestic ruminants, that there are major methodological issues with many global studies (Guatteo et al. 2011). From 51 publications analysed for the review, only five of these were considered reliable based on qualitative assessment of the study methodologies (Guatteo et al. 2011). Regardless of the qualitative assessment, there was a lot of variation in the reported prevalence of cattle globally. Meta-analysis of publications revealed animal-level prevalence in cattle to range from 0% to 100% with a median of 19.4% (interquartile range (IQR) 6.6%–39.3%; n = 36). At the herd level there were reports from 4.4% to 100%; with a median of 37.7% (IQR 19.3– 69.7%; n = 27). Within-herd prevalence reports varied from 0% to 48.7% with a median 26.3% (IQR 21.7%–38.2%; n = 7).

The majority of published investigations in cattle globally are from the dairy industry and minimal studies have investigated C. burnetii exposure or infection in beef cattle. Table 2.5 shows a summary of studies where beef cattle have been included. When dairy and beef cattle are reported in the one study, there are consistent findings of lower prevalence in beef than dairy herds (McCaughey et al. 2010; Alvarez et al. 2012; Paul et al. 2014). In the Madrid region of Spain an overall apparent prevalence of 6.76% (95% CI 5.42, 8.41) was reported. When stratified for production types, beef cattle showed 1.89% (95% CI 1.34, 2.54) seropositive and dairy cattle 2.73 % (95% CI 2.04, 3.52) seropositive (Alvarez et al. 2012). Although, when analysed for herd level prevalence, beef cattle were found to have only 24.3% of herds positive compared to dairy cattle, where 75.0% of herds tested positive. A similar pattern was identified in France; the between-herd seroprevalence was significantly higher in dairy cattle (n = 176, mean: 64.9%, 95% CI = 58.9, 70.6) compared to beef cattle (n = 87, 18.9%, 95% CI 15.4, 22.8) (Gache et al. 2017). Overall, there have been reports of animal-level prevalence in beef cattle ranging from 1.7% (in Korean native cattle) to 6.6% (in semi-extensive grazing systems in northern Spain; Ruiz-Fons et al., 2010;Table 2.5).

Table 2.5 Globally reported C. burnetii seroprevalence studies with beef cattle included Estimated Sample size apparent Estimated Country of Serological prevalence Reference Cattle type herd study Number of Number of test (reported 95% prevalence animals herds confidence interval) (Staley, Zebu Malawi MyBurgh & 200 - CFT 6.5 - (beef primarily) Chaparro 1989) 2.8 beef 2 826 152 35.5 (2.2–3.4) (McCaughey et 10.4 Ireland dairy 2 356 121 ELISA 64.5 al. 2010) (9.1–11.6) 6.2 total 5 182 273 48.4 (5.6–6.9) (Ruiz-Fons et al. Spain beef 626 46 ELISA and CFT 6.6 (± 2%) 43.0 2010) 1.89 beef 720 72 25.0 (1.34–2.54) 2.73 (Alvarez et al. dairy 200 20 75.0 Spain ELISA (2.04–3.52) 2012) * bullfighting 180 18 0 0.0 6.76 total 1 100 110 33.0 (5.42–8.41) Korean native (Lyoo et al. 3 087 362 1.7 - Korea (beef primarily) IFA 2017) dairy 1 224 171 10.5 - 4.54 beef 716 - - (3.16–6.30) (Paul et al. 15.10 Denmark dairy 84 - ELISA - 2014) * (8.53–23.64) total 800 - 5.50% - * indicates seroprevalence studies that also performed risk factor analysis; - indicates was not reported in the study 49

Overall, there is a need for studies to be performed that incorporate knowledge of diagnostic test parameters and appropriate sampling procedure for assessing the true prevalence of exposure and infection in cattle populations across different industries and geographical regions.

2.10.5. Risk factors for coxiellosis in cattle Several global studies have investigated risk factors for exposure or infection to C. burnetii through observational studies. Most utilise a cross-sectional study design with serological antibody response to anti- C. burnetii IgG (phase I and/or phase II) as the outcome variable, however some studies also used PCR testing as a criteria for infection (Paul et al. 2014, 2012; Agger & Paul 2014). Additional data at both animal and herd level have been used to assess risk factors for previous exposure, or active infection. Most studies focused on dairy cattle (Boroduske et al. 2017; Agger & Paul 2014); however, there are a few seroprevalence surveys including dairy, beef, bullfighting and mixed-use cattle that performed further analysis to identify putative risk factors (Table 2.5). There have been no studies primarily focused on identifying risk factors for coxiellosis in beef cattle.

A C. burnetii sero-survey of cattle going to slaughter in Denmark assessed risk factors using Bayesian methods to account for diagnostic test uncertainties (Paul et al. 2012). They identified that the number of animal movements, age and breed groups (cattle raised for milk production vs meat production) were risk factors for seropositivity (Paul, 2013). This finding is consistent with a report from Spain that identified production type (dairy, beef, bullfighting) and herd size as significantly associated with seropositive results at bivariate analysis (Alvarez et al. 2012). However, in the final multivariable model, production type was not found to be significantly associated with herd-level test results.

The hypothesis that dairy cattle are at a higher risk of C. burnetii infection could be explained by genotypic differences between cattle breeds. Within commercially used dairy cattle, it was identified that Danish Holstein breeds had an increased risk of seropositivity compared to Jersey breeds (Paul et al. 2012). Holstein breeds are typically higher yielding milk producers, therefore this association could be correlated with the metabolic stress experienced during lactation, enabling C. burnetii infection to thrive. However, the putative association of prevalence and production-type could be just as likely due to different management systems implemented across cattle industries. Intensive housing, controlled mating and synchronised calving periods tend to be more common in dairy production compared to beef and bullfighting production. These management factors may increase the

50 risk of transmission of infection at the time of parturition and lactation in closely confined animal sheds. In Southern Italy, farm management practices were assessed to identify if different cattle housing practices could be identified as potential risk factors for increased C. burnetii seropositivity (Capuano, Landolfi & Monetti 2001). A cross-sectional study was performed using 1 188 cattle from 53 farms. Serology was performed with an IFA test and animal-level proportion positive used as the outcome variable of interest. The farms were categorised into three groups according to housing: permanently housed cows (n = 21), cattle housed in winter and moved to graze in spring (n = 26), cattle not housed (n = 6). Results indicated that cattle housed in winter after grazing in spring had a higher seroprevalence of coxiellosis (19.6%) and non-housed cattle had the lowest seroprevalence (1.9%). Although this study did find a statistically significant difference between seropositivity and specific housing systems, there are limitations in the analysis that reduces the validity of the findings. The animal-level prevalence estimates do not take into consideration the clustered nature of the data (animals grouped within farms) and no other putative risk factors were included in the analysis. This study would benefit from multilevel, multivariable analysis that includes data on animal and herd level variables such as breed, age, size of herd and production type, to ensure associations are accurately interpreted.

A well-designed cross-sectional investigation focused on cattle in the main milking region of Ecuador identified some interesting results (Carbonero et al. 2015). Serological testing was performed on a sample of dairy and mixed (dairy-beef) cattle (animal n = 2 668, herd n = 386) using an ELISA test with known diagnostic sensitivity and specificity. Individual cattle data (age, breed and sex) were collected and farm surveys completed to assess additional explanatory variables related to nutrition, farm facilities, general farm biosecurity and animal health. Univariable analysis was performed, and all explanatory variables significant at p > 0.2 by a Chi-square test were then included in multivariable analysis. Generalised estimating equations (GEE) modelling was used to determine factors associated with seropositivity. From 26 variables retained for the multivariable analysis, only four showed significance at p < 0.05. There was a positive association between increasing C. burnetii seropositivity and increasing age, feeding calves milk replacer and the presence of bovine respiratory syncytial virus. The fourth factor, disinfection of the calves’ umbilical cord, was identified as a protective factor. Cow age or parity have been associated with seropositivity in several other studies (McCaughey et al. 2010; Böttcher et al. 2011; Paul et al. 2012). Feeding calves milk replacer and the presence of bovine respiratory syncytial virus

51 are novel putative risk factors and may be worth following up with prospective cohort studies to elucidate if these factors could be considered causal.

In Denmark, a cross-sectional study identified a decreased risk of C. burnetii seropositivity in dairy-cows from herds where the quarantine of newly purchased animals exists and where veterinarians took higher hygienic precautions (e.g. biosecurity and infection control measures; Paul et al., 2012). Another study focused on dairy-cattle in Latvia and identified purchasing cattle from abroad and the increasing number of cattle in milking sheds as associated with increased seropositivity and PCR positive milk (Boroduske et al. 2017). The movement of latently infected cattle without standard quarantine practices and increased stocking density, seem plausible factors to increase the risk of spread of C. burnetii into uninfected herds and to increase transmission between infected and susceptible animals. These studies highlight the importance that simple quarantine practices and not overstocking could have in reducing the potential spread of C. burnetii between and within cattle populations.

Currently, there have not been any studies that have investigated potential risk factors for coxiellosis in beef or dairy cattle within Australia. There would be great value in investigating recent or past C. burnetii infections across different cattle industries and geographical regions to identify risk factors that may link to animal, environmental or agent factors.

2.10.6. Impact of coxiellosis on reproductive performance in cattle Early experimental studies of the outcome of infection of cattle with C. burnetii were insufficient to support or refute C. burnetii as an important cause of abortion or subfertility in cattle (Bell, Parker & Stoenner 1949). More recent studies, examining clinical laboratory samples from cattle abortions, have detected C. burnetii bacterial DNA, however detection has also been reported from cattle that delivered healthy calves (Jones et al. 2010; Agerholm 2013). Cabassi et al. (2006), reported an association between high C. burnetii seroprevalence and increased abortion rates in dairy-cattle herds in Italy (Cabassi et al. 2006). However, there are some limitations to this study that may influence interpretation of the findings. The study involved a retrospective case-control approach using seroprevalence to evaluate the association between C. burnetii serostatus and a history of abortion (Cabassi et al. 2006). Serum samples were tested using the commercial indirect ELISA test (CHECKIT-Q-Fever, Bommeli Diagnostics, Bern, Switzerland) and reported a seroprevalence of 45% from the 650 animals with a history of abortion; this was compared

52 to a seroprevalence of 22% from the 600 animals with no history of abortion. There was a significant difference between the seroprevalence of the cattle group with a history of abortion and the control cattle group (p < 0.001). However, this study provided comparisons of crude prevalence estimates without having additional data on diagnostic test parameters, the animals or herds. This design makes it harder to establish causation. Prospective cohort designs and inclusion of additional animal and herd level factors in multivariable statistical models would provide stronger evidence of a causal association between C. burnetii infection and abortion.

Large epidemics of abortions due to coxiellosis have been recorded in small ruminant herds and there may be an association with reduced milk production in commercial dairy goats (Guatteo et al. 2011; Roest, Bossers & Rebel 2013; Bond et al. 2015; Canevari et al. 2018). By contrast, only sporadic cases of C. burnetii induced abortion tend to be reported in cattle and there are no reports of epidemics of abortions in cattle herds (Agerholm 2013). More recently, a cluster of 4 cases of C. burnetii abortion was reported by Macias-Rioseco et. al. (2019). The cases occurred within 2 months of each other in a dairy herd in Uruguay. In all cases, abortion occurred late term (approximate gestational age 240–270 days) and histopathological and molecular PCR methods confirmed C. burnetii as the most likely cause of abortion after ruling out other common aetiological pathogens (Macías-rioseco et al. 2019). Under-reporting of C. burnetii as a cause of bovine abortions is possible due to infrequent testing during routine investigations (Carpenter et al. 2006; Anderson 2011).

Although it is stated in many publications that natural infection with C. burnetii can cause placentitis, retained foetal membranes (RFM), metritis and subsequent infertility in cattle, there are mixed results from primary research to support these claims (To et al. 1998; Muskens, van Maanen & Mars 2011; Bildfell et al. 2000; Woldehiwet 2004). Agerholm (2013) concluded from a review of the literature: “evidence has not been provided that shows causation between C. burnetii and poor conception rates, subfertility/infertility, sterility, retained placenta, or endometritis/metritis neither at individual level nor at herd level.”

Primary research from Spain and Germany suggests an association with C. burnetii seropositivity and a prolonged calving to conception interval (CCI) (López-Gatius, Almeria & Garcia-Ispierto 2012; Freick et al. 2017). Firstly, Lopez-Gatius et al. (2012) reported a shorter CCI in cattle with low seropositive results and longer CCI in cattle with high seropositive results. There was a positive association with C. burnetii seropositivity and placental retention, however, no subsequent reduced fertility was identified at the herd level. Secondly, a prospective longitudinal observational study followed an endemically infected 53 dairy cattle herd in Germany and identified a similar pattern of prolonged CCI; a small difference was noted in the CCI between high-milk C. burnetii shedders and non-milk shedders, however it was not statistically significant (Freick et al. 2017). This study concluded that C. burnetii serology and positive vaginal shedding had no association with reproductive outcomes for this herd (Freick et al. 2017).

An extensive investigation performed in dairy-cattle in Spain found that infected cattle (C. burnetii IgG seropositive and PCR positive) had a lower risk of endometritis, showed earlier return to oestrus and shorter calving to conception periods (Garcia-Ispierto et al. 2013). Although these findings seem counter intuitive, it was postulated that a latent C. burnetii infection prior to pregnancy, may protect cows from an acute gestational infection, likened to the pathogenesis of congenital toxoplasmosis infection in women (Garcia-Ispierto et al. 2013). Hence, this latent infection may provide a protective immunity to acute infection during the study period and therefore these cattle showed improved reproductive performance. These results and theory highlight that thorough understanding of this infection has not yet been reached.

Hansen et. al. (2011) conducted a study to determine if C. burnetii infection in dairy cattle was associated with cotyledonary inflammatory lesions after parturition. One hundred Danish dairy cattle herds were randomly selected following participation in a previous Q fever study. These herds were then classified into C. burnetii positive, intermediate and negative herds according to bulk tank milk sampling. The farmers collected placental samples during calving, irrespective of the birthing outcome. These samples were then examined by histopathology, immunohistochemistry and PCR methods. They found that placental inflammation associated with C. burnetii infection was very rare and concluded that placentitis and subsequent RFM following coxiellosis is of little importance in cattle (Hansen et al. 2011).

A recent short communication has been the first to describe endometritis and uterine vasculitis associated with active C. burnetii infection in the uterus of cattle with a history of infertility and chronic endometritis (De Biase et al. 2018). Although the study design does not sufficiently fulfil the criteria for causality, these preliminary findings suggest that C. burnetii could be associated with uterine lesions related to progressive reproductive disorders. Experimentally infecting pregnant cattle with C. burnetii, similarly to pregnant goats in the Netherlands, would be required to further investigate this hypothesis (Roest 2013).

From the current literature, it is not clear if C. burnetii could have a significant impact on herd reproductive performance in cattle, as only a few studies published to date have investigated C. burnetii and reproductive outcomes in cattle from a population or epidemiological approach. There is evidence to support that infection with C. burnetii can lead to “sporadic cases of abortion, premature delivery, stillbirth and weak offspring” as well as the birth of normal progeny. However, there does not appear to be enough evidence to unequivocally confirm an association between C. burnetii infection and reproductive disorders such as endometritis, metriris or RFM’s that may lead to subsequent reduced fertility and conception rates.

2.10.7. Prevention and control of coxiellosis in ruminants Vaccination and antimicrobial treatments have been used in France and the Netherlands in an attempt to control and reduce the prevalence of coxiellosis in ruminants for both animal health and public health reasons (Garcia-Ispierto, Tutusaus & López-Gatius 2014; Roest, Bossers & Rebel 2013; Bontje et al. 2016). An inactivated Nine Mile strain C. burnetii vaccine, Coxevac (CEVA Sante Animale, France), is registered for use in cattle and goats. Vaccination must be implemented to target non-immune animals prior to pregnancy to have a protective effect and prevent abortions and bacterial shedding at parturition (Hogerwerf et al. 2011). Studies in dairy-cattle support the vaccination of nulliparous heifers in endemically infected dairy herds (Pinero et al. 2014; Guatteo et al. 2008). There have been mixed reports on the effectiveness of antimicrobials for preventing abortion and bacterial shedding (Garcia-Ispierto, Tutusaus & López-Gatius 2014). Although there have been suggestions that two consecutive injections of oxytetracycline during the last month of pregnancy can reduce shedding in small ruminants, a recently published veterinary consensus statement advises against the use of antimicrobials for control or treatment of coxiellosis due to a lack of evidence-based benefit (Plummer et al. 2018).

Additional control measures to reduce transmission and environmental contamination in endemically infected herds focus on the prompt removal and disposal of aborted foetuses and placentas as well as limiting the spread of manure into the environment (Arricau- Bouvery & Rodolakis 2005; Kovácová & Kazar 2002). The identification and removal of high shedding goats was found to be impractical in the Netherlands situation due to difficulties with diagnostic tests; this method would require extensive testing that may have become economically impractical (Hogerwerf et al. 2014). The presence of latently infected cattle and high infectivity of C. burnetii can make biosecurity control measures quite difficult.

There are currently no control programs specifically designed to reduce the transmission of coxiellosis in the Australian cattle industries. Australia does not have a licensed animal vaccine to prevent infection, although one is currently being developed with private industry support in Victoria for use in dairy-goats. Human vaccination and protective personal equipment are used to prevent the spread of infection to humans in high risk occupations (Gidding et al. 2009).

2.11. Conclusion

From this literature review, it is clear that while Q fever in Australia is broadly considered endemic, the incidence of human cases varies between geographic regions of the country. The states of Queensland and New South Wales may more accurately be classified as hyper-endemic, while remaining states have low incidence with sporadic cases or occasional outbreaks. Increased research into the epidemiology of human Q fever after the NQFMP, including detailed investigation of the prominent source of human infections, would help increase our understanding of why these differences exist.

After thorough examination of the current literature, there are still major queries as to whether C. burnetii transmission from beef cattle to humans is likely to contribute significantly to the relative high incidence of Q fever in Queensland, Australia. From the limited seroprevalence studies performed in beef cattle, immune response to C. burnetii is also higher in cattle from Queensland than other states of Australia, which could drive the hypothesis of causality. However, research from other countries indicate the genetic strains of C. burnetii found in cattle show little association with bacterium found in humans, suggesting cattle are not a significant source of human infection. At the time of this thesis, no such molecular strain comparisons were published in Australia to provide evidence to support or refute this hypothesis.

It is apparent that C. burnetii has a wide range of animal reservoirs that could play a role in the transmission of infection between livestock and to humans. Although domestic ruminants are often reported to be the most common source of human infections, native wildlife, feral animals, and domesticated pets can spread C. burnetii to humans. Characterisation of different bacterial strains is essential for further understanding the complex epidemiology across multiple hosts and geographic regions, and the molecular genotyping methods that exist are used for this purpose. The identification of bacterial genotypes present in Australian animal populations is crucial for outbreak investigations, in order to ascertain the source of human disease, to aid in the development of improved 56 diagnostic tests and vaccines and to research bacterial evolution and virulence factors of Coxiella.

Prevalence reports of coxiellosis in cattle in Australia are scarce and difficult to compare to international reports due to methodological disparities. In order to fill this gap, there is a need for development of improved and affordable serological diagnostic tests specifically validated for use in cattle. This may encourage surveillance or seroprevalence studies for the purpose of estimating the distribution of disease within livestock and investigating putative risk factors for coxiellosis in cattle in Australia and the role of coxiellosis in cattle as a risk for human Q fever infection.

From the literature, it has been suggested that beef cattle could be at a lower risk of coxiellosis than dairy cattle. This is an interesting hypothesis, although the association has not been thoroughly investigated. It is worth noting that during the 1970s, a significant spike in Q fever cases was seen in Queensland abattoir workers when altered market conditions resulted in an increase in the number of pregnant dairy cattle going to slaughter (Derrick, 1961). Although large abattoirs are focused on beef cattle processing for domestic and international markets, the culling of dairy cows and processing of veal occur at the same establishments. The movement and interaction of beef and dairy cattle at sale and slaughter could be a potential source of transmission of C. burnetii.

In Australia, coxiellosis is not listed as an endemic disease of significance to the meat and livestock industry. However, there has been a lack of investigative effort to determine what effect, if any, coxiellosis may have on the Australian cattle industry. Although it may not be consistently reported, it is clear from experimental and observational studies that C. burnetii has the potential to affect reproductive performance in cattle. These information gaps suggest that there is a need for further research in Australia to understand the potential impacts of C. burnetii infection on animal (ruminants and wildlife) health, disease and productivity, and the role of animal infections in human disease risk (Q fever).

Without any current control programs nor licensed animal vaccines for the prevention of coxiellosis in Australia, there is a need to further study coxiellosis in beef cattle to improve our understanding of the epidemiology and impact of infection in animals and the public health risks from ruminant-derived exposures.

The following chapter is published as a peer-reviewed original article in the Australian Journal of Communicable Diseases Intelligence as follows:

Tozer S*, Wood C*, Si D, Nissen M, Sloots T, Lambert S. (2020) The improving state of Q fever surveillance: A review of Queensland notifications, 2003–2017, Communicable Diseases Intelligence, Vol. 44; pp 1-22

* is to signify that these authors contributed equally to the work carried out for this publication.

I, Caitlin Wood, state that I have participated sufficiently in the publication to take public responsibility for the work. I was joint-first author of the manuscript and I made substantive contributions to the concept and design, analysis and interpretation of the research data on which the publication is based and in the writing and editing of the manuscript.

Chapter Three

Report of human Q fever notification surveillance data in Queensland, Australia, 2003–2017

“Disease does not occur randomly but in patterns

which reflect the operation of underlying causes…

Knowledge of these patterns constitutes the major key

to understanding causation, and hence,

devising methods of control and prevention.”

- John P. Fox, MD, MPH (1908–1987)

3. Report of Q fever notifications in Queensland, Australia, 2003–2017

3.1. Introduction

Q fever, the human disease caused by the bacterium Coxiella burnetii, has a worldwide distribution with the exception of New Zealand (Fournier, Marrie & Raoult 1998; Greenslade et al. 2003; Kaplan & Bertagna 1955; Maurin & Raoult 1999). The main route of transmission is via the inhalation of contaminated dust or bacterium-containing aerosols from infected animals or contaminated environments (Stoker & Marmion 1955). The disease in Australia has historically been considered an occupational risk, primarily for abattoir workers, veterinarians, shearers, tanners, and farmers (Garner et al. 1997). However, more recent reports suggest transmission from unknown sources may have previously been overlooked, or are becoming more common (Tozer et al. 2014; Sloan-Gardner et al. 2017). Several animal species are known to shed C. burnetii into the environment and be potential reservoirs of C. burnetii. In Australia, these include livestock, domestic mammals and native wildlife such as wallabies, dingoes and kangaroos (Cooper et al. 2013; Banazis et al. 2010; Stevenson et al. 2015; Tozer et al. 2014).

Q fever is a notifiable condition in all states and territories of Australia. Primary infection results in clinical manifestations ranging from no symptoms to acute Q fever, the classical presentation of which is an influenza-like illness, fevers, sweats, headaches, and may less commonly include other conditions, such as hepatitis and pneumonia. Of symptomatic patients, 2% are hospitalised (Maurin & Raoult 1999; Parker, Barralet & Bell 2006). Approximately 5% of acute Q fever cases develop a persistent focalised infection, recrudescing months or years after initial infection. This chronic form of Q fever can present as endocarditis in 60–70% of cases, or chronic pneumonia with or without hepatitis (Maurin & Raoult 1999). A third disease state, post- Q fever fatigue syndrome was first described as a chronic fatigue syndrome in Australian abattoir workers (Marmion et al. 1996). Acute and chronic forms of Q fever are treatable with doxycycline, which is the antibiotic of choice (Watanabe & Takahashi 2008).

In Australia, a whole-cell vaccine against Q fever (Q Vax®) was first licensed in 1989, and is manufactured by Seqirus, a subsidiary of CSL limited (Commonwealth Serum Laboratories, Victoria, Australia.). The vaccine is recommended for at-risk individuals aged 15 years or older and the Queensland Government Workplace Health and Safety department (Work Health and Safety Act 2011) monitor use of the vaccine in order to reduce Q fever acquired in the workplace. Between 2001 and 2006, Q Vax® was used in a government funded vaccination program, the

National Q Fever Management Program (NQFMP), with the aim to reduce the incidence of disease (Palmer et al. 2007).

3.1.1. Chapter objectives The aim of this chapter is to describe the extent and trends of Q fever notification data in Queensland over a 15-year period and to investigate enhanced data following changes to the surveillance of Q fever. From 2012, all public health units in Queensland were instructed to complete an enhanced Q fever surveillance form as a routine investigation step following notification of confirmed Q fever cases; providing more consistent follow up, including details on risk exposures and animal contacts. Although previous epidemiological reports reviewing the Australian national Q fever notifications have been published (Sloan-Gardner et al. 2017), to date there have been no publications reporting Q fever notifications isolated to Queensland, Australia.

3.2. Materials and Methods

3.2.1. Ethics statement Human ethics for access to Q fever notification data used in this study was approved by the Children’s Health Queensland Hospital and Health Service Human Research Ethics Committee (HREC) number: HREC/08/QRCH/66 amendment number HREC/08/QRCH/66/AM03.

3.2.2. Q fever case definition Confirmed cases of Q fever must fulfil the national case definition for Q fever notification (Australian Government, Department of Health, Australia 2004). This requires either: (i) the detection of C. burnetii via culture; (ii) the detection of C. burnetii by molecular methods; or (iii) the seroconversion or significant increase in antibody level to Phase II antigen in paired sera taken in the absence of a recent Q fever vaccination, at least 14 days apart, and screened in parallel.

3.2.3. Notification data Notification data, for all confirmed Q fever cases in Queensland, between 1 January 2003 and 31 December 2017, were extracted from the Notifiable Conditions System (NOCS) by the Communicable Disease Branch of Queensland Health. Data available in the NOCS are compiled from clinical information initiated and provided by diagnostic testing from pathology providers, with follow up from select individual public health units (PHUs) via case reporting

61 forms. From 2012 onwards, enhanced surveillance data have been captured using laboratory- initiated notifications of Q fever cases and then direct patient contact via the PHU, collecting data with the Q fever case report form (Appendix Figure A. 3-1). Public health unit investigators conduct this survey either by phone questionnaire or in person. The data is then entered into the NOCS. The data fields extracted for analysis included: year of onset, age, sex, indigenous status, local government area (LGA) of residence, occupation, whether hospitalisation was required, exposure to animal, abattoir and environmental related risks, Q fever vaccination status, and awareness of risk for Q fever and Q fever vaccination.

The Q fever case report form was structured with exposure questions within 3 broad categories: exposure to animal, abattoir and environmental related risks (Appendix Figure A. 3-2). These exposure categories were maintained and responses summarised and presented in summary tables in this chapter. The definition of “any exposure to a putative risk factor” was having indicated exposure to at least one known risk factor within one of these three categories.

3.2.4. Statistical analysis This study provides descriptive analysis, with counts, rates, and proportions of the notification data per year and using three 5-year periods from 2003–2017. Estimated Resident Population (ERP) data were attained per year from the Australian Bureau of Statistics for populations for Queensland and within LGAs (‘Australian Bureau of Statistics, Australian Government’ 2019). Queensland data were also analysed according to Hospital and Health Service (HHS) divisions of Queensland Health. There are currently 16 HHS divisions within Queensland, and these provide general health care to the community. Population data for HHS was obtained directly from open access Queensland Government website (‘Population health data and statistics | Queensland Health’ 2020). The estimated notification rate per LGA or HHS was calculated for each year accordingly:

𝑡𝑜𝑡𝑎𝑙 𝑛𝑜𝑡𝑖𝑓𝑖𝑐𝑎𝑡𝑖𝑜𝑛𝑠 𝑄 𝑓𝑒𝑣𝑒𝑟 𝑛𝑜𝑡𝑖𝑓𝑖𝑐𝑎𝑡𝑖𝑜𝑛 𝑟𝑎𝑡𝑒 100 000 𝑒𝑠𝑡𝑖𝑚𝑎𝑡𝑒𝑑 𝑟𝑒𝑠𝑖𝑑𝑒𝑛𝑡 𝑝𝑜𝑝𝑢𝑙𝑎𝑡𝑖𝑜𝑛

Where the notification rate for multiple years was calculated, the average ERP for the total period was used as an estimated denominator. Where possible, Queensland state Q fever data were compared to Australian national Q fever data, sourced from the National Notifiable Diseases Surveillance System (Australian Government 2019).

Data were cleaned and analysed using software Microsoft Excel 2011 and R (‘R: A language and environment for statistical computing’ 2019). Surveillance data included a free

62 text field for occupation. Data captured in this field were re-categorised and aggregated into occupation group categories. See the chapter appendix for details of groupings (Table A. 3-2).

Visualisation of the spatial distribution of case notification rates per LGA was performed by creating choropleth maps in Q GIS (‘QGIS Geographic Information System’ 2019).

3.3. Results

3.3.1. Q fever notification counts and rates by age and sex Between 2003 and 2017, a total of 2 838 cases of Q fever were notified in Queensland. The range of annual case notification rates was 3.0 to 5.1 per 100 000 population. Annual notification counts and rates for Queensland and Australia have been presented together in Figure 3.1. Although the notifications fluctuate over this study period, the annual Queensland notification rate remained approximately double the national notification rate. For this period, Queensland accounted for 43% (2 838/6 591) of all the national Australian Q fever notifications while the population of Queensland represented 19.6% of the Australian population during this time-frame.

700 7.0

600 6.0 year

/ 500 5.0

400 4.0 population

Counts 300 3.0 000

100

200 2.0

100 1.0 Rates

0 0.0 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 Queensland Notifications Australian Notifications Year Queensland Notification Rates Australian Notification Rates

Figure 3.1 Q fever notifications and annual rates (per 100 000 population per year) for Queensland and Australia, by year, 2003 - 2017.

In Queensland, for the whole study period, Q fever was notified in all age groups. The median age of cases was 46 years (interquartile range (IQR): 33-57;Table 3.1). The median age for males was 45 years (IQR 32–57) and for females 48 years (IQR 36–58). Across the entire study period, the highest crude case notifications were in the 55–59 years age group, followed by the 40–44 years age group. During the 15-year period analysed, 117 notifications occurred in children aged ≤15 years, with an increase in notifications occurring as age increased.

Overall, 74.6% (2 118/2 838) of all notifications were recorded as male, and 25.4% (720/2 383) were female. When stratified by sex and age (Figure 3.2), males in the 40-44 age group had the highest number of cases in Queensland over the study period. Q fever occurred more frequently in males than females with an overall ratio (M:F) of 3:1. From 2003 to 2017 the proportion of notifications recorded as female remained relatively similar (21% in 2003, 23% in 2017), however there was a noticeable increase from 18% in 2006 to a peak of 30% in 2008. From 2008 onwards, the proportion remained largely stable, gradually decreasing to 23%. A summary of Queensland notification counts and average annual notification rates/100 000 population by sex and age groups is presented, aggregated into three 5-year periods, is presented in the Appendix (Table A. 3-3).

300

Male Female 250

200

150 Notifications 100

0 0‐45‐910‐14 15‐19 20‐24 25‐29 30‐34 35‐39 40‐44 45‐49 50‐54 55‐59 60‐64 65‐69 70‐74 75+ 5 year age group

Figure 3.2 Summary of Queensland Q fever notifications by sex and 5-year age groupings, 2003 - 2017

Table 3.1 Trends in demographics of confirmed Q fever cases in Queensland between 2003 and 2017, summarised by 5-year periods and for entire period. Overall 2003 - 2007 2008 - 2012 2013 - 2017 2003-2017 No. cases 873 795 1170 2838 No. of male cases* 669 (76.6%) 575 (72.3%) 874 (74.7%) 2118 (74.6%) Notification rate (males) 6.76 5.19 7.31 6.43 No. of female cases* 204 (23.4%) 220 (27.7%) 296 (25.3%) 720 (25.4%) Notification rate (females) 2.05 1.98 2.44 2.17 Male: female ratio 3.3:1 2.6:1 3.0:1 3.0:1 Median age (IQR) 43 (31 - 55) 46 (34 - 57) 48 (34 - 59) 46 (33 - 57) Median male age (IQR) 42 (30 - 54) 46 (33 - 57) 48 (34 - 60) 45 (32 -57) Median female age (IQR) 46 (34 - 57) 48 (37 - 58) 49 (36 - 58) 48 (36 - 58) Notification rates are calculated as an annual average notification / 100,000 population * Numbers inside parenthesis are the percentage of total cases for that time-period; IQR, inter-quartile range

3.3.2. Distribution of Q fever notifications, by HHS and LGA Over the full 15 year study period, from 2003–2017, 22% (n=625/2 838) of all Queensland notifications were from the Darling Downs HHS area, with 13% (n=368/2 838) from the South West HHS, and 9% (n=261/2 838) from Townsville HHS. However, the highest average annual Q fever case notification rate/100 000 population over this time period was for the South West HHS (95.5), then Central West HHS (52.5) followed by the Darling Downs HHS divisions (15.9) Table 3.2.

Table 3.2 Notifications of Q fever by Hospital and Health Service (HHS) of residence, 2003– 2017, Queensland % of all Notification Total No. of Hospital Health Service Qld % hospitalised rate/100 000 notifications division cases population/year Darling Downs 625 22.0 33.1 15.9 South West 368 13.0 24.7 95.5 Townsville 261 9.2 47.9 7.8 Metro South 234 8.3 47.0 1.5 Cairns and Hinterland 225 7.9 31.6 6.5 West Moreton 216 7.6 50.0 6.1 Sunshine Coast 170 6.0 36.5 3.2 Mackay 161 5.7 43.5 6.5 Central Queensland 141 5.0 50.4 4.6 Metro North 131 4.6 45.8 1.0 Gold Coast 123 4.3 52.8 1.6 Central West 94 3.3 48.9 52.5 Wide Bay 70 2.5 58.6 2.3 Torres and Cape 13 0.5 30.8 3.6 North West 5 0.2 40.0 1.1 Interstate/international 1 0.0 100.0 - Total Qld notifications 2 838 100 40.0 4.9

Figure 3.3 Cumulative Q fever notification rates per 100 000 population/ 5 years, aggregated by residential local government area (LGA), from 2003–2017.

The cumulative 5-year notification rates, aggregated by residential LGA, over consecutive 5-year periods are presented in Figure 3.3 to show geographical distribution of Q fever notifications across Queensland. For each time-period, the LGA of Mareeba has the highest 5-year case notification rate. Other LGA’s that consistently had high case notification rates for the study period were Balonne, Paroo and Murweh.

3.3.3. Hospitalization and work absences, due to Q fever During 2003–2017, hospitalisation data were recorded for 2 123/2 838 Q fever patients, of which 53.3% (1 134/2 123) required hospitalisation, this figure represents 40.0% of the total Q fever cases for this period. Hospitalised patients had a length of stay ranging from 1 to 57 days. Of patients that were hospitalised, the median stay was 5 days (mean = 6.3 days; IQR 3–7 days). Collectively, days of hospitalisation for these patients amounted to 6 336 days. Number of days absent from work due to illness, was recorded for 947 patients (33.3%), with 864 (30.4%) requiring at least 1 day absent. Number of days absent per case ranged from 1 to 330 days, median 10 days (mean = 14.5 days; IQR 5–20 days).

3.3.4. Occupational groups For the years 2003–2012, the majority of notified cases (74%; 1 239/1 345) had either unanswered or unidentifiable responses for the free-text field describing their occupation. For the years 2013–2017, most cases (89%; 1 044/1 170) had a response entered for occupation; however, 26% (270/1 044) of these were classed as unknown, according to the reclassification described in Appendix Table A. 3-2.

Table 3.3 Table of Q fever cases in Queensland, Australia (2013–2017) by occupational category

Occupational categories No. cases (%) Agricultural/farming# 236 (30.5) Unemployed/retired 153 (19.8) Trades 104 (12.8) Teaching or student 46 (5.9) Transport 43 (5.6) Meat processing# 40 (5.2) Gardening/landscaping# 37 (4.8) Administrative 37 (4.8) Health/hospitality 31 (4.0) Mining 19 (2.5) Services 12 (1.6) Wildlife# 11 (1.4) Veterinary/ companion animal# 10 (1.3) Total 779 (100)

Numbers in parentheses are column percentages # Occupational groups that are considered “high risk” for Q fever within Australia.

Table 3.3 shows a summary of Q fever cases per occupational category for the years 2013– 2017. Of the 779 cases that could be reclassified into occupational categories, 30% were involved in agricultural or farming activities, 20% were unemployed or retired, 13% involved 67 working in a trade. Of these cases with identifiable occupational groups, approximately 49% (377/774) of cases would be considered at-risk occupational groups within Australia (CDNA 2018).

3.3.5. At-risk exposures within one month prior to onset of Q fever For the study period 2003–2017, at-risk exposure data fields for Q fever notifications in Queensland were summarised into three categories: abattoir, animal-related and environmental (Table 3.4). Details of at-risk exposures under these three categories are presented in Table 3.5, Table 3.6 and Table 3.7; note that multiple exposures can be recorded for an individual case. Identification of exposures that may increase risk of disease are described in reports by Queensland Government Workplace Health and Safety (‘Q fever - worksafe.qld.gov.au’ 2019) or in the Communicable Diseases Network Australia, Q fever National Guidelines for Public Health Units (CDNA 2018). Additional data for any animal contact, that may not involve at-risk activities, were only available from 2012 onwards, hence summarised for period 2013–2017 only.

Table 3.4 Summary of at-risk exposure groups for Q fever cases in Queensland, by 5-year period

5-year time frames 2003 - 2007 2008 - 2012 2013 - 2017 Total number of Q fever cases 873 795 1 170 Exposures during 1 month prior to onset of illness* Abattoir related exposure No. with any abattoir exposure 18 (2%) 35 (4%) 85 (7%) Animal-related exposures No. with any animal-related exposure 329 (38%) 299 (38%) 612 (52%) Environmental exposure No. with any environmental exposure 109 (12%) 460 (58%) 961 (82%)

No. cases with any exposure to a putative risk factor ± 512 (63%) 481 (61%) 1 002 (86%)

No. cases without information or no reports of risk 325 (37%) 314 (39%) 168 (14%) exposures±

Numbers in parentheses are column percentages * Multiple exposures can be recorded for the same individual ± As per definition of persons at increased risk of disease as reported by Queensland Government Workplace Health and Safety or in the Communicable Diseases Network Australia (CDNA) Q fever National Guidelines for Public Health Units (CDNA 2018)

For the period 2013–2017, from the risk exposures detailed in Table 3.5, environmental exposures were identified in 82% (961/1 170) of all Q fever notifications in Queensland, with 68% of responders (703/1 041) reporting exposure to dust from paddocks or animal yards in the one month prior to disease onset. Sixty-six percent of cases (684/1 039) responded yes, to living or working within 300 m of bush/scrub/forest areas. This was followed by living or working within 1 km of an abattoir/animal grazing land or saleyards (59%; 613/1 041). For the same period, at-risk animal-related exposures, as detailed in Table 3.6 were evident in 52% (612/1 170) of notified Q fever cases in Queensland. The most frequent activities were using animal manure or fertiliser, working with straw or animal bedding, followed by assisting or observing in an animal birthing. Of those who assisted in animal birthing, 65% (131/199) specified cattle (dairy and beef) calving. Of 1 170 cases notified, 7% (n = 85) reported some abattoir exposure in the month prior to disease onset, including: working inside or on the grounds of an abattoir; working as a contractor; or visiting an abattoir (Table 3.7).

An approximate response rate for surveillance of at-risk exposures was determined by taking the number of individuals that answered at least one question, divided by the total number of individual cases for that time-period. The 2013–2017 period has the highest response rates (>86%) for details of at-risk exposures during the 1 month prior to disease onset.

Table 3.5 Details of environmental exposures for Queensland Q fever cases during the one month prior to disease onset.

Time period 2003–2007 2008–2012 2013–2017 Total number of Q fever cases 873 795 1,170 Approximate response rate# 61% 62% 89%

Responders positive for exposure Responders Responders positive for exposure Responders Responders positive for exposure Responders exposure to dust from paddock or animal yards 504 68 (13%) 492 316 (64%) 1,041 703 (68%) live/work within 300 m of bush/scrub/forest area 496 70 (13%) 487 317 (65%) 1,039 684 (66%) live/work within 1 km of abattoir/animal grazing or saleyards 501 63 (12%) 482 254 (53%) 1,041 613 (59%) lived on a farm 531 51 (10%) 493 215 (44%) 1,016 488 (48%) exposure to livestock transport 521 50 (10%) 486 206 (42%) 1,018 516 (51%) had contact with untreated water 527 50 (10%) 487 147 (30%) 1,028 480 (47%) visited a farm 522 62 (12%) 488 191 (41%) 723 310 (43%) laundered clothes from someone who works with animals 521 27 (5%) 488 123 (25%) 1,020 262 (26%) contact with a Q fever infected person 528 9 (2%) 493 26 (5%) 1,040 56 (5%) consumed unpasteurised milk or milk products 527 12 (2%) 496 41 (8%) 1,042 56 (5%)

# approximate response rate; determined by taking the number of individuals that answered at least one question, divided by the total number of individual cases for that time period

Table 3.6 Details of at-risk animal-related exposures for Queensland Q fever cases during the one month prior to disease onset.

Time period 2003–2007 2008–2012 2013–2017 Total number of Q fever cases 873 795 1,170 Approximate response rate# 60% 61% 87%

Responders exposure for positive Responders Responders exposure for positive Responders Responders exposure for positive Responders worked with animal manure/fertiliser 524 41 (8%) 480 145 (30%) 1,009 353 (35%) worked with straw or animal bedding 519 18 (3%) 483 92 (19%) 1,004 223 (22%) observing or assisted in animal birthing 516 144 (28%) 482 121 (25%) 1,015 199 (20%) involved in slaughtering, skinning or meat 522 27 (5%) 483 68 (14%) 1,010 161 (16%) processing involved in shooting/hunting 525 18 (3%) 481 45 (9%) 1,007 156 (15%) attended a saleyard or animal show 520 20 (4%) 474 53 (11%) 1,006 127 (13%) worked with wool 521 10 (2%) 487 26 (5%) 1,019 70 (7%)

# approximate response rate; determined by taking the number of individuals that answered at least one question, divided by the total number of individual cases for that time period

Table 3.7 Details of abattoir exposures for Queensland Q fever cases during the one month prior to disease onset.

Time period 2003–2007 2008–2012 2013–2017 Total number of Q fever cases 873 795 1,170

Approximate response rate# 15% 34% 86%

Responders Responders positive for exposure Responders Responders positive for exposure Responders Responders positive for exposure worked inside the abattoir 133 13 (2%) 269 27 (10%) 1,011 67 (7%) worked in the grounds of the abattoir 80 4 (1%) 91 13 (14%) 333 26 (8%) visited abattoir 83 2 (0%) 86 10 (12%) 338 23 (7%) contract worker at an abattoir 81 4 (1%) 80 6 (8%) 321 9 (3%)

# approximate response rate; determined by taking the number of individuals that answered at least one question, divided by the total number of individual cases for that time period

3.3.1. Additional animal contact surveillance data (2013–2017) From 2013–2017, 99.9% of cases reported at least one direct animal contact in the one month prior to onset of disease (1169/1 170). A summary of specific animal exposures is presented in Table 3.8. Caution should be taken when interpreting this data, as it includes multiple animal co-exposures and these exposures were not confirmed as the source of Q fever. However, the most common animal contact recorded was dogs (n = 661; 56.5%) followed by native wildlife (n = 619; 52.9%) then cattle (n=569; 48.6%).

Table 3.8 Table of direct animal or arachnid contact in the one month prior to onset of Q fever, 2013-2017 % of total Responded Animals and insects cases yes (n=1 170) dogs 661 56.5% Australian native wildlife 587 50.2% kangaroos 418 small marsupials 169 cattle 569 48.6% cats 332 28.4% sheep 204 17.4% ticks 197 16.8% feral pigs 156 13.3% domestic goat 109 9.3% domestic pigs 80 6.8% feral goat 67 5.7% horses 66 5.6% other 215 18.4%

3.3.2. Q fever vaccination and awareness of risk of contracting disease A summary of the Queensland Q fever notification’s awareness of their own risk for contracting Q fever and awareness of the Q fever vaccine, is presented in Table 3.9. From those who responded, 27% of cases from 2003–2007 were aware they were at risk of Q fever; this reduced to 21% for the following time-periods 2008–2012 and 2013–2017. For awareness of the Q fever vaccine, the crude proportion of cases that responded yes, decreased slightly from 45% (2003–2007) to 42% (2013–2017). However, it is clear that response rates for surveillance fields increased over these time-periods, hence comparing crude measures may be biased. Over the entire study period, 75% (2 137/2 838) of cases had vaccination status recorded. Of these, 92% had not been previously vaccinated, 3% were listed as being vaccinated for Q fever and 5% were unknown.

Table 3.9 Details of awareness of risk and vaccination from notified Q fever cases in Queensland, by 5-year periods

Time period 2003–2007 2008–2012 2013–2017 Total number of Q fever cases 873 795 1170

Aware that they were at risk of Q fever No. of responders (%) 530 (61%) 490 (62%) 1050 (90%) Yes 142 (27%) 103 (21%) 220 (21%) No 330 (62%) 357 (73%) 774 (74%) Unknown 58 (11%) 30 (6%) 56 (5%)

Aware of Q fever vaccination No. of responders (%) 526 (60%) 488 (61%) 1048 (90%) Yes 239 (45%) 216 (44%) 445 (42%) No 213 (40%) 239(49%) 543 (52%) Unknown 74 (14%) 33 (7%) 60 (6%)

For the years 2013–2017, 86% of cases were linked to at least one previously described, at-risk exposures in the 1 month prior to disease onset (Table 3.4). However, for the same time-period, only 19% of cases were aware that they may be at risk of Q fever. Data for those that responded yes to at-risk exposures, has been cross- referenced with the data field for awareness of Q fever risk. The at-risk exposure category with the least awareness was environmental-related and the most awareness was for the notified cases that had abattoir-related exposure.

Table 3.10 Personal awareness of risk of Q fever, in cases that responded yes to an “at- risk” exposure during the one month prior to disease onset, 2013–2017

"At-risk" exposure in 1 Personal awareness of risk month prior to disease onset* No Yes Unknown Total Environmental 698 (74%) 208 (22%) 37 (4%) 943 Animal 391 (59%) 194 (29%) 18 (3%) 603 Abattoir 42 (50%) 36 (43%) 6 (7%) 84 * Multiple exposures can be recorded for the same individual Definitions of persons at increased risk of disease as reported by Queensland Government Workplace Health and Safety or in the Communicable Diseases network Australia (CDNA) Q fever National Guidelines for Public Health Units (CDNA 2018)

3.4. Discussion and conclusions

This report represents the first detailed description encompassing all data fields for the surveillance of Q fever notifications in Queensland, Australia, and provides summaries of the extent, and trends of this important notifiable disease from 2003-2017. Overall, Queensland accounted for 43% of the national Q fever notifications for this study period. The annual Queensland rates (per 100 000 population/year) showed fluctuation over time. The highest annual notification rate was observed in 2003, and the lowest rate in 2009. Minor peaks in annual notification rates appeared in 2007 and 2015. A similar pattern was noticed in the overall Australian national case notifications for the same years, though the pattern in Australian rates may be most heavily influenced by Queensland data given that Queensland reports relatively more Q fever cases per year than any other state or territory. There are multiple factors that may influence these fluctuations, including the end of the NQFMP in 2006, heightened awareness by general practitioners, changes in surveillance, testing and reporting of case notifications and the occurrence of outbreaks (Sloan-Gardner et al. 2017; Parker, Barralet & Bell 2006; Gunaratnam et al. 2014).

Other studies have reported that Q fever predominately occurs in males of working age with highest notification rates in the 40–59 years age group, this trend remains true for the Queensland data analysed in this study (Sloan-Gardner et al. 2017; Clutterbuck et al. 2018; Bond et al. 2018; Garner et al. 1997). Overall 74.6% of cases were male, with the highest notification rates reported in the 55–59 years age category. It was previously reported from Australian national data, that the average age of case notifications and the proportion of female cases had increased over time from 1991 to 2014 (Sloan-Gardner et al. 2017). In the current study, the average age of notified cases was seen to be increasing over time; however, the proportion of notifications recorded as female remained relatively constant. There was a peak identified in the proportion of female notifications immediately following the end of the NQFMP; which may be explained by a relative reduction in male notifications following the success of the program.

Although it has been previously reported that only 2% of acute Q fever cases require hospitalisation, this study showed that 40.0% of the total Q fever cases spent at least one day in hospital (Parker, Barralet & Bell 2006; Maurin & Raoult 1999). This is similar to a report from New South Wales, Australia, where 46.5% of cases from 2011–2015 were hospitalised (Clutterbuck et al. 2018). Without current nation-wide comparable data, it is uncertain if these hospitalisation rates are consistent throughout Australia or are higher for these two states. However, it may be worth noting that combined, Queensland and New 75

South Wales account for approximately 80% of all national Q fever notifications (Sloan- Gardner et al. 2017).

The overall Q fever notification rate in Queensland was approximately double the Australian national notification rate. Visual inspection of the notification rates by LGA suggests there is spatial variation in the occurrence of Q fever within Queensland. The geographical distribution of Q fever within Queensland requires more in-depth spatial and/or spatio-temporal analysis to identify clustering of disease across space and time and to identify potential risk factors that may be linked to geographical hotspots. The distribution of Q fever globally seems to vary between and within countries, likely due to the wide range of animal reservoirs, differences in C. burnetii strains and environmental factors that are not yet completely understood (Million & Raoult 2015).

When examining the surveillance data-field responses across this study period, it was clear that since the implementation of enhanced Q fever surveillance in 2012, there was an improvement in the approximated response rate to survey questions. Prior to 2012 there was no formal Q fever report form, therefore exposure data collected prior to 2012 may be difficult to interpret or make inferences from. In 2018, in order to provide nationally consistent advice and guidance to public health units, Q fever reporting was included using the Series of National Guidelines (SoNG) including a national Case Reporting Form. This will allow comparable data to be collected across multiple states so that the epidemiology of Q fever can be analysed accurately at a national and international level.

We acknowledge that there were difficulties analysing occupational data, as many entries were unidentifiable; included area codes, specific organisation names or suburbs rather than an occupation. However, the quality of data improved from 2012 onwards, likely because of the change in survey protocol. The most frequent occupational group for the 2013–2017 time-period, from respondents with an identifiable response, was agricultural/farming, followed by retired or unemployed. It highlights the poor uptake of Q Vax® or knowledge gap that still exists in some at-risk occupations. Although farmers were a target group for NQFMP, it seems that the use of prophylactic vaccination in this at-risk occupational group has not continued following the end of the national program. There is also perhaps a gap in active auditing for Q fever prevention in agricultural workplaces that would otherwise encourage increased testing and vaccination. For this time period, 51% of cases would be considered to be in a classical at-risk occupational group for Q fever in Australia and therefore should be preventable with correct vaccination (CDNA 2018). However, the remaining 49% of cases did not report an at-risk occupation; hence this study 76 reiterates that occupation is a poor proxy for Q fever risk exposure. This is in line with a New South Wales report for Q fever notifications during a similar time-period (Clutterbuck et al. 2018).

This study identified that approximately 7% of Queensland Q fever notifications have reported some abattoir exposure in the one month prior to illness. This includes working in or on the grounds of the abattoir, working as a contractor of an abattoir or visiting an abattoir. This figure is much lower that what has been reported for Victoria, Australia, where 25% of notifications were related to abattoir exposures for the years 2009-2013 (Bond et al. 2018). The lower percentage of Queensland Q fever cases with abattoir exposures is likely due to the strong workplace health and safety regulations and active auditing regarding Q fever vaccination and immunity screening checks as a risk mitigation strategy within the Queensland meat-work industry. Hence, this data does not suggest a reduced risk of exposure to C. burnetii in the meat-processing industry in Queensland; rather it may indicate the successful control and prevention of human disease through vaccination of at-risk people.

This study has identified, with improved surveillance data of Q fever notifications in Queensland (2013–2017), that 86% of notified cases identified an at-risk exposure (environmental, animal or abattoir) during the one month prior to disease onset. The most frequent at-risk exposures for that time-period were environmental exposures: exposure to dust from animal paddocks, living within 300 m of bush/scrub/forest and living or working within 1 km of an abattoir. The transmission of Q fever to humans from environmental dispersal of C. burnetii from animal holdings and abattoirs has been documented in a review article (Clark & Soares Magalhães 2018). Although the review did not include any Australian publications, it is likely that similar scenarios exist and local investigations are warranted to assess the risk of airborne geographical dispersal of C. burnetii to Queensland communities, from wildlife areas, livestock holdings and abattoirs.

During the one month prior to disease onset, 99% of notifications reported at least one animal contact. The most common animal species was dogs, followed by native wildlife and then cattle. This data, however, should be interpreted with caution. Firstly, the animal exposure recorded may not be a meaningful contact that could lead to Q fever transmission and may be an animal observed from a distance or in passing. Secondly, meaningful associations to exposure patterns cannot be concluded from this nature of surveillance data, without a control group for comparison. Finally, the notification data presented here had many co-exposures and the likely source of transmission was not investigated further with 77 in-depth household visits or additional laboratory testing of animals of environmental samples. An additional analysis of this surveillance data focused on specific animal exposures associated with Q fever in Queensland has been completed and published as a separate peer-reviewed manuscript (Clark et al. 2020).

This study has presented a descriptive summary of surveillance data in order to develop hypothesis for further investigations. Surveillance data for human Q fever notifications in Queensland has improved since the implementation of case follow up and the case report form in 2012. This has revealed that a large proportion of Q fever cases identified a known at-risk exposure; therefore, more cases should be preventable with the correct use of vaccination. Education and vaccination need to be targeted to agricultural workers as this occupational group is still reporting many cases. However, greater awareness of the potential risk of infection through indirect environmental exposures should be another focus to help reduce the impact of Q fever on public health in Queensland.

3.5. Chapter three appendix

Table A. 3-1 Queensland Government Q fever case report form

Table A. 3-2 Notifiable Conditions System (NOCS) free text field ‘place of work’ from Q fever questionaire

New field Classification Field "place of work" free text words aggregated into new categories

Occupation Administrative sales/reception/admin/recruit/office/retail/survey/account/book/politic/lawyer

Agricultural/ farm/grazier/livestock/cattle/sheep/horse/camel/goat/pig/shear/muster/stockman/station/farrier/fish/wool/drover/grower/cotton/ farming fruit/ cane

Gardening/ landscape landscape/lawn/grass/mow/grounds/earth/garden/aborist

Health/ hospitality health/therapist/beaut/carer/nurse/doctor/hosp/bar/tavern

Meat industry abattoir/meat/butcher/pet food/specific names of abattoirs in Queensland

Mining miner/mining/mine Services army/navy/airforce/soldier/barrack/police/RAAF/veteran

Teaching/ student student/college/school/teach/specific names of universities in Queensland

Trades handy/mechanic/maintain/caretaker/glazier/courier/tile/machin/baker/build/construction/paint/boiler/apprentice/concrete/builde r/ labourer/cabinet/electrician/fencer

Transport truck/transport/drive/rail

Unemployed/ retired unemployed/retired/pension

Veterinary vet/anim/cat breeder/breeder/dog

Wildlife wild/roo/shoot/ranger Unknown numbers/unknown/not stated

Table A. 3-3 Queensland Q fever notifications aggregated into three, 5-year time-periods by age grouping and sex from 2003 - 2017

2003-2007 2008-2012 2013-2017 average average average annual annual annual Sex/Age group Count Count Count rate/ rate/ rate/ 100,000 100,000 100,000 Queensland Population 0-4 4 0.31 2 0.13 7 0.44 5-9 9 0.67 7 0.49 16 0.99 10-14 22 1.57 14 0.96 19 1.25 15-19 46 3.38 48 3.19 54 3.51 20-24 58 4.16 41 2.59 58 3.42 25-29 57 4.34 43 2.70 70 4.05 30-34 63 4.39 49 3.30 70 4.17 35-39 94 6.57 61 3.82 80 5.12 40-44 111 7.56 98 6.29 113 6.76 45-49 94 6.72 94 6.09 126 7.87 50-54 93 7.21 88 6.09 132 8.43 55-59 88 7.36 103 7.92 137 9.50 60-64 62 6.67 76 6.41 110 8.57 65-69 37 5.19 38 4.17 93 8.02 70-74 15 2.65 12 1.80 46 5.36 75+ 20 1.86 21 1.73 39 2.76 Total 873 4.45 795 3.61 1170 4.89 Male 0-4 2 0.30 2 0.26 5 0.61 5-9 7 1.02 5 0.68 6 0.72 10-14 11 1.53 11 1.47 12 1.54 15-19 39 5.61 35 4.55 43 5.47 20-24 45 6.36 35 4.36 51 5.94 25-29 51 7.72 28 3.47 54 6.25 30-34 52 7.30 40 5.41 53 6.36 35-39 80 11.30 47 5.93 61 7.90 40-44 84 11.59 75 9.71 85 10.30 45-49 71 10.26 58 7.60 88 11.22 50-54 67 10.40 61 8.53 91 11.83 55-59 59 9.75 79 12.16 102 14.38 60-64 46 9.73 48 8.01 85 13.41 65-69 28 7.76 26 5.67 71 12.25 70-74 13 4.68 10 3.00 39 9.14 75+ 14 3.12 15 0.29 28 4.50 Total 669 6.84 575 5.24 874 7.35 Female 0-4 2 0.32 0 0.00 2 0.26 5-9 2 0.31 2 0.29 10 1.27 10-14 11 1.61 3 0.42 7 0.95 15-19 7 1.05 13 1.77 11 1.46 20-24 13 1.90 6 0.77 7 0.84 25-29 6 0.92 15 1.91 16 1.85 30-34 11 1.52 9 1.21 17 2.01 35-39 14 1.94 14 1.74 19 2.40 40-44 27 3.63 23 2.92 28 3.30 45-49 23 3.26 36 4.61 38 4.66 50-54 26 4.03 27 3.70 41 5.15 55-59 29 4.91 24 3.68 35 4.78 60-64 16 3.50 28 4.77 25 3.85 65-69 9 2.55 12 2.65 22 3.80 70-74 2 0.70 2 0.60 7 1.62 75+ 6 0.95 6 0.86 11 1.39 Total 204 2.08 220 2.00 296 2.46

Chapter Four

Detection of Coxiella burnetii in beef cattle at an abattoir in Queensland, Australia

“Absence of proof is not proof of absence.”

- William Cowper (1731–1800)

4. Detection of Coxiella burnetii in beef cattle at an abattoir in Queensland, Australia

4.1. Introduction

4.1.1. Background and history Q fever is the human disease caused by zoonotic infection with Coxiella burnetii, a rickettsial-like intracellular bacterium. First recognised in the 1930s in Brisbane abattoir workers, it has historically been a disease associated with livestock contact, although cases do not always report direct animal contact (Derrick 1937; Parker, Barralet & Bell 2006; Million & Raoult 2015). The epidemiology of disease in humans appears to vary within and between countries as a likely reflection of host susceptibility, reservoir prevalence and environmental factors (Million & Raoult 2015). In 1958, a sharp spike in Q fever cases was observed in Queensland abattoir workers; this was correlated with the introduction of a boneless beef exporting trade (Derrick 1961). To fill market demand, poorer conditioned beef and dairy cattle, including a high proportion of pregnant cows, were slaughtered. Mammary glands and placenta are the primary target organs for localisation and multiplication of C. burnetii in pregnant ruminants and high bacterial shedding occurs through milk, placenta and birth fluids (Hansen et al. 2011; Guatteo et al. 2006). It is likely that the shift in population traits of cattle going to slaughter, coinciding with an increased number of immunologically naïve workers employed to meet the workload, were causal factors that led to this increased incidence in Q fever (Derrick 1961).

Abattoirs are still significant risks for C. burnetii exposure as evident from a recent meta-analysis that reported an expected seroprevalence in unvaccinated meat workers to be 26% (95% Confidence Interval (CI) 17, 35%; Woldeyohannes et al., 2018). The highest risk of Q fever infection was identified in abattoirs that slaughtered cattle, sheep and goats. A review of 1991-1994 Australian national Q fever notifications revealed a strong association between notification of Q fever and number of abattoirs in the area (Garner et al. 1997). The use of human vaccination (Q-VAX® Seqirus, Australia), in addition to increased hygienic measures and personnel protective equipment, has ensured that Q fever cases directly related to abattoir work has reduced. This pattern is confirmed from analysis of Q fever notifications during the Australian National Q Fever Management Program (NQFMP) which was funded between 2002 and 2006, where a 50% decrease in national notification rates was noted (Gidding et al., 2009; Sloan-Gardner et al., 2017). Young adult males showed

85 the most significant decline of Q fever cases, which may be “consistent with the profile of the abattoir workforce, the main target of phase 1 of the program” (Gidding et al. 2009).

Although the NQFMP ceased in 2006, current Workplace Health and Safety regulations within Australian have continued requiring immunity screening checks and vaccination of all abattoir workers with an aim of reducing occupational risks for Q fever. Because of the effectiveness of the Q-VAX® (Seqirus, Australia) human vaccine, the consequent reduction in notifications of Q fever cases in people associated with abattoirs may not be reflective of prevalence of animal infection with Coxiella or potential exposure risk for susceptible (unvaccinated) people. Therefore, C. burnetii testing at abattoirs, from animal and environmental sources, is warranted to monitor the potential bacterial exposure coming from slaughtered ruminants.

4.1.2. Abattoir surveys for animal disease monitoring Disease surveillance can be defined as methods of systematically gathering, recording and analysing data on the occurrence of infectious disease, which may involve testing populations of animals (Thrusfield 2007). Large abattoirs provide invaluable opportunity for the examination, collection and testing of samples for infectious diseases of significance to animal or public health. Well-designed active, targeted surveillance can provide data with a numerator and a denominator focused on a specific disease or pathogen, thus allowing the estimation of disease frequencies within a population (Cameron 1999). This contrasts with passive surveillance, which is defined as “the secondary use of routinely collected data that was generated for another purpose” (Sergeant & Perkins 2015). Passive surveillance from abattoirs regularly includes the examination of only clinically or pathologically affected cases, thus no denominator is provided and the data is not sufficient to estimate prevalence of disease (Thrusfield 2007; Sergeant & Perkins 2015).

Australia successfully eradicated bovine brucellosis and tuberculosis in the late 1990s after nearly 30 years of control programs. This incorporated a combination of disease surveillance techniques including the testing of abattoir samples to inform and monitor the control programs (More, Radunz & Glanville 2015). Trace-back information from meat- processing establishments allows data to be collected from animals originating in different geographical regions. Surveillance for bovine tuberculosis at routine slaughter of cattle in the United Kingdom is another example of ongoing epidemiological surveillance contributing to a national eradication program (Chaintarli & Upton 2018). Abattoir surveys have also been used to investigate the prevalence of Salmonella sp. and Escherichia coli 0157 in different

86 meat products (Chapman 2000). Abattoir surveillance can provide a practical and cost effective method of monitoring the prevalence of a number of important infectious diseases in animals from spatially distant regions.

Investigations into the detection of C. burnetii infection in cattle going to slaughter have been reported in Australia and Denmark (Cooper et al. 2011; Paul et al. 2014). Serological methods are commonly used to detect immunological exposure to C. burnetii in cattle. For example, in Australia, Cooper et al. (2011) collected blood samples from a single abattoir that received cattle from north and north-western Queensland and determined a seropositivity of 10.1% (95% CI 10.1, 10.2). Paul et al. (2014) collected blood samples from six abattoirs within Denmark and estimated an overall animal seroprevalence 5.6% (95% CI 2.8, 8.4). While these surveys have provided important estimates of C. burnetii exposure, serology alone is unable to identify active infection and the potential for transmission of infection to humans or contamination of the environment. Monitoring of slaughtered animals with the purpose of identifying the aetiological agent (C. burnetii) has the potential to address these issues.

At the time of writing this thesis, only one study could be found which described an abattoir survey to detect the presence of the bacterium C. burnetii, as opposed to detecting serological exposure (Agerholm et al. 2017). This study sampled bovine heart valves showing evidence of inflammation at three Danish abattoirs over a 12 month time-period. Deoxyribonucleic acid (DNA) of C. burnetii was detected in 25% of the cardiac tissue sampled (n = 100), using polymerase chain reaction (PCR) methods. Although results were not suitable to estimate the prevalence of infection in all cattle going to slaughter, the detection of C. burnetii from inflamed heart valve tissue is an intriguing and novel finding. Valvular endocarditis is a known complication of human Q fever and although experimental infection has induced endocarditis in rabbits, there has been limited research investigating such non-reproductive infections in ruminants (Eldin et al. 2016; Bell, Parker & Stoenner 1949).

4.1.3. Clinical samples and molecular epidemiology There is published evidence to support the sampling of pregnant uteri and the female reproductive tract of ruminants to increase the chance of detecting C. burnetii. The placenta is the primary target for infection and multiplication of C. burnetii in pregnant ruminants (Agerholm, 2013; Roest et al., 2013). However, there is inadequate literature available indicating the ideal sample to detect C. burnetii in non-pregnant animals. Early experimental

87 infections in cattle found that bacteria could be isolated from liver, spleen, lung, lymph nodes, intestinal tract and mammary glands of cattle after recovery from acute infection, thus suggesting a persistence of bacteria in multiple organs (Bell, Parker & Stoenner 1949). A study in Portugal detected C. burnetii DNA in various tissue samples collected from both wildlife and domestic ruminants (Cumbassá et al. 2015). This study has highlighted that it is possible to use tissue samples for molecular detection and genotyping of C. burnetii.

Recent advances in C. burnetii genotyping techniques have enabled the identification of bacterial strain variations. Detection of C. burnetii DNA either by PCR or after bacterial culture is required for initial screening. Then, methods such as multispacer sequence typing (MST), single nucleotide polymorphism typing (SNP) or multiple locus variable number of tandem repeats analysis (MLVA) can be used to classify genomic strains of the bacteria (Arricau-Bouvery et al. 2006; van Schaik & Samuel 2012; H. Roest et al. 2013). These methods have become important for identification of the source of infection in Q fever outbreaks. Identification of the C. burnetii genotype responsible for a human Q fever outbreak (Victoria, Australia) was accomplished using MLVA techniques and identified a goat-strain responsible for the outbreak (Bond et al. 2015). Similar methods were used to identify the C. burnetii genotypes responsible for the Netherlands outbreak (Roest et al. 2011).

There is currently a gap in knowledge of specific strain variations of C. burnetii infecting cattle in Australia. Therefore, investigations into strain identification may increase knowledge of the transmission of infection within cattle and between cattle and other species. It may also provide insights into bacterial virulence factors that affect the pathogenesis of human Q fever.

4.1.4. Chapter objectives The objective of this thesis chapter was to perform a focused molecular survey for C. burnetii at an abattoir in South-east Queensland that slaughters cattle from regions across Queensland and northern New South Wales. We hypothesised that active disease surveillance through an abattoir was likely to be an effective method of 1) estimating the prevalence of C. burnetii infection in a population of beef cattle going to slaughter, and 2) enabling identification of specific genotypes of C. burnetii infecting cattle in this population.

4.2. Material and Methods

Animal ethics for this study was approved by the University of Queensland Animal Ethics Committee: ANRFA/SVS/202/16.

In order to fulfil the objectives of this chapter, two studies were conducted, Study 1 focused on placental tissue and fluid collection and Study 2 focussed on liver tissue collection. All biological samples were collected for the detection of bacterial DNA using PCR methods; due to PC3 laboratory restrictions and difficulties with intracellular bacterial culture, PCR has become a common and acceptable tool for detecting C. burnetii bacterial DNA (H. Roest et al. 2013).

4.2.1. Sample strategy and sample collection

4.2.1.1. Sample size estimates Sample size estimates were based on calculations done using online tools (‘WinEpi: Working in Epidemiology’ 2016) and commercial power analysis software (‘Sample Size Software | Power Analysis Software | PASS |’ 2016). All estimates were based on sampling individual cows within a mob (cows from one property or saleyard consignment) and with alpha set to 0.05, beta to 0.8, and with assumed test sensitivity and specificity set at 100% (perfect tests). Estimates of expected animal prevalence were based on prior research (Cooper et al. 2011).

Initial calculations were based on sample size required to be confident of detecting at least one positive animal conditional on the mob having a defined prevalence of positive animals. The number of animals required for testing in order to be 95% confident of detecting at least one positive animal was 25 per mob if the expected prevalence in the mob was 10% or higher. If prevalence was 15% or higher, then a sample of 25 animals would achieve 99% confidence of detecting at least one positive animal.

Additional sample size calculations were then based on the number of animals required to estimate prevalence with a defined precision. For an expected prevalence of 10% and precision of 5 and 10%, the required sample sizes were 554 and 140, respectively. For an expected prevalence of 15% and precision of 5 and 10%, the required sample sizes were 784 and 196, respectively.

Finally, calculations were revisited to estimate the precision that would be achieved given a fixed sample size. The aim here was to identify a minimum sample size such that the lower 95% CI was greater than zero, in order to ensure that for a defined expected 89 prevalence, a given sampling strategy would be likely to return a non-zero prevalence estimate i.e. unlikely to inadvertently report a zero prevalence. If the expected prevalence was 10%, then a sample size of 40 or more was likely to have a lower 95% CI for the prevalence that was greater than zero. If the expected prevalence was 15% or greater, then the minimum sample size that achieved a positive lower 95% CI was n = 30.

The findings from sample size estimates were used to inform a final sampling target of 30 to 50 animals per property/PIC (property identification number), between 50–150 animals per major region and a total sample size of 500–800 animals. Queensland regions were identified from publically available map such as in Figure 4.1 [Far North Queensland, North Queensland, Central Queensland, Mackay/Isaac and Whitsunday, Darling Downs South West, South East Queensland/Wide Bay Burnett].

Sample size estimations were originally done to inform sero-prevalence studies, but the findings were deemed to be applicable to studies aimed at detection of genetic material.

4.2.1.2. Study 1

4.2.1.2.1. Sampling strategy Samples were collected for the first study using a convenience sampling strategy. Practical constraints limited the number of samples that could be collected on a given sampling day. The proportion of cows arriving at the abattoir that were pregnant was unknown and expected to be low (~10%). It was considered unlikely that large numbers of pregnant animals would be purposefully shipped to slaughter. This meant that the number of pregnant uteri available for sampling from one mob was expected to be limited. Finally, there were physical and financial constraints that restricted the total number of samples that could be collected and processed for subsequent analyses. An ad-hoc sample number of 100 pregnant cattle was the initial target.

Samples were collected in a room on-site at the abattoir. Post-slaughter, all pregnant uteri were sent down a chute from the main kill-floor to this room for further processing. As we were unable to predict when and how many cows would be pregnant, all gravid uteri that presented were visually inspected and samples collected if they fit the following inclusion criteria:

1. Cattle must be mid to late term pregnant (gestational age 4–9 months) as determined by abattoir technician and sample collector (Table 4.1). 2. The uterus must be intact without damage or tears on visual inspection.

Table 4.1 Guide used to estimate gestational age of foetus

Size category* Approx. length Gestational age (crown to rump) (months) Large Over 75 cm 8–9 Medium 55–75 cm 7–8 Small 45–55 cm 5–6 Very small 20–40 cm 4–5 Under sized less than 18 cm <3

This table was adapted from guidelines in Appendix A in: Kirkbridge’s Diagnosis of Abortion and Neonatal Loss in Animals; *Size category is a sizing system used by the abattoir technician.

4.2.1.2.2. Sample collection and storage - placental fluid and tissue The uterine wall was swabbed with 70% alcohol at the site for aspiration. A 16 gauge, 1 1 /2-inch needle, attached to a 10-ml syringe was inserted into the gravid uterus and 10 ml of amniotic fluid was aspirated. A new sterile syringe and needle was used for each sample collected. The amniotic fluid was then separated into 3 x 2 ml sterile pre- labelled screw top containers and placed into a box and immediately put on ice. The uterus was then cut open by the abattoir technician. Following exteriorisation of the foetus, 2–3 cotyledons with some inter-cotyledonary membranes intact were cut en-block, with a sterile, single-use scalpel blade. These tissue samples were placed into numbered single use plastic zip lock bags and placed inside numbered firm plastic screw top containers to be transported to the University of Queensland, Gatton campus for further processing.

Samples were kept on ice for 1–3 hours. At the end of the collection time, the chilled samples were transported immediately to the University campus where they were processed and frozen within 6 hours of collection.

The placental tissue samples were processed in the PC2 biohazard hood in the Veterinary School Post Mortem wet laboratory. A longitudinal section from the cotyledons and inter-cotyledonary placental tissues were cut from the fresh samples using sterilised surgical instruments and single use scalpels. Sections were then placed in pre-labelled 5-ml screw top containers. The amniotic fluid and placental tissue samples were frozen at -80 °C for future DNA extraction and PCR testing.

4.2.1.3. Study 2

4.2.1.3.1. Tissue optimisation An initial sample collection of two different tissue types (liver and spleen) was performed in order to compare and optimise tissue collection and molecular methods for Study 2. Bovine liver and spleen tissue were collected separately as core tissue biopsies 1 with a 16 gauge, 1 /2-inch needle attached to 3-ml syringe. Core tissue samples were collected repeatedly and stored in labelled 1.5-ml Eppendorf tubes, prefilled with 200 µl, 400 µl, 600 µl, 800 µl, 1 000 µl and 1 200 µl of phosphate-buffered saline (PBS) and DNA Tissue Lysis buffer (Roche Diagnostics, Australia), separately. Samples were kept at room temperature and then frozen at -20 oC within 5 hours of collection.

4.2.1.3.2. Sampling strategy At the start of the study, communications were established with abattoir personnel to inform the study team when larger consignments of cattle were arriving at the abattoir and the region in Queensland where they were coming from. Weekly updates from abattoir personnel were then used to inform optimal collection days when a study team member would travel to the abattoir for sample collection. A systematic random sampling method was used to then collect samples from the animals within mobs being processed. A number between 1 and 10 was randomly drawn from a bag to determine the starting sample number and then every 3rd or 4th animal was sampled (depending on the total mob size). Sampling continued until the end of the mob or until the desired sample size was reached.

4.2.1.3.3. Sample collection and storage - liver tissue Tissue samples were collected from whole bovine liver after it was examined by the on-site meat inspectors. The unique kill number was identified and recorded to align with the unique sample number on pre-labelled collection containers. Core liver tissue biopsies were collected with a 16 gauge, 1 ½-inch needle attached to 3-ml syringe and then expulsed into 1.5-ml Eppendorf tubes prefilled with 400 µl of MagNA Pure 96 DNA Tissue Lysis buffer (Roche Diagnostics, Australia). Samples were kept at room temperature and then frozen at -20 oC within 5 hours of collection. At the end of each collection day, the abattoir provided printed paperwork with animal and property details aligning to the unique kill number.

Figure 4.1 Map of Queensland showing major regions defined (https://www.business.qld.gov.au/invest/queenslands-regional-locations/map-of-queensland- regions)

4.2.2. Laboratory methods

4.2.2.1. Molecular extraction from different samples

4.2.2.1.1. Amniotic fluid Deoxyribonucleic acid (DNA) was extracted from amniotic fluid samples using the commercially available DNeasy Blood and Tissue kit (QIAGEN, Texas, USA) as per the manufacturer’s instructions. Prior to DNA extraction, 5 µl of equine herpes virus (EHV; with an expected cycle threshold (Ct) value of 33; equivalent to 1x104 copies of EHV DNA) was added to the buffer AL solution prior to addition of the amniotic fluid sample, as previously described as an extraction and inhibition control (Bialasiewicz et al., 2008; Tozer et al., 2014). This step was added to monitor the efficiency of the extraction process by comparing Ct values for the EHV standard RT-PCR.

I. Sample preparation prior to extraction

Frozen amniotic fluid samples were defrosted at 4 oC prior to use. A 2 ml volume of each amniotic fluid sample was centrifuged for 5 min at 10 000 x g. The supernatant was removed carefully leaving approximately 300 µl of supernatant above the pellet. The pellet was resuspended in the remaining supernatant.

II. DNA extraction – DNeasy Blood and Tissue kit (QIAGEN)

Firstly, 200 µl buffer AL was added to 1.5-ml microcentrifuge tubes pre-labelled with sample numbers. An aliquot of 200 µl of amniotic fluid sample was added to the microcentrifuge tube with buffer AL and 5 µl EHV. An additional 3 tubes were prepared which contained water instead of amniotic fluid to confirm there was no cross contamination during the extraction process. Qiagen proteinase K (QIAGEN, Brisbane, Australia), 20 µl, was added and mixed thoroughly by vortexing, and incubated at 62 oC for 20 min. Ethanol (100%) 200µl was added to the sample and mixed thoroughly by vortexing. The entire sample mixture was then pipetted into the DNeasy Mini spin column placed in a 2-ml collection tube. This was centrifuged at >6 000 x g (8 000 rpm) for 1 min. The flow through and collection tubes were discarded. The DNeasy mini spin column was placed in a new 2- ml collection tube, 500 µl Buffer AW1 added, and centrifuged at >6 000 x g (8 000 rpm) for 1 min. The collection tubes were discarded. The DNeasy mini spin column was placed in a new 2-ml collection tube, 500 µl Buffer AW2 added, and centrifuged at >6 000 x g (8 000 rpm) for 1 min to clean the DNA of inhibitors. The collection tubes were discarded. A final “dry spin” of the column in a new collection tube for 3 mins >20 000 x g (14 000 rpm) ensured 94 the removal of all ethanol. The DNeasy mini spin column was placed in a clean 1.5-ml microcentrifuge tube, and 100 μl buffer AE pipetted directly onto the DNeasy membrane. This was incubated at room temperature for 1 min, and then centrifuged for 1 min at >6 000 x g (8 000 rpm) to elute approximately 100 µl. Final DNA extract was stored frozen at -80 oC.

4.2.2.1.2. Placental tissue I. Sample preparation prior to extraction

Frozen placental tissue samples were defrosted at 4 oC prior to use. Each sample was placed on a single-use glass slide and mechanically disrupted with a single-use scalpel blade. A small quantity (20–100 mg) of tissue was then added to 1.5-ml microcentrifuge tubes pre-filled with 180 µl of DNA Tissue Lysis buffer (Roche Diagnostics, Australia) spiked with 5 µl EHV. Samples were heated to 100 oC for 30 min. Proteinase K (ROCHE Applied Science), 40µl, was added and vortexed for 2–3 min. Samples were then incubated overnight at 56 oC.

II. DNA extraction – Semi-automated MagnaPure

Total nucleic acid was extracted from digested tissue samples using the DNA and Viral NA Small Volume Kit on the MagNA Pure 96 extraction robot (Roche Diagnostics, Australia) as per the manufacturer’s protocol (DNATissue SV2.0). Extracted total nucleic acid was eluted into 100 µl then stored at 4 oC if PCR testing was to occur immediately or frozen at -80 oC for future testing.

4.2.2.1.3. Liver and spleen tissue Following the initial sample collection of two different tissue types (liver and spleen), samples were defrosted at 4 oC prior to use. Each sample was divided into duplicates for optimisation of extraction and tissue sample comparison. A spike, consisting of a standard volume of reconstituted commercial (Virion/Serion, DKSH, Victoria, Australia) C. burnetii Nine Mile strain whole cell bacteria, was then added to one of each duplicates of liver and spleen tissue at each dilution, respectively. The remaining duplicates were left without a C. burnetii spike as shown in Figure 4.2.

Liver and spleen samples collected for the tissue optimisation process and liver samples collected for Study 2 underwent the same total nucleic acid extraction as described below.

Volume of solution 1,200 µl 1,000 µl 800 µl 600 µl 400 µl 200 µl No C.b No C.b No C.b No C.b No C.b No C.b

PBS C.b C.b C.b C.b C.b C.b Spike Spike Spike Spike Spike Spike Spleen No C.b No C.b No C.b No C.b No C.b No C.b

Lysis C.b C.b C.b C.b C.b C.b Spike Spike Spike Spike Spike Spike

No C.b No C.b No C.b No C.b No C.b No C.b

PBS C.b C.b C.b C.b C.b C.b Spike Spike Spike Spike Spike Spike Liver No C.b No C.b No C.b No C.b No C.b No C.b

Lysis C.b C.b C.b C.b C.b C.b Spike Spike Spike Spike Spike Spike

Figure 4.2 Physical layout of samples for tissue optimisation for liver and spleen samples prior to DNA extraction Key: C.b, Coxiella burnetii

I. Sample preparation prior to extraction

Samples were heated to 100 oC for 30 min. Proteinase K (Roche Applied Science), 40 µl was added and vortexed for 2–3 min. Samples were then incubated overnight at 56 oC.

II. DNA extraction – Semi-automated MagnaPure

Total nucleic acid was extracted from digested tissue samples using the DNA and Viral NA Small Volume Kit on the MagnaPure 96 extraction robot (Roche Diagnostics, Australia) the same as the placental tissue samples described above.

4.2.2.2. Real-time PCR methods

4.2.2.2.1. Quality control of test samples

I. Equine Herpes Virus PCR

Real-time PCR for the detection of EHV was performed on all extracted samples to monitor the extraction process. The PCR mix consisted of 1µl of working primer mix (forward and reverse) at 10pmol concentration, 0.2 µl of the probe at 4 pmol concentration (Table 4.2) and 12.5 µL of Quantitect Probe master mix (Qiagen, Brisbane, Australia). This assay was performed in a ViiA™ 7 Real-Time PCR System (Thermo Fisher Scientific) using the following cycling conditions: 15 min denaturation incubation at 95 °C, followed by 50 cycles of 95 °C for 15 sec and 60°C for 1 min (Tozer et al. 2014).

II. Housekeeping genes

Real-time PCR for the detection of mammalian housekeeping genes was performed on all extracted samples. Two mammalian genes GAPDH (glyceraldehyde-3-phosphate dehydrogenase) and β-actin were chosen to detect DNA of mammalian-cells in samples. Primers for GAPDH and β-actin were based on published sequences (Table 4.2;

Lisowski et al., 2008). Briefly, the PCR mix consisted of 8.5 µL H2O, 12.5 µL SYBR-green (Quantitect), 1µl of working primer mix and 3.0 µL of sample DNA. This assay was performed in a ViiA™ 7 Real-Time PCR System (Thermo Fisher Scientific). Standard cycling conditions were used with a 95 °C denaturation step, 30 s at 60 °C (for annealing) and 40 sec at 72 °C for elongation. The melt curve analysis consisted of 2 sec at 95 °C, 5 sec at 58 °C and slow heating at rate of 0.1 °C per sec up to 95°C, for product dissociation (Lisowski et al. 2008).

4.2.2.2.2. Coxiella burnetii testing from test samples Three different individual real-time (RT) PCR assays were performed for the detection of C. burnetii DNA in test samples. Specific primers and probes targeting three different genes were used to increase test sensitivity. The first target was the IS1111 element of the transposase gene that has been found to be repeated up to 20 times through the C. burnetii genome (Klee et al. 2006). The second target was the com1 outer membrane gene (Lockhart et al. 2011). The final target was the heat shock operon (htpAB) recently used in Australian goat studies as published by Bond et al. (2015). Complete sequences of gene targets are provided in Table 4.2. All primers and probes were synthesised by GeneWorks Pty Ltd (Hindmarsh, Australia). Lyophilised primers and probes were reconstituted to a

97 standard 200 µM stock concentration with sterile, distilled water and were stored in a dedicated PCR set up laboratory in a -20 °C freezer.

Positive control samples used for RT-PCR assays consisted of reconstituted C. burnetii antigen (Virion/Serion, DHSH) Nine Mile strain commercially available for complement fixation test (Tozer et al. 2014). Serial dilutions to end-point detection of this Nine Mile strain whole cell bacterial suspension were used as quality control standards to optimise RT-PCR assays. Negative control samples were included in all PCR runs.

Each PCR mix consisted of 1 µl of respective working primer mixes (IS1111, com1 or htpAB) at 10 pmol concentration, 0.2 µl of specific probes at 5 pmol concentration, 12.5 µL of Quantitect Probe master mix (Qiagen) and 5 µL of extracted DNA in a final volume of 25 µL. Amplification was performed in a ViiA™ 7 Real-Time PCR System, using the following cycling conditions: 15 min incubation at 95°C, 50 cycles of 95°C for 15 sec and 60 °C for 1 min.

Test samples were considered positive if one or more targets for C. burnetii tested positive for bacterial DNA. The cycle threshold (Ct) value was defined as ≤ 29.9 indicating a strong positive reaction, Ct of 30.0–38.9 as a positive reaction and 39.0–44.9 as weak positive, Ct >45.0 indicated a test negative result. For a positive sample to be suitable for further molecular genotype testing a Ct ≤ 30.0 was required (Vincent et al. 2016).

4.2.2.2.3. PCR testing for tissue optimisation During the tissue optimisation process, housekeeping genes and C. burnetii targets were used to test the multiple duplicate sample solutions as described above in section 4.2.2.1.3. Firstly, the housekeeping genes GAPDH and β-actin were analysed for both tissue types at each concentration of lysis buffer and PBS and results compared. Secondly, in the tissue samples spiked with C. burnetii, the Ct values for targets IS1111 and heat shock operon at each volumes of lysis buffer and PBS were analysed and compared with the known positive control. These targets were monitored for assay performance to determine which tissue sample, liver or spleen, would be used as the collection method for Study 2.

4.2.2.3. Statistical analysis The overall animal level apparent prevalence (AP) of C. burnetii PCR positive liver samples was calculated in STATA® by running a negative binomial model with a cluster variable for property. The model was run as an intercept only model and then confidence intervals were estimated for the intercept. For regional summaries, AP estimates were calculated as the proportion of samples positive per region and state with the mean and 98

95% confidence intervals (Wilson-Score method) presented (Brown, Cai & Das Gupta 2001). The regional estimates do not account for clustering of animals within properties. Apparent prevalence results were not adjusted for diagnostic sensitivity or specificity of the tests used.

Table 4.2 List of oligonucleotide sequences for gene target primers and probe sequences for PCR assays Gene Target Primers Probe Reference

Forward - GTC TTA AGG TGG GCT GCG TG C. burnetii FAM - AGC GAA CCA TTG transposase GTA TCG GAC GTT TAT GG - (Klee et al. 2006) gene “IS1111” BHQ Reverse - CCC CGA ATC TCA TTG ATC AGC

Forward - AAA ACC TCC GCG TTG TCT TCA C. burnetii FAM - AGA ACT GCC CAT TTT (Lockhart et al. outer membrane TGG CGG CCA - BHQ-1 2011) gene “com1” Reverse - GCT AAT GAT ACT TTG GCA GCG TAT TG

Forward - GTG GCT TCG CGT ACA TCA GA C. burnetii FAM - AGC CAG TAC GGT heat shock (Bond et al. 2015) CGC TGT TGT GGT- BHQ1 protein “htpAB” Reverse - CAT GGG GTT CAT TCC AGC A

Forward - GAT GAC ACT AGC GAC TTC GA Equine Herpes FAM-TTT CGC GTG CCT CCT (Tozer et al. Virus CCA G-BHQ-1 2014) Reverse - AGG GCA GAA ACC ATA GAC A

Forward - GAG CGG GAA ATC GTC CGT GAC (Lisowski et al. β-actin No probe required 2008) Reverse - GTG TTG GCG TAG AGG TCC TTG C

Forward - ACC ACT TTG GCA TCG “GAPDH” TGG AG glyceraldehyde- (Lisowski et al. No probe required 3-phosphate 2008) dehydrogenase Reverse - GGG CCA TCC ACA GTC TTC TG

100

4.3. Results

4.3.1. Summary of cattle sampled Samples for Study 1 were collected between July 2016 and August 2016. Amniotic fluid was collected from 92 mid to late term pregnant cattle post-slaughter. Of the 92 pregnant cattle that were sampled, 80 also had placental tissue collected. Samples for study 2 (799 liver samples) were collected between April 2018 and June 2018. Inclusive of both studies, 891 cattle going to slaughter at one abattoir were sampled. Of these, 843 originated from known regions of Queensland, 48 from known regions of New South Wales and 8 from unknown origins (Table 4.3).

Table 4.3 Regional origins of cattle sampled during abattoir surveillance Region Study 1 Study 2 Amniotic fluid/ Liver placenta South East Queensland/ Wide Bay Burnett 5 20 Darling Downs South West 32 285 Central Queensland 55 217 Mackay Isaac and Whitsunday 0 83 North Queensland 0 127 Far North Queensland 0 11 New South Wales 0 48 Unknown origin 0 8 TOTAL 92 799

4.3.2. Results from Study 1 From the first study, 92 amniotic fluid and 80 placental tissue samples underwent individual DNA extraction processing. All samples tested positive for EHV using RT-PCR; indicating the extraction processing was successful. All samples tested positive for GAPDH housekeeping gene indicating that DNA from mammalian cells were present in the samples. All 172 samples from 92 individual cows tested negative for C. burnetii targets com1 and IS1111 and htpAB (AP = 0.0% (95% CI 0.0, 4.0%). All positive controls for C. burnetii showed positive results for all target genes. All negative controls tested negative for C. burnetii targets.

101

4.3.3. Results from Study 2

4.3.3.1. Tissue optimisation All tissue samples used for the optimisation process tested positive for GAPDH and β-actin housekeeping genes at each dilution, indicating that DNA from mammalian cells were present and detectable in every optimisation sample. As there was no noticeable difference between the performances of the two housekeeping gene assays, GAPDH was arbitrarily chosen as the general housekeeping gene for all study samples.

Table 4.4 Results for the tissue optimisation of liver and spleen samples

C. burnetii Cycle threshold Solution Tissue type genome target (average*)

PBS Spleen IS1111 37.7 PBS Spleen htpAB ¥44.1 PBS Liver IS1111 25.8 PBS Liver htpAB 28.9

Lysis Spleen IS1111 27.1 Lysis Spleen htpAB 33.8 Lysis Liver IS1111 24.6 Lysis Liver htpAB 26.4

KEY: *Reported cycle thresholds are an average of 6 samples; ¥ This result is an average of 4 samples as 2 samples did not produce a detectable response; IS1111, C. burnetii transposase gene; htpAB, C. burnetii heat shock protein gene; PBS, phosphate-buffered saline; Lysis, DNA Tissue Lysis buffer (Roche Diagnostics, Australia)

The average RT-PCR Ct value for both C. burnetii gene targets, IS1111 and htpAB, was better for liver samples compared to the average Ct value for spleen samples. For the spleen samples collected in PBS, only 4/6 samples produced a detectable Ct for the htpAB target. The average Ct value for liver collected in lysis was closer to the positive control sample Ct than liver collected in PBS (Table 4.4). It was therefore determined that liver collected in lysis buffer had better assay performance. Finally, the RT-PCR Ct values for each individual liver sample collected in different volumes of lysis buffer (200 µl, 400 µl, 600 µl, 800 µl, 1 000 µl and 1 200 µl) were inspected and compared to the C. burnetii positive control sample with no tissue present. The liver tissue collected into 400 µl tissue lysis buffer

102 had the closest Ct values with suitable curves to the positive control sample, therefore this was chosen as the collection protocol for Study 2.

4.3.3.2. Liver collection From Study 2, 799 liver samples were collected from animals originating from 45 individual PICs (property identifying codes). All 799 liver core biopsy tissue samples underwent individual total nucleic acid extraction processing. All samples tested positive for EHV using RT-PCR; indicating the extraction processing was successfully performed. All samples tested positive for GAPDH housekeeping gene to indicate that mammalian cells were present in each sample.

From 799 samples tested using RT-PCR for C. burnetii gene targets IS1111 and htpAB, 6 samples were positive to the IS1111 gene. Positive samples came from animals originating from 4/45 properties. The overall animal-level apparent prevalence (accounting for animals clustered within properties) was 0.75% (95% CI 0.148, 1.35). Breakdown of the percentage of positive samples according to geographical location of originating cattle with 95% CIs were presented in Table 4.5. Please note, the table showed crude test positive results with estimated 95% CI that were calculated without accounting for clustering at the property. The model would not converge to estimate regional prevalence estimates accounting for clustering of animals within properties. The Ct values for the positive PCR samples were shown in Table 4.6. All known positive controls for C. burnetii showed positive Ct values for gene targets IS1111 and htpAB.

Cycling threshold values for the PCR positive liver samples were not adequate to enable further laboratory testing or molecular genotyping analysis to be performed.

103

Table 4.5 Prevalence of C. burnetii from bovine liver samples according to geographical origin of cattle

PCR Total Percentage 95% Confidence State/Region positive sample positive (%) Interval (%) samples number Lower limit Upper limit South East Queensland/ Wide Bay Burnett 0 20 0 0.0 16.1 Darling Downs South West 3 285 1.1 0.4 3.0 Central Queensland 1 217 0.5 0.1 2.6 Mackay, Isaac and Whitsunday 0 83 0 0.0 4.4 North Queensland 2 127 1.6 0.4 5.6 Far North Queensland 0 11 0 0.0 25.9 New South Wales 0 48 0 0.0 7.4 Unknown origin 0 8 0 0.0 32.4 TOTAL 6 799 0.8 0.3 1.6

Table 4.6 Cycle threshold (Ct) values of C. burnetii PCR positive targets; limit of Ct value set that >45 is negative. C. burnetii Origin of target cattle IS1111 htpAB Barcaldine 39.0 nd Mt Isa 38.4 nd Mt Isa 37.6 nd Roma 38.6 nd Roma 38.7 nd Roma 42.6 nd KEY: nd = not detected

104

4.4. Discussion and conclusions

The high risk of Q fever for abattoir workers has long been reported in Australia, however current workplace health and safety policies incorporating vaccination have reduced this occupational hazard for the industry’s personnel (Gidding et al. 2009; Woldeyohannes et al. 2018). While human vaccination has successfully reduced the incidence of abattoir-related Q fever over recent years, Q fever notifications associated with farming and agricultural workers are still apparent. In Australia, there have not been any control programs to reduce C. burnetii infection in cattle; therefore, it seemed plausible that C. burnetii bacteria could be circulating within abattoirs. Cattle-only facilities therefore provided an opportunity to further investigate C. burnetii infection in cattle in Australia.

During this project, 2 surveys were performed to detect C. burnetii in cattle going to slaughter at a large abattoir in south-east Queensland. Cattle originating from several Queensland regions and northern New South Wales were successfully sampled and analysed using molecular laboratory methods. C. burnetii DNA was detected, although at a lower than expected frequency, from the liver of cattle originating from Darling Downs South West, Central Queensland and North Queensland regions. There was however, no evidence of C. burnetii infection in any placental tissue or amniotic fluid samples collected from pregnant cattle post-mortem.

Although C. burnetii was not detected in placental tissue or fluid samples from pregnant cattle going to slaughter in the current study, this may be explained by limitations in the sampling strategy and small sample size. However, it is important to highlight the upper CI of this prevalence estimate is ~4%, and acknowledge that if the clustering of the samples were incorporated into the analysis then this upper estimate could be slightly higher. Thus, results presented here indicate that the rate of infection may be zero or moderately low, but certainly not extremely high.

There were constraints imposed by logistics that limited the number of samples collected and processed on a given sampling day. Collection of placental tissue and amniotic/allantoic fluid was labour intensive and required additional hours of processing once leaving the abattoir. The proportion of cows arriving at the abattoir pregnant was also low and unpredictable. Although there are some circumstances (such as during severe, prolonged drought) where pregnant cows may be sent to slaughter for welfare reasons, during Study 1 collection period, the national herd was going through a re-building phase of

105 re-stocking and growth, recovering from the previous drought salvage sales (Meat and Livestock Australia 2019a).

Exposure to infected pregnant ruminants is a well-established risk factor for the transmission of Q fever to humans (Arricau-Bouvery & Rodolakis 2005; Roest et al. 2010). The bacterium, C. burnetii, has an affinity to the placenta and placental membranes have been reported to contain up to 109 organisms/g of tissue (Sobotta et al. 2016). In dairy cattle, PCR methods have been used for the detection of C. burnetii in abortive tissue with more organisms detected from cotyledons than foetal abdominal and thoracic fluid (Agerholm 2013). Experimental infection of goats also revealed that C. burnetii has a strong affinity for the trophoblasts of the placenta, and PCR methods were able to detect C. burnetii from ruminants that experienced abortions as well as some that delivered healthy kids or calves (Agerholm 2013; Roest, Bossers & Rebel 2013). Although it is mentioned extensively in the literature that a large number of bacteria can be found in the birthing fluids of infected cattle, no primary research could be found that described testing of amniotic or allantoic fluid from cattle. It has been reported that C. burnetii DNA was identified in allantoic and amniotic fluid in experimentally infected pregnant goats (Roest 2013).

The initial sampling strategy was targeted towards detecting disease using a herd-level, risk-based sampling method to identify C. burnetii suitable for molecular genotyping. The sampling strategy was reviewed following completion of PCR testing on an initial run of samples from 92 pregnant uteri. A decision was made to revise the sampling protocols and revise sample size estimates that incorporated precision at an animal-level for study 2. However, as this sampling strategy did not incorporate clustering of animals within properties/mobs, the required sample numbers are likely smaller than desired for an unbiased estimating of animal-level prevalence. Caution should be made to avoid making strong inferences from findings in the study.

During Study 2, C. burnetii DNA was detected in liver samples from cattle originating from three out of six geographical regions within Queensland. The frequency detected within the positive regions was very low, ranging from 0.5% (95% CI 0.1, 2.6%) to 1.6% (CI 0.4, 5.6%) and the overall animal prevalence was 0.8% (CI 0.3, 1.6%). Although these estimates may be biased due to deficits in the sampling strategy, the results are similar to published seroprevalence reports of beef cattle from Korea, Ireland and Spain (ranging from 1.7–1.9%; McCaughey et al., 2010; Alvarez et al., 2012; Lyoo et al., 2017). While seroprevalence is not necessarily an ideal indication of infection status, it is worth noting this similarity.

106

C. burnetii DNA was not detected in liver samples from cattle in the remaining three regions of Queensland (Far North Queensland, Mackay Isaac and Whitsunday and South East Queensland/ Wide Bay Burnett) or New South Wales; however, this may be due to limitations of sample size. Smaller numbers of cattle were available for testing from these regions. This may be because they are areas that do not typically run beef cattle or they are geographically distant and may send cattle to different abattoirs. It is therefore, plausible that the apparent negative results in the regions may be attributed to the inability to reach required numbers to detect infection with confidence, rather than a true zero prevalence of infection.

The sampling of liver for bacterial detection may be a major limitation of the study. There is insufficient information to unequivocally define specific tissues as being more likely to test positive for Coxiella in chronically infected cattle. In humans, haematogenous spread of bacteria results in infection of the liver, spleen, bone marrow, reproductive tract and other tissues (Eldin et al. 2016). In ruminants, C. burnetii has been isolated or detected using molecular methods from liver, spleen and other tissues (Cumbassá et al. 2015; Bell, Parker & Stoenner 1949). These factors, along with results from the laboratory tissue optimisation described in this chapter, were considered when planning the study. Although the tissue optimisation process was able to provide information regarding the ability to extract and detect DNA in bovine liver and spleen samples spiked with C. burnetii in the laboratory, it was not designed to test which tissues may have a higher natural affinity for a latent infection. Future studies that investigate C. burnetii presence and quantity in tissues of naturally or experimentally infected cattle would help to establish reliable sampling protocols. The higher Ct values (lower DNA yield) observed in this study for both C. burnetii PCR targets in the spleen samples may have been a result of PCR inhibition. The spleen is high in red blood cells, lymphatic tissue and immunoglobulins that are recognized as PCR inhibitors. Although we used an endogenous control (GAPDH) to confirm bovine cells were present in each sample, an exogenous internal amplification control (such as EHV) was not used in the tissue optimisation experiment. This would be important for future work and the extraction method could potentially be further optimised to utilise spleen as a useful tissue sample for diagnostic development.

Three separate C. burnetii gene targets were used for RT-PCR detection in this study. The com 1 and IS1111 have commonly been used for the detection of C. burnetii DNA in Q fever patients and have also been reported for animal and environmental detection in Australia (Tozer et al. 2014; Klee et al. 2006; Lockhart et al. 2011). Both targets were 107 reported as suitable assays to detect C. burnetii infection in clinical samples of aborted placental cotyledon and foetal samples submitted to a laboratory in the United Kingdom (Jones et al. 2010). The IS1111 assay was found to have a high analytic sensitivity, “with a limit of detection of 10 copies of template, theoretically equating to a single bacterium” (Jones et al. 2010). More recently, the heat shock operon htpAB was identified as a suitable gene target for PCR detection C. burnetii in goat and human samples during a Q fever outbreak in Victoria, Australia (Bond et al. 2015). Abattoir samples in this study were only found RT-PCR positive using the IS1111 gene target. All samples tested with targets com1 or htpAB returned negative results. Although these targets and PCR assays are well established, the diagnostic sensitivity and specificity of the methods for testing specific cattle samples are unknown.

An abattoir surveillance system can be suitable as a cost-effective way to estimate prevalence and monitor infections (such as coxiellosis) that are unlikely to influence presentation at an abattoir (Cowled et al. 2013). For example, infections that may have subclinical or asymptomatic disease. However, it must be acknowledged that biases may exist because cattle sub-populations presenting for slaughter tend to be either older animals being culled or alternatively younger, healthy animals and may not be truly representative of the broader cattle populations (Cowled et al. 2013). Within this survey, additional biases may exist given that larger consignments of cattle were focused for sampling and the sample strategy implemented may not have been appropriate to estimate regional animal level prevalence of infection.

Abattoir surveillance may not be an ideal method for quick and effective identification of C. burnetii strains infecting cattle in Australia. Although in other studies, clinical samples have been used to identify bacterial strains using molecular genotyping techniques, this relies on a good quality PCR positive sample (Jado et al. 2012; Cumbassá et al. 2015). It is possible that risk-based sampling strategies may increase the likelihood of detecting the organism. One option might be the collection of samples from cattle on-farm, at time of calving (or abortion) or during the immediate peri-parturient period. It is known that highest bacterial shedding can be limited to time of parturition, however the dynamics of infection in cattle are not completely understood (Roest, Bossers & Rebel 2013; Guatteo et al. 2006). Another possibility is to identify suspected infected herds from serological-surveys and then follow-up with appropriate sampling for identification of C. burnetii genotypes. These methods may increase the sensitivity of the surveillance system, thus increasing the likelihood of achieving a good quality sample. Overall, monitoring of genomic strains of 108

C. burnetii from an abattoir surveillance survey appear to be extremely difficult and impractical.

The studies described in this chapter were conducted at two distinct time-points, therefore, we can only estimate a cross-sectional prevalence at those particular times. An extension of this study could implement collection of samples for surveillance continuously throughout a year or at several times over multiple years. This may allow temporal variations in infection to be monitored and permit analysis for putative seasonal variation.

While findings from this current chapter did not identify a high prevalence of C. burnetii in beef cattle going to slaughter, it cannot be confidently stated that there is a low risk for abattoir workers. It was identified that non-reproductive organs (such as liver) harbouring the bacteria during a latent infection could risk transmission to abattoir workers. The infective dose of C. burnetii is extremely low, therefore even a low bacterial load could pose a threat (Sobotta et al. 2016). These results support the current Australian risk mitigation strategy of prophylactic vaccination of abattoir workers against Q fever.

109

The following chapter is published as a peer-reviewed original article in Preventive Veterinary Medicine as follows:

I, Caitlin Wood, state that I have participated sufficiently in the publication to take public responsibility of the work. I was first author of the manuscript and I made substantive contributions to the concept and design, analysis and interpretation of the research data on which the publication was based and in the writing and editing of the manuscript.

110

Chapter Five

Optimisation and validation of an indirect immunofluorescence assay (IFA) for the detection of IgG antibodies against Coxiella burnetii in bovine serum

“Measurement is the first step that leads to control and eventually to improvement. If you can’t measure something, you can’t understand it. If you can’t understand it, you can’t control it. If you can’t control it, you can’t improve it.”

‐ H. James Harrington (1929–present)

111

5. Optimisation and validation of an indirect immunofluorescence assay (IFA) for the detection of IgG antibodies against Coxiella burnetii in bovine serum

5.1. Introduction

Coxiella burnetii is a zoonotic bacterium able to infect multiple species, including but not limited to livestock and humans. It is the causative agent of Q fever in humans and coxiellosis in animals (Woldehiwet 2004). Domesticated ruminants, including cattle, sheep and goats are considered to be important sources of human Q fever (Maurin & Raoult 1999). Infection with C. burnetii may cause reproductive disease in ruminants and has been associated with bovine abortion, premature delivery and birth of weak neonates in dairy cattle in Europe (Agerholm 2013; Hansen et al. 2011). Subclinical infection in cattle may persist for weeks to months, with bacterial shedding possible through multiple secretions including placental fluids, vaginal mucus, milk, urine and faeces (Guatteo et al. 2011). Estimations of the true prevalence and distribution of coxiellosis in dairy and beef cattle populations in Australia are limited due to minimal surveillance and minimal standardisation of diagnostic test methods.

The C. burnetii bacterium exhibits two phase variations, phase I and phase II, with both being used as antigens for serological assays. Antibodies against C. burnetii can be detected in bovine serum and milk and serology is considered a reliable diagnostic method for herd-level investigations (OIE 2018; H. Roest et al. 2013). Identification of IgG antibodies against C. burnetii provides evidence of past exposure and/or recent infection (Natale et al. 2012). In human studies, detection of elevated levels of phase I and phase II antibodies are used to differentiate between the acute and chronic or persistent forms of Q fever disease (Tozer et al. 2011; Eldin et al. 2016). The detection of phase-specific antibody patterns could have potential to identify ruminants that are chronically shedding bacteria (Lucchese et al. 2015). There is also preliminary evidence that a phase I IgG seroconversion may indicate protective immunity in goats following previous exposure to C. burnetii (Muleme et al. 2017; Canevari et al. 2018). However, a clear association between C. burnetii phase-specific antibody response and clinical disease in cattle has not been well established.

The validation of affordable and reliable serological tests may encourage surveillance and epidemiological studies of bovine coxiellosis in Australia. The most commonly used serological methods are the complement fixation test (CFT), indirect immunofluorescence

112 assay (IFA) and enzyme-linked immunosorbent assay (ELISA) (Natale et al. 2012; Porter et al. 2011; Brom et al. 2015). The only commercially available serological test for use in cattle in Australia is a multi-species ruminant ELISA kit (Q Fever Ab Test IDEXX Laboratories, United States of America). This kit can be prohibitively expensive for large-scale cattle screening and there are no current validation data available from the manufacturer for the test’s performance in bovines. While the CFT has been standardised and validated for use in veterinary laboratories, it is consistently reported to have very low sensitivity (ranging from 29.8% - 36.7%), is laborious to perform and prone to non-specific reactions resulting in inconclusive test results (Emery, Ostlund & Schmitt 2012; Horigan et al. 2011; Kittelberger et al. 2009; Muleme et al. 2016; OIE 2018). Currently, there is no gold standard for sero- diagnosis of coxiellosis in livestock.

The IFA is the reference test for human Q fever diagnostics in Australia and was found to have greater sensitivity than the CFT and provide a more cost-effective option to monitor antibody dynamics of patients (Worswick & Marmion 1985). While there are no commercially available IFA kits for use in ruminants, Muleme et al. (2016) recently validated an IFA for detecting IgG and IgM antibodies against C. burnetii in goat serum. They estimated the diagnostic sensitivity (DSe) of the IFA, at a 1:160 serum dilution cut-off, to be 94.8% (Credible Interval (CrI) 80.3–99.6); this was higher than the DSe of the ELISA (70.1%; CrI 52.7–91.0) and CFT (29.8%; CrI 17.0–44.8). The diagnostic specificity (DSp) for detecting IgG was found to be similar across all three serological tests (92.5%, 96.2%, 96.8%, respectively). Therefore, in goats, the IFA appeared to have superior sensitivity compared to both the IDEXX ELISA kit and CFT test methods (Muleme et al. 2016).

5.1.1. Chapter objectives The objectives of this study were to modify, optimise and validate an IFA for the detection of phase-specific IgG antibodies against C. burnetii in cattle serum for the purpose of estimating prevalence of exposure in epidemiological investigations. Additionally, a commercially available ELISA was evaluated for application in seroprevalence studies in Australian cattle.

113

5.2. Materials and Methods

Initially, we modified and optimised a human IFA test to detect IgG antibodies against C. burnetii in bovine serum. Then, a direct comparison of the modified IFA with the commercial ELISA kit (Q Fever Ab Test IDEXX Laboratories, United States of America) was accomplished by testing serum samples from distinct cattle populations across Australia and New Zealand. Finally, we used Bayesian latent class analysis to estimate the diagnostic test accuracy (DSe and DSp) of the serological tests in the absence of a gold standard (OIE 2016a).

5.2.1. Serum samples

5.2.1.1. Sample collection, storage and animal ethics Serum samples tested during this validation study were originally collected as a part of either university research projects (The University of Queensland Animal Ethics approval SVS/115/11/MLA (NF) and The University of Sydney Animal Ethics approval 593), or by government department staff for regulatory testing that did not require animal ethics approval (Victorian Government Department of Economic Development, Jobs, Transport and Resources and the New Zealand Ministry of Primary Industries). The use of all samples for the current study was approved by The University of Queensland animal ethics ANRFA/SVS/100/16. Samples were stored frozen at -20 oC or -80 oC depending on the intended length of storage.

5.2.1.2. Control sera for IFA testing Positive bovine control sera (freeze-dried) for the IFA were sourced from the Australian National Quality Assurance Program (ANQAP; AgriBio, DEDJTR, Bundoora, Victoria). The negative control sera, were obtained from New Zealand, which is considered by the World Organisation for Animal Health (OIE) to be free of C. burnetii (OIE 2018). The Animal Health Laboratory, Wallaceville, New Zealand Ministry of Primary Industries had tested the serum samples using the ELISA (Q Fever Ab Test, IDEXX) to confirm negative status which we subsequently repeated.

5.2.1.3. Samples for proficiency panel Ten bovine sera were sourced from the Australian Rickettsial Reference Laboratory (ARRL), Geelong, Victoria, to be used as a proficiency panel to compare inter-laboratory IFA test results. These sera were from individual cattle tested by the ARRL using both IFA and ELISA methods. .

114

5.2.1.4. Samples for the diagnostic test comparison of IFA and ELISA A total of 458 bovine sera were used to compare the IFA and the ELISA test methods. Serum samples were sourced from four distinct populations. The first population was represented by serum samples (n = 156) collected from a suspected C. burnetii-infected dairy-cattle herd in New South Wales, Australia. At the time of sample collection, this herd recorded C. burnetii positive individual cow composite milk samples using PCR (C. burnetii DNA target IS1111a) with 34 out of 155 (21.9%) samples confirmed positive for C. burnetii on sequencing (K. Bosward, unpublished data).

The second population was represented by samples (n = 159) previously collected for the purpose of infectious disease testing within a longitudinal epidemiological study investigating causes of poor reproductive performance in beef-cattle (McGowan et al. 2014). Systematic sampling methods were used to collect blood samples from cattle within the three properties enrolled in central Queensland, Australia. Central Queensland is a rural region of Queensland with endemic human Q fever. Although these herds did not have C. burnetii PCR testing performed, beef-cattle from this region have previously been reported to have a seroprevalence of approximately 16% (Cooper 2011).

The third population were samples from a dairy-cattle herd (n = 96) with no history of suspected infection in Victoria, Australia. Although there have been limited published studies, cattle from Victoria have reported a low prevalence (0.5%) from previous and recent serological surveys (Tan 2018; Hore & Kovesdy 1972). The fourth population were the negative-reference sera (n = 47) from New Zealand.

5.2.2. Serological test methods

5.2.2.1. Development of the IFA slides Menzel-Glaser 40-well microscope slides (Tasman Scientific, Belgrave Heights, Victoria, Australia) were cleaned with 100% methanol. Slides were then air dried and labelled accordingly. Phase I and phase II C. burnetii, Nine Mile strain, antigen were reconstituted as per manufacturer’s instructions with 1 mL of distilled water (Virion/Serion, 97076 Würzburg, Germany). Briefly, reconstituted antigen was diluted with 0.5% chicken yolk sac into three different working stock solutions (Tozer et al. 2011), phase I, phase II and a mixed-phase (combined phase I and phase II) solution. These solutions were spotted onto the wells of separate slides and allowed to air dry. Slides were fixed by immersion in 100% methanol for 5-10 min, then air dried. Slides were stored at -20 oC for long term storage. Anti-bovine IgG polyclonal (whole molecule) fluorescein isothiocyanate (FITC) 115 antibody, raised in rabbit (Sigma- Aldrich, St Louis, MO 63103 USA, product number: F7887) was used to detect IgG antibody-antigen complexes.

Each new batch of slides were quality controlled using the IFA protocol outlined below including positive and negative control sera. The batch was considered to have passed quality control if the positive control test sera fluoresced to +/- one titre of its known titre and the negative control showed minimal or no background fluorescence.

5.2.2.2. Brief overview of IFA method The IFA method for detecting human phase specific antibodies against C. burnetii (Tozer et al. 2011) was modified for testing bovine serum. Briefly, IFA slides, test sera and reagents were brought to room temperature. Test sera were diluted in 2% casein- phosphate-buffered saline (PBS) to minimise non-specific binding (Muleme et al. 2016). Diluted test sera were individually placed on the slides in duplicate and incubated in a humidity chamber for a 30 min at 37 oC. If C. burnetii antibodies were present in the test serum, they would adhere to the antigen coated to the slide during this initial incubation (Figure 5.1). The slides were washed in 10% PBS for 5 min, three times and allowed to air dry. Anti-bovine IgG-FITC conjugate diluted in 0.05% Evans blue dye, was added and again incubated in a humidity chamber for 30 min at 37 oC. If IgG antibody-antigen complexes are present, the FITC conjugate would couple with the immunoglobulin complex during the second incubation period. The slides were washed three times removing any excess unbound reagents in 10% PBS for 5 min each time. The slides were air dried and coverslips mounted. Examination of the slides was performed with an immunofluorescent microscope (Nikon Eclipse E600) under a 40x lens (total 400x magnification), with oil immersion under a 100x lens (total 1000x magnification) used for closer inspection. If the test serum contained IgG antibodies against C. burnetii, the immune complex would produce an apple green fluorescence of the individual bacteria adhered to the slide, signifying a positive result (Figure 5.2).

Checkerboard dilutions were used to determine optimum cut-off points for both serum and conjugate dilution. A two-fold serial dilution of 1:40 to 1:2560 was used with negative and positive control sera in duplicate with a range of IgG FITC-conjugate dilutions from 1:300 to 1:600. All negative control sera (n = 47) were then individually titered from 1:40 to 1:2560; the lowest dilution that produced no background fluorescence was considered as the lowest possible cut-off (the sera screening dilution).

116

Figure 5.1 Depiction of the principle of how IFA works.

Figure 5.2 Photograph of an IFA slide - fluorescence observed on C. burnetii combined phase I and phase II antigen staining with positive bovine IgG antibodies (under 100x lense with oil immersion).

5.2.2.3. Screening of sera for the diagnostic test comparison Individual sera from all four populations (n = 458) were screened in duplicate at 1:80 dilution with 2% casein buffer loaded onto slides coated with mixed phase I and II antigens. 117

Samples fluorescing at this initial screening dilution on mixed-phase slides, were then re- screened and titered at doubling dilutions from 1:80 to 1:1280 on individual slides prepared with separate phase I and phase II antigens respectively.

5.2.2.4. Brief overview of ELISA method The IDEXX ELISA Q Fever Ab Tests were performed as per the manufacturer’s guidelines (IDEXX 2017). All reagents and control sera for the ELISA were provided as a part of the commercial kit. Briefly, micro-well plates coated with combined phase I and phase II C. burnetii antigen were provided in the commercial kit. Test sera and control sera were diluted at 1:400 and loaded into the wells in duplicate. Dilution optimisation was not performed for the ELISA because it was a commercial kit with a stated dilution required. After incubation and washing steps, a peroxidase labelled anti-ruminant IgG conjugate was applied to the wells. After further incubation and washing steps, tetramethylbenzidine (TMB) substrate was applied to the wells allowing the development of colour. The amount of C. burnetii IgG antibodies present in the serum sample was directly proportional to the final colour outcome. Test results were achieved by reading the ELISA plate through a microplate spectrometer (SpectraMax 340PC384 Microplate Reader) at 450 nm. The optical density (OD) of the test and control samples were compared using the formula provided to determine the sample to positive (S/P) ratio expressed as a percentage:

𝑆 𝑂𝐷 𝑠𝑎𝑚𝑝𝑙𝑒 𝑂𝐷 𝑛𝑒𝑔𝑎𝑡𝑖𝑣𝑒 𝑐𝑜𝑛𝑡𝑟𝑜𝑙 % 100 𝑃 𝑂𝐷 𝑝𝑜𝑠𝑖𝑡𝑖𝑣𝑒 𝑐𝑜𝑛𝑡𝑟𝑜𝑙 𝑂𝐷 𝑛𝑒𝑔𝑎𝑡𝑖𝑣𝑒 𝑐𝑜𝑛𝑡𝑟𝑜𝑙

Cut-off for S/P% values were taken from the kit insert as follows: negative result S/P% < 30%; suspect results 30 % > S/P% < 40%; positive result S/P % > 40%.

5.2.2.5. Analytical performance of the IFA The proficiency panel of 10 bovine sera were screened using the IFA method at doubling dilutions beginning at 1:40 to 1:5120. End-point titres were determined for IgG against both phase I and phase II C. burnetii antigen individually. Screening test results from two independent readers were recorded and compared to those from the reference laboratory to determine the reliability of interpretation of IFA method and hence to investigate subjectivity between operators.

5.2.2.6. Comparison of phase variation on IFA slides on a subset of samples An additional comparison was performed with a subset of samples (the New South Wales dairy herd; n = 146). Test sera were screened, in duplicate, on slides coated with 118 combined-phase antigen (a mixed preparation of phase I and II antigen) and also in duplicate, on slides individually coated with phase I and phase II antigen separately. Serum results obtained for the IFA slide preparation of combined-phase antigen were compared with serum results from IFA slide preparations of phase I and phase II separately. A positive to either phase I or phase II or both phases was considered positive for the separate slide preparations. These cross-classified test results were then analysed.

5.2.3. Statistical analysis Cohen’s kappa test statistic (ĸ) and prevalence and bias adjusted kappa (PABAK) test statistic were calculated using STATA/SE 15.0 ® (Stata Statistical Software, Stata Corporation, College Station, TX, USA) and WinPepi (Abramson 2011) to determine the agreement beyond chance of paired test results (Byrt, Bishop & Carlin 1993). Interpretation of kappa statistics were as follows: > 0.8 excellent agreement; > 0.6–0.8 substantial agreement; > 0.4–0.6 moderate agreement; > 0.2–0.4 slight agreement; 0–0.2 poor agreement (Thrusfield 2007).

Bayesian latent class modelling was used to evaluate the performances of the IFA and the ELISA for the detection of IgG antibodies against C. burnetii in bovine serum in the absence of a gold standard reference test (Branscum, Gardner & Johnson 2005). This analysis assumes the exposure status of each sample as unknown and hence “latent”; the model estimates the probability that each of the four possible test combinations (T1+ T2+ ; T1+ T2- ; T1- T2+ and T1- T2-) represents a true positive sample, thus allowing inference on the diagnostic accuracy of both tests to be made (Branscum, Gardner & Johnson 2005). Correlation between the two tests was assumed given both were serological assays targeting the same antigens of C. burnetii. The conditional dependence model for two tests over two populations, as described by Branscum et al. (2005) was adapted to account for four populations. DSe and DSp of the two tests were assumed constant across the four populations; prevalence of C. burnetii exposure was assumed distinct in the four populations.

Priors for DSe and DSp of the IFA and ELISA were incorporated into the model using independent and informative unimodal beta distributions based on published literature. Unimodal beta distributions were estimated using the “epi.betabuster” function implemented within the “epiR” library (Stevenson 2017) in R (‘R: A language and environment for statistical computing’ 2019) as shown in Table 5.1. The IFA was estimated to have a DSe with mode of 94.8% and DSp with a distribution centred on 92.5% using data published from

119 goats in Australia (Muleme et al. 2016). The ELISA was estimated to have DSe with mode of 81.3% and DSp with mode of 87.4%, as published by Horigan et al. (Horigan et al. 2011). Prior information about the assumed prevalence estimates of C. burnetii exposure in the four cattle populations were also taken from published literature available at the time or knowledge of infection status of the herd (Table 5.1). Diffuse priors centred on the point estimates from published sources were included in the model. These priors were made sufficiently diffuse so that the data could overwhelmingly inform the posterior distributions. The New Zealand population was assumed free of infection; hence, a mixture distribution derived from a Bernoulli distribution and Beta prior distribution was used to account for the high probability that the population was not infected at all (Branscum, Gardner & Johnson 2005).

Bayesian inferences were based on the joint posterior distribution, approximated using the computer software JAGS (version 4.3.0, citation), implemented with R2jags package in the R statistical package (See Appendix of chapter for R code) . The model was run with two independently initiated chains, each of 101000 iterations, discarding the first 1000, then thinning by 10 to reduce autocorrelation (based on assessment of convergence using the Gelman-Rubin test statistic and autocorrelation; Gelman & Rubin 1992; Plummer et al. 2006). Diagnostic sensitivity and specificity of the two tests, true prevalence of the four populations, and Youden’s index were selected for monitoring. The Markov Chain Monte Carlo (MCMC) accuracy was assessed by visual inspection of the autocorrelation function and estimation of the effective sample size (Kruschke 2015).

Final estimates were reported as the median and 95% credible interval (CrI) of the joint posterior distribution. Modelling was repeated assuming different dilution cut-offs (1:80, 1:160, 1:320, 1:640, 1:1280) for the IFA. This allowed estimation of a two-way receiver operator characteristic (ROC) curve, which interpreted with Youden’s index allowed identification of the IFA cut-off with the highest combined DSe and DSp (i.e. the globally optimal cut-off). Youden’s index (J) is defined as the cut-point that achieves the maximum sensitivity and specificity; J = maxc {Se (c) + Sp (c) − 1}. To test the influence of the priors on the final model outputs, less-informed priors were used to perform sensitivity analysis Table 5.2 (Kostoulas et al. 2017).

120

Table 5.1 Table of prior distributions used in the Bayesian latent class model

Detailed prior specifications

Test specifications

Mode 5th Percentile Parameters for prior distribution Reference

ELISA DSe 0.813 0.54 Beta (8.97, 2.833) (Horigan et al. 2011) ELISA DSp 0.874 0.55 Beta (7.252, 1.901) (Horigan et al. 2011) IFA DSe 0.948 0.8 Beta (4.793, 1.208) (Muleme et al. 2016) IFA DSp 0.925 0.7 Beta (5.053, 1.329) (Muleme et al. 2016)

Population prevalence

Mode 95th Percentile 99th Percentile Parameters for prior distribution Reference

NSW 0.32 0.6 Beta (3.773, 6.892) (Kittelberger et al. 2009) QLD 0.2 0.5 Beta (2.637, 7.548) (Cooper et al. 2011) VIC 0.01 0.2 Beta (1.136, 14.521) (Hore & Kovesdy 1972) NZa 0.00 0.001 Beta (1,500) (OIE 2018) 0.00 0.01 Beta (1, 458.211)

Key: IFA, indirect immunofluorescence assay; ELISA, Enzyme-linked immunosorbent assay; NSW, New South Wales; QLD, Queensland; VIC, Victoria; NZ, New Zealand; DSe, diagnostic sensitivity; DSp, diagnostic specificity; a A mixture distribution with two priors was used to allow for the possibility of zero infection prevalence. The first prior modelled the probability of the population being infected (akin to a zero-inflation term), and the second prior modelled the prevalence if the population was infected.

121

Table 5.2 Table of less informed prior specifications used for sensitivity analysis

Vague prior specifications

Test specifications

mode 5th percentile parameters for Beta distribution ELISA DSe 0.75 0.25 (2.659,1.553) ELISA DSp 0.75 0.25 (2.659,1.553) IFA DSe 0.75 0.25 (2.659,1.553) IFA DSp 0.75 0.25 (2.659,1.553)

Population

prevalence

mode 95th percentile 99th percentile parameters for Beta distribution NSW 0.32 0.80 (1.616, 2.309) QLD 0.2 0.70 (1.512, 3.048) VIC 0.01 0.30 (1.079, 8.821) NZa 0.00 0.001 (1,500) 0.00 0.001 (1, 458.211)

KEY: IFA, indirect immunofluorescence assay; ELISA, Enzyme-linked immunosorbent assay; NSW, New South Wales; QLD, Queensland; VIC, Victoria; NZ, New Zealand; Se, sensitivity; Sp, specificity; a A mixture distribution with two priors was used to allow for zero infection prevalence. The first prior modelled the probability of the population being infected, and the second prior modelled the prevalence if the population is infected.

122

5.3. Results

5.3.1. Optimisation of the IFA The optimal dilution of anti-bovine IgG FITC conjugate for was found to be 1:600 with 0.05% Evans blue counterstain. At a serum dilution > 1:160 with 2% casein buffer, no fluorescence was observed from the 47 known-negative samples. Twenty-six of the 47 (57%) negative samples did show fluorescence at the 1:40 serum dilution. At the 1:80 dilution, 12 (26%) of these samples showed fluorescence against phase I, and 17 (36%) against phase II, as shown in (Figure 5.3). A serum dilution of 1:160 was therefore considered optimal, this eliminated false positive results due to background fluorescence.

5.3.2. Analytical test performance of the IFA Titration results of the proficiency panel sera for phase I and phase II IgG against C. burnetii are presented in Table 5.3. Sample titrations for all 10 sera were in agreement with the reference laboratory within +/- one titre. When considered dichotomised using the 1:160 dilution cut-off to categorise samples as positive or negative, results were in complete agreement.

Table 5.3 Indirect immunofluorescence assay (IFA) end point titre results for IgG antibodies against C. burnetii for a proficiency panel of 10 bovine sera across two laboratories.

Proficiency IFA Titre IFA Titre

panel Phase I IgG Phase II IgG

Internal Reference Internal Reference Sample number laboratory laboratory laboratory laboratory

1 0 0 0 0 2 0 0 0 0 3 0 0 0 0 4 *40 *0 *40 *0 5 *640 *320 160 160 7 *640 *320 *2560 *1280 10 640 640 2560 2560 8 1280 1280 1280 1280 6 1280 1280 2560 2560 9 2560 2560 > 5120 > 5120

Key: * Indicates where a difference of + one titre was identified between laboratories.

123

100%

80%

60%

Phase 1 40% Phase 2

20% detectable fluorescence Proportin of samples with

0% 1:40 1:80 1:160 1:320 1:640 1:1280 1:2560 Serum dilution

Figure 5.3 Serial dilution of 47 known negative bovine sera screened for antibodies against C. burnetii IgG phase I (P1) and phase II (P2) indicated that background fluorescence was visible using the IFA at serum dilutions of 1/40 and 1/80.

In total, 477 individual IFA wells were read by two independent readers within the one laboratory to assess the subjectivity of operator readings. Overall observed agreement between the two independent readers was 97.9% (K = 0.91; 95% Confidence Interval (CI) 0.83, 1.00), representing almost perfect agreement beyond chance. When adjusted for prevalence and bias, the agreement beyond chance increased with a PABAK = 0.95 (95% CI 0.92, 1.00).

5.3.3. Comparison of phase variation on IFA slides for screening sera Samples from 146 cattle were screened in duplicate on slides coated with combined phase I and II antigen, and in duplicate on slides coated separately with individual phase I and phase II antigens. Screening sera results obtained from both groups of slides were compared. Analysis showed excellent overall agreement with 95.2%; agreement beyond chance represented by ĸ = 0.85 (95% CI: 0.74, 0.96) and PABAK = 0.90 (95% CI: 0.88, 0.98). It was therefore concluded that the combined antigen coated slides were highly comparable with separate phase antigen prepared slides to detect anti-bovine C. burnetii IgG.

124

5.3.4. Comparison of the IFA and ELISA Youden’s index was highest for the IFA at the 1:160 dilution (Figure 5.4), confirming this to be the optimal global cut-off dilution for a positive serum sample. A comparison of test results attained with the developed IFA and the commercial ELISA across four cattle populations, are presented in Table 5.4.

Figure 5.4 Estimates from the Bayesian latent class model for the diagnostic sensitivity and diagnostic specificity of the IFA at different cut-off titres of 1:80, 1:160, 1:320, 1:640, 1:1280. Solid black line represents diagnostic sensitivity, dashed line represents diagnostic specificity and shaded areas represent 95% credible intervals. Red reference line at 160 indicates the cut-off with the highest Youden’s index (J).

Where J = maxc {Se (c) + Sp (c) − 1}

At a cutoff of 1:160, Bayesian estimates of the DSe of the IFA (73.6%; 95% CI 61.1, 85.9) was lower than the ELISA (87.9%; 95% CI 73.9, 96.4) in detecting IgG phase I and/or phase II antibodies to C. burnetii in bovine serum. The DSp of the IFA and ELISA were both very high, 98.2% (95% CI 95.1, 99.7) and 97.7% (95% CI 93.2, 99.7), respectively. Plots of prior and posterior distributions for DSe and DSp of the ELISA and IFA tests are provided in 125

Figure 5.5. Table 5.5 shows a summary of the median and 95% CrI posterior estimates of test performance variables and the estimated true prevalence within the four cattle populations. Posterior estimates of the correlation coefficient (ρ) of test outcomes for positive samples was ρ+ = 0.287 (95% CrI 0.008, 0.701) and for negative samples was ρ- = -0.013 (95% CrI -0.199, 0.353). Model diagnostics were all satisfactory. Results from the sensitivity analysis demonstrated that the prior estimate of the IFA DSe appears to have the most influence on the model. When the model was re-run using less informed priors (Table 5.2) the posterior estimate for the IFA DSe was reduced from 73.6% to 66.5% (Table 5.6).

Figure 5.5 Prior and posterior distributions for diagnostic sensitivity and diagnostic specificity of enzyme-linked immunosorbent assay (ELISA) and indirect immunofluorescence assay (IFA) for the detection of IgG antibodies against C. burnetii in bovine serum; dotted line represents the prior distribution, solid line represents the posterior distribution of each variable.

126

Table 5.4 Cross classified test results from the IFA (1:160 cut-off) and ELISA (IDEXX) across four cattle populations

Cattle populations

Population 1 Population 2 Population 3 Population 4 (n=156) (n=159) (n=96) (n=47) ELISA+ ELISA - ELISA + ELISA - ELISA + ELISA - ELISA + ELISA - IFA + 31 2 16 7 7 0 0 0 IFA - 16 107 6 130 2 87 0 47 Key: IFA, indirect immunofluorescence assay; ELISA, Enzyme-linked immunosorbent assay; +, test positive; -, test negative; Population 1, sera samples from New South Wales dairy cattle; Population 2, sera samples from Queensland beef cattle; Population 3, sera samples from Victorian dairy cattle; Population 4, negative control sera from New Zealand

Table 5.5 Bayesian estimates of the diagnostic sensitivity and specificity for the IFA and ELISA and estimated true prevalence for the four cattle populations for the detection of IgG antibodies against C. burnetii in bovine serum

Sensitivity Specificity True prevalence (%) Test Population 1 Population 2 Population 3 (95% CrI) (95% CrI) (95% CrI) (95% CrI) (95% CrI) 0.736 0.982 IFA* (0.611, 0.859) (0.951, 0.997) 30.3 15.8 7.5 (22.0, 39.3) (9.4, 23.4) (2.2, 14.0) ELISA 0.879 0.977 (IDEXX) (0.739, 0.964) (0.932, 0.997)

Key: IFA, indirect immunofluorescence assay; ELISA, enzyme-linked immunosorbent assay; CrI, Credible Interval; Population 1, represent sera samples from a New South Wales dairy herd; Population 2, represent sera samples from >1 beef cattle herds in Queensland and insufficient data was available for multilevel estimation of herd- and animal-level true prevalence; Population 3, represent sera samples from a Victorian dairy herd; * IFA results taken using the 1:160 dilution cut-off; point estimates are the posterior medians from the model output

127

Table 5.6 Results from sensitivity analysis: Bayesian estimates of diagnostic sensitivity and specificity for the IFA and ELISA for the detection of IgG antibodies against C. burnetii in bovine serum and true prevalence estimates for three cattle populations, using less informative priors.

Sensitivity Specificity True prevalence (%) Test Population 1 Population 2 Population 3 (95% CrI) (95% CrI) (95% CrI) (95% CrI) (95% CrI)

0.665 0.973 31.9 15.7 7.7 IFA* (0.425, 0.799) (0.928, 0.994) (22.4, 51.1) (7.9, 27.7) (1.5, 16.4)

ELISA 0.861 0.977

(IDEXX) (0.540, 0.981) (0.928, 0.994)

KEY: IFA, indirect immunofluorescence assay; ELISA, enzyme-linked immunosorbent assay; CrI, Credible Interval; Population 1, represent sera samples from a New South Wales dairy herd; Population 2, represent sera samples from >1 beef cattle herds in Queensland and insufficient data was available for multilevel estimation of herd- and animal-level true prevalence; Population 3, represent sera samples from a Victorian dairy herd; * IFA results taken using the 1:160 dilution cut-off; point estimates are the posterior medians from the model fitted with less-informed priors as stated in Table S1.

128

5.4. Discussion and conclusions

In this study, an IFA, which is the reference diagnostic test for human Q fever serology, was modified and evaluated for the detection of IgG antibodies against C. burnetii in cattle sera alongside an existing commercially available ELISA. Cut-off values for the IFA were determined during the test optimisation process and the DSe and DSp at this optimal cut-off value was then estimated. This study reports the diagnostic test performance of the IFA and a commercial (IDEXX) ELISA, for the detection of C. burnetii IgG antibodies in cattle sera. Knowledge of diagnostic test performance enables correct interpretation of serological test results by allowing them to be adjusted appropriately to account for diagnostic test imperfections.

While the OIE (World Organisation for Animal Health) Q fever guidelines acknowledge the ELISA and IFA as both suitable tests for herd level prevalence studies, many laboratories continue to use the CFT despite consistent reports in the literature of low DSe in both cattle and goats (ranging from 29.8% to 36.7%) (Horigan et al. 2011; Kittelberger et al. 2009; OIE 2018; Muleme et al. 2016). There is no lack of analytical specifications of test performance and standardised serological methods for C. burnetii serology in ruminants. Although, prior to this study, there were few diagnostic specifications that could be validly utilised to reliably adjust apparent prevalence estimates or inform screening and surveillance activities in cattle in the Australian context.

The IFA sera dilution cut-off point (1:160) determined using positive and negative controls was in agreement with a recent study that validated an IFA for use in goats (Muleme et al. 2016). Diagnostic characterisation of the assay at multiple cut-off points indicated the highest Youden’s index at the same cut-off, indicating this to have the highest combined DSe and DSp. These results suggest that the IFA, modified to detect bovine IgG antibodies against C. burnetii, may not be reliable for serum dilutions below 1:160. At this screening cut-off, it was determined that the estimated DSe of the IFA was lower than the ELISA in detecting IgG against phase I and/or phase II C. burnetii in bovine serum; however, DSp was comparable between the tests.

At the time of this study, no published literature was available reporting the DSe and DSp of an IFA for the detection of anti- C. burnetii immunoglobulins specifically in cattle. There are publications that report crude seroprevalence results in cattle using IFA methods, however the DSe and DSp of the tests were not evaluated or reported (Lyoo et al., 2017; Vaidya et al., 2010). In fact, in one of the studies the IFA was assumed to have 100% DSe

129 and 100% DSp and is used as a reference test to compare other diagnostic methods (ELISA and PCR) (Vaidya et al. 2010). Such an approach introduces bias and leads to inaccurate interpretations of test parameters and prevalence estimates.

Several publications have performed test comparisons using different ELISA kits; results for test DSe and DSp were similar to findings from this study. In Denmark, Bayesian latent class analysis was used to analyse the diagnostic performance of the IDEXX ELISA kit for use in bovine milk and blood samples; the ELISA, based on serum samples, was reported to have a DSe of 84.0% and DSp 99.0% (Paul et al. 2013). Horigan et al. (2011) also investigated the diagnostic accuracy of three ELISA tests for use in cattle serum, comparing multiple populations with the ELISAs and a CFT. They report the DSe and DSp of an unspecified commercial ELISA to be 81.3% and 87.4%, respectively (Horigan et al. 2011). In another study, the DSes of two ELISAs for cattle sero-diagnosis were estimated to be significantly higher than CFT; ELISA-1 97%, ELISA-2 97% and CFT 34%; although the specificity of the CFT was superior to both ELISAs (Lucchese et al. 2016).

Rousset et al. (2007) published similar results in goats; with an ELISA and an IFA having overall good agreement for detecting C. burnetii antibodies in goat sera, however the ELISA was able to detect more positives than the IFA. Unfortunately, that study only looked at crude comparisons of the two tests and did not attempt to estimate the test performance of either assay against a gold standard or by using alternative methods such as latent class analysis.

Bayesian latent class methods were used in this study to validate the modified IFA and the commercial ELISA, by estimating the DSe and DSp of the tests. In the absence of a gold standard reference test, results from multiple tests can be used to evaluate test performances (Toft, Jørgensen & Højsgaard 2005; Tu, Kowalski & Jia 1999). This method of test validation has been applied increasingly in the last decade in veterinary diagnostics and is recommended by the OIE when insufficient reference samples are available that are representative of the population where the test is intended to be used (Horigan et al. 2011; Paul et al. 2014; OIE 2018). Latent class analysis was similarly used to estimate the DSes and DSps of two ELISA tests and a CFT for the sero-diagnosis of coxiellosis in cattle, confirming other reports that ELISAs have much higher DSes than the CFT (Lucchese et al. 2016).

Prior estimates of IFA and ELISA test sensitivity and specificity were based on the best available published literature. After performing sensitivity analyses with less informed

130 priors, it became apparent that the posterior DSe estimate for the IFA was most influenced by the high prior IFA DSe estimate. The informed priors came from a validation study based on goats (Muleme et al. 2016). It seems plausible that the IFA test specifications for use in goats may not be applicable for cattle. This may be important when considering serological tests for use in ruminants; species-specific test validation is not always reported. As discussed in the standard reporting guidelines for Bayesian latent class models (STARD- BLCM), the final model estimates were derived from the models using informed priors that were defined prior to the analysis and sensitivity analysis was performed (Kostoulas et al. 2017). The results from the sensitivity analysis were included in Table 5.6.

The model outputs of true prevalence estimates presented in this study represent crude test results adjusted for the estimated DSes and DSps of the tests. However, insufficient data were available for multi-level estimation of herd- and animal-level true prevalence in populations that had more than one herd. It should be noted that due to the non-representative structure of sampling from some of the study populations, the external validity of the true prevalence estimates is limited. Therefore, the true prevalence estimates presented here are not intended to be used for inference of regional or state-wide prevalence.

An additional limitation of the current study is that the IFA only detects IgG and needs to be modified and validated to detect IgM antibodies, as previously undertaken in validation for goats (Muleme et al. 2016), therefore current or very recent infections may be underreported. A further study, to validate the IFA for detection of anti- C. burnetii IgM in cattle sera, is warranted.

While these results found the IFA to be less sensitive than the ELISA, it may have other benefits that make this serological method appealing for research or surveillance. Firstly, when testing large numbers of cattle sera, a validated in-house IFA such as described here, has lower consumable costs than the commercial IDEXX ELISA kit. Crude budget calculations at the time of the study were performed and it was estimated that the consumable costs for testing one serum sample (in duplicate) with this ELISA was AUD$8.30 (Australian dollars) compared to AUD$0.70 per sample with the in-house IFA test. However, the IFA method is more labour intensive (labour costs were not included in consumable price above) and does not currently have an automated reading method. It has been demonstrated in this study, that screening sera on IFA slides coated with a combined preparation of phase I and phase II antigens can accurately detect bovine IgG antibodies,

131 further reducing costs. If positive at screening, samples can then be tested on separate, individual antigen coated slides to determine phase-specific antibody titres.

For this Bayesian latent class analysis, the DSe and DSp of the IFA was estimated using combined test results from separate phase I and phase II IFA slides. At present, the IDEXX ELISA cannot distinguish between different phase-specific antibody responses as it is pre-coated with a combined antigen preparation of phase I and phase II. The separated phase I and phase II IFA may therefore be a useful tool for investigating phase-specific serological patterns in relation to disease status in cattle with the potential to identify chronically shedding animals (Lucchese et al. 2015). Phase specific serological patterns have long been used to aid in the diagnosis and interpretation of different stages of infection in humans (Eldin et al. 2016).

As per the OIE (World Organisation for Animal Health) guidelines, this IFA is suitable for the purpose of estimating prevalence of C. burnetii exposure in cattle, for further analysis to identify risk factors and possibly as a confirmatory test (OIE 2016b). This IFA may not be appropriate for testing cattle herds for freedom of disease, as the sensitivity may be sufficiently low to result in unacceptable risk of false negatives (Thrusfield 2007). Overall, having knowledge of the estimated test performance of each serological method enables accurate interpretation of test results, which reduces bias and misleading conclusions to be drawn from either under or over estimation of infection.

132

5.5. Chapter five appendix

R code for IFA/ELISA Latent Class Model

------

# Latent class modelling, iteratively for each cut-off in a vector of potentials

cutoffs<-c(80,160,320,640,1280) #note 1280 is >=1280

#create list to store outputs

jfit.list<-list()

for(i in 1:length(cutoffs))

{

### SETTING CUTOFF for test2 and making apparent 2x2s ###

cutoff<-cutoffs[i]

# test-pop

test11<-dat$elisa[dat$pop==1]

test21<-(dat$ifa.igg[dat$pop==1]>=cutoff)*1

test12<-dat$elisa[dat$pop==2]

test22<-(dat$ifa.igg[dat$pop==2]>=cutoff)*1

test13<-dat$elisa[dat$pop==3]

test23<-(dat$ifa.igg[dat$pop==3]>=cutoff)*1

test14<-dat$elisa[dat$pop==4]

test24<-(dat$ifa.igg[dat$pop==4]>=cutoff)*1

#check

#cbind(test22, dat$ifa.igg[dat$pop==2])

# load data (APPARENT 2 X 2)

133

#following Branscum (2005), for each pop y[1:2] are

#y[1] +,+

#y[2] +,-

#y[3] -,+

#y[4] -,- y1<-y2<-y3<-y4<-rep (0, 4) y1<-c(sum((test11==1 & test21==1)*1, na.rm=T),

sum((test11==1 & test21==0)*1, na.rm=T),

sum((test11==0 & test21==1)*1, na.rm=T),

sum((test11==0 & test21==0)*1, na.rm=T)) y2<-c(sum((test12==1 & test22==1)*1, na.rm=T),

sum((test12==1 & test22==0)*1, na.rm=T),

sum((test12==0 & test22==1)*1, na.rm=T),

sum((test12==0 & test22==0)*1, na.rm=T)) y3<-c(sum((test13==1 & test23==1)*1, na.rm=T),

sum((test13==1 & test23==0)*1, na.rm=T),

sum((test13==0 & test23==1)*1, na.rm=T),

sum((test13==0 & test23==0)*1, na.rm=T)) y4<-c(sum((test14==1 & test24==1)*1, na.rm=T),

sum((test14==1 & test24==0)*1, na.rm=T),

sum((test14==0 & test24==1)*1, na.rm=T),

sum((test14==0 & test24==0)*1, na.rm=T)) n1=sum(y1); n2=sum(y2);n3=sum(y3); n4=sum(y4)

134

j.data <- list("n1","n2","n3","n4","y1","y2","y3","y4")

j.parameters.to.monitor <- c("se1", "se2", "sp1", "sp2","rhopos", "rhoneg", "pi1","pi2","pi3","pi4","youden")

#=initialising values 2 chains======#

j.inits<-list(

list(pi1=0.2, pi2=0.5, pi3=0.01, prev4=0.1, tau4=0.05, se1=0.60, sp1=0.60, lambdase2=0.60, gammase2=0.60, lambdasp2=0.60, gammasp2=0.60,.RNG.name="base::Super-Duper", .RNG.seed=14),

list(pi1=0.5, pi2=0.2, pi3=0.05, prev4=0.05, tau4=0.01, se1=0.80, sp1=0.80, lambdase2=0.80, gammase2=0.80, lambdasp2=0.80, gammasp2=0.80, .RNG.name="base::Super-Duper", .RNG.seed=14)

)

### run model ########

jfit <- jags(data=j.data, inits=j.inits, parameters.to.save=j.parameters.to.monitor,

n.chains=2, n.iter=101000, n.thin=1, n.burnin=1000, model.file="model_2tests_4pop_flat.txt")

jfit.list[[i]]<-jfit

cat("iteration ", i, " out of ", length(cutoffs), "\n", sep=""); flush.console() # print progress of a loop

}

135

The following chapter was published as a peer-reviewed original article in Preventive Veterinary Medicine as follows:

136

Chapter Six

Prevalence and spatial distribution of Coxiella burnetii seropositivity in northern Australian beef cattle adjusted for diagnostic test uncertainty

“More is missed by not looking than by not knowing.”

- Thomas McCrae (1870 – 1935)

137

6. Prevalence and spatial distribution of Coxiella burnetii exposure in northern Australian beef cattle adjusted for diagnostic test uncertainty

6.1. Introduction

Coxiella burnetii has been recognised globally as an important zoonotic infection by both human and animal health authorities (OIE 2018; Eldin et al. 2016). Coxiellosis in cattle is often described as subclinical, however there is evidence of associated reproductive problems such as sporadic abortion, premature birth and birth of weak calves (Agerholm 2013). Human Q fever can occur as both acute and chronic forms with a range of clinical symptoms, and may initially be misdiagnosed as influenza (Eastwood et al. 2018). Persistent focalised C. burnetii infections can result in chronic hepatitis, gestational complications, paediatric osteomyelitis and endocarditis (Parker, Barralet & Bell 2006; Eldin et al. 2016).

In Australia, Q fever is a notifiable disease in humans with national case notifications that fluctuated between 317 to 868 cases per year between 1991 and 2018 (Australian Government 2019). Australia consistently reports high annual notifications rates of Q fever, which are likely an underestimation of the true disease incidence (Tozer 2015; Gidding et al. 2020). When comparing the 10-year average annual Q fever notification rates from 2009 to 2018, stratified by states and territories, there were marked differences between geographic regions of Australia. The state of Queensland reported an average 10-year rate of 4.3 cases per 100 000 population per year, New South Wales 2.5 and the remaining states reported rates of less than 1 case per 100 000 population per year (Australian Government 2019). When comparing the states and territories of Australia, Queensland consistently reported the highest annual case notification rates over the last 20 years.

Although Q fever is listed by the OIE (World Organisation for Animal Health) as an important animal infection, it is not a nationally notifiable animal disease in Australia. Without current monitoring or surveillance of this infection in ruminants and with minimal incentive for research, the true prevalence and distribution of coxiellosis in Australian ruminant populations is unknown.

Beef cattle production is important to Australia’s economy and represented approximately 20% of the value of agricultural production in 2018–19, with a gross value of $AUD19.6 billion for cattle and calf production including live cattle exports (‘Australian Bureau of Statistics, Australian Government’ 2019). The northern Australian beef industry

138 accounts for about 61% of the total Australian beef cattle herd. Approximately 42% of the total Australian cattle population in Queensland (10.5 million head), 9% (2.2 million head) in the Northern Territory and 3% (0.8 million head) in Western Australia (north of the Tropic of Capricorn; Meat and Livestock Australia 2019).

Quantitative data on coxiellosis in cattle in Australia is limited. Between 1954 and 2018, southern states of Australia reported an animal-level prevalence of C. burnetii exposure in cattle of less than 1% (Tan 2018; Cronin 2015; Hore & Kovesdy 1972; Forbes, Wannan & Keast 1954). Only two publications were identified that reported C. burnetii prevalence in cattle in northern Australia; using complement fixation testing, cows (n = 330) from 11 dairy herds in north Queensland were found to have 0.0% seropositive (Pitt 1997). However, testing of beef cattle from a sample of properties across Queensland using an in- house enzyme-linked immunosorbent assay (ELISA,) found C. burnetii IgG exposure of 16.8% (95% CI 16.7, 16.8 %; n = 1 835) at the animal level, with 78.5% of properties having at least one positive animal (Cooper et al., 2011). This estimate was higher than other Australian studies and higher than reports for beef cattle internationally. However, the diagnostic sensitivity and specificity of the ELISA test used was not determined, and thus what has been reported is the apparent prevalence of C. burnetii exposure. As most serological test methods are imperfect, test results should be adjusted to estimate true rather than apparent prevalence to avoid biased and overly confident estimates (Dohoo, S. W. Martin & Stryhn 2009).

6.1.1. Chapter objectives The objectives of this study were to investigate the prevalence of C. burnetii exposure in beef cattle breeding herds across northern Australia and estimate the true prevalence and spatial distribution at that time point. Based on differences in human notification rates between regions, it was hypothesised that there may be an associated regional difference in cattle C. burnetii exposure between cattle farmed in the Northern Territory and Queensland.

6.2. Materials and Methods

6.2.1. Ethics statements Animal ethics for this study were approved by the University of Queensland Animal Ethics Committee: SVS/115/11/MLA (NF) and ANRFA/SVS/100/16.

139

6.2.2. Study design A cross-sectional study design, using previously collected samples, was used to estimate the prevalence of C. burnetii exposure in a sample of beef cows from properties in Queensland and the Northern Territory, Australia. A total of 2 012 sera samples, collected between December 2010 and December 2011, were tested for serological evidence of C. burnetii exposure. Blood samples were collected as part of a Meat and Livestock Australia-funded epidemiological study investigating factors affecting the reproductive performance of beef cattle in northern Australia: Northern Australian Beef Fertility project: CashCow (McGowan et al. 2014).

6.2.3. Property selection criteria and blood sampling protocol The target population for the CashCow project was all commercial beef breeding properties within Queensland, Northern Territory and northern Western Australia. The source population for the project included all beef breeding properties serviced by 31 cattle veterinarians located across the region. Study sites were identified using a convenience sampling approach, selected on the basis that owners/managers were prepared to muster enrolled cattle at least twice a year and then stratified to ensure all major beef breeding regions of northern Australia were adequately represented (McGowan et al. 2014). Consenting properties were enrolled for a 3–4 year period from 2008 to 2011. Seventy-eight properties with approximately 56 000 electronically identified breeding females managed in approximately 165 mobs were enrolled in the study. For many properties, a mob of mated heifers and a mob of mated cows were initially enrolled in the project. A mob was defined as a discrete management group of cattle.

Blood samples were collected from cattle to determine the impact of infectious diseases that may affect reproduction. A systematic sampling method was used to select individual animals within mobs for collection of blood samples at the time of mustering (McGowan et al. 2014). Between 15 and 30 blood samples were collected from each mob. For example if the mob size was 150 then a sample was collected from every 10th female as they presented in the crush. Blood samples were collected by coccygeal venepuncture, into 9ml plain collection tubes. Sera were decanted approximately 24 hours after collection and either frozen or shipped chilled to the University of Queensland for storage. Samples were stored frozen at -20 oC until laboratory testing. During the CashCow project, serum samples underwent a maximum of three freeze/thaw cycles prior to C. burnetii testing in 2017.

140

Figure 6.1 Geographical distribution of participating cattle properties from northern Australia (n = 60) with serum samples tested for C. burnetii IgG exposure.

141

For this current study, all remaining serum samples from the 2011 blood sampling, having sufficient volume to perform laboratory testing, were utilised. Of the original 78 properties enrolled, only 60 properties were included in this study. Figure 6.1 shows the geographical locations of the 60 properties included in this study.

6.2.4. Serological methods The sera were tested for IgG antibodies against C. burnetii using an indirect immunofluorescent assay (IFA) modified and validated for use in cattle as described in Chapter 5 (Wood et al. 2019). The IFA slides were produced in-house, coated with C. burnetii phase I and/or phase II antigen as previously described, with best estimates of the diagnostic sensitivity (DSe) and specificity (DSp) of this assay being, 73.6% and 98.2%, respectively (Wood et al. 2019). All test sera were initially screened at a dilution of 1:160, on slides with a combined phase I and phase II antigen coating. Test sera that tested positive at this screening dilution were then titered out to endpoint on separate phase I and phase II slides. Sera were considered positive if they showed reactivity to either phase I or phase II antigen at the 1:160 cut-off dilution.

Briefly, IFA slides, test sera and reagents were brought to room temperature before use. Serum samples were diluted with 2% casein- phosphate-buffered saline (PBS) to minimise non-specific binding. Diluted test sera were placed on the slides in duplicate and incubated in a humidity chamber for a 30 min incubation period at 37 oC; if the serum contained specific C. burnetii antibodies, they adhered to the antigen during this time. The slides were washed in 10% PBS for 5 min, three times and allowed to air dry. Anti-bovine IgG-FITC conjugate, diluted in 0.05% Evans blue dye, was added and incubated in a humidity chamber for 30 min at 37 oC. If IgG antibody-antigen complexes were adhered on the slide, the FITC conjugate would attach to it during this incubation period. The slides were again washed in 10% PBS for 5 min, three times and allowed to air dry. Coverslips were mounted to slides with mounting media. An immunofluorescent microscope (Nikon Eclipse E600) was used to view the slides at 40x magnification and then with oil immersion at 100x magnification. If the test serum contained IgG antibodies against C. burnetii, there was an apple green fluorescence indicating a positive result.

6.2.5. Statistical analysis

6.2.5.1. Apparent prevalence Crude diagnostic test results were reported for the serological testing as an apparent prevalence (AP); equal to the percentage of test-positive animals from the total samples

142 tested. Results were calculated at the individual animal level for state, region and properties with Wilson-Score 95% Confidence Intervals (CI). Property-level AP was also presented with 95% Wilson-Score CI; a property was considered positive if one or more individual animals on the property were test positive.

6.2.5.2. Bayesian latent class models Bayesian hierarchical latent class modelling (LCM) was utilised to estimate the true prevalence (TP) of C. burnetii exposure based on this sample of beef cattle within properties

(TPproperty) and within regions (TPregion), with the unit of interest being the individual animal. The test samples used in this study had a hierarchical data structure, with cattle clustered in properties and properties clustered within regions. Due to this complex data structure, statistical approaches accounting for the clustering were required. The true prevalence model was developed to account for the data structure and included random effects for region and for properties nested in the regions. The IFA serological test is not a perfect test; therefore the DSe and DSp of the IFA were also accounted for in the model using the formula to estimate TP as a function of AP:

𝐴𝑃 𝐷𝑆𝑒 𝑇𝑃 1 𝑇𝑃1 𝐷𝑆𝑝

th Where, APij is the apparent prevalence, i = 1, ..., 60 is the i property and j = 1, ..., 5 is the th j region. Let yij represent counts of positive sera for property i in region j, then it can be shown that the counts follow a binomial distribution as: yij~dbinom (APij,nij)

Random effects for properties (uij) nested in regions (i.e., wj ) and the regions themselves

(wj) were included in the model to address the data structure of the sampling method in this study. Hence, the true prevalence (TPij) could be expressed with a logit link as follows:

logitTPij𝛽 𝑤 𝑢

th th where TPij is the true prevalence for the i property in the j region, uij and wj represent the prevalence difference of property and region from the overall prevalence and:

β0 ~ dnorm (0, 0.1)

uij ~ dnorm(wj, 1/𝜎 u)

wj ~ dnorm(0, 1/𝜎 w )

Uniform hyper-priors were assumed for the region and property random effects, i.e.,

143

𝜎u ~ dunif(0,1)

𝜎w ~ dunif(0,1)

The true prevalence of C. burnetii exposure in each property within each region were inferred according to the random effects as follows:

expβ 𝑤 ∗𝑟𝑒𝑔𝑖𝑜𝑛 𝑢 ∗𝑝𝑟𝑜𝑝𝑒𝑟𝑡𝑦 𝜋 1 expβ 𝑤 ∗𝑟𝑒𝑔𝑖𝑜𝑛 𝑢 ∗ 𝑝𝑟𝑜𝑝𝑒𝑟𝑡𝑦

The true prevalence of C. burnetii exposure in each region were calculated as follows:

expβ 𝑤 ∗ region 𝜋 1 expβ 𝑤 ∗ region

Informed priors for DSe and DSp of the IFA were incorporated into the model using unimodal beta distributions based on published diagnostic test parameters (Wood et al. 2019). Unimodal beta distributions were estimated using the “epi.betabuster” function implemented within the “epiR” library (Stevenson 2017) in R (‘R: A language and environment for statistical computing’ 2019). Sensitivity analysis was performed by re- running the model with alternative priors for IFA DSe and with hyper-priors for the random effects (σu and σw) of dunif(0, b) for b = 1.5, 2 and 2.5 ().

The Bayesian hierarchical true prevalence model was coded in OpenBUGS (version 3.2.3, rev 1012). See supplementary materials for example data and code from these analyses. Bayesian inferences were based on the joint posterior distribution, approximated using the computer software JAGS (version 4.3.0, citation), implemented with R2jags package (Yu-Sung & Masanao 2015) in the R statistical package (‘R: A language and environment for statistical computing’ 2019). This implementation makes use of a Markov Chain Monte Carlo (MCMC) sampling algorithm to acquire Monte Carlo (MC) samples from the posterior distribution (Paul et al. 2014). The MCMC model was initiated with three independent chains and run for 250 000 iterations, burning the first 20 000. The MCMC convergence was assessed by visual inspection of the chains, and Gelman-Rubin plots and test statistics. Autocorrelation and effective sample size were also estimated (Kruschke 2015).

Posterior estimates were used as TPproperty estimates for the 60 cattle properties located within distinct regional areas of northern Australia. TPregion were calculated from model outputs for each of the five geographical regions specified in the model: north Queensland, central Queensland, southern Queensland, south-east Queensland and the

144

Northern Territory. TPregion should be interpreted as the predicted true prevalence of C. burnetii exposure on a typical property selected within that specific region. All true prevalence estimates were reported as the median and 95% Credible Intervals (CrI) of the posterior distribution. The Queensland TPregion estimates were compared with the Northern

Territory TPregion.

For regions with low prevalence estimates (median TPregion < 5%), the likelihood of C. burnetii disease freedom was inferred as the probability of C. burnetii exposure being less than a pre-specified design prevalence (θ). Two values of this design prevalence were tested (5% and 1%) using the Boolean step() function. The step () function created a Boolean value for each simulations in which the TP < θ. The mean value of a Boolean node is a probability; hence the Monte-Carlo estimate of P (TP < θ) represented the probability of the prevalence being less than the specified design prevalence for each region under observation.

6.2.5.3. Spatial visualisation Enrolled beef cattle properties were geo-referenced with latitude and longitude at the time of the original study; each property was then referenced with a code for “region” within Queensland, according to property geo-referenced points from open-access Australian government regional boundary maps. Maps were produced to visualise the distribution of true prevalence model output estimates as point data (TPproperty). All maps were created using software QGIS 2.18.26 (‘QGIS Geographic Information System’ 2019).

6.3. Results

6.3.1. Laboratory IFA testing of bovine samples Approximately 80% (n = 1 602) of the cattle tested were located in Queensland and 20% (n = 410) in the Northern Territory. The median number of sera tested per property was 31 with a range of 4 to 112 (inter-quartile range 17–41). A descriptive summary including a breakdown of the number of properties per region are shown in Table 6.1.

Overall, 5.2% (95% CI 4.3, 6.2 ; n = 104/2 012) of test sera returned a positive result using the IFA test at the 1:160 cut-off for either phase I or phase 2 IgG antibodies against C. burnetii. From the test sera, 4.4% (95% CI 3.6, 5.4; n=89/2 012) were positive for phase 1, 3.5% (95% CI 2.8, 4.4; n=71/2 012) were positive for phase 2 and 2.7% (95% CI 2.1, 3.5; n=55/2 012) were positive for both phase 1 and phase 2.

145

Table 6.1 Descriptive summary of crude indirect immunofluorescent assay (IFA) results for Coxiella burnetii exposure in bovine serum

Animal level Property level

No. of No. of Total no. Percentage Total no. Percentage positive positive positive samples positive (95% CI) properties (95% CI) samples properties

Overall Northern Australia 104 2012 5.2 (3.9, 6.8) 32 60 53.3 (40.9, 65.4)

State Queensland 102 1602 6.4 (4.9, 8.2) 30 49 61.2 (60.1, 61.5)

Northern Territory 2 410 0.5 (0.1, 1.8) 2 11 18.2 (14.6, 22.0)

Region Northern Queensland 41 606 6.8 (5.0, 9.1) 12 20 60.0 (58.8, 60.8) Central Queensland 44 628 7.0 (5.2, 9.2) 10 14 71.4 (69.7, 71.7)

Southern Queensland 16 240 6.7 (4.1, 10.6) 7 9 77.8 (74.9, 77.2) South-east Queensland 1 128 0.8 (0.1, 4.3) 1 6 16.7 (9.9, 23.7)

Northern Territory 2 410 0.5 (0.1, 1.8) 2 11 18.2 (14.6, 22.0)

KEY: CI, Confidence Interval

146

Apparent prevalence of C. burnetii exposure varied between cattle in Queensland and the Northern Territory; 6.4% (95% CI 5.3, 7.7) of individual sera from Queensland cattle were test positive, whereas 0.5% (95% CI 0.1, 1.8) of individual sera from Northern Territory cattle were test positive. From Queensland, 61.2% (95% CI 47.2, 73.6; n= 30/49) of the properties tested, had at least one test positive animal identified. From the Northern Territory, 18.2% (95% CI 5.1, 47.7; n= 2/11) of properties tested, had at least one animal IFA positive.

A descriptive summary of the IFA test results and AP at the animal and property level, stratified by state/territory and geographic regions are shown in Table 6.1. Overall, 53.3% (32/60) of properties had evidence of previous exposure to C. burnetii. From the positive properties, the average within-property AP was 8.7% with a range of 0.9–22.2%.

6.3.2. Bayesian latent class modelling True prevalence estimates from the Bayesian hierarchical LCM are presented as a caterpillar plot (Figure 6.2) and as geo-referenced data points on a map of Queensland and the Northern Territory (Figure 6.3). The caterpillar plot indicates the TPproperty as a median posterior prevalence estimate for each property with 95% CrI. The median

TPproperty estimates within Queensland ranged from 0.1 to 15.9%. For georeferenced data- points on the map, the median estimates are displayed as proportions without CrIs for better visualisation. When mapped, the distribution of Queensland TPproperty did not suggest areas of clustering. However, the 95% credible intervals of these estimates are very wide and the highest upper limit reached 35.5%.

Regional predicted true prevalence estimates indicate that cattle in this study, originating from northern, central and southern Queensland had very similar TPregion; median: 7.2%, 7.5% and 6.8% respectively (Table 6.2), however, the Northern Territory and south-east Queensland cattle had lower TPregion (median 1.1% and 1.9%). Cattle in this study, from the Northern Territory had lower TPregion of C. burnetii exposure than cattle from all regions of Queensland except south-east Queensland. The south-east region of Queensland has the lowest regional estimate for the state. Inspection of Figure 6.3 highlights that all sampled properties in the Northern Territory returned true prevalence estimates <1%. Much more variance in true prevalence is apparent in properties from Queensland.

147

Table 6.2 Predicted regional true prevalence estimates and 95% Credible Intervals, derived from the Bayesian hierarchical latent class model Estimated true Regions 95% Credible Interval prevalence Northern Queensland 7.3% (3.8, 13.6) Central Queensland 7.6% (4.0, 14.2) Southern Queensland 6.8% (2.9, 14.9) South-east Queensland 1.9% (0.4, 6.2) Northern Territory 1.0% (0.3, 3.3) Regional true prevalence estimate can be interpreted as the predicted prevalence of C. burnetii exposure on a typical property within that specific region. The point estimate is the median of the posterior distribution with the 2.5% and 97.5% percentiles presented as 95% credible intervals.

Table 6.3 Posterior probabilities of estimated true prevalence being less than a specified design prevalence

Design prevalence Regions 1% 5% South-east Queensland 0.19 0.94 Northern Territory 0.48 0.99

For samples from the Northern Territory, the probabilities that the posterior estimate of true prevalence is less than design prevalence (5% and 1%) are 0.99 and 0.46, respectively. For samples from south-east Queensland the probabilities are 0.94 and 0.18 for the same design prevalence (5% and 1%), respectively (Table 6.3). Model diagnostics were all satisfactory. Changes in the model outputs from the sensitivity analyses were robust to the use of different priors (Table A. 6-1,Table A. 6-2).

148

Figure 6.2 Caterpillar plot of model estimates for property level true prevalence as a proportion; the point estimate is the median of the posterior distribution of predicted prevalence and the solid line indicates 95% Credible Intervals.

149

Figure 6.3 Map of the Northern Territory and Queensland showing the distribution of within-property true prevalence estimates of exposure to C. burnetii (n=60). Point values are the median posterior true prevalence estimates.

150

6.4. Discussion and conclusion

This study revealed that the apparent prevalence or unadjusted serological exposure of C. burnetii in a large sample of beef cattle from Northern Australia (the Northern Territory and Queensland), using an IFA, validated for use in cattle, was 5.2% (95% CI 4.3, 6.2). When stratified by state/territory, the apparent prevalence of cattle sampled from the Northern Territory was 0.5% (95% CI 0.1, 1.8) and 6.4% (95% CI 5.3, 7.7) for cattle sampled from Queensland. Although the majority of published prevalence studies investigating C. burnetii exposure in cattle have focused on dairy cattle, some international publications included beef cattle. It was reported that beef cattle going to slaughter in Denmark had an AP of 4.5% (95% CI 3.2, 6.3) using a commercial ELISA Q fever kit (Paul et al. 2014). These crude serological test results are similar to reports from beef cattle in other countries with reported APs ranging from 1.7% - 6.6% (Lyoo et al. 2017; Alvarez et al. 2012; McCaughey et al. 2010; Ruiz-Fons et al. 2010).

A Bayesian hierarchical LCM was developed to provide more robust prevalence estimates in order to compare coxiellosis exposure from this sample of cattle across five distinct regions of northern Australia (the Northern Territory, northern Queensland, central Queensland, southern Queensland and south-east Queensland). The model was designed to account for the IFA test DSe and DSp and incorporate the hierarchical structure of the data into the analysis. From the posterior outputs of the final model, the predicted probability of exposure in a typical beef cattle property from the Northern Territory was 1.1% (95% CrI 0.3, 3.2). This was lower than 3 out of the 4 regions within Queensland. However, from the data presented, it cannot be concluded that the Northern Territory cattle are free of disease. Nonetheless, we estimated that there was >95% probability that these cattle have a true prevalence of <5%. A diagnostic test with higher sensitivity and using a study design specific for this purpose would be required in order to contribute towards demonstration of regional freedom of disease.

This study is the first to report any prevalence estimates for C. burnetii exposure in cattle from the Northern Territory. Results provide evidence to support the hypothesis that cattle managed in the Northern Territory have a lower true prevalence of C. burnetii exposure than cattle managed in most regions of Queensland with the exception of south- east Queensland. We noticed from national Q fever surveillance data that human case notifications have historically been different between Queensland and the Northern Territory. Prior to 2002, the Northern Territory had never reported a case of Q fever. From

151

2002 to 2018, the annual notification rates fluctuated from 0 to 2.4 cases/100 000 with an average annual rate of 0.8 cases/100 000 and very low case numbers. The Queensland data from the same time-period shows an average annual notification rate of 4.7 cases/100 000; with significantly higher notification rates prior to the National Australian Q Fever Management Program (Palmer et al. 2007). Although this pattern is interesting, it does not imply causality between cattle exposure and human Q fever cases. As C. burnetii is known to survive in the environment and is able to transmit between many reservoir animals, this pattern may suggest less C. burnetii in the general environment in the Northern Territory, or that the conditions do not favour the persistence and spread of the bacterium. In this respect, cattle exposure may function as a sentinel marker for C. burnetii within this regional ecosystem.

For this current study, there were no samples available for IFA testing from northern Australian beef cattle properties based in Western Australia. Although, one previous publication has reported the AP of C. burnetii from beef cattle in Western Australia, the specific geographical location of sampling within the state was not specified (Banazis et al., 2010). Cattle in that study (n = 329) had a seroprevalence of 0.6% using an IDEXX CHEKiT Q Fever ELISA (IDEXX Laboratories Inc., Switzerland) and 7.9% tested PCR positive from urine and faecal samples.

Within the cattle sampled from Queensland, the predicted prevalence of C. burnetii exposure in a typical property within three broad regions: northern, central, southern, was similar. Cattle from the south-east Queensland region had the lowest TPregion. However, this region also had the smallest number of properties enrolled and small sample sizes within properties. Which may be indicative of less farming in the region and smaller property sizes.

In general, there was a lot of variation in TPfarm estimates for Queensland cattle farms. It should also be mentioned that in general most beef farming across these regions follow an all-year round calving pattern, hence there is not the same seasonal pattern of pregnancy and calving as there would be in other more temperate areas that could influence the transmission of coxiellosis. Although beyond the scope of this analysis, the influence of farming practices and seasonal patterns on regional prevalence could provide insights into putative protective factors from coxiellosis. An in-depth analysis of risk factors for coxiellosis in beef cattle of northern Australia would be beneficial to explore in further research.

Within Australia, only one preliminary study from Victoria investigated the regional TP of C. burnetii exposure in beef and dairy cattle taking into account the performance of the diagnostic test used (Tan 2018). From Goulburn Valley, Victoria, cattle had a TP 0.0% (95% 152

CrI 0.0, 0.0%; n = 278) and from Gippsland, Victoria 0.4% (95% CrI 0.0, 3.5%; n = 247). True prevalence estimates were calculated using similar Bayesian methods as described here, however without the additional multilevel modelling to account for clustering of animals within properties. The TP results reported from Victorian cattle was very similar to the cattle from the Northern Territory and south-east Queensland in this study. Most Queensland regions, however, had higher results than Victorian regions. Therefore, cattle in Queensland have a higher level of exposure to C. burnetii than cattle in southern states as suggested from a previous study (Cooper et al. 2011).

The statistical model presented in this study incorporates both imperfections of the diagnostic test used and the hierarchical data structure of the cattle populations into the final TP estimates. The structure of this model may be useful to provide assistance with infectious disease prevalence estimates in future studies. It could also be utilised during passive surveillance of C. burnetii (or other infections) within animal populations, thus enabling a more accurate interpretation of serological test results from serum banks or samples collected for other purposes. Statistical methods applied in this study may enable improved comparisons between C. burnetii prevalence estimates both within and between regions / countries and allow further analysis into putative risk factors of C. burnetii in cattle in Australia.

There are several limitations to the present study. Firstly, it should be noted that the samples tested during this study might not be representative of the broader regions of northern Australia due to the structure of sampling; therefore the external validity of the true prevalence estimates may be limited. Although the properties enrolled in the original CashCow study were stratified to ensure a balanced sample across major beef-cattle breeding regions, it did not constitute a random sampling method from a complete sample frame. There may be selection bias, as the properties enrolled in the original study may represent the better-managed properties from northern Australia. Secondly, only serum samples from 60 of the 78 properties were available to be tested for C. burnetii exposure. For the final analysis, the clustering of mobs within properties was not incorporated into the model, as not all properties had multiple mobs; however, some properties had several mobs. Hence, all results were aggregated at the property level. Although we do not think these factors would bias the C. burnetii prevalence results in any particular direction, these factors are acknowledged to ensure awareness of potential limitations with respect to any inferences made from the true prevalence estimates.

153

We also acknowledge that IgG antibody exposure may not be a reliable indicator of active C. burnetii infection in cattle. Although serology can be useful to detect past exposure or recent infection, it is not known exactly how long C. burnetii immunoglobulins remain detectable in serum of cattle (Natale et al. 2012). There are recent publications investigating different phase specific serological patterns to identify cattle C. burnetii infection status and shedding patterns (Lucchese et al. 2015). Therefore, interpretation of C. burnetii IgG exposure should be done with caution.

Results from this study have provided baseline true prevalence estimates and described the geographic distribution of C. burnetii exposure in a sample of extensively- farmed beef cattle from areas of northern Australia (Queensland and the Northern Territory). From this study, there was not enough evidence or appropriate data to suggest cattle hotspots within different regions of Queensland; however, we noted a lower level of exposure in the cattle sampled from the Northern Territory and south-east Queensland. Representative sampling of cattle across broad regions of Australia for C. burnetii testing, followed by in-depth geo-statistical and spatiotemporal analyses are warranted and should help to investigate putative risk factors for coxiellosis such as livestock density, wildlife density, environmental conditions (including rainfall, humidity and wind) and animal movements.

154

6.5. Chapter six appendix

Table A. 6-1 Table of prior specifications used for sensitivity analysis Prior specifications

Test specifications Sensitivity 5th parameters for prior mode distribution analysis percentile distribution

1 IFA DSe 0.88 0.96 dbeta (8.726, 2.053545) 2 IFA DSe 0.50 0.79 dbeta (3.26, 3.26)

3 σu dunif (0,1.5)

σw dunif (0,1.5)

4 σu dunif (0,2)

σw dunif (0,2)

5 σu dunif (0,2.5) σw dunif (0,2.5)

KEY: IFA, indirect immunofluorescent assay; DSe, Diagnostic sensitivity;

σu,, random effects for property; σw random effects for region

155

Table A. 6-2 Results from sensitivity analysis: Bayesian estimates of regional true prevalence of C. burnetii exposure using alternative model priors.

Posterior Estimates with 95% Credible Interval

Regions

Northern Central Southern South-east Northern Queensland Queensland Queensland Queensland Territory Model 7.3% (3.8, 13.6) 7.6% (4.0, 14.2) 6.8% (2.9, 14.9) 1.9% (0.4, 6.2) 1.0% (0.3, 3.3)

Sensitivity analysis 1 6.3% (3.3,11.6) 6.6% (3.4, 12.1) 5.9% (2.4, 12.7) 1.7% (0.3, 5.3) 0.9% (0.2, 2.8)

Sensitivity analysis 2 11.3% (4.7, 35.1) 11.6% (4.7, 35.9) 10.5% (3.5, 36.1) 2.9% (0.6, 13.5) 1.6% (0.4, 7.3)

Sensitivity analysis 3 7.3% (3.5, 14.1) 7.6% (3.7, 14.8) 6.8% (2.6, 15.2) 1.3% (0.2, 5.7) 0.7% (0.1, 2.7)

Sensitivity analysis 4 7.4% (3.5, 14.3) 7.6% (3.7, 14.7) 7.2% (2.6, 15.9) 1.0% (0.1, 5.0) 0.5% (0.0, 2.3)

Sensitivity analysis 5 7.4% (3.7, 14.2) 7.7% (3.7, 15.1) 7.3% (2.6, 16.9) 0.8% (0.0, 4.7) 0.4% (0.0, 2.2)

156

Table A. 6-3 R Code with data # Model: region and property as random effects model <- function(){ for(i in 1:60){y[i] ~ dbin(ap[i],n[i]) ap[i] <- se*pi[i] + (1-pi[i])*(1-sp) logit(pi[i]) <- b + u[Property[i]] } # Priors # priors for the fixed effects (constant only) b~dnorm(0,0.1)

# Priors for each property parameter: u1,...,u60 for(k in 1:60){ u[Property[k]] ~dnorm(w[region[k]], 1/(sigmau^2)) # random effect for property, random intercepts are centred on the random effect for the region within which each property is nested. } # Priors for each region parameter: w1,w2,w3,w4,w5 for( j in 1:5){ w[j] ~dnorm(muw,1/(sigmaw^2)) # random effect for region }

se~dbeta(10.542, 4.405083); ## 95% sure it is greater than 0.61 and mode of 0.737 sp~dbeta(19.811, 1.344804); ## 95% sure it is greater than 0.9 and mode of 0.98 muw <- 0 sigmau ~ dunif(0,1) sigmaw ~ dunif(0,1)

#Calculated outputs for(c in 1:5){ # prediction for a typical farm in each region xb[c] <- b + w[c] + 0

157

# predicted prevalence for a typical farm in each region xb_pi[c] <- exp(xb[c])/(1 + exp(xb[c]))

#indices to calculate complementary probabilities to the exceedance probabilities of C. burnetii disease freedom at a prespecified level (here 1%) for each region. DisFreePvalue[c] <- 1 - step(xb_pi[c] - 0.01) # }

} # Parameters to be monitored parameters <- c("se", "sp","pi","ap", "u", "w", "xb","xb_pi", " DisFreePvalue ")

# Initials inits1 <- list("se"=0.80,"sp"=0.90) inits2 <- list("se"=0.70,"sp"=0.99) inits3 <- list("se"=0.90,"sp"=0.95) inits <- list(inits1, inits2, inits3)

#Data Property <-c(1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60) region<-c(4, 1, 4, 3, 1, 4, 2, 2, 5, 5, 1, 1, 3, 4, 2, 2, 1, 2, 5, 1, 5, 2, 2, 2, 1, 1, 2, 4, 4, 3, 4, 1, 3, 1, 1, 1, 5, 2, 2, 3, 3, 3, 4, 2, 5, 2, 3, 1, 1, 1, 2, 4, 1, 1, 1, 1, 3, 3, 1, 3) y<-c(0, 2, 2, 1, 5, 2, 8, 2, 0, 1, 0, 2, 0, 4, 4, 4, 5, 2, 0, 0, 0, 1, 0, 0, 0, 5, 0, 1, 0, 0, 4, 3, 0, 0, 2, 0, 0, 8, 0, 0, 0, 0, 2, 6, 0, 5, 1, 5, 0, 0, 4, 1, 1, 6, 0, 1, 0, 0, 4, 0) n<-c(6, 16, 25, 112, 25, 29, 64, 54, 39, 35, 22, 35, 91, 22, 84, 52, 69, 16, 15, 23, 7, 28, 17, 27, 4, 41, 59, 20, 15, 31, 53, 53, 13, 8, 9, 35, 15, 44, 33, 34, 17, 24, 41, 56, 17, 55, 38, 39, 9, 36, 39, 29, 32, 61, 12, 36, 9, 31, 41, 10) data_jags <- list(y=y, region=region, n=n, Property= Property)

158

### Fitting the model using R Jags library(R2jags) bayes.mod.fit <- jags(data = data_jags, inits = inits, parameters.to.save = parameters, n.chains = 3, n.iter = 250000, n.burnin = 20000, model.file = model) bayes.mod.fit

#Followed with appropriate assessment of chains, burn-in and posterior distribution statistics.

159

Chapter Seven

Is Coxiella burnetii exposure associated with reduced reproductive performance in beef cattle in northern Australia?

“The work of epidemiology is related to unanswered questions,

but also to unquestioned answers.”

- Patricia Happ Buffler (1938–2013)

160

7. Is Coxiella burnetii exposure associated with reduced reproductive performance in beef cattle in northern Australia?

7.1. Introduction

Coxiella burnetii is an obligate intracellular bacterium known to cause the zoonotic disease Q fever. Parturient or aborting domestic ruminants are reported to be important sources of human infection. Coxiellosis has been associated with sporadic cases of abortion, premature delivery, stillbirth and weak offspring in cattle, sheep and goats (Agerholm 2013). However, there has not been any primary research into coxiellosis as a potential cause of reduced reproductive performance in beef cattle in Australia.

The beef cattle industry in Australia has a significant impact on the country’s economy. It was reported in 2017-2018, that beef cattle contributed approximately “20% ($AUD12.3 billion) of the total gross value of farm production” (‘Australian Bureau of Statistics, Australian Government’ 2019). Broadly, the beef cattle industry can be divided into northern (northern Western Australia, the Northern Territory and Queensland) and southern (southern Western Australia, South Australia, New South Wales, Victoria and Tasmania) regions (Martin et al., 2013). Approximately 60% of Australian beef cattle are produced in the northern region, with Queensland contributing 44% of national beef production (Meat and Livestock Australia 2014).

Reproductive performance is an important measure for determining the productivity and profitability of beef breeding businesses. It has been reported that the southern Australian beef herds have higher reproductive performance compared to those of northern Australia (McGowan et al. 2015; Gleeson, Martin & Mifsud 2012). Within the northern beef industry, there is marked variability in reproductive performance due to complex and multifactorial causes. Reduced reproductive performance in the north has been attributed to animal, herd and property level factors including harsh environmental conditions, genotypic variations, nutritional variations, time of calving, duration of lactation, animal age and the presence of infectious disease (McCosker 2016).

C. burnetii can cause sporadic abortions in cattle, however, the role of this infection in other reproductive disorders in cattle is more controversial (Agerholm 2013; De Biase et al. 2018). Although it is stated in many publications that natural infection with C. burnetii can cause bovine placentitis, retained foetal membranes (RFM), metritis and subsequent cow infertility, there is little primary research to support this (To et al. 1998; Muskens, van 161

Maanen & Mars 2011; Bildfell et al. 2000; Woldehiwet 2004). A histopathological study reported that bovine placental lesions associated with C. burnetii infection are very rare and that placentitis and subsequent RFM following infection is unlikely in cattle (Hansen et al. 2011). However, another histopathological study described endometritis and uterine vasculitis associated with C. burnetii infection in the uterus of cattle with a history of infertility; although the study design does not sufficiently fulfil criteria for causality (De Biase et al. 2018). A study of high-producing dairy herds identified a possible link with seropositivity and RFM, but no association with fertility (López-Gatius, Almeria & Garcia-Ispierto 2012). Further, a prospective longitudinal observational study of an endemically infected dairy herd in Germany found no association between positive serology and vaginal shedding of C. burnetii and reproductive outcomes (Freick et al. 2017). There was a small difference noted in the calving to conception interval between high milk shedders and non-milk shedders, but it was not statistically significant (Freick et al. 2017). Finally, an extensive investigation performed in dairy-cattle in Spain found that infected cattle (seropositive and PCR positive) actually had a lower risk of endometritis, returned to oestrus earlier postpartum and had shorter calving to conception intervals (Garcia-Ispierto et al. 2013).

To date, there are no published reports describing the relationship between coxiellosis in dairy or beef cattle in Australia and reproductive performance. The aim of this chapter was to examine the putative relationship between C. burnetii infection and reproductive performance in a large population of commercial beef breeding cattle in northern Australia.

7.2. Material and Methods

A retrospective analysis was conducted to assess and describe any putative relationship between C. burnetii infection and the annual pregnancy rate of beef cows in northern Australia.

7.2.1. Animal ethics statement Animal ethics for this study were approved by the University of Queensland Animal Ethics Committee approval: SVS/115/11/MLA (NF) and ANRFA/SVS/100/16.

7.2.2. Study design and study population The beef cattle reproductive data, analysed for this thesis chapter, was generated during an industry-funded epidemiology study investigating causes of poor reproductive performance in the north Australian beef industry: “The Northern Beef Fertility Project;

162

CashCow” (McGowan et al. 2014). The CashCow project was run as a longitudinal study from 2008 to 2011 with the aim to “determine and quantify the major associations between herd management, nutritional and environmental, and individual cow attributes and measures of reproductive performance” (McCosker 2016) in a sample of commercial beef breeding herds of northern Australia.

7.2.2.1. Background and methodology from CashCow The target population for CashCow was all commercial cattle properties within the north Australian beef industry, inclusive of Queensland, Northern Territory and northern Western Australia. The source population for the original study were all properties from this major beef breeding area that were serviced by cattle veterinarians who were members of the Australian Cattle Veterinarians’ National Cattle Pregnancy Diagnosis scheme. Study sites were then identified using a convenience sampling approach. The study population consisted of seventy-eight commercial beef cattle properties with approximately 78,000 individually identified females, managed in approximately 165 mobs, enrolled for a 3-4 year period. A mob was defined as a group of cattle that were managed as a unit within a herd on a property. The experimental unit was property and the analytical unit was animal.

7.2.3. Data and sample collection Data and biological samples were collected during the original study using a combination of methods briefly described below. A comprehensive description of data collection methods can be found in Chapter three of McCosker’s 2016 University of Queensland thesis (McCosker 2016).

7.2.3.1. Infectious disease monitoring A systematic random sampling method was used to collect biological samples for infectious disease monitoring at the time of the two annual mustering events on each enrolled property (McGowan et al. 2014). Between 15 to 30 individual blood and vaginal mucus samples were collected from each mob. For example if the mob size was 150 then a sample was collected from every 10th female as they presented in the crush to be examined. Blood samples were collected by coccygeal venepuncture into 9ml plain serum collection vials. Sera was decanted approximately 24 hrs after collection and either frozen or shipped chilled to the University of Queensland for storage.

In the CashCow project, bovine sera were tested for Bovine Viral Diahorrea Virus (BVDV/ bovine pestivirus), Neosporosis, Leptospirosis (Leptospira borgpetersenii serovar Hardjo and Leptospira interrogans serovar Pomona) and Bovine Ephemeral Fever (BEF) as 163 described by McGowan et. al (2014). Aggregated test results were used as mob level explanatory variables for each infectious disease respectively.

7.2.3.1.1. Coxiella burnetii serology Laboratory serological testing for C. burnetii IgG exposure was performed on 2 012 sera samples retrospectively for this thesis and is described in detail in chapter six. All blood samples from the 2011 CashCow sample collection, having sufficient volume remaining to perform laboratory testing were utilised. Hence, of the original 78 properties enrolled, 60 properties with viable samples were included in this study.

7.2.3.1.2. Coxiella burnetii testing from vaginal swabs I. Sample preparation and DNA extraction

Using molecular methods, 1 570 vaginal swabs were tested for evidence of C. burnetii exposure or bacterial deoxyribonucleic acid (DNA). The frozen vaginal swab samples were brought to room temperature and vortexed. A 20 µl sample of Tween 20 PBS solution from 10 individual samples (from the same property) was pooled into one Eppendorf; making up a total volume of 200 µl pooled sample. Prior to DNA extraction a known volume of equine herpes virus (EHV) standard, equivalent to 1x104 copies of EHV DNA was added to each pooled sample as an exogenous control. Samples was vortexed for 2-3 min and incubated at 100 °C for 20 minutes. Samples were vortexed again after incubation and spun down briefly. The 200 µl pooled sample was then transferred to MagNA Pure processing cartridge. Total nucleic acid was extracted using the DNA and Viral NA Small Volume Kit on the MagnaPure 96 extraction robot (Roche Diagnostics, Australia) as per the manufacturer’s protocol (DNATissue SV2.0) (Wang et al. 2018). Extracted total nucleic acid was eluted to 100µL (Wang et al. 2018).

II. Real-time polymerase chain reaction (PCR) methods a) Monitoring the precision of DNA extraction process

Real-time PCR for detection of the exogenous internal control EHV was performed on all samples to monitor precision of the extraction process, as published by Tozer et al. 2014. The PCR mix consisted of 1 µl of working primer mix (forward and reverse) at 10 pmol concentration, 0.2 µl of the probe at 4 pmol concentration and 12.5 µL of Quantitect Probe master mix (Qiagen, Brisbane, Australia; Table 7.1). This assay was performed in a ViiA™ 7 Real-Time PCR System (Thermo Fisher Scientific) using the following cycling conditions: 15 min incubation at 95 °C, followed by 50 cycles of 95 °C for 15 sec and 60 °C for 1 min (Bialasiewicz et al. 2008; Tozer et al. 2014). 164

An endogenous internal control was monitored to ensure that sample (vaginal mucous) was present in each reaction. The mammalian gene GAPDH (glyceraldehyde-3- phosphate dehydrogenase) was chosen to detect DNA of mammalian-cells in samples. Real-time PCR for the detection of GAPDH was performed on all extracted samples. Primers for GAPDH were based on published sequences (Table 7.2; Lisowski et al., 2008). Briefly, the PCR mix consisted of 8.5 µL H2O, 12.5µL SYBR-green (Quantitect), 1 µL of working primer mix and 3.0 µL of sample DNA. This assay was performed in a ViiA™ 7 Real-Time PCR System (Thermo Fisher Scientific). Standard cycling conditions were used with a 95 °C denaturation step, 30 s at 60 °C (for annealing) and 40 sec at 72 °C for elongation. The melt curve analysis consisted of 2 s at 95 °C, 5 s at 58 °C and slow heating at rate of 0.1 °C per sec up to 95 °C, for product dissociation (Lisowski et al. 2008).

Table 7.1 Detailed oligonucleotide sequences used for Equine Herpes Virus PCR assay Target Name Oligonucleotide sequence

EQHSV - Forward GAT GAC ACT AGC GAC TTC GA

Equine Herpes EQHSV - Reverse AGG GCA GAA ACC ATA GAC A Virus

EQHSV - Probe FAM-TTT CGC GTG CCT CCT CCA G-BHQ-1

KEY: Reference: Tozer et al., 2014

Table 7.2 Oliglonucleotide sequences for gene target primers for house-keeping gene Name Oligonucleotide sequence

Forward - ACC ACT TTG GCA TCG TGG AG “GAPDH” glyceraldehyde-3-phosphate dehydrogenase Reverse - GGG CCA TCC ACA GTC TTC TG

KEY: Reference (Lisowski et al. 2008)

165 b) Coxiella burnetii PCR

Two individual real-time PCR assays were performed for the detection of C. burnetii. Specific primers and probes targeting two different genes were used to increase sensitivity (Table 7.3). The first target was the IS1111 element of the transposase gene that has been found to be repeated up to 20 times through the C. burnetii genome (Klee et al. 2006). The second target was the com1 outer membrane gene (Lockhart et al. 2011). Each PCR mix consisted of 1 µl of respective working primer mixes (IS1111, com1 or htpAB) at 10 pmol concentration, 0.2 µl of specific probes at 5 pmol concentration, 12.5 µL of Quantitect Probe master mix (Qiagen) and 5 µL of extracted DNA in a final volume of 25 µL.

Positive control samples used for PCR assays consisted of reconstituted C. burnetii antigen (Virion/Serion, DHSH) Nine Mile strain commercially available for complement fixation test (Tozer et al. 2014). Serial dilutions from 1/10 to end-point detection of this Nine Mile strain whole cell bacterial suspension were used as quality control standards to optimise RT-PCR assays. Negative control samples were included in all PCR runs.

Amplification was performed in a ViiA™ 7 Real-Time PCR System (Thermo Fisher Scientific) using the following cycling conditions: 15 min incubation at 95 °C, 50 cycles of 95 °C for 15 sec and 60 °C for 1 min (Tozer et al. 2014). All primers and probes were synthesised by GeneWorks (Hindmarsh, Australia). Lyophilised primers and probes were reconstituted to a standard 200 µM stock concentration with sterile, distilled water and were stored in a dedicated PCR set up laboratory in a -20 °C freezer.

Table 7.3 Detailed oligonucleotide sequences for the two gene targets used for C. burnetii PCR assays Gene Target Primers Probe

F - GTC TTA AGG TGG GCT GCG TG IS1111 - FAM - AGC GAA CCA TTG GTA TCG Transposase GAC GTT TAT GG - BHQ gene R - CCC CGA ATC TCA TTG ATC AGC

F - AAA ACC TCC GCG TTG TCT TCA com1 - Outer FAM - AGA ACT GCC CAT TTT TGG membrane CGG CCA - BHQ-1 gene R - GCT AAT GAT ACT TTG GCA GCG TAT TG

166

Test samples were considered positive if one or more targets for C. burnetii tested positive for bacterial DNA. The cycle threshold (Ct) value was defined as ≤ 29 indicating a strong positive reaction, Ct of 30–38 as a positive reaction and Ct 39–44 as weak positive, Ct ≤45 indicated a test negative result.

7.2.3.2. Property and herd data Property and herd level management factors were derived from a series of farmer surveys (McCosker 2016). Data gained from these surveys included: details of property infrastructure and management, existing animal management protocols for specific classes of breeding females and mating management, annual pasture management and use of vaccination protocols for prevention of infectious diseases. Environmental data were then derived from the Australian Bureau of Meteorology with reference to GPS locations of paddocks or homesteads (McCosker 2016). Property and herd level explanatory variables have been described in Appendix (Table A. 7-1).

All enrolled properties were allocated to one of four country types according to a subjective assessment of the production potential of the grazing land and cross referenced with pasture and vegetation descriptions reported by the herd owners/managers (McCosker 2016). These country types are Central Forest, Northern Downs, Northern Forest and Southern Forest. The location of properties, coded according to this country type categorisation are presented in Figure 7.1.

7.2.3.1. Animal data All animals enrolled in the original study had unique National Livestock Identification Scheme (NLIS) radio frequency identification device (RFID) ear tags. Using the RFID as unique identifiers, data on individual animal variables were recorded for each enrolled heifer or cow with the commercially available crush-side electronic data capture system AgInfoLink’s BeeflinkData™. Data capture was conducted twice a year for the entire study, once at the main branding or weaning muster and again at the pregnancy diagnosis muster (McGowan et al. 2013).

Definitions for individual animal data has been described in Appendix Table A. 7-2 and includes age, breed, body condition score, lactation status, pregnancy status (and foetal age estimates of pregnancy), liveweight and hip height.

167

Figure 7.1 Map of enrolled properties across the Northern Territory and Queensland categorised by country type.

7.2.3.2. Nutritional data A pooled sample of faeces from each enrolled mob was collected in January, March, May, August and November. Metabolisable energy and crude protein were determined by faecal near infrared spectroscopy (NIRS), wet chemistry was used to determine faecal phosphorous (FP) and producer pasture assessments were used to estimate available pasture dry matter.

7.2.4. Annual pregnancy status For the logistic regression analysis, the outcome variable of interest was the annual pregnancy status of an individual animal in the herd. Annual pregnancy status was coded 168 for each female that was enrolled in the study and exposed to mating in a given year (McGowan et al. 2014). Cows were coded as 1 = pregnant, if they were detected pregnant or 0 = not detected pregnant, within one annual production cycle. For this study, data for the production year between 1st September 2010 to 31st August 2011 was used as this corresponded with the period during which blood samples were collected.

7.2.5. Preparation of the analysis dataset A master CashCow dataset was provided for this study as a stata (.dta) file. Data collected within the CashCow project from January 2010 to December 2011 were extracted from the dataset using R version 3.5.1 (‘R: A language and environment for statistical computing’ 2019). The dataset was then reduced to contain only observations from animals within the 60 properties with valid C. burnetii test results for 2011. C. burnetii serology results were linked by property code to the Cashcow reproductive dataset.

7.2.6. Statistical methods

7.2.6.1. C. burnetii laboratory results Crude diagnostic test results are reported for the C. burnetii serological and molecular test methods as an apparent prevalence (AP); equal to the proportion of test positive animals from the total sample tested. Results were calculated at the individual animal level for state, region and properties with Wilson-Score 95% Confidence Intervals (CI). The within-property proportion positive for C. burnetii exposure was used as a property level explanatory variable for the analysis described in this chapter.

7.2.6.2. Exploratory data analysis Statistical analysis was completed using R. Exploratory and descriptive data analysis was performed to evaluate the Cashcow dataset and identify missing data. All categorical variables were examined with tabulations, histograms or box and whisker plots and summary statistics reported for continuous variables.

Initial examination of the relationship between the explanatory C. burnetii property status and reproductive outcome measure was examined by cross tabulating and calculating the crude odds ratio of successfully becoming pregnant on a C. burnetii infected property compared to not-infected property. This relationship was further explored by plotting and visual examination of the reproductive outcome variable against the property level explanatory variable proportion positive for C. burnetii exposure.

169

7.2.6.3. Logistic regression Linear mixed-effects logistical regression models with random intercepts were used to explore the association between putative risk factors and the reproductive outcome (the probability of an animal to be pregnant in a specific mob within a specific property, equivalent to the pregnancy rate). Two levels, “mob” and “property” were fitted as random intercepts to account for clustering in the data. All potential explanatory variables were considered for biological plausibility e.g. age, body condition score, previously raised a calf. Model fitting was carried out into stages. First, putative risk factors were screened by univariable analysis; Wald test p values were used to compare categories of the explanatory variable with a reference category, and the likelihood ratio test (LRT) p values and Akaike information criterion (AIC) were used to test the overall goodness of model fit of each univariable model. All continuous variables screened were assessed visually for a linear relationship with the outcome variable on the logit scale. Each continuous variable was divided into quantiles and plotted against the estimated logit values for that variable from the univariable model. The plot was visually assessed and if the relationship between the quantiles and the logit were deemed linear, that variable was used as a continuous variable. If no linear relationship was identified, the variable was fitted as a categorical variable in the modelling process. In the second modelling stage, explanatory variables with a LRT p value ≤ 0.15 were then carried forward for the multivariable analysis. Collinearity between explanatory variables was assessed using Spearman’s χ2 tests. If two variables were highly correlated, the variable that had the most biological plausibility was retained. This was performed to avoid potential instability due to collinearity within the multivariable model.

A multivariable mixed-effects logistic regression model was fitted for the reproductive outcome variable: annual pregnancy status. The structure of the model was as follows:

logit 𝑌 1

𝛽 𝐶𝑜𝑢𝑛𝑡𝑟𝑦 𝑡𝑦𝑝𝑒 𝑃𝑟𝑒𝑣_𝑝𝑟𝑒𝑔 𝐿𝑎𝑐𝑡𝑐𝑜𝑑𝑒_𝑝𝑡 𝐵𝐶𝑆

𝑆𝑢𝑝𝐷𝑅𝑌 ℎ𝑒𝑟𝑑𝑠𝑖𝑧𝑒 𝑏𝑒𝑓_𝑠𝑒𝑟𝑜𝑝 𝑣𝑎𝑐_𝑙𝑒𝑝𝑡𝑜2 𝛽𝑃𝑟𝑜𝑝_𝑝𝑜𝑠 𝛼

𝑢 𝑒

Where Yijk is the probability of pregnancy for animali, from farm property and animal mobk, 0 is the regression coefficient (intercept) of pregnancy success at the animal level, fixed effects of: country typel, previously pregnantm, lactation coden, body condition scoreo supplement feeding during the dry seasonp, herd sizeq, Bovine Ephermeral Fever

170 seroprevalencer, vaccinated for Leprospirosiss, proportion positive C. burnetiit. 0j, and u0k are random regression coefficients describing the deviation of animali’s probability of pregnancy from that of the animals in the mob and property, and eijk is the random residual error (ith animal in the jth mob in the kth property) assumed to be have a normal distribution with a mean of 0 and a variance σ2.

The final model was built following the forward manual step-wise model building approach (Dohoo, W. Martin & Stryhn 2009a). Beginning with a null model with no explanatory variables, all explanatory variables carried forward from the univariable analyses were then introduced, starting with the term with the largest Wald test statistic. Variables individually added to the multivariable model were retained if they were found to be statistically significant using the LRT p value of 0.05 or less. Explanatory variables that did not reach statistical significance were then reintroduced one-by-one to the model and tested for significance at an alpha level of ≤ 0.05. Property-level C. burnetii exposure, as a proportion positive, was fitted in the final model as a polynomial continuous variable. The order of the polynomial variable was determined using AIC. Potential confounders were examined in the final model. Variables were considered a confounder if, when introduced to the final model, they changed the estimated coefficients by 25% or greater. Confounding variables were declared and retained in the final model if present. Biologically plausible two- way interactions were considered and retained in the model if they reached statistical significance at an alpha level of ≤ 0.05. To assess the model fit, the residual error distribution assumption was evaluated by a normal probability plot of residuals. Normalised residuals were evaluated graphically against predicted values across all properties to assess the homogeneity of variance of the error terms. The overall fit of the final model was evaluated using Hosmer and Lemeshow goodness of fit test. Further diagnostics, including the calculation of leverage and delta-betas, were used to identify any outliers or highly influential observations (implemented within the influence.ME package in R; Nieuwenhuis, Grotenhuis and Pelzer, 2012). The mixed model was fitted using the maximum likelihood (Laplace Approximation) procedure within the lme4 package (Pinheiro et al. 2019) in R.

An attempt was made to explore the association between C. burnetii exposure and annual pregnancy status from resulting model output, while accounting for other explanatory variables in the original model. Bootstrap methods were performed to resample the data (1 000 replicates) across different levels of C. burnetii exposure. The multivariable model was then refit on the resampled data to estimate the predicted marginal pregnancy probabilities and their associated 95% confidence intervals. 171

7.3. Results

7.3.1. Serology - IFA Sera (n = 2 012) from 60 properties in Queensland (n = 1 602) and the Northern Territory (n = 410) were tested. The median number of sera tested per property was 31 with a range of 4 to 112 (Interquartile range (IQR) 17–41). Overall, 5.2% (95% CI 4.3, 6.2) of sera tested returned a positive result. The IFA test results and proportion positive with 95% CIs at the animal level, stratified by state/territory and country type are shown in Table 7.4. For Queensland, 61.2% (30/49) of the properties tested, had at least one IFA positive animal, whereas in the Northern Territory, only 18.2% (2/11) of properties tested had at least one IFA positive animal. Overall 53.3% (32/60) of properties had serological evidence of exposure to C. burnetii. For the 32 positive properties, the average within-property proportion positive was 8.7% with a range of 0.9%–22.2%.

Table 7.4 Descriptive summary of crude IFA results for Coxiella burnetii exposure in bovine test sera aggregated by country type

No. of Total no. Proportion positive samples positive samples tested (Wilson 95% CI) Overall Northern Australia 104 2 012 5.2 (4.3–6.2)

State/Territory Queensland 102 1 602 6.4 (5.3–7.7) Northern Territory 2 410 0.5 (0.1–1.8)

Country Type Central Forest 36 489 7.4 (5.4–10.0) Northern Downs 17 520 3.3 (2.1–5.2) Northern Forest 40 653 6.1 (4.5– 8.2) Southern Forest 11 350 3.1 (1.8–5.5)

7.3.2. Vaginal mucus PCR Overall, 1 570 individual vaginal swabs were pooled into 157 pooled samples and screened for C. burnetii DNA using PCR methods. From these pools, 1/157 returned a “weak positive” result using genome target IS1111. Therefore, PCR results indicates 1.6% (95$ CI 0.8, 2.5%; n = 1/60) of properties showed evidence of reproductive tract shedding. The PCR results were not included as an explanatory variable in the multivariable model due to the very low proportion detection.

172

7.3.3. Population annual pregnancy status From the Cashcow reproductive dataset, there were 22 374 valid individual animal observations from 113 mobs on 60 properties for the outcome variable, annual pregnancy status. The number of valid cattle observations on each property ranged from 62 to 3,912 with a median of 263 (IQR 191–334). Overall, of the 22 374 observations, 73.9% (n = 16– 549) were found to be pregnant within one annual production cycle and 26.0% (n = 5 825) failed to become pregnant in the same production cycle. The within-property proportion pregnant, ranged from 24.5% to 96.4% (median 79.4%; IQR 61.1%–91.2%).

7.3.4. Factors associated with annual pregnancy status One hundred and twenty seven candidate explanatory variables were considered during univariable screening. Table 7.5 presents summary statistics and estimated coefficients for explanatory variables that were significantly associated with annual pregnancy status and were subsequently included in the multivariable modelling process.

173

Table 7.5 Results from univariable analysis for explanatory variables retained for the multivariable model building

Variable count 95% CI Number Estimated pregnant Total Coefficient Explanatory variables within year number per (Standard lower upper LRT p- (%) category error) Pr (>|z|) Odds Ratio limit limit value

Age (years) 0.09 ≤ 4 1992 (71%) 2820 reference > 4 13858 (75%) 18558 0.19 (0.11) 0.08 1.21 0.97 1.51

Previous body condition score (previous branding/weaning muster) <0.001 1 - 2 1218 (64%) 1899 reference 2.5 2215 (75%) 2952 0.18 (0.08) 0.02 1.19 1.03 1.38 3 4051 (71%) 5726 0.08 (0.07) 0.31 1.08 0.93 1.25 3.5 2341 (71%) 3229 0.20 (0.08) 0.01 1.22 1.05 1.41 ≥ 4.0 1973 (78%) 2514 0.34 (0.09) <0.001 1.40 1.18 1.67

Body condition score (wet/dry muster)

1 - 2 640 (43%) 1489 reference <0.001 2.5 1484 (69%) 2139 0.41 (0.08) <0.001 1.51 1.28 1.78 3 4274 (78%) 5454 0.81 (0.08) <0.001 2.25 1.93 2.62 3.5 3403 (82%) 4127 0.93 (0.09) <0.001 2.54 2.15 3.00 4 - 5 3097 (81%) 3822 1.21 (0.09) <0.001 3.35 2.82 3.96

Body condition score (pregnancy diagnosis muster) 1 - 2 963 (36%) 2680 reference <0.001 2.5 1730 (55%) 3133 0.80 (0.07) <0.001 2.2333673 1.956508 2.55 3 3968 (73%) 5424 1.58 (0.07) <0.001 4.8711057 4.268003 5.56 3.5 4001 (87%) 4601 2.26 (0.08) <0.001 9.5559924 8.222384 11.1059 ≥ 4.0 5593 (91%) 6128 2.75 (0.08) <0.001 15.6634344 13.43826 18.25706

Change in BCS (from branding/weaning muster to pregnancy diagnosis muster) <0.001 Gained 4391 (75%) 5869 reference Lost 4913 (76%) 6479 0.30 (0.06) <0.001 1.348394 1.209258 1.50 174

Maintained 2953 (77%) 3857 0.06 (0.06) 0.322 1.058845 0.945654 1.19

Previous lactation status <0.001 not lactating 3246 (62%) 5243 reference lactating 11920 (78%) 15221 0.20 (0.05) <0.001 1.217288 1.10886 1.336318

Previously pregnant <0.001 not pregnant 2054 (86%) 2378 reference pregnant 13703 (72%) 19047 -2.03 (0.08) <0.001 0.13 0.11 0.15

Lactation status at pregnancy diagnosis muster <0.001 not lactating 10097 (82%) 12310 reference lactating 6215 (64%) 9778 -2.26 (0.05) <0.001 0.10 0.09 0.11

Previously raised a calf <0.001 no 932 (77%) 1205 reference yes 11646 (75%) 15508 -1.18 (0.10) <0.001 0.31 0.25 0.37

Previous month of calving <0.001 Apr-Jun 468 (45%) 1043 reference Dec-Jan 5112 (76%) 6726 1.52 (0.09) <0.001 4.55 3.84 5.40 Empty (not previously pregnant) 2054 (86%) 2377 3.41 (0.11) <0.001 30.37 24.42 37.78 Feb-Mar 1173 (59%) 2005 0.89 (0.10) <0.001 2.43 2.02 2.93 Jul-Sep 1035 (88%) 1182 2.24 (0.14) <0.001 9.35 7.17 12.20 Oct-Nov 5023 (79%) 6326 2.03 (0.09) <0.001 7.59 6.33 9.09

Country type <0.001 Central forest 2708 (86%) 3149 reference Northern downs 6433 (85%) 7540 -0.22 (0.36) 0.54 0.80 0.39 1.63 Northern forest 4634 (55%) 8376 -1.56 (0.31) <0.001 0.21 0.11 0.39 Southern forest 2774 (84%) 3309 -0.07 (0.34) 0.84 0.93 0.48 1.83

Genotype/breed <0.001 ≥ 75% Bos indicus 13446 (77%) 17365 reference < 75% Bos indicus 3103 (62%) 5009 -1.07 (0.28) <0.001 0.34 0.20 0.60

Supplementation during dry season 0.00 no supplementation 4957 (86%) 5795 reference 175

supplementation provided 11592 (70%) 16579 -0.89 (0.29) 0.002 0.41 0.24 0.71

Property area 0.05 < 41000 ha 6880 (79%) 8686 reference ≥41000 ha 7363 (77%) 9580 -0.59 (0.30) 0.05 0.55 0.31 1.00

Cow herd size 0.02 > 500 15169 (73%) 20812 reference ≤ 500 1380 (88%) 1562 0.99 (0.44) 0.02 2.69 1.14 6.34

Major source of property income 0.05 Bullocks 4391 (76%) 5762 reference Feeder 8931 (76%) 11706 0.24 (0.33) 0.48 1.27 0.66 2.44 Weaners 2435 (65%) 3746 -0.65 (0.36) 0.07 0.52 0.26 1.06

Mustering methods 0.07 aerial assisted 9448 (70%) 13419 reference ground 5920 (81%) 7327 0.65 (0.29) 0.03 1.91 1.08 3.35 trapping 245 (20%) 1204 0.72 (0.57) 0.21 2.05 0.67 6.25

Replacement bull selection policy 0.02 little 6947 (67%) 10416 reference most 6090 (80%) 7570 0.84 (0.32) 0.01 2.33 1.25 4.34 some 2605 (81%) 3214 0.97 (0.40) 0.02 2.63 1.20 5.77

Bull management policy 0.01 little 6105 (67%) 9120 reference most 6214 (79%) 7820 1.08 (0.36) 0.00 2.94 1.45 5.95 some 3323 (78%) 4260 1.21 (0.40) 0.00 3.36 1.54 7.33

Farm policy for Bovine viral diarrhoea virus (BVDV) prevention 0.0407 vaccination not used 14770 (73%) 20331 reference vaccination used 234 (14%) 1711 0.97 (0.47) 0.04 2.65 1.06 6.64

Farm policy for Leptospirosis prevention 0.011 vaccination not used 11923 (76%) 15750 reference vaccination used 4324 (69%) 6292 0.81 (0.32) 0.01 2.26 1.21 4.19

Farm policy for Bovine ephemeral fever (BEF) prevention <0.001 176

vaccination not used 12720 (71%) 17994 reference vaccination used 3527 (87%) 4048 1.26 (0.29) <0.001 3.52 1.99 6.21

Duration of wet season 0.003 normal 3117 (57%) 5457 reference long 13162 (80%) 16554 1.03 (0.36) 0.00 2.81 1.40 5.63

Description of wet season (onset and duration) 0.0089 early - long 13162 (80%) 16554 reference early - normal 2754 (55%) 5015 -1.13 (0.35) 0.00 0.32 0.16 0.64 normal - normal 363 (82%) 442 -0.00 (1.03) 1.00 1.00 0.13 7.53

BVDV serological exposure - recent 0.04 high 864 (88%) 986 reference low 6601 (73%) 9024 -0.38 (0.38) 0.32 0.69 0.32 1.45 moderate 8079 (74%) 10971 0.10 (0.38) 0.80 1.10 0.52 2.33

BEF serological exposure 0.1175 high 12457 (74%) 16721 reference moderate 2601 (70%) 3731 -0.38 (0.24) 0.12 0.69 0.43 1.10

7.3.4.1. Multivariable analysis The final model for annual pregnancy status, included data representing 15 038 animal observations from 92 mobs on 51 properties. In the final multivariable model (Table 7.6) there was an effect of country type, previous pregnancy status, lactation status at pregnancy diagnosis muster, BCS at the wet/dry muster, supplementation during the dry season, herd size, BEF serological exposure and farm policy for leptospirosis vaccination. All diagnostics for this final model were satisfactory and the overall goodness of fit was good with a P value > 0.05 (Hosmer & Lemeshow 2000).

177

Table 7.6 Estimated regression coefficients for the final multivariable mixed-effects logistic regression model for reproductive outcome “annual pregnancy status”

No. of Standard Variable Coefficients OR (95% CI) p value cattle Error Country type all other country types 13998 reference Northern Forest 8376 -1.47 0.33 0.23 (0.12 - 0.43) <0.001

Previously pregnant not pregnant 2378 reference pregnant 19047 -1.23 0.13 0.29 (0.22 - 0.37) <0.001 Lactation status at pregnancy diagnosis muster not lactating 12310 reference lactating 9778 -2.33 0.072 0.09 (0.08 - 0.11) <0.001 Body condition score (wet/dry muster) 1 - 2 1489 reference 2.5 2139 0.62 0.10 1.85 (1.54 - 2.24) <0.001 3 5454 1.00 0.09 2.72 (2.29 - 3.24) <0.001 3.5 4127 1.04 0.10 2.84 (2.33 - 3.45) <0.001 4 - 5 3822 0.83 0.11 2.29 (1.86 - 2.82) <0.001 Supplementation during dry season no supplementation 5795 reference supplementation provided 16579 -1.40 0.33 0.25 (0.13 - 0.47) <0.001 Herd size > 500 20812 reference ≤ 500 1562 1.17 0.39 3.22 (1.49 - 6.97) 0.00

Bovine Ephemeral Fever (BEF) serological exposure high 16721 moderate 3731 1.02 0.31 2.77 (1.52 - 5.05) <0.001 Farm policy for Leptospirosis prevention vaccination not used 15750 vaccination used 6292 0.88 0.32 2.41 (1.29 - 4.51) 0.01 C. burnetii exposure Polynomial – 1st order N/A -4.02 5.99 0.02 (0.00 - 2248) 0.50 Polynomial – 2nd order N/A -36.46 8.93 0.00 (0.00 - 0.00) <0.001

Random effects (Standard Deviation)

Property 0.5811 (0.762)#

Mob 0.4117 (0.642) #

2 Overall Hosmer and Lemshow GoF χ 9.038282 df = 8 ; P value = 0.34 # The figures in the parentheses are the standard deviations of variance for the random effects Key: OR; Odds Ratio, N/A not applicable

178

7.3.4.2. Estimated predicted probability Figure 7.2 was generated from the final multivariable model to explore the relationship between C. burnetii proportion positive and the outcome (annual pregnancy status), adjusted for the impacts of other explanatory factors included in the model. When examining the plotted graph, a curvilinear relationship was identified. Initially a relatively flat line exists from C. burnetii exposure 0.0% to approximately 10.0%, then there is a moderate negative slope of the regression line once property exposure is greater than approximately 10.0%.

Figure 7.2 This figure was generated post-hoc from the final multivariable model to explore the relationship between Coxiella burnetii (C. burnetii) proportion positive and the outcome (predicted probability of pregnancy) adjusted for the impacts of other explanatory factors included in the model. Shaded area represent 95% confidence intervals.

179

7.4. Discussion and conclusions

In this study, the putative relationship between the reproductive performance of a large population of commercial beef breeding cattle in northern Australia and serological C. burnetii IgG results was examined. To our knowledge, it is the first time that this relationship has been examined in beef cattle in Australia or other countries. It is the first report to identify that C. burnetii may have an association with reduced reproductive performance in beef cattle in Australia.

The present study utilised data from a large industry-based project and used mixed- effects logistic regression models to account for multi-factorial variables on reproductive performance. The models were then used to examine the relationship between C. burnetii exposure and the reproductive outcome “annual pregnancy status”. The main finding from this analysis revealed a curvilinear relationship between the property C. burnetii serological exposure level (as a proportion positive) and annual pregnancy status. The pattern indicates that there may be little effect of C. burnetii exposure on predicted annual pregnancy rate until a threshold is exceeded, then a moderate negative association once the property exposure is greater than 10%. These results indicate that there is potentially a negative association with beef cattle conception rates on properties with high C. burnetii seroexposure. Higher prevalence may reflect active infection within the herd, while low levels of exposure may be more indicative of past infection or sporadic infection. Although the study design does not allow causality to be inferred, it provides evidence for a hypothesis that warrants further research and investigation.

The curvilinear relationship identified in this study is interesting and may explain why international publications have described mixed reports of evidence describing reproductive disorders in cattle associated to C. burnetii infection (Agerholm 2013). While some literature states that there is no relationship evident between C. burnetii infection and reduced reproductive performance in cattle, others have presented evidence to support both positive and negative associations. A report from an extensive investigation performed in dairy-cattle in Spain identified that infected cattle (C. burnetii IgG seropositive and PCR positive) had a lower risk of endometritis, showed earlier return to oestrus and shorter calving to conception periods (Garcia-Ispierto et al. 2013). Although this seems surprising, one explanation is that C. burnetii immunoglobulin response could indicate a current protective immunity. At low IgG levels, there may be indications of a prior active infection, hence cows may have current protective immunity and therefore show improved reproductive performance. However, this type of interpretation is made with caution, as the exact dynamics of C. burnetii sero-

180 response after acute infection and details of protective immunity are not completely understood or well described in the literature (Garcia-Ispierto, Tutusaus & López-Gatius 2014).

This highlights an important limitation to be considered when interpreting results from the current study. Firstly, aggregated herd-level serology results were used as the explanatory variable because the sample identifier codes from the previous study could not be matched to the individual animal reproductive dataset. Ideally, this analysis would have been performed with animal level IFA results. When considering C. burnetii infection status, IFA IgG serological test results have been used alone without concurrent interpretation of PCR results. Although vaginal swab samples were tested, the results were very low and potentially biased, therefore they were excluded from the main statistical analysis of this chapter. Interpretation of serological results is difficult and there are reports that serologically negative animals can test PCR-positive, thus, indicating a sero-negative animal with a current, active infection. Serological IgG exposure may be an indicator of past infection that has resolved or become a chronic infection. While there are studies based on serology alone, the OIE does suggest a combination of serological and PCR results to more accurately determine infected status (OIE 2018). Serology is appealing for research studies due to the ease and accessibility of blood samples and affordability of laboratory testing. Thorough PCR testing to determine C. burnetii infection status would require multiple excretion routes to be sampled from cattle as they are reported to shed C. burnetii inconsistently across vaginal mucous, milk, and faeces (Guatteo et al. 2007). For complete samples to be collected from all potential shedding routes, the study would become very expensive and impractical in a commercial beef cattle setting. These factors should be considered when designing further studies to confirm this putative pattern and allow greater depth of understanding with the ability to make inferences from future findings.

The low prevalence PCR results may be biased due to selection bias and misclassification bias. Firstly, selection bias is likely because the vaginal swab samples were collected during a previous study and timing was not aimed for optimal C. burnetii identification. Vaginal shedding of C. burnetii in cattle appears to be very specific around the time of parturition and in the days to weeks just after calving (Guatteo et al. 2007). Secondly, “misclassification bias results from a rearrangement of the study individuals into incorrect categories because of error in classifying exposure, outcome or both” (Dohoo, W. Martin & Stryhn 2009b). Therefore misclassification bias may have occurred if truly infected animals were categorised as “test negative” because they were not shedding at the time of sample

181 collection. Therefore, the low PCR positive results from vaginal swab samples may not be good indication of true infection status of cattle in this study.

However, in a large-scale commercial beef cattle setting, like northern Australia, pregnancy detection is the time of a reproductive cycle when farmers and veterinarians will have closest contact with pregnant cattle. Cows will then calve out in the paddocks and would not commonly have human contact during parturition. We can therefore extrapolate two main things from this; firstly that vaginal swab samples collected at time of pregnancy diagnosis could commonly provide false negative results in infected cows. Secondly, from a public health perspective, we can suggest that close contact with pregnant cattle during early pregnancy testing appears to have a low risk for Q fever transmission as evident by very low PCR shedding from cattle at this time of their reproductive cycle.

It has been reported that C. burnetii infection in cattle may cause sporadic abortions during gestation (Agerholm 2013); therefore, it would have been ideal to examine an outcome of abortion and/or foetal calf loss in order to assess putative relationships with C. burnetii exposure and these outcomes. However, this was not possible due to logistical constraints of data from the original study. For this study, the outcome variable for reproductive performance (annual pregnancy rate) represents a conception rate and measures the animal’s ability to become pregnant within one annual reproductive cycle. It has been recognised elsewhere that the reproductive performance of beef cattle in northern Australia is lower with greater variability than beef cattle in southern Australia (McCosker 2016; Gleeson, Martin & Mifsud 2012; Entwistle 1983). However, to the best of our knowledge, C. burnetii as a potential cause of reduced reproductive performance within this broader context of commercial beef production in Australia has not been considered previously. The distribution of C. burnetii exposure in Australian cattle could be one element of this puzzle that has not yet been well investigated.

The extent of several reproductive performance parameters in a population of commercial beef cattle was measured and described during a large, longitudinal study: “The Northern Beef Fertility Project; CashCow” (McGowan et al. 2014). Data from this original project generated a large and unique dataset that was originally analysed to identify important risk factors at the animal, mob and property level to advise beef producers and industry on potential management strategies for improving reproductive performance and hence productivity (McGowan et al. 2013). Although we utilised data collected from the CashCow project, we did not analyse the complete dataset, as only a subset of the properties had serum samples available for C. burnetii testing. We investigated explanatory

182 risk factors for reproductive performance from this subset of beef cattle in northern Australia with the specific purpose to assess and describe any putative relationship between C. burnetii exposure and the outcome of successfully becoming pregnant during one production cycle. Therefore, the models built during this study were not intended to determine additional risk factors or to be used to advise general management practices for commercial beef cattle enterprises.

We acknowledge that the analysis presented here should be interpreted with caution, and should serve as a hypothesis generating process that will require additional examination of individual explanatory variables for reproductive performance and C. burnetii exposure. For example, it was previously recognised that cattle in the Northern Forest region showed lower annual pregnancy rates than cattle in the other country types (McCosker 2016). This was also observed in the current study, however, it could be beneficial to further investigate and examine what additional risk factors or confounders may be present in the Northern Forest that could explain these results. It is known that Australian wildlife and feral animals are common vectors for C. burnetii and may cause increased exposure and transmission of C. burnetii to cattle farmed in certain regions or in specific ecosystems (Cooper et al. 2012, 2013; Potter et al. 2011). Therefore ecological, environmental and herd property level factors may need further assessment to firstly identify if C. burnetii exposure is merely a proxy itself for harsh geographical landscape or if it can be causally associated with reduced reproductive performance. An additional study designed to investigate property level risk factors associated with differences in C. burnetii exposure is also warranted and may require in-depth spatio-temporal analysis that incorporates geographical and seasonal environmental factors.

183

7.5. Chapter seven appendix

Table A. 7-1 List with brief details of explanatory risk variables at the mob or property level

Variable Brief explanation/details Southern Forest / Central Forest / Northern Downs / Country type Northern Forest.

Property Size (ha) Continuous variable

Property management structure Owner manager / Private manager / Company manager / Leasee or Agistee

<10 years; 10-<20 years; ≥20 years Duration managing enrolled property

continuous variable Average annual rainfall (mm) <500 cows, 500 to <1000 cows, 1000 to <3000 cows Cow herd size and ≥3000 cows. Average size of breeding ≤100 cattle, >100-400 cattle or >400 cattle management groups

Females deliberately exposed to bulls for a period less than 4 months (Control mated for ≤3 months); Females deliberately exposed to bulls for 4-7 months, typically bulls removed at pregnancy diagnosis muster and reintroduced early New Year (Control mated between 4-7 Mating management months); Females deliberately exposed to bulls for >7m of a year (Continuously (>7m) mated without segregation); Females deliberately exposed to bulls for >7m of a year with cows segregated on the basis of either lactation status or foetal age (Continuously >7m mated with segregation).

Use of aerial vehicles such as a helicopter or aeroplane Primary mustering technique used (Air);

The number of times an enrolled female was absent throughout the study calculated as a percentage of the Mustering efficiency cumulative number of animals attempted to be mustered was categorised as <5%, 5-<10% and ≥10%. An absent animal was defined as an animal that failed to be mustered but was known to be alive because it subsequently turned up at a later scheduled muster.

All weaned female calves retained and later exposed to bulls (No Selection); Heifer selection based on visual Heifer selection protocol appraisal (Visual appraisal); Heifer selection based on estimated live weight or live weight (with or the without the use of scales) (Liveweight) gain prior to first mating.

Provision of supplemental nitrogen Provided in some years; Provided in all years; Not during the dry season (May to provided. October) Provided in some years; Provided in all years; Not

provided.

184

Provision of supplemental phosphorus during the wet season (November to April) Cows not culled on age; Cows culled at ≤10 years of Age cows were culled for age age; Cows culled at >10 years of age

The number of females culled though out the year as a Culling rate for breeding females percentage of those mated in the previous breeding season was calculated and categorised as <10%, 10- <15% and ≥15%.

Sale of weaners; Sale of feeder cattle; Sale of Major source of income cows/bulls and bullocks. More than one source was allowed to be identified

Bull to female mating ratio used <2:100 females; 2-3:100 females; ≥4:100 females

Some best practice: at least 2 of the following used – replacement bulls vaccinated for tick fever [if required] and bovine ephemeral fever (BEF), BCS managed prior Replacement bull selection policy to first mating, introduced to property in cooler months, allowed ≥2 months to acclimatise prior to first mating. Note no bull breeding soundness examination (BBSE) used. Most best practice: replacement bulls selected on basis of having passed a veterinary BBSE and at least 2 of the following; vaccinated for tick fever [if required] and bovine ephemeral fever (BEF), BCS managed prior to first mating, introduced to property in cooler months, allowed ≥2 months to acclimatise prior to first mating Nil best practice: Did not meet criteria for either ‘Some’ or ‘Most best practice’.

Some best practice: at least 2 of the following – same age bulls mated together, vaccinated for BEF annually, Herd bulls management policy BCS managed prior to mating, treated for external and internal parasites annually, bulls culled at ≥8years of age. Note BBSE not included. Most best practice: bulls were selected on the basis of having passed BBSE and at least 3 of the following: same age bulls mated together, vaccinated for BEF annually, BCS managed, treated for external and internal parasites annually, bulls culled at ≥8years of age. Nil best practice: Did not meet criteria for either ‘Some’ or ‘Most best practice’

Herd vaccination policy – bovine Heifers only vaccination; Whole herd vaccination; Not viral diarrhoea virus (BVDV) vaccinated

Herd vaccination policy - Vibriosis, Categorised as either vaccinated or not vaccinated for Leptospirosis, Botulism each disease

185

The wet season onset was derived using interpolated daily rainfall information that was downloaded from the Australian Bureau of Meteorology (BOM) using the GPS Wet season onset location for the paddock or homestead. The wet season onset was defined as the date at which an accumulation of 50 mm of rainfall was reached in 14 days or fewer, starting from any day after September 1 (but before March 31).

Using interpolated daily rainfall information for the GPS location of a paddock or homestead from the Australian Days after wet season onset to Bureau of Meteorology (BOM), the number of days follow up rain following the wet season onset until another major rainfall event was derived. A major rainfall event was defined as an accumulation of 50 mm of rainfall in 14 days or less.

The reported average estimated minimum quantity of Minimum dry season biomass pasture available between May 1 to October 31 was categorised as either <2000 kg/ha or ≥2000 kg/ha

The average CP content of the pasture for the wet (November 1 to April 30) and dry seasons (May 1 to October 31) was recorded on the continuous scale. Seasonal crude protein(CP) content Documented threshold value of 6-8% CP were of the pasture inspected based on Minson (1990) and Winks et al. (1979). However, these were not found to be discriminatory and subsequently re-categorised as <5% or ≥5%.

The average DMD of the pasture for the wet (November Seasonal dry matter digestibility 1 to April 30) and dry seasons (May 1 to October 31) (DMD) of pasture was recorded on the continuous scale and subsequently categorised as either <55% or ≥55% (Jackson 2012).

The ratio of the dietary crude protein content to dry Average ratio of dry matter matter disability was calculated and averaged across digestibility to dietary crude protein both wet (November 1 to April 30) and dry (May 1 to during dry season October 31) seasons and categorised as either <8:1 and ≥8:1 DMD:CP (Dixon and Coates 2005).

The ME of the diet was calculated from the DMD using Average wet season ratio of faecal the equation ME=0.172 x DMD−1.707 (CSIRO 2007). phosphorus (FP) to metabolisable The average FP:ME ratio for all samples collected energy (ME) during November 1 – April 30 was categorised as ≥500 mg P : 1 MJ ME or <500 mg P : 1 MJ ME.

NB: The above table was re-created from tables and information in the thesis: “Risk factors affecting the reproductive outcome of beef breeding herds in North Australia” (McCosker 2016).

186

Table A. 7-2 Table of animal level data collected for the CashCow project Variable Brief description/details

Explanatory data Breed was categorised according to the estimated Bos indicus contect: <75% Bos indicus or ≥75% Genotype/breed Bos indicus.

Year brand Represented the year the animal was branded.

Hip height was measured manually with a Hip Height retractable measuring tape (Fordyce et al. 2013).

Integer variable, calculated by taking the year of Age observation minus the year the animal was branded plus 1.

Heifers; 1st lactation cows; mature cows (≤9 years old excluding heifers and 1st lactation Year class cows); aged cows >9 years.

Assessed and scored on a 1 to 5 scale using 0.5 Body condition score increments (Hunt 2006). 1 = very thin; 5 = very (BCS) fat.

The change in body condition score between the pregnancy diagnosis muster and the subsequent Change in BCS weaning or branding muster was calculated and categorised as either ‘lost condition’, ‘maintained condition’ or ‘gained condition’.

Live weight Live weight was measured using electronic scales and was a continuous variable.

Based on a muster taking place between 1 month Mustered within 2 prior to and 2 months after a females expected months of calving month of calving, heifers or cows were categorised as either 1=mustered and 0=not mustered.

Using interpolated temperature and humidity data Average Temperature from the Australian Bureau of Meteorology (BOM) Humidity Index (THI) for the GPS location of either a paddock or during expected month property homestead the THI was estimated for of calving each day using the equation sited by Hahn et al. (2009):

THI=0.8 x Ambient temperature+{((Relative humidity)÷100 x (Ambient temperature-14.4))+46.4}

187

The average THI during the expected month of calving recorded on the continuous scale. Documented threshold value of 74 and 79 were inspected based on (Thom 1959). However these were not found to be discriminatory and subsequently categorised as <72 and ≥72.

The cumulative number of days the estimated THI exceeded 79 (Thom Categorised as <14 days and ≥14 days. 1959) during the expected month of calving.

The cumulative number Using interpolated maximum temperature data of days the maximum from the BOM for the GPS location of either a temperature exceeded paddock or homestead the number of days 40oC during the exceeding or equal to 40oC was categorised as expected month of <14 days and ≥14 days. calving.

Proportion of the Categorised as <40%, 40-<70%, 70 to <90% and paddock within 2.5km of ≥90%. permanent water at the time of calving

Outcome data

Lactation status Determined by visual assessment of the udder or attempted expression of milk

Determined by manual rectal palpation of the Pregnancy diagnosis reproductive tract by experienced cattle veterinarian. Females were defined as pregnant or not pregnant.

All animals confirmed pregnant then had foetal age estimations made following manual rectal Foetal age estimates palpation of the reproductive tract. These estimates were used to then calculate the date of conception and the predicted date of calving.

NB: The above table was re-created from tables and information in the thesis: “Risk factors affecting the reproductive outcome of beef breeding herds in North Australia” (McCosker 2016).

188

Chapter eight

General discussion

Old Walt Whitman Went finding and seeking, Finding less than sought Seeking more than found, Every detail minding Of the seeking or the finding.

Pleasured equally In seeking as in finding, Each detail minding, Old Walt went seeking And finding.

- Langston Hughes (1902–1967)

189

8. General discussion

Research carried out during this thesis was aimed at improving our understanding of the epidemiology of coxiellosis in beef cattle in northern Australia and to gain insights into the potential risk to public health. To explore these aims, this thesis was written as five primary research chapters. Initially, Queensland human Q fever notification surveillance data over 15 years was analysed to describe the extent and trends of human notifications in this state. Subsequent studies were carried out to investigate the molecular epidemiology, laboratory diagnostics, prevalence, geographical distribution, and production effects of coxiellosis in beef cattle in Northern Australia. This final discussion chapter will bring together the key findings from the research, highlight how results offer significant contributions to the current literature, discuss limitations and suggest where future work should be focused.

Human Q fever was first discovered in Queensland, Australia during the 1930s amongst abattoir workers and dairy farmers and more recently has emerged globally as a potential threat for significant human epidemics associated with C. burnetii outbreaks in ruminants (Delsing, Kullberg & Bleeker-Rovers 2010; Bond et al. 2015; Eldin et al. 2016). Queensland has consistently reported between 40–55% of the Australian annual Q fever cases and high case notification rates in regional areas (Tozer et al. 2020). Although the National Q fever Management Program (NQFMP) appeared to have some success in reducing Q fever notifications within Australia, the program ended in 2006 and notification rates and counts have continued to fluctuate with an upward trend since 2009. From the analysis of Queensland Q fever notifications from 2003–2017, in chapter three, it was evident that many Q fever cases are still reported from farming and agricultural occupational groups. However, the surveillance data available for occupational groups may be subject to bias as there was missing data and inconsistencies with the free-text data field. This is in agreement with other reports from Australia that suggest using occupational group as a proxy for risk of transmission has major limitations (Clutterbuck et al. 2018).

In Chapter three, from enhanced surveillance data collected consistently from 2012 onwards, it was clear that the majority of notified Q fever cases reported at least one known environmental, animal or abattoir “at-risk” exposure within the one month prior to disease onset. The “at-risk” exposures included contacts that may increase the risk of Q fever as described previously in reports by Queensland Government Workplace Health and Safety

190

(‘Q fever - worksafe.qld.gov.au’ 2019) or in the Communicable Diseases Network Australia (CDNA) Q fever National Guidelines for Public Health Units (CDNA 2018). While only 7% of cases reported an abattoir-related exposure, 52% reported an at-risk animal-related exposure (Table 3.6). However, most common were at-risk environmental-related exposures (82%); including exposure to dust from paddocks or animal yards, living or working within 300 m of bush/scrub/forest areas and living or working within 1 km of an abattoir, animal grazing or saleyards. The transmission of human Q fever from exposure to dust at a saleyard in South Australia was identified during a case-control study following an outbreak investigation in 2004 (O’Connor, Tribe & Givney 2015). Additionally, the role of distance from ruminant holdings as a risk factor for human Q fever outbreaks was the topic of a recent systematic review (Clark & Soares Magalhães 2018). They identified the greatest risk for Q fever outbreaks was within a 5 km radius of infected farms in rural areas and within 2-4 km distance from source farms in urban outbreaks; although this would likely be affected by wind, timing of outbreaks and presence of landscape features such as vegetative barriers. It was interesting that this review only identified sheep and goat holdings as risks for human Q fever outbreaks and found no evidence to support cattle as having a major contribution (Clark & Soares Magalhães 2018). However, the review did not include any Australian publications, most likely because of the lack of publications available from Australia. Considerations for future government-funded Q fever control programs should incorporate broader educational and vaccination schemes that aim to increase the awareness of potential Q fever transmission from environmental exposures and extend the focus beyond the traditional high-risk occupational groups.

Although the analysis performed in chapter three was descriptive and has potential for bias due to the nature of the surveillance data, it has provided detailed tables of exposure frequencies that can inform future survey designs. This study revealed that nearly 50% of Q fever cases notified from 2013–2017 had direct contact with cattle in the one month prior to disease onset. This information was not broken down into beef cattle vs dairy cattle exposures, and therefore has limited value when considering the causal relationship between human Q fever and beef cattle C. burnetii infection. The current reporting of human Q fever cases still has some very broad exposure groups that could be further refined in order to gain a more thorough understanding of Q fever risks specifically from beef cattle. While we have identified an improvement in the surveillance of human Q fever cases within Queensland over time, coxiellosis remains absent from the list of nationally notifiable animal

191 diseases and there has been limited work towards increasing the knowledge of the epidemiology in ruminants or other animals at a local or national level.

Zoonotic diseases, such as coxiellosis, that may challenge animal and public health, firstly require identification and quantification to allow thorough investigations assessing potential effects, and causal disease determinants. Although there is awareness of acquiring human Q fever disease through exposure at abattoirs and ruminant farming, at the time of this thesis, there was minimal knowledge of the extent of C. burnetii infection in beef cattle in Australia. Research for this thesis was therefore primarily focused to identify and quantify C. burnetii infecting beef cattle going to slaughter and on commercial beef breeding properties in Queensland and the Northern Territory. In this context, “identification” included the direct detection and molecular characterisation of specific C. burnetii bacterial isolates and the identification of recent or previous infection indirectly through serological testing. Therefore, both molecular and serological methods were explored during this project to gain insights into the identification and quantification of C. burnetii in beef cattle in northern Australia.

During this thesis, I attempted to identify C. burnetii bacterial DNA that could then be used for further molecular genotyping to characterise strain variation of C. burnetii within Australian beef cattle. Due to the weakly positive results, unfortunately I was unable to use samples for molecular genotyping. In order to improve the chances of success, I would suggest using a more targeted animal testing approach that could focus around notified human Q fever cases and outbreaks. For example, if a case of Q fever was notified and mentioned directly helping at a calving within the month prior to disease onset, this property could be visited and samples collected from humans, animals and environmental sources (dust from paddocks, dust inside buildings and shed, samples from animal bedding,). Molecular genotyping in combination with traditional epidemiological investigations were critical during the Q fever epidemic in the Netherlands, where one predominant genotype of C. burnetii was identified from dairy goat, sheep and human cases (Roest et al. 2011). Additionally, the identification of C. burnetii genotypes causing human Q fever in Cayenne, French Guiana identified the three-toed sloth as the putative reservoir (Million & Raoult 2015). Molecular genotyping to distinguish C. burnetii detected in Australian animal and human samples would help identify common transmission pathways and may provide insights into virulence and elucidate differences in clinical presentations of human Q fever (Vincent et al. 2016).

192

Chapter four described an active surveillance survey at a large, single-site abattoir in South-east Queensland. The objectives for this study were to estimate the prevalence of C. burnetii infection in a population of cattle going to slaughter and obtain bacterial samples of C. burnetii that would enable identification of specific genotypes of C. burnetii infecting cattle in this population. Although serological testing could have been a suitable method to investigate seroprevalence, this would only provide immunological evidence of recent or previous exposure to C. burnetii and would not allow the identification of specific bacterial strains. Therefore, collecting samples for DNA extraction and polymerase chain reaction (PCR) methods was chosen in order to fulfil this objective. Unfortunately, a minimal quantity of C. burnetii DNA was identified during the survey from abattoir samples thus genotyping was not possible. If abattoir surveillance was attempted in another study, I would suggest collecting samples from mammary gland tissue of breeding beef cattle. The mammary glands are reported to be a prime excretion route; therefore, this could increase the chances of retrieving more C. burnetii DNA.

Placental tissue and liver samples were collected and tested for C. burnetii bacterial markers using well-established gene targets (Lockhart et al. 2011; Tozer et al. 2014; Klee et al. 2006; Bond et al. 2015). Initial molecular testing of placental samples returned no positives. Therefore, the study design was revised and liver samples were then collected for testing. To the best of our knowledge, this is the first time that liver samples have been used for a focused surveillance of C. burnetii infection in cattle. Although there was minimal literature reporting where the bacteria persist in non-gravid animals, in human patients with chronic forms of Q fever, C. burnetii is commonly identified in the liver (Eldin et al. 2016; Million & Raoult 2015). There has been preliminary evidence from wildlife and roadkill samples to hypothesise liver as a suitable tissue sample for detection of C. burnetii in animals (Cumbassá et al. 2015). In light of the above-mentioned limitations, it was difficult to confidently determine if the low positive test results were a reflection of a true low prevalence of infection or that liver was not the ideal tissue choice for sampling. All except one of the 6 positive samples had Ct values greater than 38.0 for gene target IS1111 and none of the samples were detected positive using heat shock protein target. These samples could be false positives and one could argue that none of the animals sampled were convincingly positive. Therefore, I am cautious in making strong inferences on prevalence from these results.

As described in chapter seven of this thesis, the same PCR methods were used to test vaginal swab samples that were collected from breeding beef cattle during a previous 193 study (McGowan et al. 2014). Similarly, like the abattoir study, any positive PCR samples had extremely low C. burnetii DNA content and were not suitable for additional genotyping techniques. If strong positive PCR results (Cycle threshold (Ct) < 30) had been identified, these would have undergone further laboratory analysis for molecular genotyping. In this particular situation, the likely reason that the vaginal swab samples had a low bacterial load could be due to the timing of sample collection relative to parturition. Therefore, this could be described as a form of misclassification bias and/or selection bias (Dohoo, W. Martin & Stryhn 2009b). The swabs were collected at the time of pregnancy diagnosis, in order to screen for a range of infectious organisms that may cause reduced reproductive performance in cattle. However, it has been reported that even if cattle are infected with C. burnetii, vaginal excretion of the bacteria may often not occur until the time of parturition or abortion and may then persist for subsequent weeks to months (Guatteo, Joly & Beaudeau 2012). Thus, using samples collected for other purposes, and not timing sample collection to provide an optimal ability to detect C. burnetii may have led to biased false negative results. Identification and investigation into the local genetic make-up of C. burnetii bacteria in Australian cattle is still needed in order to improve our knowledge of disease transmission, aid in investigating outbreaks and to determine virulence factors that may contribute to animal and human clinical presentations.

It has been highlighted in global literature that there are discrepancies with the testing and interpretation of immunological exposure to C. burnetii in ruminants (Kittelberger et al. 2009; Rousset et al. 2007). Different sampling protocols and minimal standardisation of diagnostic test methods validated specifically for use in cattle have led to difficulties with comparing seroprevalence results between regions and countries. Chapter four of this thesis described the optimisation and validation of an in-house indirect fluorescent antibody test (IFA) for the serological detection of C. burnetii antibodies in cattle. The validation of affordable and reliable serological tests was intended to encourage surveillance and epidemiological studies of bovine coxiellosis in Australia and other countries. Although the IFA was found to be less sensitive than originally expected, it was a sufficient method to allow large scale testing to be completed at an affordable price. There is much room for more work to be focused on improving diagnostic test methods for exposure to C. burnetii in ruminants. If this component of the thesis was to be restarted, I would investigate developing an in-house ELISA, coated separately with phase 1 and phase 2 antigen. The IFA method does not have an automated reading method, thus a lot of time was spent performing individual readings of the IFA sample wells in this assay. An ELISA testing

194 approach could reduce time through automation of test readings with a microplate reader. With an easily automated serological test that was able to determine separate phase 1 and phase 2 antibody responses, more research could be focused on interpretation of serological profiles in relation to infection status and shedding patterns could be performed.

Within chapter six, an extensive collection of breeding beef cattle serum samples was tested for anti- C. burnetii IgG antibodies using the recently validated IFA. The objectives of this study were to investigate the prevalence of C. burnetii exposure in commercial beef cattle breeding herds across northern Australia and estimate the true prevalence and spatial distribution at that time point. Based on differences in historical human notification rates between states/territories, it was hypothesised that there may be an associated regional difference in cattle C. burnetii exposure between cattle farmed in the Northern Territory and Queensland. Hierarchical Bayesian latent class models were used to estimate the true prevalence of C. burnetii exposure adjusted for diagnostic test uncertainty. To our knowledge, this is the first time that cattle within the Northern Territory have been tested for C. burnetii exposure. The estimated true prevalence of cattle on a typical farm in the Northern Territory was found to be lower than a typical farm in most regions within Queensland, although cattle in the south-east Queensland region were similar to the Northern Territory. While these results are interesting and informative, they should not be used to infer causation or directly link beef cattle as the source of human Q fever. I have identified that cattle tested from the Northern Territory farming regions showed minimal exposure to C. burnetii. It is possible that they were regionally free of disease; however, considering the DSe (approximately 74%) of the IFA and the non-representative sampling, I cannot confidently conclude freedom of disease from this study.

The statistical model presented in this study incorporated both imperfections of the diagnostic test used and the hierarchical data structure of the cattle populations into the final true prevalence estimates. The structure of this model may be useful to provide assistance with infectious disease prevalence estimates in future studies. It could also be utilised during passive surveillance of C. burnetii (or other infections) within animal populations, thus enabling a more accurate interpretation of serological test results from serum banks or samples collected for other purposes. In order to estimate true prevalence of infectious disease, study design, sampling methods and diagnostic test inaccuracies need to be considered during the analysis (Rogan & Gladen 1978; Suess, Gardner & Johnson 2002). If the hierarchical structure of the sample population is not acknowledged, as is evident in many laboratory-based research studies, results may be reported with tighter confidence 195 intervals than are reasonable. This may lead to overconfidence in the precision of the estimated prevalence and spurious associations or significance attributed where it is not warranted (Dohoo, S. W. Martin & Stryhn 2009). Statistical methods applied in this study may enable improved comparisons between prevalence estimates both within and between regions/countries and may allow further analysis into putative risk factors of C. burnetii in cattle in Australia.

When considered together, chapter four and chapter five have provided tools and baseline data that will enable greater surveillance and understanding of the epidemiology of C. burnetii in beef cattle. This is required in order to make informed decisions about the potential effect of coxiellosis on animal health, industry and public health in Australia. Coxiellosis has not previously been examined as an infectious disease of importance by the Australian red meat industry (Lane et al. 2015). Therefore, there has not been any investigation into the potential effect of C. burnetii on production or reproductive performance in beef cattle. This may be an important consideration, bearing in mind that more Australian farmers are involved in beef cattle farming than any other agricultural activity and the significance of this industry to the national economy. The cattle industries (both beef and dairy) revolve around each female animal producing one calf per year, therefore the entire population of breeding females are expected to be pregnant annually in order to contribute either a new animal for meat production or lactation for milking.

The final research chapter in this thesis (chapter seven) was carried out to investigate if C. burnetii exposure was associated with reduced reproductive performance in a sample of breeding cattle from Queensland and the Northern Territory. International publications have presented mixed evidence describing the clinical effects in cattle ranging from asymptomatic disease to both improved and reduced reproductive performance in cattle. In this chapter, a curvilinear relationship between C. burnetii exposure and annual pregnancy rate was identified; this pattern may fit with the varied reports. These preliminary results are hypothesis forming and prospective studies that could examine this relationship in more depth should be designed. Although the preliminary results from chapter seven are interesting and novel in Australia, they should be interpreted with caution, as there are unknown risk factors and potential confounders not considered within this analysis that may interfere with the noted pattern. Ecological, environmental and additional property level factors may need further assessment to firstly identify if C. burnetii exposure is merely a proxy itself for harsh geographical landscape or poor management or exposure to wildlife or if it can be causally associated with reduced reproductive performance. It is known that 196

Australian wildlife and feral animals have tested positive for C. burnetii DNA and high seroexposure and may play a role in the increased exposure and transmission of C. burnetii to cattle farmed in certain regions or in specific ecosystems (Cooper et al. 2012, 2013; Potter et al. 2011). An additional study designed to investigate risk factors associated with differences in C. burnetii exposure is also needed and may require in-depth spatio-temporal analysis that incorporates geographical and seasonal environmental factors.

The detailed mechanisms of spread of infection between reservoir hosts, livestock and humans have not been well documented in Australia and there is no clear or convincing model of transmission. However, it is plausible that the natural reservoir of C. burnetii in Australia is native wildlife, including bandicoots and macropods, as suggested by Edward Derrick himself during early investigations and supported by more recent studies identifying high seroexposure and positive bacterial DNA in wildlife (Derrick, Pope & Smith 1959; Cooper et al. 2012, 2013). It is also plausible that environmental conditions and the presence of multiple hosts have produced a nosogenic territory for hyperendemic C. burnetii in Queensland. Changes in ecosystems from human activity can modify nosogenic patterns resulting in the emergence of infectious disease. Beef cattle in the northern regions of Australia are extensively managed, therefore they have increased opportunity for interaction with native reservoir hosts and C. burnetii in the environment; thus enabling infectious contact to occur. The northern beef cattle industry rapidly increased in size during the early 1900’s, preceding the emergence of Query fever in the 1930s. While there is no doubt that human Q fever can be attributed to direct contact with cattle, in many parts of the world, cattle are not acknowledged as a significant source of human Q fever. Therefore, are cattle in Queensland a more significant source of human Q fever than in other areas regions, or should cattle be considered merely a sentinel to indicate increased environmental contamination or risk of C. burnetii exposure in the region? It may be that Queensland is in fact a nosogenic territory that has ecological, social and environmental conditions that support C. burnetii, and cattle play a role in this.

During this thesis, I have separately investigated the distribution of human Q fever cases in Queensland and the serological exposure of C. burnetii in a sample of beef cattle, I did not perform formal analyses of combined human and animal datasets. It was considered inappropriate to perform statistical testing for associations between the two different datasets. Firstly, the human case notification data was available across a wide- timeframe, whereas the cattle data presented was from a single point in time over 2011.

197

Figure 8.1 Choropleth map displaying human Q fever notification rate/5-years, for the time-period of 2009–2013, aggregated by local government area; overlayed with within- property beef cattle true prevalence estimates as points.

The human data was of case notifications of disease, whereas the cattle data was serological exposure to C. burnetii. Incidence of human disease was estimated using notification rate per year or per 5 year, using estimated residential population data as the denominator and Q fever cases aggregated by local government areas (LGA) as the numerator. There were some concerns of the external validity of the cattle data to provide an unbiased estimate for the broader beef cattle populations. However, I would still like to present here in the general discussion, a basic overlay of the cattle property true prevalence results in chapter seven combined with a choropleth map of Queensland human notification rates drawn from records analysed in chapter three (Figure 8.1). It is interesting to see that the LGA with the highest accumulative 5-year incidence (2009–2013) of Q fever contains a property with high property true prevalence of exposure. However, at the state level, there 198 is no obvious pattern as the properties show great variability in true prevalence. There are many LGAs that did not include cattle properties enrolled in the CashCow study, therefore there was no serological evidence of cattle exposure within these LGAs.

In summary, higher serological exposure identified in cattle from areas of Queensland compared to the Northern Territory followed the same crude pattern as human Q fever notifications, which could be misinterpreted to suggest a causal association. Molecular genotyping of cattle derived samples from northern Australia was not successful and therefore unable to provide additional evidence of the nature of this relationship. Queensland human public health surveillance data suggested some contact with cattle was evident in half of the notified Queensland Q fever cases, however there were confounders and multiple exposures that need to be considered and accounted for before making inferences from this dataset. Although low PCR in vaginal swabs collected at pregnancy diagnosis could suggest that close contact with pregnant cattle during early stages of gestation would appear to have a low risk of direct transmission to humans, this would need to be confirmed with prospective longitudinal studies. Additionally, very low amounts of bacterial DNA were found in abattoir samples, again suggesting that beef cattle may not pose a serious threat for human Q fever. Although I was not able to find evidence to confirm beef cattle as significant sources of human Q fever, there is still a need for more in-depth landscape epidemiology and studies to determine ecological factors that affect the occurrence, maintenance and transmission of disease in northern Australia.

Research undertaken during this thesis has provided a solid foundation for the examination of coxiellosis in beef cattle of northern Australia; however, more research is required to provide deeper insights into specific areas of this topic. The final part of this discussion presents suggestions of several areas where additional research would be beneficial. There is still an information gap of the prevalence and distribution of coxiellosis in other livestock and ruminants within Australia. There is a paucity of current research into coxiellosis specifically in the Australian dairy-cattle industry. I would suggest more work to assess animal health and welfare, potential production effects and improve our understanding of the public health risks for dairy workers. To begin with, basic seroprevalence studies in dairy-cattle, goats and sheep would help improve the general knowledge of exposure and identify suspected infected herds across geographical regions. Knowledge of exposure would then enable identification of areas that may have increased risk of infection. From this current research, I have identified a regional disparity of prevalence between beef cattle in the Northern Territory and regions of Queensland. It would 199 be possible to design studies that purposefully use the regional disparity to identify differences that may in turn shed light on factors driving exposure risk. A prospective case- control study where notified human cases are compared to suitable controls and both are investigated to measure their contamination and risk profiles could be interesting. However, care would be required to define appropriate controls in this context. In cattle herds, prospective longitudinal studies, designed to follow naturally infected herds within these regions, could additionally allow assessment of immunological responses and bacterial shedding patterns. Subsequent animal and environmental sampling, for molecular genotyping, could also be focused towards these areas identified as having higher animal exposure. It may be possible to design studies that incorporate molecular genotyping techniques, to monitor the direction of transmission of infection between wildlife, domestic animals, and the environment and thus endeavour to elucidate the common reservoirs for human Q fever.

Additional work could focus towards areas with high incidence of human Q fever or surrounding family/community clusters. Within Queensland, each district Public Health Unit investigates notified Q fever cases with an enhanced surveillance questionnaire in order to gain further insight into risk factors for Coxiella exposure. Contact with animals and high- risk exposures are identified as likely transmission points and recorded, however no further field investigations take place to identify potential sources of infection with laboratory testing. This constitutes a significant gap in the current investigation protocol, particularly in the context of family clusters where source attribution is paramount. There is an opportunity to work with the current Queensland public health teams and develop a framework for a One Health surveillance of Q fever. As highlighted by Zinsstag et al. (2015), well designed integrated investigations or surveillance activities will provide more significant value than studies performed separately by veterinary or human health. This reference promotes One Health by discussing its application for surveillance and control of Brucellosis, another important zoonotic disease. While this is likely not-financially practical as a government funded surveillance system, research-driven funding could be sourced to provide proof of concept and develop this platform. This proposed project would utilise an integrated approach, which incorporates human, animal, and environmental outbreak investigations within identified familial clusters of Q fever in regions of Queensland. This would include follow-up sample collection from animals and the environment within the household of family clusters to identify serological exposure in animals and identify C. burnetii bacteria in animals and the environment.

200

Finally, there is a need to further investigate the hypothesis that C. burnetii infection can cause reduced reproductive performance in cattle. There is global evidence that this infectious disease could affect the reproductive performance in beef and dairy cattle. Results from this thesis have identified a potential association with sero-exposure and a reduction in fertility in beef cattle. However, additional work such as a well-designed cohort study or case-control study is necessary to further explore this research question. Within the context of northern Australia, these studies may be better suited to dairy herds as they are not extensively managed and there is a better ability to collect good observational data. Most beef properties will only physically observe cattle at 6 or 12-monthly musters and there is no close observation during the time of calving. For example, it is difficult to assess if abortions have occurred in an extensive beef cattle herd. Foetal calf loss can be extrapolated by the absence of a calf at the weaning muster from a cow that was previously identified as pregnant, although there are many assumptions within this reasoning. Although it is possible that the foetus was aborted, it could have been born healthy and then lost to predators as a neonate. Therefore, the ability to collect suitable data to assess reproductive performance in beef cattle in northern Australia is logistically difficult.

Although Q fever is not a new disease, it is still an important disease in Australia and experts in the field highlight it as a re-emerging zoonotic infection globally (Eldin et al. 2016; Gwida, El-Ashker & Khan 2012; Million & Raoult 2015; Arricau-Bouvery & Rodolakis 2005). Coxiellosis is an underreported and under-investigated infection in animals within Australia. Research outcomes from this thesis have explored epidemiological aspects of the infection within the Australian context that may encourage One Health research in the field, ideally using collaborative cross-disciplinary approaches that are well suited to zoonotic diseases. Current control practices are isolated to human vaccination without any control programs nor licensed vaccines in animals. However, further study of coxiellosis in farmed livestock and wildlife as potential reservoirs of infection will improve our ability to reduce the impact of Q fever to public health and improve health and productivity within the agricultural industries.

201

9. List of References Abramson, JH 2011, ‘WINPEPI updated: computer programs for epidemiologists, and their teaching potential’, Epidemiologic Perspectives & Innovations, vol. 8, no. 1, pp. 1–9.

Agerholm, JS 2013, ‘Coxiella burnetii associated reproductive disorders in domestic animals-a critical review.’, Acta Veterinaria Scandinavica, vol. 55, no. 13, pp. 1–13.

Agerholm, JS, Jensen, TK, Agger, JF, Engelsma, MY & Roest, HIJ 2017, ‘Presence of Coxiella burnetii DNA in inflamed bovine cardiac valves’, BMC Veterinary Research, vol. 13, no. 69, pp. 1–7.

Agger, JF & Paul, S 2014, ‘Increasing prevalence of Coxiella burnetii seropositive Danish dairy cattle herds’, Acta Veterinaria Scandinavica, vol. 56, no. 46, pp. 1–4.

Alvarez, J, Perez, A, Mardones, FO, Pérez-Sancho, M, García-Seco, T, Pagés, E, Mirat, F, Díaz, R, Carpintero, J & Domínguez, L 2012, ‘Epidemiological factors associated with the exposure of cattle to Coxiella burnetii in the Madrid region of Spain’, The Veterinary Journal, vol. 194, no. 1, pp. 102–7.

Alvarez, J, Whitten, T, Branscum, AJ, Garcia-Seco, T, Bender, JB, Scheftel, J & Perez, A 2018, ‘Understanding Q fever risk to humans in Minnesota through the analysis of spatiotemporal trends’, Vector Borne Zoonotic Diseases, vol. 18, pp. 89–95.

Amara, A, Bechah, Y & Mege, JL 2012, ‘Immune response and Coxiella burnetii invasion’, in Coxiella burnetii: Recent Advances and New Perspectives in Research of the Q Fever Bacterium, Springer Netherlands, Dordrecht, pp. 287–298.

Anderson, M 2011, ‘Disorders of cattle’, in Kirkbrides’s Diagnosis of Abortion and Neonatal Loss in Animals, Wiley-Blackwell, Oxford, pp. 13–48.

Archer, B, Hallahan, C, Stanley, P, Seward, K, Lesjak, M, Hope, K & Brown, A 2017, ‘Atypical outbreak of Q fever affecting low risk residents of a remote rural town in New South Wales’, Communicable Diseases Intelligence, vol. 41, no. 2, pp. 125–133.

Armstrong, M, Francis, J, Robson, J, Graves, S, Mills, D, Ferguson, J & Nourse, C 2019, ‘Q fever vaccination of children in Australia: Limited experience to date’, Journal of Paediatrics and Child Health, vol. 55, no. 9, pp. 1099–1102.

Arricau-Bouvery, N, Hauck, Y, Bejaoui, A, Frangoulidis, D, Bodier, CC, Souriau, A, Meyer,

202

H, Neubauer, H, Rodolakis, A & Vergnaud, G 2006, ‘Molecular characterization of Coxiella burnetii isolates by infrequent restriction site-PCR and MLVA typing’, BMC Microbiology, vol. 6, no. 38, pp. 1–14.

Arricau-Bouvery, N & Rodolakis, A 2005, ‘Is Q fever an emerging or re-emerging zoonosis?’, Veterinary Research, vol. 36, no. 3, pp. 327–349.

Astobiza, I, Tilburg, JJ, Piñero, A, Hurtado, A, García-Pérez, AL, Nabuurs-Franssen, MH & Klaassen, CH 2012, ‘Genotyping of Coxiella burnetii from domestic ruminants in northern Spain’, BMC Veterinary Research, vol. 8, no. 241, pp. 1–8.

‘Australian Bureau of Statistics, Australian Government’ 2019, retrieved April 17, 2019, from .

Australian Government 2019, ‘National Notifiable Diseases Surveillance System’, retrieved February 20, 2019, from .

AusVet 2006, ‘A review of the structure and dynamics of the Australian beef cattle industry’, Department of Agriculture, Forest and Fisheries, pp. 1–93.

Babudieri, B 1959, ‘Q fever: a zoonosis’, Advances in Veterinary Sciences, vol. 5, pp. 82– 154.

Banazis, MJ, Bestall, AS, Reid, SA & Fenwick, SG 2010, ‘A survey of Western Australian sheep, cattle and kangaroos to determine the prevalence of Coxiella burnetii’, Veterinary Microbiology, vol. 143, no. 2–4, pp. 337–345.

Barralet, JH & Parker, NR 2004, ‘Q fever in children: an emerging public health issue in Queensland’, Medical Journal of Australia, vol. 180, no. 11, pp. 596–597.

Bell, E., Parker, R. & Stoenner, H. 1949, ‘Q fever; Experimental Q fever in cattle’, The American Journal of Public Health, vol. 39, pp. 478–484.

Bialasiewicz, S, Whiley, DM, Buhrer-Skinner, M, Bautista, C, Barker, K, Aitken, S, Gordon, R, Muller, R, Lambert, SB, Debattista, J, Nissen, MD & Sloots, TP 2008, ‘A novel gel- based method for self-collection and ambient temperature postal transport of urine for PCR detection of Chlamydia trachomatis’, Sexually Transmitted Infections, vol. 85, no. 2, pp. 102–105.

De Biase, D, Costagliola, A, Piero, F Del, Palo, R Di, Coronati, D, Galiero, G, Uberti, BD, Lucibelli, MG, Fabbiano, A, Davoust, B, Raoult, D & Paciello, O 2018, ‘Coxiella burnetii in

203 infertile dairy cattle with chronic endometritis’, Veterinary Pathology, vol. 55, no. 4, pp. 539–542.

Bildfell, RJ, Thomson, GW, Haines, DM, McEwen, BJ & Smart, N 2000, ‘Coxiella burnetii infection is associated with placentitis in cases of bovine abortion’, Journal of Veterinary Diagnostic Investigations, vol. 12, no. 5, pp. 419–25.

Bond, KA, Franklin, L, Sutton, B, Stevenson, MA & Firestone, SM 2018, ‘Review of 20 years of human acute Q fever notifications in Victoria, 1994–2013’, Australian Veterinary Journal, vol. 96, no. 6, pp. 223–230.

Bond, KA, Vincent, G, Wilks, CR, Franklin, L, Sutton, B, Stenos, J, Cowan, R, Lim, K, Athan, E, Harris, O, Macfarlane-Berry, L, Segal, Y & Firestone, SM 2015, ‘One Health approach to controlling a Q fever outbreak on an Australian goat farm’, Epidemiology and Infection, vol. 144, no. 6, pp. 1–13.

Bontje, DM, Backer, JA, Hogerwerf, L, Roest, HI & van Roermund, HJ 2016, ‘Analysis of Q fever in Dutch dairy goat herds and assessment of control measures by means of a transmission model’, Preventive Veterinary Medicine, vol. 123, pp. 71–89.

Boroduske, A, Trofimova, J, Kibilds, J, Papule, U, Sergejeva, M, Rodze, I & Grantina- Ievina, L 2017, ‘Coxiella burnetii (Q fever) infection in dairy cattle and associated risk factors in Latvia’, Epidemiology and Infection, vol. 145, no. 10, pp. 2011–2019.

Böttcher, J, Vossen, A, Janowetz, B, Alex, M, Gangl, A, Randt, A & Meier, N 2011, ‘Insights into the dynamics of endemic Coxiella burnetii infection in cattle by application of phase-specific ELISAs in an infected dairy herd’, Veterinary Microbiology, vol. 151, no. 3– 4, pp. 291–300.

Branscum, AJ, Gardner, IA & Johnson, WO 2005, ‘Estimation of diagnostic-test sensitivity and specificity through Bayesian modeling’, Preventive Veterinary Medicine, vol. 68, pp. 145–163.

Britton, PN, Macartney, K, Arbuckle, S, Little, D & Kesson, A 2015, ‘A rare case of Q fever osteomyelitis in a child from regional Australia’, Journal of the Pediatric Infectious Diseases Society, vol. 4, no. 3, pp. e28–e31.

Brom, R Van den, Engelen, E van, Roest, HIJ, Hoek, W van der & Vellema, P 2015, ‘Coxiella burnetii infections in sheep or goats: an opinionated review’, Veterinary Microbiology, vol. 181, no. 1–2, pp. 119–129.

204

Brown, LD, Cai, TT & Das Gupta, A 2001, ‘Interval estimation for a binomial proportion’, Statistical Science, vol. 16, no. 2, pp. 101–117.

Burnet, FM & Freeman, M 1937, ‘Experimental studies on the virus of “Q” fever’, Medical Journal of Australia, vol. 2, no. 8, pp. 299–305.

Byrt, T, Bishop, J & Carlin, JB 1993, ‘Bias, prevalence and kappa’, Journal of Clinical Epidemiology, vol. 46, no. 5, pp. 423–429.

Cabassi, CS, Taddei, S, Donofrio, G, Ghidini, F, Piancastelli, C, Flammini, CF & Cavirani, S 2006, ‘Association between Coxiella burnetii seropositivity and abortion in dairy cattle of Northern Italy’, New Microbiologica, vol. 29, pp. 211–214.

Cameron, AR 1999, Survey toolbox : a practical manual and software package for active surveillance of livestock diseases in developing countries, Australian Centre for International Agricultural Research, Canberra.

Canevari, JT, Firestone, SM, Vincent, G, Campbell, A, Tan, T, Muleme, M, Cameron, A & Stevenson, MA 2018, ‘The prevalence of Coxiella burnetii shedding in dairy goats at the time of parturition in an endemically infected enterprise and associated milk yield losses’, BMC Veterinary Research, vol. 14, no. 353, pp. 1–9.

Capuano, F, Landolfi, MC & Monetti, DM 2001, ‘Influence of three types of farm management on the seroprevalence of Q fever as assessed by an indirect immunofluorescence assay’, Veterinary Record Record, vol. 149, pp. 669–671.

Carbonero, A, Guzman, LT, Montano, K, Torralbo, A, Arenas-Montes, A, Saa, LR, Guzmán, LT, Montaño, K, Torralbo, A, Arenas-Montes, A & Saa, LR 2015, ‘Coxiella burnetii seroprevalence and associated risk factors in dairy and mixed cattle farms from Ecuador’, Preventive Veterinary Medicine, vol. 118, no. 4, pp. 427–435.

Carpenter, TE, Chrièl, M, Andersen, MM, Wulfson, L, Jensen, AM, Houe, H & Greiner, M 2006, ‘An epidemiologic study of late-term abortions in dairy cattle in Denmark, July 2000– August 2003’, Preventive Veterinary Medicine, vol. 77, no. 3–4, pp. 215–229.

CDNA 2018, ‘Q fever: CDNA National Guidelines for Public Health Units’, Communicable Diseases Network Australia, pp. 1–31.

Ceglie, L, Guerrini, E, Rampazzo, E, Barberio, A, Tilburg, JJHC, Hagen, F, Lucchese, L, Zuliani, F, Marangon, S & Natale, A 2015, ‘Molecular characterization by MLVA of Coxiella burnetii strains infecting dairy cows and goats of north-eastern Italy’, Microbes and 205

Infection, vol. 17, no. 11–12, pp. 776–781.

Chaintarli, K & Upton, P 2018, ‘Analysis of bovine tuberculosis surveillance at routine slaughter of cattle in Great Britain 2013-2016’, Department of Epidemiological Sciences, APHA (Weybridge), pp. 1–100.

Chapman, PA 2000, ‘Sources of Escherichia coli O157 and experiences over the past 15 years in Sheffield, UK’, Symposium series (Society for Applied Microbiology), no. 29, pp. 51S-60S.

Clark, NJ & Soares Magalhães, RJ 2018, ‘Airborne geographical dispersal of Q fever from livestock holdings to human communities: a systematic review and critical appraisal of evidence’, BMC Infectious Diseases, vol. 18, no. 1, pp. 1–9.

Clark, NJ, Tozer, S, Wood, C, Firestone, SM, Stevenson, M, Caraguel, C, Chaber, A, Heller, J & Soares Magalhães, RJ 2020, ‘Unravelling animal exposure profiles of human Q fever cases in Queensland, Australia, using natural language processing’, Transboundary and Emerging Diseases, vol. 67, no. 5, pp. 2133–2145.

Clutterbuck, HC, Eastwood, K, Massey, PD, Hope, K & Mor, SM 2018, ‘Surveillance system enhancements for Q fever in NSW, 2005-2015’, Communicable diseases intelligence, vol. 42, no. 18, pp. 2005–2015.

Collins, J & Huynh, M 2014, ‘Estimation of diagnostic test accuracy without full verification: a review of latent class methods’, Statistics in Medicine, vol. 33, no. 24, pp. 4141–4169.

Cooke, RA 2008, ‘Q fever. Was Edward Derrick’s contribution undervalued?’, Medical Journal of Australia, vol. 189, no. 11–12, pp. 660–662.

Cooper, A, Goullet, M, Mitchell, J, Ketheesan, N & Govan, B 2012, ‘Serological evidence of Coxiella burnetii exposure in native marsupials and introduced animals in Queensland, Australia’, Epidemiology and Infection, vol. 140, no. 7, pp. 1304–1308.

Cooper, A, Hedlefs, R, McGowan, M, Ketheesan, N & Govan, B 2011, ‘Serological evidence of Coxiella burnetii infection in beef cattle in Queensland’, Australian Veterinary Journal, vol. 89, no. 7, pp. 260–264.

Cooper, A, Layton, R, Owens, L, Ketheesan, N & Govan, B 2007, ‘Evidence for the classification of a crayfish pathogen as a member of the genus Coxiella’, Letters in Applied Microbiology, vol. 45, no. 5, pp. 558–563.

206

Cooper, A, Stephens, J, Ketheesan, N & Govan, B 2013, ‘Detection of Coxiella burnetii DNA in wildlife and ticks in northern Queensland, Australia’, Vector Borne Zoonotic Diseases, vol. 13, no. 1, pp. 12–16.

Cooper, AE 2011, ‘Identification of potential reservoirs of Q fever in Queensland, Australia’, Thesis (PhD) James Cook University, Australia, pp. 1–340.

Cowled, B, Sergeant, E, Kennedy, D, Gordon, R & Limited, M& LA 2013, ‘Endemic diseases scoping project’, Meat and Livestock Australia (report B.AHE.0226), pp. 1–114.

Cronin, N 2015, An investigation into the serological evidence of Coxiella burnetii infection in beef cattle in the Lachlan Livestock Health and Pest Authority in NSW, Forbes, http://www.flockandherd.net.au/cattle/reader/q-fever.html.

Cumbassá, A, Barahona, MJ, Cunha, M V., Azórin, B, Fonseca, C, Rosalino, LM, Tilburg, J, Hagen, F, Santos, AS & Botelho, A 2015, ‘Coxiella burnetii DNA detected in domestic ruminants and wildlife from Portugal’, Veterinary Microbiology, vol. 180, no. 1–2, pp. 136– 141.

Davis, GE, Cox, HR, Parker, RR & Dyer, RE 1938, ‘A filter-passing infectious agent isolated from ticks’, Public Health Reports, vol. 53, no. 52, pp. 2259–2282.

Delsing, C., Kullberg, B. & Bleeker-Rovers, C. 2010, ‘Q fever in the Netherlands from 2007 to 2010’, The Netherlands Journal of Medicine, vol. 68, no. 12, pp. 382–387.

Derrick, E., Smith, DJ. & Brown, H. 1942, ‘Studies in the epidemiology of Q fever: The role of the cow in the transmission of human infection’, The Journal of Experimental Biology and Medical Science, vol. 20, pp. 105–110.

Derrick, EH 1937, ‘“ Q” Fever, a new fever entity: clinical features, diagnosis and laboratory investigation’, Medical Journal of Australia, vol. 2, pp. 287–299.

Derrick, EH 1961, ‘The changing pattern of Q fever in Queensland’, Pathologia et microbiologia, vol. 24(Suppl), pp. 73–79.

Derrick, EH, Pope, J & Smith, D 1959, ‘Outbreak of Q fever in Queensland associated with sheep’, Medical Journal of Australia, vol. 2, pp. 585–588.

Dohoo, I, Martin, SW & Stryhn, H 2009, ‘Screening and Diagnostic Tests’, in Veterinary Epidemiologic Research, VER Inc, Charlottetown, Prince Edward Island, pp. 91–134.

Dohoo, I, Martin, W & Stryhn, H 2009a, ‘Model-building strategies’, in Veterinary

207

Epidemiologic Research, VER Inc, Charlottetown, Prince Edward Island, pp. 365–393.

Dohoo, I, Martin, W & Stryhn, H 2009b, ‘Validity in observational studies’, in Veterinary Epidemiologic Research, VER Inc, Charlottetown, Prince Edward Island, pp. 243–270.

Durham, PJK & Paine, GD 1997, ‘Serological survey for antibodies to infectious agents in beef cattle in northern South Australia’, Australian Veterinary Journal, vol. 75, no. 2, pp. 139–140.

Duron, O 2015, ‘The IS1111 insertion sequence used for detection of Coxiella burnetii is widespread in Coxiella-like endosymbionts of ticks’, Federation of European Microbiological Societies Microbiology Letters, vol. 362, no. 17, pp. 1–8.

Duron, O, Noël, V, Mccoy, KD, Bonazzi, M, Sidi-Boumedine, K, Morel, O, Vavre, F, Zenner, L, Jourdain, E, Durand, P, Arnathau, C, Renaud, F, Trape, J-F, Biguezoton, AS, Cremaschi, J, Dietrich, M, Léger, E, Appelgren, A, Dupraz, M, Gómez-Díaz, E, Diatta, G, Dayo, G-K, Adakal, H, Zoungrana, S, Vial, L, Chevillon, C, Dicko, D & Koulodo, D 2015, ‘The recent evolution of a maternally-inherited endosymbiont of ticks led to the emergence of the Q fever pathogen, Coxiella burnetii’, PLos Pathogens, vol. 11, no. 5.

Eastwood, K, Graves, SR, Massey, PD, Bosward, K & Hutchinson, P 2018, ‘Q fever: A rural disease with potential urban consequences’, Australian Journal of General Practice, vol. 47, no. 3, pp. 112–116.

El-Mahallawy, HS, Kelly, P, Zhang, J, Yang, Y, Zhang, H, Wei, L, Mao, Y, Yang, Z, Zhang, Z, Fan, W & Wang, C 2016, ‘High seroprevalence of Coxiella burnetii in dairy cattle in China’, Tropical Animal Health and Production, vol. 48, no. 2, pp. 423–426.

Eldin, C, Mélenotte, C, Mediannikov, O, Ghigo, E, Million, M, Edouard, S, Mege, J-L, Maurin, M & Raoult, D 2016, ‘From Q fever to Coxiella burnetii infection: a paradigm change’, Clinical Microbiology Reviews, vol. 30, no. 1, pp. 115–190.

Elsa, J, Duron, O, Severine, B, Gonzalez-Acuna, D & Sidi-Boumedine, K 2015, ‘Molecular methods routinely used to detect Coxiella burnetii in ticks cross-react with Coxiella-like bacteria’, Infection Ecology and Epidemiology, vol. 5, pp. 1–6.

Emery, MP, Ostlund, EN & Schmitt, BJ 2012, ‘Comparison of Q fever serology methods in cattle, goats, and sheep’, Journal of Veterinary Diagnostic Investigations, vol. 24, no. 2, pp. 379–382.

Entwistle, KW 1983, ‘Factors influencing reproduction in beef cattle in Australia’, Australian 208

Meat Reseach Committee, pp. 1–30.

Fenner, F 2010, ‘Deliberate introduction of the European rabbit, Oryctolagus cuniculus, into Australia’, Revue scientifique et technique (International Office of Epizootics), vol. 29, no. 1, pp. 103–11.

Flannagan, RS, Cosío, G & Grinstein, S 2009, ‘Antimicrobial mechanisms of phagocytes and bacterial evasion strategies’, Nature Reviews Microbiology, vol. 7, no. 5, pp. 355–366.

Forbes, BR V, Wannan, JS & Keast, JC 1954, ‘A serological survey of cattle in New South Wales for Q fever infection: A preliminary report’, Australian Veterinary Journal, vol. 30, no. 9, pp. 266–268.

Fordyce, G, Anderson, A, McCosker, K, Williams, P, Holroyd, R, Corbet, N & Sullivan, M 2013, ‘Liveweight prediction from hip height, condition score, fetal age and breed in tropical female cattle.’, Animal Production Science, vol. 53, pp. 276–282.

Foster, M 2014, ‘Emerging animal and plant industries: their value to Australia’, Rural Industries Research and Development Corporation, pp. 1–194.

Fournier, PE, Marrie, TJ & Raoult, D 1998, ‘Diagnosis of Q fever’, Journal of Clinical Microbiology, vol. 36, no. 7, pp. 1823–34.

Freick, M, Enbergs, H, Walraph, J, Diller, R, Weber, J & Konrath, A 2017, ‘Coxiella burnetii: Serological reactions and bacterial shedding in primiparous dairy cows in an endemically infected herd—impact on milk yield and fertility’, Reproduction in Domestic Animals, vol. 52, no. 1, pp. 160–169.

Gache, K, Rousset, E, Perrin, JB, De Cremoux, R, Hosteing, S, Jourdain, E, Guatteo, R, Nicollet, P, Touratier, A, Calavas, D & Sala, C 2017, ‘Estimation of the frequency of Q fever in sheep, goat and cattle herds in France: results of a 3-year study of the seroprevalence of Q fever and excretion level of Coxiella burnetii in abortive episodes’, Epidemiology and Infection, vol. 145, pp. 3131–3142.

Garcia-Ispierto, I, López-Helguera, I, Tutusaus, J, Serrano, B, Monleón, E, Badiola, J & López-Gatius, F 2013, ‘Coxiella burnetii shedding during the peripartum period and subsequent fertility in dairy cattle’, Reproduction in Domestic Animals, vol. 48, no. 3, pp. 441–446.

Garcia-Ispierto, I, Tutusaus, J & López-Gatius, F 2014, ‘Does Coxiella burnetii affect reproduction in cattle? A clinical update’, Reproduction in Domestic Animals, vol. 49, no. 4, 209 pp. 529–535.

Garner, M, Longbottom, H, Cannon, R & Plant, A 1997, ‘A review of Q fever in Australia 1991-1994’, Australian and New Zealand Journal of Public Health, vol. 21, no. 7, pp. 722– 730.

Gelman, A & Rubin, D 1992, ‘Inference from iterative simulation using multiple sequences’, Statistical Science, vol. 7, no. 4, pp. 457–511.

Gidding, HF, Peng, CQ, Graves, S, Massey, PD, Nguyen, C, Stenos, J, Quinn, HE, McIntyre, PB, Durrheim, DN & Wood, N 2020, ‘Q fever seroprevalence in Australia suggests one in twenty people have been exposed’, Epidemiology and Infection, vol. 148, pp. 0–4.

Gidding, HF, Wallace, C, Lawrence, GL & McIntyre, PB 2009, ‘Australia’s national Q fever vaccination program’, Vaccine, vol. 27, no. 14, pp. 2037–2041.

Glazunova, O, Roux, V, Freylikman, O, Sekeyova, Z, Fournous, G, Tyczka, J, Tokarevich, N, Kovacava, E, Marrie, TJ & Raoult, D 2005, ‘Coxiella burnetii genotyping’, Emerging Infectious Diseases, vol. 11, no. 8, pp. 1211–1217.

Gleeson, T, Martin, P & Mifsud, C 2012, ‘Northern Australian beef industry: assessment of risks and opportunities’, ABARES report to client prepared for the Northern Australia Ministerial Forum, no. May, p. 168.

González-Barrio, D, Hagen, F, Tilburg, JJHC & Ruiz-Fons, F 2016, ‘Coxiella burnetii genotypes in Iberian wildlife’, Microbial Ecology, vol. 72, no. 4, pp. 890–897.

Gonzalez-Barrio, D, Maio, E, Vieira-Pinto, M & Ruiz-Fons, F 2015, ‘European rabbits as reservoir for Coxiella burnetii’, Emerging Infectious Diseases, vol. 21, no. 6, pp. 1055– 1058.

Greenslade, E, Jennings L, Woodward, A & Weinstein, P 2003, ‘Has Coxiella burnetii (Q fever) been introduced into New Zealand?’, Journal of Clinical Microbiology, vol. 9, no. 1, pp. 2001–2003.

Guatteo, R, Beaudeau, F, Berri, M, Rodolakis, A, Joly, A & Seegers, H 2006, ‘Shedding routes of Coxiella burnetii in dairy cows: implications for detection and control’, Veterinary Research, vol. 37, no. 6, pp. 827–833.

Guatteo, R, Beaudeau, F, Joly, A & Seegers, H 2007, ‘Coxiella burnetii shedding by dairy

210 cows’, Veterinary Research, vol. 38, no. 6, pp. 849–860.

Guatteo, R, Joly, A & Beaudeau, F 2012, ‘Shedding and serological patterns of dairy cows following abortions associated with Coxiella burnetii DNA detection’, Veterinary Microbiology, vol. 155, pp. 430–433.

Guatteo, R, Seegers, H, Joly, A & Beaudeau, F 2008, ‘Prevention of Coxiella burnetii shedding in infected dairy herds using a phase I C. burnetii inactivated vaccine’, Vaccine, vol. 26, no. 34, pp. 4320–4328.

Guatteo, R, Seegers, H, Taurel, AF, Joly, A & Beaudeau, F 2011, ‘Prevalence of Coxiella burnetii infection in domestic ruminants: a critical review’, Veterinary Microbiology, vol. 149, no. 1–2, pp. 1–16.

Gunaratnam, P, Massey, P, Eastwood, K, Durrheim, D, Graves, S, Coote, D, Fisher, L, Durrhein, D, Graves, S, Coote, D & Fisher, L 2014, ‘Diagnosis and management of zoonoses - A tool for general practice’, Australian Family Physician, vol. 43, no. 3, pp. 124–128.

Gwida, M, El-Ashker, M & Khan, I 2012, ‘Q fever: A re-emerging disease?’, Journal of Veterinary Science and Technology, vol. 3, no. 5, pp. 1–5.

Hackstadt, T 1990, ‘The role of lipopolysaccharides in the virulence of Coxiella burnetii’, Annals of the New York Academy of Sciences, vol. 590, pp. 27–32.

Halliday, I 2018, ‘Australian Dairy Industry - In Focus 2018 report’, Dairy Australia, pp. 1– 56, retrieved from .

Hansen, MS, Rodolakis, A, Cochonneau, D, Agger, JF, Christoffersen, AB, Jensen, TK & Agerholm, JS 2011, ‘Coxiella burnetii associated placental lesions and infection level in parturient cows’, Veterinary Journal, vol. 190, no. 2, pp. 135–139.

Harman, JB 1949, ‘Q fever in Great Britain; clinical account of eight cases’, The Lancet, pp. 1028–1030.

Harris, P, Eales, KM, Squires, R, Govan, B & Norton, R 2013, ‘Acute Q fever in northern Queensland: variation in incidence related to rainfall and geographical location’, Epidemiology and Infection, vol. 141, no. 5, pp. 1034–1038.

Harris, RJ, Storm, PA, Lloyd, A, Arens, M & Marmion, BP 2000, ‘Long-term persistence of

211

Coxiella burnetii in the host after primary Q fever’, Epidemiology and Infection, vol. 124, no. 3, pp. 543–549.

Hendrix, LR & Chen, C 2012, ‘Antigenic analysis for vaccines and diagnostics’, in Coxiella burnetii: Recent Advances and New Perspectives in Research of the Q Fever Bacterium, Springer Netherlands, Dordrecht, pp. 299–328.

Herremans, T, Hogema, BM, Nabuurs, M, Peeters, M, Wegdam-Blans, M, Schneeberger, P, Nijhuis, C, Notermans, DW, Galama, J, Horrevorts, A, van Loo, IHM, Vlaminckx, B, Zaaijer, HL, Koopmans, MP, Berkhout, H, Socolovschi, C, Raoult, D, Stenos, J, Nicholson, W & Bijlmer, H 2013, ‘Comparison of the performance of IFA, CFA, and ELISA assays for the serodiagnosis of acute Q fever by quality assessment’, Diagnostic Microbiology and Infectious Disease, vol. 75, no. 1, pp. 16–21.

Hogerwerf, L, van den Brom, R, Roest, HIJ, Bouma, A, Vellema, P, Pieterse, M, Dercksen, D & Nielen, M 2011, ‘Reduction of Coxiella burnetii prevalence by vaccination of goats and sheep, The Netherlands’, Emerging infectious diseases, vol. 17, no. 3, pp. 379–86.

Hogerwerf, L, Koop, G, Klinkenberg, D, Roest, HIJ, Vellema, P & Nielen, M 2014, ‘Test and cull of high risk Coxiella burnetii infected pregnant dairy goats is not feasible due to poor test performance’, The Veterinary Journal, vol. 200, no. 2, pp. 343–5.

Hoover, TA, Culp, DW, Vodkin, MH, Williams, JC & Thompson, HA 2002, ‘Chromosomal DNA deletions explain phenotypic characteristics of two antigenic variants, phase II and RSA 514 (crazy), of the Coxiella burnetii nine mile strain’, Infection and Immunity, vol. 70, no. 12, pp. 6726–6733.

Hore, DE & Kovesdy, L 1972, ‘A serological survey of dairy cattle in Victoria for antibody to Coxiella burnetii’, Australian Veterinary Journal, vol. 48, no. 2, p. 71.

Horigan, MW, Bell, MM, Pollard, TR, Sayers, AR & Pritchard, GC 2011, ‘Q fever diagnosis in domestic ruminants: comparison between complement fixation and commercial enzyme- linked immunosorbent assays’, Journal of Veterinary Diagnostic Investigations, vol. 23, no. 5, pp. 924–931.

Hosmer, DW & Lemeshow, S 2000, Applied Logistic Regression, John Wiley & Sons, Inc., Hoboken, NJ, USA.

IDEXX 2017, ‘Q Fever (C. burnetii) Test Kit’, retrieved November 28, 2017, from .

212

Jado, I, Carranza-Rodríguez, C, Félix Barandika, J, Toledo, Á, García-Amil, C, Serrano, B, Bolaños, M, Gil, H, Escudero, R, García-Pérez, AL, Sonia Olmeda, A, Astobiza, I, Lobo, B, Rodríguez-Vargas, M, Luis Pérez-Arellano, J, López-Gatius, F, Pascual-Velasco, F, Cilla, G, Rodríguez, NF & Anda, P 2012, ‘Molecular method for the characterization of Coxiella burnetii from clinical and environmental samples: variability of genotypes in Spain’, BMC Microbiology, vol. 12, no. 91, pp. 1–10.

Jones, RM, Twomey, DF, Hannon, S, Errington, J, Pritchard, GC & Sawyer, J 2010, ‘Detection of Coxiella burnetii in placenta and abortion samples from British ruminants using real-time PCR’, Veterinary Record, vol. 167, no. 25, pp. 965–967.

Joulié, A, Sidi-Boumedine, K, Bailly, X, Gasqui, P, Barry, S, Jaffrelo, L, Poncet, C, Abrial, D, Yang, E, Leblond, A, Rousset, E & Jourdain, E 2017, ‘Molecular epidemiology of Coxiella burnetii in French livestock reveals the existence of three main genotype clusters and suggests species-specific associations as well as regional stability’, Infection, Genetics and Evolution, vol. 48, pp. 142–149.

Kaplan, MM & Bertagna, P 1955, ‘The geographical distribution of Q fever’, Bulletin of the World Health Organization, vol. 13, no. 5, pp. 829–60.

Kittelberger, R, Mars, J, Wibberley, G, Sting, R, Henning, K, Horner, GW, Garnett, KM, Hannah, MJ, Jenner, JA & Pigott, CJ 2009, ‘Comparison of the Q-fever complement fixation test and two commercial enzyme-linked immunosorbent assays for the detection of serum antibodies against Coxiella burnetii (Q-fever) in ruminants: Recommendations for use of serological tests on importe’, New Zealand Veterinary Journal, vol. 57, no. 5, pp. 262–268.

Klee, SR, Tyczka, J, Ellerbrok, H, Franz, T, Linke, S, Baljer, G & Appel, B 2006, ‘Highly sensitive real-time PCR for specific detection and quantification of Coxiella burnetii’, BMC Microbiology, vol. 6, no. 2, pp. 1–8.

Kopecny, L, Bosward, KL, Shapiro, A & Norris, JM 2013, ‘Investigating Coxiella burnetii infection in a breeding cattery at the centre of a Q fever outbreak’, Journal of Feline Medicine and Surgery, vol. 15, no. 12, pp. 1037–1045.

Kosatsky, T 1984, ‘Household outbreak of Q-fever pneumonia related to a parturient cat’, Lancet, vol. 2, no. 8417–8418, pp. 1447–1449.

Kostoulas, P, Nielsen, SS, Branscum, AJ, Johnson, WO, Dendukuri, N, Dhand, NK, Toft, N & Gardner, IA 2017, ‘STARD-BLCM: Standards for the Reporting of Diagnostic accuracy 213 studies that use Bayesian Latent Class Models’, Preventive Veterinary Medicine, vol. 138, pp. 37–47.

Kovácová, E & Kazar, J 2002, ‘Q fever-still a query and underestimated infectious disease’, Acta Virologica, vol. 46, no. 4, pp. 193–210.

Kruschke, JK 2015, Doing Bayesian data analysis : a tutorial with R, JAGS, and Stan 2nd edn, Elsevier, London.

Kruszewska, D & Tylewska-Wierzbanowska, S 1997, ‘Isolation of Coxiella burnetii from bull semen’, Research in Veterinary Science, vol. 62, pp. 299–300.

Ladbury, GF, Van Leuken, JG, Swart, A, Vellema, P, Schimmer, B, Ter Schegget, R & Van der Hoek, W 2015, ‘Integrating interdisciplinary methodologies for One Health: goat farm re-implicated as the probable source of an urban Q fever outbreak, the Netherlands, 2009’, BMC Infectious Diseases, vol. 15, no. 372, pp. 1–11.

Lane, J, Jubb, T, Shephard, R, Webb-Ware, J & Fordyce, G 2015, Priority list of endemic diseases for the red meat industries, Meat & Livestock Australia Limited, Sydney, Australia.

Lang, GH 1990, ‘Coxiellosis (Q fever) in animals’, in TJ Marrie (ed), Q fever: Volume I: The disease, CRC Press, Boston, pp. 22–48.

Lisowski, P, Pierzchała, M, Gościk, J, Pareek, CS & Zwierzchowski, L 2008, ‘Evaluation of reference genes for studies of gene expression in the bovine liver, kidney, pituitary, and thyroid’, Journal of Applied Genetics, vol. 49, no. 4, pp. 367–372.

Lockhart, MG, Graves, SR, Banazis, MJ, Fenwick, SG & Stenos, J 2011, ‘A comparison of methods for extracting DNA from Coxiella burnetii as measured by a duplex qPCR assay’, Letters in Applied Microbiology, vol. 52, no. 5, pp. 514–520.

López-Gatius, F, Almeria, S & Garcia-Ispierto, I 2012, ‘Serological screening for Coxiella burnetii infection and related reproductive performance in high producing dairy cows’, Research in Veterinary Science, vol. 93, no. 1, pp. 67–73.

Lucchese, L, Capello, K, Barberio, A, Ceglie, L, Guerrini, E, Zuliani, F, Mion, M, Stegeman, A, Rampazzo, E, Marangon, S & Natale, A 2016, ‘Evaluation of serological tests for Q fever in ruminants using the latent class analysis’, Clinical Research in Infectious Diseases, vol. 3, no. 2, pp. 1–5.

214

Lucchese, L, Capello, K, Barberio, A, Zuliani, F, Stegeman, A, Ceglie, L, Guerrini, E, Marangon, S & Natale, A 2015, ‘IFAT and ELISA phase I/phase II as tools for the identification of Q fever chronic milk shedders in cattle’, Veterinary Microbiology, vol. 179, no. 1–2, pp. 102–108.

Lyoo, K-S, Kim, D, Jang, HG, Lee, S-J, Park, MY & Hahn, T-W 2017, ‘Prevalence of antibodies against Coxiella burnetii in Korean native cattle, dairy cattle, and dogs in South Korea’, Vector-Borne and Zoonotic Diseases, vol. 17, no. 3, pp. 213–216.

Macías-rioseco, M, Riet-correa, F, Miller, MM, Sondgeroth, K, Fraga, M, Silveira, C, Uzal, FA & Giannitti, F 2019, ‘Bovine abortion caused by Coxiella burnetii: report of a cluster of cases in Uruguay and review of the literature’, Journal of Veterinary Diagnostic Investigations, vol. 31, no. 4, pp. 634–639.

Malo, JA, Colbran, C, Young, M, Vasant, B, Jarvinen, K, Viney, K & Lambert, SB 2018, ‘An outbreak of Q fever associated with parturient cat exposure at an animal refuge and veterinary clinic in southeast Queensland’, Australian and New Zealand Journal of Public Health, vol. 5, pp. 1–5.

Marenzoni, ML, Stefanetti, V, Papa, P, Casagrande Proietti, P, Bietta, A, Coletti, M, Passamonti, F & Henning, K 2013, ‘Is the horse a reservoir or an indicator of Coxiella burnetii infection? Systematic review and biomolecular investigation’, Veterinary Microbiology, vol. 167, no. 3–4, pp. 662–669.

Marmion, B 2007, ‘Q fever: The long journey to control by vaccination’, Medical Journal of Australia, vol. 186, no. 4, pp. 164–166.

Marmion, BP, Ormsbee, RA, Kyrkou, M, Wright, J, Worswick, DA, Izzo, AA, Esterman, A, Feery, B & Shapiro, RA 1990, ‘Vaccine prophylaxis of abattoir-associated Q fever: eight years’ experience in Australian abattoirs’, Epidemiology and Infection, vol. 104, no. 2, pp. 275–287.

Marmion, BP, Shannon, M, Maddocks, I, Storm, P & Penttila, I 1996, ‘Protracted debility and fatigue after acute Q fever’, The Lancet, vol. 347, no. 9006, pp. 977–978.

Marrie 1990, ‘Epidemiology of Q fever’, in TJ Marrie (ed), Q fever: Volume I: The disease, CRC Press, Boston, pp. 49–70.

Marrie, T, Williams, J, Schlech, W & Yates, L 1986, ‘Q fever pneumonia associated with exposure to wild rabbits’, The Lancet, vol. 327, no. 8478, pp. 427–429.

215

Martin, P, Phillips, P, Leith, R & Caboche, T 2013, ‘Australian beef: Financial performance of beef cattle producing farms, 2010–11 to 2012–13’, Australian Government, Department of Agriculture, Fisheries and Forestry, ABARES, pp. 1–61.

Massung, RF, Cutler, SJ & Frangoulidis, D 2012, ‘Molecular typing of Coxiella burnetii (Q fever)’, in Coxiella burnetii: Recent Advances and New Perspectives in Research of the Q Fever Bacterium, pp. 381–396.

Maurin, M & Raoult, D 1999, ‘Q fever’, Clinical Microbiology Reviews, vol. 12, no. 4, pp. 518–553.

McCaughey, C, Murray, LJ, McKenna, JP, Menzies, FD, McCullough, SJ, O’neill, HJ, Wyatt, DE, Cardwell, CR & Coyle, P V 2010, ‘Coxiella burnetii (Q fever) seroprevalence in cattle’, Epidemiology and Infection, vol. 138, no. 01, pp. 21–27.

McCaul, TF & Williams, JC 1981, ‘Developmental cycle of Coxiella burnetii: Structure and morphogenesis of vegetative and sporogenic differentiations’, Journal of Bacteriology, vol. 147, no. 3, pp. 1063–1076.

McCosker, KD 2016, ‘Risk factors affecting the reproductive outcome of beef breeding herds in North Australia’, Thesis (PhD) The University of Queensland, Australia, pp. 1– 341.

McGowan, M, Fordyce, G, O’Rourke, P, Barnes, T, Morton, J, Menzies, D, Jephcott, S, McCosker, K, Smith, D, Perkins, N, Marquart, L, Newsome, T & Burns, B 2014, Northern Australian beef fertility project: CashCow, Meat and Livestock Australia B.NBP.0382, Sydney, 1-297.

McGowan, M, McCosker, K, Fordyce, G, Smith, D, O’Rouke, P, Perkins, N, Barnes, T, Marquart, L, Jephcott, J, Morton, J, Newsome, T, Menzies, D & Burns, B 2015, Technical synopsis : CashCow findings Insights into the productivity and performance of northern breeding herds, Meat and Livestock Australia, North Sydney, 1-36.

McGowan, MR, McCosker, KD, Fordyce, G, Smith, DR, Perkins, NR, O’Rourke, PK, Barnes, T, Marquart, L, Menzies, D, Newsome, T, Joyner, D, Phillips, N, Burns, BM, Morton, JM & Jephcott, J 2013, ‘Factors affecting performance of beef breeding herds: findings from the CashCow project’, in Australian Cattle Veterinarians Conference, pp. 112–116.

Meat and Livestock Australia 2014, ‘Australia’s beef industry’, MLA Fast Facts, p. 2,

216 retrieved April 6, 2020, from .

Meat and Livestock Australia 2019a, ‘Industry projections 2019’, Industry report, pp. 1–8, retrieved from .

Meat and Livestock Australia 2019b, ‘Australia’s beef industry’, MLA Fast Facts, p. 2, retrieved April 6, 2020, from .

Milazzo, A, Hall, R, Storm, P, Harris, RJ, Winslow, W & Marmion, BP 2001, ‘Sexually transmitted Q fever.’, Clinical infectious diseases : an official publication of the Infectious Diseases Society of America, vol. 33, no. 3, pp. 399–402.

Million, M & Raoult, D 2015, ‘Recent advances in the study of Q fever epidemiology, diagnosis and management’, Journal of Infection, vol. 71 Suppl 1, pp. S2-9.

More, SJ, Radunz, B & Glanville, RJ 2015, ‘Review: Lessons learned during the successful eradication of bovine tuberculosis from Australia’, Veterinary Record, vol. 177, no. 9, pp. 224–232.

Muleme, M, Stenos, J, Vincent, G, Campbell, A, Graves, S, Warner, S, Devlin, JM, Nguyen, C, Stevenson, MA, Wilks, CR & Firestone, SM 2016, ‘Bayesian validation of the indirect immunofluorescence assay and its superiority to the enzyme-linked immunosorbent assay and the complement fixation test for detecting antibodies against Coxiella burnetii in goat serum’, Clinical and Vaccine Immunology, vol. 23, no. 6, pp. 507– 514.

Muleme, M, Stenos, J, Vincent, G, Wilks, CR, Devlin, JM, Campbell, A, Cameron, A, Stevenson, MA, Graves, S & Firestone, SM 2017, ‘Peripartum dynamics of Coxiella burnetii infections in intensively managed dairy goats associated with a Q fever outbreak in Australia’, Preventive Veterinary Medicine, vol. 139, pp. 58–66.

Muskens, J, van Maanen, C & Mars, MH 2011, ‘Dairy cows with metritis: Coxiella burnetii test results in uterine, blood and bulk milk samples’, Veterinary Microbiology, vol. 147, no. 1–2, pp. 186–189.

Narasaki, CT & Toman, R 2012, ‘Lipopolysaccharide of Coxiella burnetii’, in Coxiella

217 burnetii: Recent Advances and New Perspectives in Research of the Q Fever Bacterium, Springer Netherlands, Dordrecht, pp. 65–90.

Natale, A, Bucci, G, Capello, K, Barberio, A, Tavella, A, Nardelli, S, Marangon, S & Ceglie, L 2012, ‘Old and new diagnostic approaches for Q fever diagnosis: correlation among serological (CFT, ELISA) and molecular analyses’, Comparative Immunology, Microbiology and Infectious Diseases, vol. 35, no. 4, pp. 375–379.

Nieuwenhuis, R, Grotenhuis, M Te & Pelzer, B 2012, ‘influence.ME: tools for detecting influential data in mixed effects models’, R Journal, vol. 4, no. 2, pp. 38–47.

‘Notifiable Animal Diseases’ 2018, Biosecurity Tasmania, Department of Primary Industries, Park, Water and Environement, retrieved from .

Nourse, C, Allworth, A, Jones, A, Horvath, R, McCormack, J, Bartlett, J, Hayes, D & Robson, JM 2004, ‘Three cases of Q fever osteomyelitis in children and a review of the literature’, Clinical Infectious Diseases, vol. 39, no. 7, pp. e61-6.

O’Connor, BA, Tribe, IG & Givney, R 2015, ‘A windy day in a sheep saleyard: an outbreak of Q fever in rural South Australia’, Epidemiology and Infection, vol. 143, no. 2, pp. 391– 398.

OIE 2016a, ‘Chapter 3.6.5 Statistical Approaches to Validation’, in Manual of Diagnostic Tests and Vaccines for Terrestrial Animals 2016, Paris, France, pp. 1–12.

OIE 2016b, ‘Validation guideline 3.6.1 Development and optimisation of antibody detection assays’, in Manual of Diagnostic Tests and Vaccines for Terrestrial Animals 2016, OIE - World Organisation for Animal Health, Paris, France, pp. 1–13.

OIE 2018, ‘Chapter 2.1.16 - Q Fever’, in Manual of Diagnostic Tests and Vaccines for Terrestrial Animals 2018, OIE - World Organisation for Animal Health, Paris, France, pp. 1–15.

Omsland, A, Cockrell, DC, Howe, D, Fischer, ER, Virtaneva, K, Sturdevant, DE, Porcella, SF & Heinzen, R a 2009, ‘Host cell-free growth of the Q fever bacterium Coxiella burnetii’, Proceedings of the National Academy of Sciences of the United States of America, vol. 106, no. 11, pp. 4430–4434.

Oyston, PCF & Davies, C 2011, ‘Q fever: The neglected biothreat agent’, Journal of 218

Medical Microbiology, vol. 60, no. 1, pp. 9–21.

Palmer, C, McCall, B, Jarvinen, K, Krause, M & Heel, K 2007, ‘“The dust hasn’t settled yet”: the National Q fever Management Program, missed opportunities for vaccination and community exposures’, Australian and New Zealand Journal of Public Health, vol. 31, no. 4, pp. 330–332.

Parker, NR, Barralet, JH & Bell, AM 2006, ‘Q fever’, Lancet, vol. 367, no. 9511, pp. 679– 688.

Paul, S, Agger, JF, Agerholm, JS & Markussen, B 2014, ‘Prevalence and risk factors of Coxiella burnetii seropositivity in Danish beef and dairy cattle at slaughter adjusted for test uncertainty’, Preventive Veterinary Medicine, vol. 113, no. 4, pp. 504–511.

Paul, S, Agger, JF, Markussen, B, Christoffersen, A-B & Agerholm, JS 2012, ‘Factors associated with Coxiella burnetii antibody positivity in Danish dairy cows.’, Preventive Veterinary Medicine, vol. 107, no. 1–2, pp. 57–64.

Paul, S, Toft, N, Agerholm, JS, Christoffersen, A-B & Agger, JF 2013, ‘Bayesian estimation of sensitivity and specificity of Coxiella burnetii antibody ELISA tests in bovine blood and milk’, Preventive Veterinary Medicine, vol. 109, pp. 258–263.

Pearson, T, Hornstra, HM, Hilsabeck, R, Gates, LT, Olivas, SM, Birdsell, DM, Hall, CM, German, S, Cook, JM, Seymour, ML, Priestley, RA, Kondas, A V, Clark Friedman, CL, Price, EP, Schupp, JM, Liu, CM, Price, LB, Massung, RF, Kersh, GJ & Keim, P 2014, ‘High prevalence and two dominant host-specific genotypes of Coxiella burnetii in U.S. milk’, BMC Microbiology, vol. 14, no. 1, pp. 1–9.

Pinero, A, Barandika, JF, Garcia-Perez, AL & Hurtado, A 2015, ‘Genetic diversity and variation over time of Coxiella burnetii genotypes in dairy cattle and the farm environment’, Infection, Genetics and Evolution, vol. 31, pp. 231–235.

Pinero, A, Ruiz-Fons, F, Hurtado, A, Barandika, JF, Atxaerandio, R & Garcia-Perez, AL 2014, ‘Changes in the dynamics of Coxiella burnetii infection in dairy cattle: an approach to match field data with the epidemiological cycle of C. burnetii in endemic herds’, Journal of Dairy Science, vol. 97, no. 5, pp. 2718–2730.

Pinheiro, J, Bates, D, DebRoy, S, Sarkar, D & Team, RC 2019, ‘Linear and nonlinear mixed effects models: nlme’, R package version 3.1-148.

Pinsky, RL, Fishbein, DB, Greene, CR & Gensheimer, KF 1991, ‘An outbreak of cat- 219 associated Q fever in the United States’, The Journal of Infectious Diseases, vol. 164, no. 1, pp. 202–204.

Pitt, D 1997, ‘Structured animal health surveillance 1996’, Queensland Department of Primary Industries, pp. 1–19.

Plommet, M, Capponi, M, Gestin, J, Renoux, G & Marly, J 1973, ‘Fievre Q Experimentale Des Bovins’, Annals de Reserches Veterinarires, vol. 4, no. 2, pp. 325–346.

Plummer, M, Nicky, B, Cowles, K & Vines, K 2006, ‘CODA:convergence diagnosis and output analysis for MCMC’, R News, vol. 6, no. 1, pp. 7–11.

Plummer, PJ, McClure, Jt, Menzies, P, Morley, PS, Van den Brom, R, Van Metre, DC & Paul Plummer, CJ 2018, ‘Management of Coxiella burnetii infection in livestock populations and the associated zoonotic risk: A consensus statement’, Journal of Veterinary Internal Medicine, vol. 32, pp. 1481–1494.

‘Population health data and statistics | Queensland Health’ 2020, Queensland Governement, retrieved February 19, 2020, from .

Porter, SR, Czaplicki, G, Mainil, J, Guattéo, R & Saegerman, C 2011, ‘Q fever: Current state of knowledge and perspectives of research of a neglected zoonosis’, International Journal of Microbiology, vol. 2011, pp. 1–22.

Potter, AS, Banazis, MJ, Yang, R, Reid, SA & Fenwick, SG 2011, ‘Prevalence of Coxiella burnetii in western grey kangaroos (Macropus fuliginosus) in Western Australia’, Journal of Wildlife Diseases, vol. 47, no. 4, pp. 821–828.

Pritchard, GC, Smith, RP, Errington, J, Hannon, S, Jones, RM & Mearns, R 2011, ‘Prevalence of Coxiella burnetii in livestock abortion material using PCR’, Veterinary Record, vol. 169, no. 15, p. 391.

‘Q fever - worksafe.qld.gov.au’ 2019, retrieved February 11, 2020, from .

‘QGIS Geographic Information System’ 2019, Open Source Geospatial Foundation Project http://qgis.org.

‘R: A language and environment for statistical computing’ 2019, R Foundation for

220

Statistical Computing.

Rahal, M, Tahir, D, Eldin, C, Bitam, I, Raoult, D & Parola, P 2018, ‘Genotyping of Coxiella burnetii detected in placental tissues from aborted dairy cattle in the north of Algeria’, Comparative Immunology, Microbiology and Infectious Diseases, vol. 57, no. October 2017, pp. 50–54.

Raoult, D, Marrie, T & Mege, J 2005, ‘Natural history and pathophysiology of Q fever’, The Lancet - Infectious Diseases, vol. 5, no. 4, pp. 219–226.

Roest, Bossers, A & Rebel, JMJ 2013, ‘Q fever diagnosis and control in domestic ruminants’, Developments in Biologicals, vol. 135, pp. 183–189.

Roest, H, Bossers, A, Van Zijderveld, F & Rebel, J 2013, ‘Clinical microbiology of Coxiella burnetii and relevant aspects for the diagnosis and control of the zoonotic disease Q fever’, Veterinary Quarterly, vol. 33, no. 3, pp. 148–160.

Roest, HI., Tilburg, JJ., Van Der Hoek, W, Vellema, P, Zijderveld, F., Klaassen, CHW & Raoult, D 2010, ‘The Q fever epidemic in The Netherlands : history, onset, response and reflection’, Epidemiology and Infection, vol. 139, pp. 1–12.

Roest, HIJ 2013, ‘Coxiella burnetii in pregnant goats’, Thesis (PhD), Utrecht University, Netherlands, pp. 1–200.

Roest, HIJ, Ruuls, RC, Tilburg, JJHC, Nabuurs-Franssen, MH, Klaassen, CHW, Vellema, P, van den Brom, R, Dercksen, D, Wouda, W, Spierenburg, MAH, van der Spek, AN, Buijs, R, de Boer, AG, Willemsen, PTJ & van Zijderveld, FG 2011, ‘Molecular epidemiology of Coxiella burnetii from ruminants in Q fever outbreak, The Netherlands’, Emerging Infectious Diseases, vol. 17, no. 4, pp. 668–675.

Roest, HIJ, van Solt, C, Tilburg, J, Klaassen, C, Hovius, E, Roest, F, Vellema, P, van den Brom, R & van Zijderveld, F 2013, ‘Search for possible additional reservoirs for human Q fever, the Netherlands’, Emerging Infectious Diseases, vol. 19, no. 5, pp. 834–835.

Rogan, WJ & Gladen, B 1978, ‘Estimating prevalence from the results of a screening test’, American Journal of Epidemiology, vol. 107, no. 1, pp. 71–76.

Rousset, E, Berri, M, Durand, B, Dufour, P, Prigent, M, Delcroix, T, Touratier, A & Rodolakis, A 2009, ‘Coxiella burnetii shedding routes and antibody response after outbreaks of Q fever-induced abortion in dairy goat herds’, Applied and Environmental Microbiology, vol. 75, no. 2, pp. 428–433. 221

Rousset, E, Durand, B, Berri, M, Dufour, P, Prigent, M, Russo, P, Delcroix, T, Touratier, A, Rodolakis, A & Aubert, M 2007, ‘Comparative diagnostic potential of three serological tests for abortive Q fever in goat herds’, Veterinary Microbiology, vol. 124, no. 3–4, pp. 286–97.

Ruiz-Fons, F, Astobiza, I, Barandika, JF, Hurtado, A, Atxaerandio, R, Juste, RA & García- Pérez, AL 2010, ‘Seroepidemiological study of Q fever in domestic ruminants in semi- extensive grazing systems’, BMC Veterinary Research, vol. 6, no. 1, pp. 1–6.

Safety in laboratories. Part 3, Microbiological safety and containment. 2010, Report AS/NZS 2243.3:2010, Standards Australia/New Zealand, Sydney, 1-176.

‘Sample Size Software | Power Analysis Software | PASS |’ 2016, https://www.ncss.com/software/pass/, retrieved from .

Samuel, JE & Hendrix, LR 2009, ‘Laboratory maintenance of Coxiella burnetii’, in Current Protocols in Microbiology, John Wiley & Sons, Inc. van Schaik, EJ & Samuel, JE 2012, ‘Phylogenetic diversity, virulence and comparative genomics’, in Coxiella burnetii: Recent Advances and New Perspectives in Research of the Q Fever Bacterium, Springer Netherlands, Dordrecht, pp. 13–38.

Sellens, E, Norris, JM, Dhand, NK, Heller, J, Hayes, L, Gidding, HF, Willaby, H, Wood, N & Bosward, KL 2016, ‘Q Fever Knowledge, Attitudes and Vaccination Status of Australia’s Veterinary Workforce in 2014’, PLoS ONE, vol. 11, no. 1, p. e0146819.

Sergeant, E & Perkins, N 2015, Epidemiology For Field Veterinarians: An Introduction, CABI International, Croyden, United Kingdom.

Seshadri, R, Paulsen, IT, Eisen, JA, Read, TD, Nelson, KE, Nelson, WC, Ward, NL, Tettelin, H, Davidsen, TM, Beanan, MJ, Deboy, RT, Daugherty, SC, Brinkac, LM, Madupu, R, Dodson, RJ, Khouri, HM, Lee, KH, Carty, HA, Scanlan, D, Heinzen, RA, Thompson, HA, Samuel, JE, Fraser, CM & Heidelberg, JF 2003, ‘Complete genome sequence of the Q-fever pathogen Coxiella burnetii’, Proceedings of the National Academy of Sciences, vol. 100, no. 9, pp. 5455–5460.

Shapiro, AJ, Bosward, KL, Heller, J & Norris, JM 2015, ‘Seroprevalence of Coxiella burnetii in domesticated and feral cats in eastern Australia’, Veterinary Microbiology, vol. 177, no. 1–2, pp. 154–161.

Shapiro, AJ, Norris, JM, Heller, J, Brown, G, Malik, R & Bosward, KL 2016, ‘Seroprevalence of Coxiella burnetii in Australian dogs’, Zoonoses Public Health, vol. 63, 222 no. 6, pp. 458–466.

Sivabalan, P, Saboo, A, Yew, J & Norton Mackay Base Hospital, R 2017, ‘Q fever in an endemic region of North Queensland, Australia: A 10 year review’, One Health, vol. 3, pp. 51–55.

Sloan-Gardner, TS, Massey, PD, Hutchinson, P, Knope, K & Fearnley, AE 2017, ‘Trends and risk factors for human Q fever in Australia’, Epidemiology and Infection, vol. 145, pp. 787–795.

Sobotta, K, Hillarius, K, Mager, M, Kerner, K, Heydel, C & Menge, C 2016, ‘Coxiella burnetii infects primary bovine macrophages and limits their host cell response’, Infection and Immunity, vol. 84, no. 6, pp. 1722–34.

Staley, G., MyBurgh, J. & Chaparro, F 1989, ‘Serological evidence of Q fever in cattle in Malawi’, Onderstepoort Journal of Veterinary Research, vol. 56, pp. 205–206.

Stenos, J, Graves, S & Lockhart, M 2010, ‘Coxiella burnetii’, in PCR for Clinical Microbiology, Springer Netherlands, Dordrecht, pp. 145–148.

Stevenson, M 2017, ‘epiR: Tools for the Analysis of Epidemiological Data’, R package version 0.9-93.

Stevenson, S, Gowardman, J, Tozer, S & Woods, M 2015, ‘Life-threatening Q fever infection following exposure to kangaroos and wallabies’, BMJ Case Reports, vol. 2015.

Stoker, MG & Marmion, BP 1955, ‘The spread of Q fever from animals to man; the natural history of a rickettsial disease’, Bulletin of the World Health Organization, vol. 13, no. 5, pp. 781–806.

Suess, EA, Gardner, IA & Johnson, WO 2002, ‘Hierarchical Bayesian model for prevalence inferences and determination of a country’s status for an animal pathogen’, Preventive Veterinary Medicine, vol. 55, no. 3, pp. 155–171.

Tan, T 2018, ‘A pilot study of the seroprevalence of Q fever in cattle, sheep and goats in Victoria’, Thesis (MPhil) The University of Melbourne, Australia, pp. 1–71.

Thrusfield, M V. 2007, ‘Chapter 17: Diagnostic testing’, in Veterinary Epidemiology, Blackwell Science, Oxford, pp. 327–328.

Tissot-Dupont, H, Amadei, M-A, Nezri, M & Raoult, D 2004, ‘Wind in November, Q fever in December’, Emerging Infectious Disease Journal, vol. 10, no. 7, pp. 1264–1269.

223

Tissot Dupont, H, Raoult, D, Brouqui, P, Janbon, F, Peyramond, D, Weiller, PJ, Chicheportiche, C, Nezri, M & Poirier, R 1992, ‘Epidemiologic features and clinical presentation of acute Q fever in hospitalized patients: 323 French cases’, The American Journal of Medicine, vol. 93, no. 4, pp. 427–434.

To, H, Khin, K, Htwe, N, Kako, H, Jib, K, Yamaguchi, T, Fukushi, H & Hirai, K 1998, ‘Prevalence of Coxiella burnetii infection in dairy cattle with reproductive disorders’, Journal of Veterinary Medical Science, vol. 60, no. 7, pp. 859–861.

Toft, N, Jørgensen, E & Højsgaard, S 2005, ‘Diagnosing diagnostic tests: evaluating the assumptions underlying the estimation of sensitivity and specificity in the absence of a gold standard’, Preventive Veterinary Medicine, vol. 68, no. 1, pp. 19–33.

Toman, R, Heinzen, RA, Samuel, JE & Mege, J-L (eds) 2012, Coxiella burnetii: Recent Advances and New Perspectives in Research of the Q Fever Bacterium, Springer Netherlands, Dordrecht.

Tozer, S 2015, ‘Epidemiology, diagnosis and prevention of Q fever in Queensland’, Thesis (PhD) The University of Queensland, Australia, pp. 1–309.

Tozer, S, Wood, C, Si, D, Nissen, M, Sloots, T & Petrie, S 2020, ‘The improving state of Q fever surveillance: A review of Queensland notifications , 2003 – 2017’, Communicable diseases intelligence, vol. 44, pp. 1–22.

Tozer, SJ, Lambert, SB, Sloots, TP & Nissen, MD 2011, ‘Q fever seroprevalence in metropolitan samples is similar to rural/remote samples in Queensland, Australia’, European Journal of Clinical Microbiology and Infectious Disease, vol. 30, no. 10, pp. 1287–1293.

Tozer, SJ, Lambert, SB, Strong, CL, Field, HE, Sloots, TP & Nissen, MD 2014, ‘Potential animal and environmental sources of Q fever infection for humans in Queensland’, Zoonoses Public Health, vol. 61, no. 2, pp. 105–112.

Tu, XM, Kowalski, J & Jia, G 1999, ‘Bayesian analysis of prevalence with covariates using simulation-based techniques: Applications to HIV screening’, Statistics in Medicine, vol. 18, no. 22, pp. 3059–3073.

Vaidya, VM, Malik, SVS, Bhilegaonkar, KN, Rathore, RS, Kaur, S & Barbuddhe, SB 2010, ‘Prevalence of Q fever in domestic animals with reproductive disorders’, Comparative Immunology, Microbiology and Infectious Diseases, vol. 33, no. 4, pp. 307–21.

224

Vincent, G, Stenos, J, Latham, J, Fenwick, S & Graves, S 2016, ‘Novel genotypes of Coxiella burnetii identified in isolates from Australian Q fever patients’, International Journal of Medical Microbiology, vol. 306, no. 6, pp. 463–470.

Vincent, GA, Graves, SR, Robson, JM, Nguyen, C, Hussain-Yusuf, H, Islam, A, Fenwick, SG & Stenos, J 2015, ‘Isolation of Coxiella burnetii from serum of patients with acute Q fever’, Journal of Microbiology Methods, vol. 119, pp. 74–78.

Wade, AJ, Cheng, AC, Athan, E, Molloy, JL, Harris, OC, Stenos, J & Hughes, AJ 2006, ‘Brief report: Q fever outbreak at a cosmetics supply factory’, Clinical Infectious Diseases, vol. 42, pp. 50–52.

Walter, MC, Vincent, GA, Stenos, J, Graves, S & Frangoulidis, D 2014, ‘Genome sequence of Coxiella burnetii strain AuQ01 (Arandale) from an Australian patient with acute Q fever’, Genome Announcements, vol. 2, no. 5, pp. e00964-14.

Wang, CYT, McCarthy, JS, Stone, WJ, Bousema, T, Collins, KA & Bialasiewicz, S 2018, ‘Assessing Plasmodium falciparum transmission in mosquito-feeding assays using quantitative PCR’, Malaria Journal, vol. 17, no. 1, pp. 1–11.

Watanabe, A & Takahashi, H 2008, ‘Diagnosis and treatment of Q fever: attempts to clarify current problems in Japan’, Journal of Infection and Chemotherapy, vol. 14, no. 1, pp. 1–7.

Willems, H, Jager, C, Jager, J & Baljer, G 1998, ‘Physical and genetic map of the obligate intracellular bacterium Coxiella burnetii’, Journal of Bacteriology, vol. 180, no. 15, pp. 3816–3822.

Williams, JC & Thompson, HA 1991, Q fever : the biology of Coxiella burnetii, CRC Press, Boston.

‘WinEpi: Working in Epidemiology’ 2016, http://www.winepi.net/uk/index.htm, retrieved from .

Woldehiwet, Z 2004, ‘Q fever (coxiellosis): epidemiology and pathogenesis’, Research in Veterinary Science, vol. 77, no. 2, pp. 93–100.

Woldeyohannes, SM, Gilks, CF, Baker, P, Perkins, NR & Reid, SA 2018, ‘Seroprevalence of Coxiella burnetii among abattoir and slaughterhouse workers: A meta-analysis’, One Health, vol. 6, pp. 23–28.

Wood, C, Muleme, M, Tan, T, Bosward, K, Gibson, J, Alawneh, J, McGowan, M, Barnes,

225

TS, Stenos, J, Perkins, N, Firestone, SM & Tozer, S 2019, ‘Validation of an indirect immunofluorescence assay (IFA) for the detection of IgG antibodies against Coxiella burnetii in bovine serum’, Preventive Veterinary Medicine, vol. 169, p. 104698.

Worswick, D & Marmion, BP 1985, ‘Antibody responses in acute and chronic Q fever and in subjects vaccinated against Q fever’, Journal of Medical Microbiology, vol. 19, pp. 281– 296.

Yu-Sung, S & Masanao, Y 2015, ‘R2Jags: Using R to run “JAGS”’, R package version 0.6- 1.

Zinsstag, J, Dean, A, Baljinnyam, Z, Roth, F, Kasymbekov, J & Schelling, S 2015, ‘Brucellosis surveillance and control: a case for One Health’, in One Health: the theory and practice of integrated health approaches, CABI, Wallingford, UK, pp. 153–162.

226

10. Appendix

10.1. Ethics approval certificates

227

228

229