Understanding the Ecology of Ross River Virus; Novel Approaches and Insights Into Non-Human Reservoirs

Home , Aedes camptorhynchus, Ross River (Queensland)

Understanding the ecology of Ross River virus; novel approaches and insights into non-human reservoirs

Author Stephenson, Eloise

Published 2019-11-19

Thesis Type Thesis (PhD Doctorate)

School School of Environment and Sc

DOI https://doi.org/10.25904/1912/570

Downloaded from http://hdl.handle.net/10072/389694

Griffith Research Online https://research-repository.griffith.edu.au

Understanding the ecology of Ross River virus; novel approaches and insights into non-human reservoirs.

by Eloise Birgitta Stephenson BSc, MSc

Submitted in fulfilment of the requirements of the degree Doctor of Philosophy School of Environment and Science Griffith University

August 2019

Acknowledgements

This dissertation does not represent the work of a sole individual; instead it is the product of generous support, guidance and inspiration from many people whom have woven into the fabric of my life.

I would like to start off by thanking my supervisors, Prof. Hamish McCallum, Dr. Alison Peel, Dr. Cassie Jansen, Dr. Lara Herrero, and Ass. Prof. Simon Reid, who have been incredibly supportive of both my research and my overall growth throughout my PhD. I have a large supervisory team and have been very fortunate to receive advice, guidance and support from each of them in many shapes and forms.

Very special thanks to Hamish, not just for sharing his vast knowledge on R, analyses and models with me, but for his quick wit and willingness to expand my networks through introductions. I owe a large thank you also to Ali, who has inspired and challenged me throughout my PhD. Ali is an incredible researcher who taught me to formulate and present my research questions clearly, and broadened my critical analysis skills. Ali is also a great mentor, and I am inspired by her work, her family and her fabulous earrings. I am also very grateful to Cassie, whose faith in me over the years has been unflappable! Thank you to Lara for her laboratory supervision and willingness to think outside the box across the multiple disciplines covered in this project, and to Simon who helped guide this project.

Amanda Murphy has been a shining light during this PhD; and together we have tackled fieldwork, RRV notification data and many wines across Brisbane. Amanda is a fabulous sounding board and is always bubbling with ideas and enthusiasm for the disease system we both work on. She has kept me motivated, sane and excited about my research. I thank her, and her colleagues, for all of this.

I am very lucky to work among two research groups at Griffith and am grateful for the support, laughter and feedback from the members of these groups; Lara Herreros’ group (Penny, Joyce, Aroon, Elina, Elisa, Eugene and Julie), and Hamish McCallums’ group (Doug, Laura, Tamika, Thais, Will, Mandy, Manuel, Remy and the bat catching team!). Griffith has taught me much about research, but in particular, the importance of a good support team. So I would also like to thank Dian, Isaac and Christina at EFRI for their assistance with much of the paperwork and applications that are required throughout a PhD.

My project has led me to meet, and collaborate with many other researchers and institutions all over the world. I have had the great pleasure of discussing this project and ideas with many people, and ultimately sometimes these have led to my best analyses or interpretations.

My family is a big part of my life, and throughout this PhD they offered to proof read sections, cook dinners, or just simply distract me. My family have been with me for all the moments of triumph and angst along the way. I am so fortunate to be able to speak with my Mum, Dad, sisters, grandparents and parents-in-law almost every week, and am always offered support, empathy and encouragement. The love and support of my family extends far beyond the duration of this project, and I wouldn’t have committed to a PhD without their encouragement to pursue my passions.

To my friends, whether in academia or not, I am grateful for being able to share ideas, beers and tears with you along the way. I am inspired and motivated by so many of my friendships. My PhD friends, Jenna, Pat and Steph have been especially empathetic to the process and I can only wish that my future colleagues are as supportive and inspirational as you have been.

Finally, I would like to thank my husband and best friend, Tim. Somewhere in between the lines of these pages we managed to travel, move house, get married and start a family. You have kept me grounded during all of this, recognising the little moments – like tea to keep going, a break to stop working, or even just space – as well as the big moments. Thank you for not just for supporting me to follow my dreams, but for being a part of them too. I am just as much excited about our future, as I am when I look back on our past.

iii

This work has not previously been submitted for a degree or diploma in any university. To the best of my knowledge and belief, the dissertation contains no material previously published or written by another person except where due reference is made in the dissertation itself. --- Eloise B. Stephenson

Abstract

Arboviruses contribute a significant burden to human and animal health. Circulation of arboviruses comprises three components; blood-sucking arthropods, vertebrate hosts, and viruses that can infect vertebrates and invertebrates. Interaction of these components is dependent on ecological factors (such as species distributions and climate), epidemiological factors (including vector and host immunity) and behavioural determinants (such as vector feeding host preference or host defensive behaviours). Identifying these drivers of disease emergence can be complicated but informs efforts to mitigate on-going circulation. This dissertation generates new perspectives on the transmission dynamics of Australia’s most common arbovirus, Ross River virus (RRV), with a particular focus on non-human reservoirs. Specifically, I (a) critically analysed current and historic knowledge for non-human reservoirs and vector feeding patterns, (b) examined the natural exposure of RRV in human, free-living and domestic vertebrate populations, and (c) assessed the vertebrate and vector community ecology across areas of varying human notification rates.

I began by undertaking a systematic literature review (Chapter 2) assessing evidence for non- human reservoirs of RRV. This chapter synthesised published serological, virus isolation and experimental infection studies in light of the long-held dogma that marsupials are the primary reservoir of RRV. A key finding of this chapter was emerging evidence that placental mammals and birds were also capable of transmitting RRV to mosquito vectors, suggesting a broader reservoir potential than marsupials alone. To further assess the current and historic knowledge, a meta-analysis was performed on mosquito blood meal analysis studies (Chapter 3). It was evident from this chapter that Australian mosquitoes have highly varied feeding patterns which did not reflect their taxonomic classification or larval ecology.

To understand the natural exposure of RRV in vertebrate populations I used both existing literature and performed serological surveys. Although humans are largely thought to be RRV dead-end hosts (species which are incapable of pathogen amplification), circulation of the virus in human populations is well documented and provides insights into spatial-temporal patterns of transmission relevant to the understanding of non-human reservoirs. In Chapter 4, I assessed spatial-temporal patterns of seroprevalence in human populations reported from across the natural distribution of RRV in Australia and the Pacific Island Countries from 1958 to present day. A key finding was that RRV circulated in human populations at least since 1975 when

v human seropositivity between 20 and 34% was reported in Indonesia and Papua New Guinea. This is important because these countries have different vertebrate fauna to Australia, suggesting different transmission cycles.

I then assessed RRV exposure in non-human vertebrate species, focussing on South East Queensland, a RRV endemic area (Chapter 5). Samples were collected through a network of veterinary clinics over a 12-month period, and a total 595 samples from 31 species were obtained. The results showed that taxonomic relatedness is not an important determinant for seropositivity, but rather the ecology and physiology of a species including diet, body size and longevity is most important for exposure to RRV. This study not only tests the greatest diversity of vertebrate species in a single RRV seroprevalence study, but is also novel because it provides methodological advances for analysing seroprevalence data for other zoonotic pathogens.

In my last data chapter (Chapter 6), I assessed the vertebrate-vector communities across sites with varied RRV human notification rates. Field surveys assessing ‘abundance’ and ‘diversity’ in light of human notifications were undertaken for six months. Human notifications were positively correlated with vertebrate biomass and total mosquito abundance. Although informative in highlighting variables for ongoing investigations, the data from this chapter was insufficient to determine whether this pattern was unique to the specific habitats in which the data was collected, or a true relationship between disease in humans and non-human reservoirs. Nevertheless, this result lends support to the complex transmission dynamics of RRV and the need to apply multi-disciplinary approaches to understand transmission.

Throughout the dissertation, I present compelling new evidence that species other than marsupials are involved in the transmission ecology of RRV. My analyses support the notion that transmission is locally dependent on the availability, diversity and ecology of vertebrate hosts, in combination with the diversity and abundance of mosquito vectors. Moreover, by uncovering the feeding patterns of vectors, identifying exposure rates of vertebrate species, in combination with abundance and availability of both vectors and vertebrates, the research in this dissertation provides much of the information required to build a next-generation matrix model for RRV. Throughout the dissertation I have addressed my research aims and present important new methodological techniques (such as combining serological and ecological analyses) for assessing vertebrate reservoirs of arboviruses, whilst carrying out investigations of disease ecology. vi

Publications arising from this thesis

Published Stephenson, E., Murphy, A., Jansen, C.C., Peel, A.J. & McCallum, H. (2019) Interpreting mosquito feeding patterns in Australia through an ecological lens; an analysis of blood meal studies. Parasites and Vectors. 12: 156 Stephenson, E., Peel, A.J., Reid, S. A., Jansen, C.C. & McCallum, H. (2018) The non- human reservoirs of Ross River virus: a systematic review of the evidence. Parasites and Vectors. 11(1), 188.

Submitted Ong, O., Stephenson, E.B., Old, J. (2019) Reviewing the mosquito-disease interface for non- human vertebrates in Australia. PLoS Pathogens. Submitted 17th May 2019. Stephenson, E.B., Rudd, P., Peel, A.J., McCallum, H. & Herrero, L. (2019) Ross River virus seroprevalence and associated life-history traits in non-human vertebrates across South-East Queensland. Transactions of the Royal Society of Tropical Medicine and Hygiene. Submitted 27th July 2019. Stephenson, E. B., Murphy, A., Jansen, C., Shivas, M., & McCallum, H., & Peel, A. J. (2019) Associations between Ross River virus disease and vector-vertebrate community ecologies in Brisbane, Australia. Vector-Borne and Zoonotic Diseases. Submitted 11th November 2019.

In preparation Stephenson, E.B., Madzokere, E., Webster, J., Harley, D. & Herrero, L. (2019) A systematic review of human seroprevalence for Dengue, Ross River, and Barmah Forest viruses in Australia and the Pacific. PLoS Neglected Tropical Diseases.

Other publications arising during the course of this work Lee, W-S., Webster, J., Madzokere, E., Stephenson, E.B. & Herrero, L. (2019) Mosquito antiviral defense mechanisms: a delicate balance between innate immunity and persistent viral infection. Parasites and Vectors. 12:165 Lau, C., Townell, N., Stephenson, E., van den Berg, D. & Craig, S. (2018) Leptospirosis – an important zoonoses acquired in work, play and travel. The Australian Journal of General Practice. 40(3): 105-110 Herrero, M., Thornton, P. K., Power, B., Bogard, J. R., Remans, R., Fritz, S., ... Stephenson, E., & Watson, R. A. (2017). Farming and the geography of nutrient production for human use: a transdisciplinary analysis. The Lancet Planetary Health, 1(1), e33-e42. DOI: 10.1016/S2542-5196(17)30007-4

vii

Contents

Acknowledgements ...... ii Abstract ...... v Chapter 1: Introduction ...... 1 1.1 Zoonoses; global threats, local challenges ...... 1 1.2 Identifying vectors & reservoirs of zoonotic arboviruses ...... 3 1.3 Ross River virus; an important zoonotic arbovirus...... 6 1.4 Thesis objectives ...... 14 1.5 References ...... 16 Chapter 2: The non-human reservoirs of Ross River virus: a systematic review of the evidence ...... 23 2.1 Abstract ...... 24 2.2 Background ...... 24 2.3 Methods...... 25 2.4 Results ...... 26 2.5 Discussion ...... 30 2.6 Conclusion ...... 34 2.7 References ...... 34 2.8 Supplementary Information ...... 37 Chapter 3: Interpreting mosquito feeding patterns in Australia through an ecological lens; an analysis of blood meal studies ...... 43 3.1 Abstract ...... 44 3.2 Background ...... 44 3.3 Methods...... 45 3.4 Results ...... 46 3.5 Discussion ...... 49 3.6 Conclusion ...... 52 3.7 References ...... 53 3.8 Supplementary Information ...... 55 Chapter 4: A systematic review of human seroprevalence for Dengue, Ross River, and Barmah Forest viruses in Australia and the Pacific ...... 59 4.1 Abstract ...... 61 4.2 Author summary: ...... 62 4.3 Introduction ...... 63 4.4 Methods...... 64

viii

4.5 Results ...... 66 4.6 Discussion ...... 75 4.7 Conclusion ...... 78 4.8 References ...... 79 4.9 Supplementary Information ...... 84 Chapter 5: Species traits and hotspots associated with Ross River virus infection in non- human vertebrates in South-East Queensland ...... 88 5.1 Abstract: ...... 90 5.2 Introduction ...... 91 5.3 Methods...... 92 5.4 Results ...... 97 5.5 Discussion ...... 101 5.6 Conclusion ...... 105 5.7 References ...... 107 5.8 Supplementary Information ...... 110 Chapter 6: Associations between Ross River virus disease and vector and vertebrate community ecology in Brisbane, Australia ...... 116 6.1 Abstract: ...... 117 6.2 Introduction ...... 118 6.3 Methods...... 119 6.4 Data analysis ...... 123 6.5 Results ...... 124 6.6 Discussion ...... 131 6.7 Conclusion ...... 135 6.8 References ...... 136 6.9 Supplementary Information ...... 139 Chapter 7: Discussion and conclusions ...... 145 7.1 Research summary ...... 145 7.2 Non-human reservoirs of RRV ...... 145 7.3 A multidisciplinary approach for arboviral disease ecology ...... 148 7.4 Towards a multi-host, multi-vector model for RRV ...... 152 7.5 Final conclusions and future directions ...... 157 7.6 References ...... 159

List of Figures 1.1 Epidemic and enzootic arbovirus transmission cycles...... 3 1.2 Number of human RRV notifications in each state between 1993 and 2018 ...... 9 1.3 Locations of reported local transmission of RRV to humans ...... 10 1.4 Schematic graph of project aims and corresponding chapters, publications and manuscripts associated with each ...... 14 2.1 Mean peak titre and duration of viraemia measured in different animals experimentally infected with Ross River virus ...... 29 2.2 The number and types of method used to test Ross River virus seroprevalence by decade ...... 30 2.3 Species, sample size, viraemia and antibody response to experimental infection of vertebrate species with RRV ...... 30 3.1 Bioregions in which blood meal studies took place across Australia ...... 49 3.2 Feeding associations between Australian mosquito species and vertebrate taxa ...... 50 3.3 Shannon’s diversity (h-index) of blood meal origins for Australian mosquitoes ...... 50 4.1 Flow diagram of the systematic search strategy used to identify and include studies ...... 67 4.2 Forest plots for reported seroprevalence of dengue, RRV and Barmah Forest virus ... 68-70 4.3 The spatiotemporal trends in reported for dengue, RRV and Barmah Forest virus seroprevalence in Australia and the Pacific Island Countries ...... 72 5.1 The prevalence of RRV antibodies grouped by birds, marsupials and placental mammals ...... 96 5.2 Variable importance as determined by the Random Forest model ...... 97 5.3 Conditional inference tree based on the six traits most strongly associated with seropositivity ...... 98 5.4 Maps of South-East Queensland representing locations and results of samples tested for RRV ...... 100 6.1 Distribution of study suburbs in which data was collected from in Brisbane ...... 121 6.2 Mean annual notification rate (per 100,000) for suburbs in Brisbane ...... 125 6.3 Boxplots of vertebrate biomass and mosquito abundance per suburb ...... 126 6.4 RRV mosquito community composition within each site ...... 128 6.5 Vertebrate community biomass (kg) categorised by taxonomic class and compared across sites ...... 129 6.6 Results of NMDS analysis for the vertebrate biomass, vertebrate abundance and mosquito abundance ...... 130 6.S1 Variance within and between monthly surveys across suburbs for vector-vertebrate communities ...... 140-144 7.1 Steps necessary to build a NGM model, and the data appropriate for RRV ...... 154

List of Tables 1.1 Australian arboviruses known to cause disease in humans ...... 6 2.1 Study types included in the literature review of RRV reservoir studies ...... 26 2.2 RRV experimental infection studies included in the review ...... 27 2.3 Species, sample size, viraemia and antibody response to experimental infection of vertebrate species with RRV ...... 28 2.4 The number, study and study sample size for isolates of RRV collected from non-human vertebrates ...... 29 2.S1 Combinations of search terms used to collect papers for review ...... 37 2.S2 Detailed summary of included RRV reservoir studies ...... 37-42 3.1 Location and methods for mosquito collection and blood meal analysis for the studies included in this review ...... 47-8 3.S1 Derivation of 2x2 contingency tables for mosquito feeding references ...... 55 3.S2 Reported blood meal results for Australian mosquito species...... 56-8 4.1 Article selection criteria ...... 65 4.2 Serological methods used to test samples for each virus ...... 71 4.3 Studies reporting risk factors ...... 74 4.S1 Summary of data extracted from publications ...... 84-7 5.S1 Species and their associated traits assessed in the Random Forest model ...... 110-13 5.S2 Serology results for each species, including total number tested, apparent prevalence and confidence intervals ...... 114-5 6.1 Ecological, geographical and social characteristics of suburbs used within this study ... 122 6.S1 Reported RRV notifications for SSCs included within this study ...... 139 6.S2 Reported significance for notification rates across variables ...... 139 7.1 Summary of the aims, methods and main findings from this dissertation ...... 150-1 7.2 Host and vector parameters needed for a NGM model for RRV ...... 155-6

Chapter 1

General Introduction

The overarching objective of this dissertation is to investigate the non-human reservoirs of Ross River virus using different methods and, ultimately, identify particular species or taxa which may play a key role in amplifying or diluting transmission of the pathogen to mosquitoes. Investigating this issue is vital because, globally, the majority of communicable diseases have a non-human vertebrate reservoir (Taylor, Latham, and Woolhouse 2001), yet means to identify these species and their role in transmission remain vague (Viana et al. 2014).

Identifying non-human vertebrate species involved in transmission is particularly challenging for mosquito-borne diseases, as the processes driving vector-host contact, host and vector immunity, and host and vector populations are complex. This literature review highlights the importance of arboviruses to vertebrate health and describes both mosquito and vertebrate biology and ecology in light of disease transmission. Following this, evidence used to identify reservoirs of arboviruses is comprehensively reviewed. Finally, a detailed overview of Ross River virus transmission ecology and aetiology are described, and this information is drawn together to emphasise the importance of this current study.

1.1 Zoonoses; global threats, local challenges

Zoonoses are infectious pathogens which circulate between non-human vertebrates and humans. Zoonoses are a significant cause of morbidity and mortality globally; more than 75% of human emerging infectious diseases are zoonotic, and at least 60% of existing infectious pathogens of humans have a non-human vertebrate origin (Taylor, Latham, and Woolhouse 2001). The emergence of zoonotic pathogens has been attributed to anthropogenic environmental change, and the globalisation of agriculture, commerce and human travel (Jones et al. 2008; Daszak, Cunningham, and Hyatt 2001; Taylor, Latham, and Woolhouse 2001). The serious impact that zoonoses can have on global health has been witnessed in pandemics over the last 20 years, which have included Ebola virus (Ross, Olveda, and Yuesheng 2014), Zika virus (Lucey and Gostin 2016) and severe acute respiratory syndrome (SARS) (Lau et al. 2008).

A critical challenge for managing and preventing zoonotic spillover locally is to identify the source of infection. This can be complicated because there are often a number of contributing factors that result in a spillover event, reliant on the host-pathogen dynamics, host ecology, and human contact with infected individuals (Plowright et al. 2017). Of the large diversity of zoonotic pathogens (viral, bacterial, fungal or eukaryotic) and transmission routes (such as airborne, direct contact or vector-borne), understanding the complex population biology of multi-host pathogens has been named as one of the major challenges to biomedical science in the 21st century (Woolhouse, Taylor, and Haydon 2001).

1.1.1. Zoonotic arboviruses

Zoonotic arboviruses (viruses transmitted by arthropod vectors) are particularly important from a global health perspective. More than 100 arboviruses cause disease in humans (LaBeaud, Bashir, and King 2011), with a handful of these responsible for more than 400 million human infections annually (Liang, Gao, and Gould 2015). While current efforts to eliminate the global threat posed by some arboviruses may be progressing rapidly (via vaccines and targeted vector and virus control (Moreira et al. 2009; Hoffmann et al. 2011; Villar et al. 2015)), minimising the global burden of all arboviruses is far from achievable (Patz and Olson 2006).

The causes of arboviral infections are multifactorial: the immediate cause is a bite by an infected vector, however, underlying factors relate to vector-host interactions, climatic and environmental variables enabling transmission, and immune responses of host populations (Weaver and Barrett 2004; Kuno and Chang 2005). Different transmission cycles are recognised as contributing to arboviral infections in humans, including epidemic and enzootic cycles (Figure 1.1). Zoonotic arboviruses are specifically maintained in enzootic cycles, with frequent transmission between vectors and non-human vertebrate species (Weaver and Barrett 2004). Viral emergence in humans usually coincides with the uptake of an enzootic virus by anthropozoophilic arthropods (those which bite both humans and non-human vertebrates). These complex interactions underpin the mechanisms which drive arboviral maintenance cycles and spillover events in humans. While much of the management of arboviruses has been concentrated in the diagnostics and treatment (Webster et al. 2014), the emergence of Zika virus in South America, and West Nile virus (WNV) in North America has served to heighten awareness of potential outbreaks and spillover of zoonotic arboviruses (Lanciotti et al. 1999; Lucey and Gostin 2016). The multifactorial drivers of infection require management of

arboviruses to also be multifactorial, necessitating collaboration from a variety of disciplines working at the direct and underlying causes of arboviral exposure.

Figure 1.1. Epidemic and enzootic arbovirus transmission cycles involving single or multiple hosts and vectors

The transmission ecology of arboviruses can vary in its complexity. Some arboviruses are considered specialists and persist in a limited number of vectors and/or host species (such as dengue virus; Figure 1.1) (Teixeira et al. 2002). Zoonotic arboviruses must have at least two host species (human and non-human vertebrates) which adds to the complexity of transmission and mitigation strategies. For arboviruses with more complex transmission cycles, treatment of humans alone is insufficient to disrupt the maintenance cycle (Garchitorena et al. 2017), as the virus is frequently maintained in non-human vertebrates (Figure 1.1). Multi-host, multi- vector arboviruses are, therefore, generalist viruses which can persist in a range of vertebrate hosts or mosquito vectors (Figure 1.1). Conceptually, spillover of these viruses is particularly challenging to predict, as it is reliant on the immunity of different vertebrate and vector species at both the individual (such as whether a species is competent at transmitting a virus) and the community level (such as whether there is a high herd immunity against a virus).

1.2 Identifying vectors and reservoirs of zoonotic arboviruses 1.2.1 Mosquito vectors Arthropoda is the most diverse phylum in the animal kingdom, consisting of more than one million species. Mosquitoes comprise only a fraction of these species with an estimated 3500

identified to date (Klowden 2007). The majority of mosquito species are not considered vectors of disease. To be a vector of a specific virus, a mosquito species must; (i) be hematophagous (blood feeding), (ii) be capable of supporting viral replication, and (iii) have host feeding behaviour(s) relevant to the given virus (Figure 1.1).

Vector competence studies are often used to determine whether a mosquito is capable of permitting viral replication (Lee et al. 2019). In these studies, mosquito species are exposed to known doses of a given virus. The proportion of mosquitoes that become infected, often followed by the proportion of mosquitoes that can transmit the virus, is measured (Doggett, Klowden, and Russell 2001). Mosquitoes have several antiviral mechanisms that may prohibit viral infection, progression and replication. Following the up-take of an infected blood meal, a virus will infect the epithelial cells of the mosquito midgut (Lee et al. 2019). For a mosquito to transmit the virus, it must disseminate to, and replicate in, various tissues before infection of the salivary glands occurs. Mosquito species differ in their antiviral mechanisms, and this is one of the critical factors that mediate the vector competence of a given species.

Vector-host feeding behaviour can be assessed both under controlled or natural conditions. Under controlled conditions, mosquitoes are released in an environment in which they can choose a host. These experiments may be conducted in a laboratory, for example with a ‘Y’ shaped tube, in which different hosts (or host cues) are placed at the end of the tube and a mosquito is released at the opposite end (Williams, Kokkinn, and Smith 2003). Alternatively, in outdoor settings, mosquitoes may be released in an enclosure housing caged or tethered vertebrates, with feeding mosquitoes aspirated from the vertebrate. A more common method to assess vector feeding patterns is blood meal analysis (Takken and Verhulst 2013). This involves capturing blood fed mosquitoes and identifying the vertebrate source of the bloodmeal via either serological or genetic techniques. This method is very informative, however, capturing blood fed mosquitoes is challenging and often yields low numbers (Stephenson et al. 2019). Each of these methods are associated with caveats in determining mosquito feeding patterns. In particular, it is recognised that patterns can vary across generations, locations and seasonally for the same species of mosquito (Takken and Verhulst 2013).

1.2.2 Vertebrate reservoirs A critical challenge in identifying vertebrate reservoirs is that different vertebrate species may play different roles in transmission. Species may act as amplifiers, diluters, or dead-end hosts. An ‘amplifier’ is a species, which following the bite of infected vector, is capable of replicating

a pathogen to a viraemia high enough that a susceptible vector may become infected. Collectively these species increase transmission in a community (Keesing, Holt, and Ostfeld 2006). In contrast, ‘diluter’ species are those which following the bite of an infected vector, are poor amplifiers of a pathogen, and collectively can reduce transmission of a pathogen at a community level (Ostfeld and Keesing 2012; Luis, Kuenzi, and Mills 2018). In multi-host disease systems, it is challenging, but critically important to identify whether a species may be an amplifier or a diluter for a given pathogen (Woolhouse, Taylor, and Haydon 2001).

A reservoir can be defined as a set of epidemiologically connected populations (such as mosquito vectors and vertebrate species) that permanently maintain a pathogen and transmit infection to target populations (such as humans) (Haydon et al. 2002; Viana et al. 2014). Several frameworks and criteria have been developed to identify a species as a reservoir of a pathogen. For arboviruses, the criteria commonly used to identify a vertebrate reservoir have been developed from virological principles. The criteria require a potential vertebrate reservoir to have demonstrated; (i) viraemia under controlled experimental infection conditions, (ii) successful virus isolation under natural conditions, and (iii) relatively high antibody prevalence in free-living specimens (Kuno and Chang 2005). While these criteria have strengths in identifying the capability of a species to amplify a pathogen to mosquito vectors, they are limited in identifying broader disease ecology, and in particular, demonstrating a natural association between potential vertebrate reservoir and vector.

Methods used to identify vertebrates as potential reservoirs include experimental infection, serosurveys, virus isolation and ecological surveys. Use of any defined method alone can be misleading if applied generically (Viana et al. 2014). Incorrect identification of reservoirs hosts could result in counter-productive control measures. Investigations of non-human reservoirs of a given pathogen are best when using a combination of conceptual frameworks and empirical evidence which identify a species as a capable amplifier and a reservoir under natural conditions. One example of using a combination of frameworks to identify reservoirs of vector- borne diseases is in the identification of Leishmania reservoirs (Silva, Gontijo, and Melo 2005; Chaves et al. 2007). Leishmania is a genus of trypanosomes transmitted by sandflies (Phlebotamus spp.), which cause a variety of human leishmaniasis diseases (Burza, Croft, and Boelaert 2018). The criteria used to identify reservoir hosts required a vertebrate species to have; (i) an overlapping geographical and temporal distribution with vectors, (ii) presence of the pathogen in non-human vertebrates, (iii) persistence of the pathogen at sufficient

parasitaemia to infect vectors, (iv) seroprevalence higher than 20% and, (v) high species survival of sufficient duration to ensure potential transmission to vectors (Silva, Gontijo, and Melo 2005). Although these criteria add to the framework more commonly used for arboviruses (described above), they has also been criticised for not distinguishing between dead-end or incidental hosts and reservoir species (Chaves et al. 2007).

1.3 Ross River virus; an important zoonotic arbovirus In Australia, at least twelve arboviruses are implicated in human disease (Table 1.1) (Russell 2009, 1995). With the exception of dengue virus, the majority of these arboviruses are maintained in enzootic transmission cycles (Russell 1995). While non-human species are critical for the persistence of these arboviruses in Australia, management of zoonotic arboviruses are currently either vector centric or focus on personal protection measures. In general, the vector-host interactions of Australian arboviruses are not well understood.

Table 1.1. Summary of Australian mosquito-borne viruses known to cause disease in humans Virus Distribution of human cases Transmission cycle Alphaviruses Ross River Nationwide Enzootic Barmah Forest Nationwide Enzootic Sindbis Nationwide, excluding Enzootic Tasmania Flaviviruses Murray Valley encephalitis Endemic in northwest, Enzootic epidemic in southeast West Nile (subtype Kunjin) Endemic in northwest Enzootic Japanese encephalitis Torres Strait islands, far north Enzootic Queensland Dengue Northern coastal Queensland Epidemic Edge Hill Northern Territory, Unknown Queensland, Western Australia and New South Wales Kokobera Northern Territory, Unknown Queensland, Western Australia and New South Wales Stratford New South Wales, Western Unknown Australia Bunyaviruses Gan Gan Endemic in southeast Enzootic Trubanaman Endemic in southeast Enzootic

Ross River virus (RRV) is responsible for the greatest number of reported mosquito-borne infections in Australia, averaging 4800 human notifications each year (Australian Government 2018). The virus is a multi-host, multi-vector Alphavirus of the Togoviridae family. Ross River virus is one of six Alphaviruses known to cause human joint disorders (Lwande et al. 2015), and circulates endemically in Australia and the Pacific. Ross River virus is unique because it has been isolated in more than 40 species of mosquitoes and persists in several non-human vertebrate species (Harley, Sleigh, and Ritchie 2001). This, in combination with the large public health and economic burden it has on human populations has led to the suggestion that RRV could be an important emerging infectious disease globally, with pandemic potential (Shanks 2019; Flies et al. 2018; Levi and Vignuzzi 2019).

1.3.1. Ross River virus disease and treatment First isolated in 1959, RRV is now widely believed to have been responsible for historic outbreaks referred to as “epidemic polyarthritis” in 1928 in Narrandera, New South Wales in 1928, and possibly earlier in 1886 in Natimuk, western Victoria (Kelly-Hope, Purdie, and Kay 2004). Suspicions that the virus was vector-borne were confirmed following the isolation of RRV from an Aedes vigilax (Skuse) mosquito near the Ross River in Townsville, Queensland (Doherty et al. 1963).

Ross River virus is not associated with mortality, however it is associated with various clinical symptoms and signs in humans including rash (lasting 5-10 days) (>50% of infections), fever (30-50%) and fatigue (>50%), myalgia (58%), and arthritis (especially in knees, wrists, ankles and fingers; (Mudge 1977; Fraser 1986; Barber, Denholm, and Spelman 2009; Condon and Rouse 1995; Harley et al. 2002). Symptoms typically resolve in 3-6 months. However, in some individuals joint pain has been reported to persist for prolonged periods of time (up to years). Following a bite from an infected mosquito, the intrinsic incubation period typically lasts 7-9 days (but has been reported between 3-21 days; (Kay and Aaskov 1989) before the display of signs in humans (Harley, Sleigh, and Ritchie 2001).

Notifications of RRV disease occur more frequently in adults (20-60 years old), with cases peaking in 30-50 year olds (Australian Government 2018). Disease in children is less frequently reported, and tends to be milder and shorter in duration than in adults (Fraser 1986). Children and adults exposed to RRV are also generally believed to develop life-long protective antibodies (Harley and Suhrbier 2013). Estimates of the proportion of asymptomatic cases in humans vary greatly, however the most widely accepted ratio is from 1.2:1 (Russell et al. 1998)

to 3.0:1 (Aaskov et al. 1981) asymptomatic to symptomatic cases. The highest estimate is 50:1 asymptomatic to symptomatic (Aaskov et al. 1981) but is largely considered a miscalculation as the study population included a group of non-resident soldiers which may have acquired infections elsewhere (Harley, Sleigh, and Ritchie 2001).

At present there is no specific treatment for RRV and only the symptoms can be managed supportively. For prevention, an inactive whole-virus Vero cell vaccine has been developed for RRV and underwent phase 3 testing (Wressnigg et al. 2015). A total of 1755 healthy younger adults (16-59 years old) and 209 healthy older adults (aged 60+) were involved in the study across 17 locations in Australia. The results from this trial found that this vaccine was safe and effective. This vaccine, however, has not been widely adopted or supported as a profitable investment by pharmaceutical companies, and is not currently commercially available (Aichinger et al. 2011; Holzer et al. 2011; Wressnigg et al. 2015).

Aside from human populations, horses are the only species reported to develop clinical signs with RRV (Dhama et al. 2014). Early evidence for RRV causing clinical disease in horses was conflicting. Numerous anecdotal reports suggested that RRV infection caused clinical disease in horses (Azuolas 1997; El-Hage, McCluskey, and Azuolas 2008), however experimental infection of 14 horses reported no clinical signs in any individuals (Kay et al. 1987). More recently, it is accepted that, similar to humans, RRV is associated with clinical disease in horses but that it does not occur in all infected individuals (Kay et al. 1987; Azuolas 1998; Gummow et al. 2018). Disease in horses manifests as fever, lameness, stiffness, swollen joints, reluctance to move and mild colic (Dhama et al. 2014). Clinical signs have been reported to persist for weeks to months in horses and, in some instances, can reoccur (Azuolas 1998; Studdert et al. 2003).

1.3.2 Economic burden of RRV The economic burden of RRV includes per person costs associated with diagnosis, treatment of symptoms, loss of income and inability to care for dependents. Estimates of costs also need to include vector control and social costs such as the potential loss of tourism when outbreaks occur. Only conservative estimates for diagnostic costs have been calculated for RRV, estimating $US 10 million is spent each year based on an average of 4745 cases per year (Aaskov et al. 1998) with the direct cost estimated at $AU 1018 per patient (Mylonas et al. 2002).

The economic burden of RRV may also be greatly underestimated when considering the impact of the disease on the horse racing industry. The horse racing industry is considered a large contributor to national GDP valued at more than $AU 6.3 billion in 2001, and employs more than 250,000 people (Russell 2002). At present, linkages between clinical signs and under performance are largely anecdotal, but correlations of RRV IgM antibodies with reduced racing performance have been observed (Johnston 2019). To date, no formal assessments of economic impact have been made for the racehorse industry, which also incorporates gambling turn-over.

1.3.3 Distribution and recent outbreaks of RRV Human notifications of RRV are reported from every state and territory in Australia, with the greatest number of cases consistently reported from Queensland (Figure 1.2) (Australian Government 2018). Until the last decade, RRV was predominantly associated with rural areas (Hill, Power, and Deane 2009). More recently, increasing incidence has been recorded in major urban centres of Australia including Brisbane, Adelaide, Melbourne and Sydney (Jansen et al. 2009). The largest outbreak of RRV occurred in Queensland in the summer of 2014-2015 with over 5,000 cases reported (Jansen et al. 2019), and, more recently, an epidemic of RRV was reported in the Melbourne region in 2016-2017 with human cases totalling more than state notifications of RRV for the previous three years combined (Australian Government 2018).

7000

6000

5000

4000

3000

2000 Numbernotifications of 1000

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 2019 Year

ACT NSW NT QLD SA TAS VIC WA

Figure 1.2. Number of human RRV notifications in each state between 1993 and 2018 (Source: National Notifiable Disease Surveillance System)

Outside Australia, RRV has been detected in several countries in the Pacific Islands (Figure 1.3). Between 1979 and 1980, RRV was responsible for major epidemics involving Fiji, American Samoa, New Caledonia, Wallis & Futuna, Vanuatu and the Cook Islands (Aaskov et al. 1981; Bennett et al. 1980; Goettel, Toohey, and Pillai 1980). Since this outbreak in the Pacific Islands, there has been limited information published on the persistence of RRV in these areas, but it seems likely that RRV is still circulating, supported by case reports of travellers returning from the Pacific Islands with locally acquired RRV (Lau, Weinstein, and Slaney 2012; Klapsing et al. 2005; Artsob and Spence 1991). In 2014, the known distribution of RRV was expanded to include French Polynesia (Figure 1.3), following a serosurvey which found 34% of blood donors to have positive serology for RRV antibodies (Aubry et al. 2015b, 2015a). There have been no previous reports of RRV in French Polynesia, suggesting the disease may be silently circulating other Pacific Islands, despite limited data. More empirical research is needed in these areas to determine the prevalence of RRV in the Pacific region.

Figure 1.3. Locations of reported local transmission of RRV to humans

1.3.4. Ross River virus transmission ecology

Although complicated, a detailed understanding of RRV transmission is critically important to prevent and predict localised outbreaks. To date, much of the research on RRV transmission has been vector-centric, with particular attention placed on common or abundant mosquito

species. A significant gap in understanding, however, is how important non-human vertebrates are for the maintenance and spillover of RRV.

RRV Vectors

Ross River virus has been isolated from 42 mosquito species; successful transmission from ten of these species has been confirmed under laboratory conditions (Russell 2002). Generally the main vectors are considered to be Aedes vigilax and Ae. camptorhynchus in coastal, saltwater regions and Culex annulirostris inland, with Ae. notoscriptus as the probable vector in urban areas (van den Hurk and Jansen 2016). Other species such as Coquillettidia linealis and Ae. procax are also thought to be important vectors in some regions (Ryan, Do, and Kay 2000) or during outbreaks (Jansen et al. 2019), but broadly, the relative importance of different mosquito species for maintaining RRV in the environment, and transferring infection from reservoirs to humans, is unknown.

Under vector competence experiments, Aedes vigilax and Aedes camptorhynchus have demonstrated some of the highest levels of competency by having a short extrinsic incubation period (as little as 4-5 days) and with the highest proportion of females able to transmit (or become infected) at 9-10 days (Ballard and Marshall 1986; Ryan, Do, and Kay 2000). These species have also demonstrated high transmission rates of RRV with 50-57% (Ae. vigilax) and 100% (Ae. camptorhynchus) capable of transmission following a feed from an infected bloodmeal (Kay et al. 1986; Ballard and Marshall 1986; Watson and Kay 1999). Comparatively, Ae. notoscriptus are infectious 4 days following a feed on an infectious host, but not capable of transmitting RRV until 9 days post-infection and with transmission rates highest between 12-14 days post-feeding (Ryan, Do, and Kay 2000). Aedes notoscriptus also have relatively low peak transmission rates of 13%. Despite this, given the high abundance of Ae. notoscriptus in urban areas, and their frequent feeding on both humans and non-human vertebrates, they are considered an important bridge vector in urban areas (Boyd and Kay 2001). Although understudied in vector competence studies, other species including Ae. procax, Cx. australicus and Cx. globocoxitus have been implicated as potential vectors based on correlations with their increasing abundance and outbreaks of RRV in humans (Dhileepan, Azuolas, and Gibson 1996; Jansen et al. 2019).

Vertical transmission (transmission from breeding female to larvae) across different species of mosquitoes is yet to be quantified to determine its potential importance more broadly in the transmission dynamics of RRV. RRV has been isolated from pools of male adult Ae. vigilax,

Ae. tremulus and Ae. camptorhychus mosquitoes under natural conditions (Dhileepan, Azuolas, and Gibson 1996; Lindsay et al. 1993). As males of these species do not blood feed, this provides some evidence for vertical transmission of RRV, which could be an important over- wintering strategy of the virus, but is currently unexplored.

RRV Reservoirs

The long-held dogma for RRV reservoirs has been that marsupials (macropods in particular) are better reservoirs of RRV than placental mammals, which in turn are better reservoirs than birds. The evidence for this largely arose from experimental infection studies which demonstrated that marsupials had one of the highest, longest-lived viraemias compared to other species (Kay et al. 1986). In the last decade, however, this dogma has been called into question following the discovery of endemic RRV circulation in the Pacific islands where marsupials are absent (Flies et al. 2018; Aubry et al. 2015b). Presently, there is increasing speculation that transmission dynamics are locally unique (Claflin and Webb 2015) with different vectors and, more than likely, different reservoirs playing important roles across inland, metropolitan and coastal areas.

The earliest investigations of the non-human reservoirs of RRV were in response to the isolation of the virus in Townsville in 1959. This research included experimental infection studies (Whitehead 1969), serosurveys of wild and domestic species (Whitehead et al. 1968; Doherty et al. 1966), serosurveys of humans (Doherty et al. 1966) and virus isolation attempts (Doherty et al. 1968; Doherty et al. 1971). The results from the studies so far had indicated that the seroprevalence in mammals was typically higher compared to birds (Marshall, Woodroofe, and Hirsch 1982; Doherty et al. 1966). These studies yielded the first three isolates of RRV in vertebrates which surprisingly came from three bird species; a magpie-lark Grallina cyanoleuca, a flycatcher Myiagra rubecula and a masked finch Poephila personata. Ross River virus has since been isolated from fifteen horses Equus callabus and two agile wallabies Macropus agilis (Azuolas 1998; Pascoe, George, and Cybinski 1978; Doherty et al. 1971).

With a greater focus now on mammals, epidemiological studies began in Tasmania in 1974 and assessed the seroprevalence of RRV in marsupials and cattle. A high seroprevalence was noted for marsupials, particularly brushtail possums Trichosurus vulpecula, pademelons Thylogale billardierii and red-necked wallabies Macropus rufogriseus (McManus and Marshall 1986). Around the same time brushtail possums were recognised for their ability to maintain and spread bovine tuberculosis in New Zealand (Coleman 1988). Investigations into

the diseases of brushtail possums were launched in Australia, and a study on Kangaroo Island identified RRV seropositive brushtail possums (O'Callaghan and Moore 1986). With the spotlight on brushtail possums in previous studies and their high abundance in urban areas, they were further investigated in an experimental infection study. Brushtail possums were found to have a higher titre than other experimentally infected species, and were capable of infecting susceptible mosquitoes (Boyd et al. 2001). Since this study, brushtail possums were implicated as a key reservoirs of RRV in urban environments (Kay et al. 2007). However, this has been questioned in more recent years after a study assessing the seroprevalence of possums in urban Sydney all returned negative results (Hill, Power, and Deane 2009), despite noted human incidence in the area.

Seroprevalence studies have been useful to determine vector-host interactions and the presence of RRV in a given location, but are limited in their ability to implicate a given species as a reservoir as they only determine previous exposure to RRV. Seroprevalence studies have identified RRV exposure in a broad range of species. In general, marsupials have reported the highest seropositivity, followed by horses (Stephenson et al. 2018). Investigations into the seroprevalence of bird species have been limited with only 13 species assessed, of which only two species (Tawny frogmouth Podargus strigoides and Australian gannet Morus serrator) had seropositivity. Only one study has investigated the relationship between seroprevalence, age and sex, across multiple populations for a single species, Western grey kangaroos Macropus fuliginosus. The results from this study found that location was a driving factor in seroprevalence, suggesting local transmission dynamics, including vector availability, are important for RRV (Potter et al. 2014).

In addition to experimental infection and serological studies, models have been developed to understand the transmission dynamics of RRV, and predict disease outbreaks (Jacups et al. 2015; Williams 2015; Williams, Fricker, and Kokkinn 2009; Koolhof and Carver 2017; Yu et al. 2014; Ng et al. 2014; Jacups, Whelan, and Harley 2011; Hu et al. 2010). Most of these are reliant on climatic factors such as temperature, rainfall and tidal patterns. Yet, despite climatic conditions conducive to an outbreak of mosquito-borne disease, it does not always occur when predicted (Lindsay 1995). Crucially, only a handful of these models incorporate the host ecology, and it is likely that reservoir dynamics (and thereby force of infection) are dampened due to immunity in reservoir populations (Lindsay 1995; Carver, Bestall, et al. 2009; Carver, Spafford, et al. 2009). The fragmentary understanding of the role of non-human vertebrate

species in the spread and maintenance of RRV limits our ability to accurately predict and mitigate outbreaks of RRV.

1.4 Thesis objectives In response to public health concerns of RRV in Australia, and limited understanding of the role of non-human reservoirs in the maintenance and transmission of this disease, this thesis develops a multifaceted research agenda to address key priorities and knowledge gaps related to 1) identifying key vertebrate hosts and their vector associations; 2) characterising host availability and their exposure to RRV in South East Queensland; and 3) identifying host variables most important for human notifications of RRV. Figure 1.4 provides a schematic guide to these research aims and their associated chapters and publications or manuscripts in this dissertation.

Figure 1.4. Schematic graph of project aims and corresponding chapters, publications and manuscripts associated with each.

Chapters 2 and 3 address the first aim of this dissertation. Before undertaking field and laboratory investigations on the non-human reservoirs it was critical to review all existing evidence for species. Both chapters used systematic literature review methods to collect and filter relevant literature. In Chapter 2, evidence for reservoirs was reviewed in light of the long- held dogma that marsupials are better reservoirs than placental mammals, which in turn are better than birds. Assessments were made on the methods used to identify potential reservoirs,

the species selected and limitations within studies. Chapter 3 considered Australian mosquito species and their association with potential vertebrates by assessing mosquito feeding patterns. A meta-analysis of the literature was undertaken and mosquito feeding patterns were analysed. The findings suggested that mosquito species in Australia each have unique feeding patterns, which did not correspond to their taxonomic relatedness or larval ecology. It was apparent that only one of the mosquito blood meal studies in Australia undertook a vertebrate survey at the time of mosquito collection. As preference is a function of usage relative to availability it was not possible to determine whether the patterns observed for mosquitoes in Australia reflected a preference for a given vertebrate, or more so what was most abundant. These chapters have both been published in Parasites & Vectors.

Chapter 4 and chapter 5 considered potential RRV hosts, with humans and other vertebrates examined, respectively. Outside of epidemics, humans are broadly not considered to be important reservoirs of RRV. Yet, as a species that experiences clinical signs and symptoms and is frequently tested for arboviruses, human serology provides a large amount of data and information on RRV transmission in Australia and the Pacific. Literature describing human exposure to RRV, dengue and BFV in Australia and the Pacific was collated and assessed using a meta-analysis in Chapter 4. In Chapter 5, non-human vertebrates were assessed in a highly endemic area for RRV, South-East Queensland.

Finally, to address the third aim of this dissertation a field study was undertaken in Brisbane, the capital city Queensland, located in the southeast of the state (Chapter 6). Data were collected across a gradient of human notification rates and the composition of both vertebrate and vector communities assessed. A number of variables were analysed, including species diversity, species abundance and vertebrate biomass.

1.5 References

Aaskov, J. G., J. Y. Chen, N. T. H. Hanh, and P. M. Dennington. 1998. 'Surveillance for Ross River virus infection using blood donors', American Journal of Tropical Medicine and Hygiene, 58: 726-30. Aaskov, J. G., J. U. Mataika, G. W. Lawrence, V. Rabukawaqa, M. M. Tucker, J. A. R. Miles, and D. A. Dalglish. 1981. 'An epidemic of Ross River virus infection in Fiji', American Journal of Tropical Medicine and Hygiene, 30: 1053-59. Aichinger, G., H. J. Ehrlich, J. G. Aaskov, S. Fritsch, C. Thomasser, W. Draxler, M. Wolzt, M. Mueller, F. Pinl, P. van Damme, A. Hens, J. Levy, D. Portsmouth, G. Holzer, O. Kistner, T. R. Kreil, and N. P. Barrett. 2011. 'Safety and immunogenicity of an inactivated whole virus Vero cell-derived Ross River virus vaccine: A randomized trial', Vaccine, 29: 9376-84. Artsob, H., and L. Spence. 1991. 'Imported arbovirus infections in Canada 1974-89', The Canadian journal of infectious diseases, 2: 95-100. Aubry, M., J. Finke, A. Teissier, C. Roche, J. Broult, S. Paulous, P. Despres, V-M. Cao- Lormeau, and D. Musso. 2015a. 'Seroprevalence of arboviruses among blood donors in French Polynesia, 2011-2013', International Journal of Infectious Diseases, 41: 11-12. Aubry M., J. Finke, A. Teissier, C. Roche, J. Broult, S. Paulous, P. Despres, V-M. Cao- Lormeau, and D. Musso. 2015b. 'Silent Circulation of Ross River Virus in French Polynesia', International Journal of Infectious Diseases, 37: 19-24. Australian Government. 2018. 'National Notifiable Diseases Surveillance System'. http://www9.health.gov.au/cda/source/cda-index.cfm. Azuolas, J. K. 1997. 'Arboviral diseases of horses and possums', Arbovirus Res. Aust, 7: 5-7. Azoulas, J. K. 1998. 'Ross River virus disease of horses', Australian Equine Veterinarian, 16: 56-58. Ballard, J. W. O., and I. D. Marshall. 1986. 'An investigation of the potential of Aedes camptorhynchus (Thom) as a vector of Ross River virus', Australian Journal of Experimental Biology and Medical Science, 64: 197-200. Barber, Bridget, Justin T. Denholm, and Denis Spelman. 2009. 'Ross River virus', Australian family physician, 38: 586-89. Bennett, N. M., A. L. Cunningham, J. R. E. Fraser, and B. R. Speed. 1980. 'Epidemic polyarthritis acquired in Fiji ', Medical Journal of Australia, 67-1: 316-17. Boyd, A. M., R. A. Hall, R. T. Gemmell, and B. H. Kay. 2001. 'Experimental infection of Australian brushtail possums, Trichosurus vulpecula (Phalangeridae : Marsupialia), with Ross River and Barmah Forest viruses by use of a natural mosquito vector system', American Journal of Tropical Medicine and Hygiene, 65: 777-82. Boyd, A. M., and B. H. Kay. 2001. 'Solving the urban puzzle of Ross River and Barmah Forest viruses', Arbovirus Research in Australia, 8: 14-22. Burza, S., S. L. Croft, and M. Boelaert. 2018. 'Leishmaniasis.', Lancet, 392: 951-70. Carver, S., A. Bestall, A. Jardine, and R. S. Ostfeld. 2009. 'Influence of Hosts on the Ecology of Arboviral Transmission: Potential Mechanisms Influencing Dengue, Murray Valley Encephalitis, and Ross River Virus in Australia', Vector-Borne and Zoonotic Diseases, 9: 51-64. Carver, S., H. Spafford, A. Storey, and P. Weinstein. 2009. 'Dryland Salinity and the Ecology of Ross River Virus: The Ecological Underpinnings of the Potential for Transmission', Vector-Borne and Zoonotic Diseases, 9: 611-22. Chaves, L. F., M. J. Hernandez, A. P. Dobson, and M. Pascual. 2007. 'Sources and sinks: revisiting the criteria for identifying reservoirs for American cutaneous leishmaniasis', Trends Parasitology, 23: 311-6.

Claflin, S. B., and C. E. Webb. 2015. 'Ross River Virus: Many Vectors and Unusual Hosts Make for an Unpredictable Pathogen', Plos Pathogens, 11. Coleman, JD. 1988. 'Distribution, Prevalence, and Epidemiology of Bovine Tuberculosis in Brushtail Possums, Trichosurus-Vulpecula, in the Hohonu Range, New-Zealand', Wildlife Research, 15: 651-63. Condon, R. J., and I. L. Rouse. 1995. 'Acute symptoms and sequelae of Ross River virus infection in South Western Australia. A follow up study. ', Clinical and Diagnostic Virology, 3: 273-84. Daszak, P., A. A. Cunningham, and A. D. Hyatt. 2001. 'Anthropogenic environmental change and the emergence of infectious diseases in wildlife', Acta Tropica, 78: 103-16. Dhama, K., S. Kapoor, R. V. S. Pawaiya, S. Chakraborty, R. Tiwari, and A. K. Verma. 2014. 'Ross River Virus (RRV) infection in horses and humans: a review', Pakistan Journal of Biological Sciences, 17: 768-79. Dhileepan, K., J. K. Azuolas, and C. A. Gibson. 1996. 'Evidence of vertical transmission of Ross River and Sindbis viruses (Togaviridae: Alphavirus) by mosquitoes (Diptera: Culicidae) in southeastern Australia', Journal of Medical Entomology, 33: 180-82. Doggett, S. L., M. J. Klowden, and R. C. Russell. 2001. 'Are vector competence experiments competent vector experiments?', Arbovirus Research in Australia, 8: 126-30. Doherty, R. L., B. M. Gorman, R. H. Whitehead, and J. G. Carley. 1966. 'Studies of arthropod- borne virus infections in Queensland .V. Survey of antibodies to Group A arboviruses in man and other animals ', Australian Journal of Experimental Biology and Medical Science, 44: 365-&. Doherty, R. L., H. A. Standfast, R. Domrow, E. J. Wetters, R. H. Whitehead, and J. G. Carley. 1971. 'Epidemiology of arthropod-borne virus infections at Mitchell River Mission, Cape York Peninsula, North Queensland. IV. Arbovirus infections of mosquitoes and mammals, 1967-1969.', Transactions of the Royal Society of Tropical Medicine and Hygiene, 65: 504-&. Doherty, R. L., Standfast. H., E. J. Wetters, R. Whitehead, G. J. Barrow, and B. M. Gorman. 1968. 'Virus isolation and serological studies of arthopod-borne virus infections in a high rainfall area of North Queensland ', Transactions of the Royal Society of Tropical Medicine and Hygiene, 62: 862-&. Doherty, R. L., R. H. Whitehead, B. M. Gorman, and A. K. O'Gower. 1963. 'The isolation of a third group A arbovirus in Australia, with preliminary observations on its relationship to epidemic polyarthritis', Australian Journal of Science, 26: 183-84. El-Hage, C. M., M. J. McCluskey, and J. K. Azuolas. 2008. 'Disease suspected to be caused by Ross River virus infection of horses', Australian Veterinary Journal, 86: 367-70. Flies, E. J., C. L. Lau, S. Carver, and P. Weinstein. 2018. 'Another emerging Mosquito-Borne disease? Endemic Ross River virus transmission in the absence of marsupial reservoirs', BioScience, 68: 288-93. Fraser, J. R. E. 1986. 'Epidemic polyarthritis and Ross River virus disease', Clinics in Rheumatic Diseases, 12: 369-88. Garchitorena, A., S. H. Sokolow, B. Roche, C. N. Ngonghala, M. Jocque, A. Lund, M. Barry, E. A. Mordecai, G. C. Daily, and J. H. Jones. 2017. 'Disease ecology, health and the environment: a framework to account for ecological and socio-economic drivers in the control of neglected tropical diseases', Philosophical Transactions of the Royal Society B: Biological Sciences, 372: 20160128. Goettel, M. S., M. K. Toohey, and J. S. Pillai. 1980. 'The urban mosquitoes of Auva, Fiji - Seasonal incidence and ecaluation of environmental sanitation and ULV spraying for their control', Journal of Tropical Medicine and Hygiene, 83: 165-71.

Gummow, B., R. H.H. Tan, R. K. Joice, G. Burgess, and J. Picard. 2018. 'Seroprevalence and associated risk factors of mosquito‐borne alphaviruses in horses in northern Queensland', Australian Veterinary Journal, 96: 243-51. Harley, D., D. Bossingham, D. M. Purdie, N. Pandeya, and A. C. Sleigh. 2002. 'Ross River virus disease in tropical Queensland: evolution of rheumatic manifestations in an inception cohort followed for six months', Medical Journal of Australia, 177: 352-55. Harley, D., A. Sleigh, and S. Ritchie. 2001. 'Ross River virus transmission, infection, and disease: a cross-disciplinary review', Clinical Microbiology Reviews, 14: 909-32. Harley, D., and A. Suhrbier. 2013. 'A Ross River virus disease.' in, Hunter's Tropical Medicine and Emerging Infectious Disease, 9th Edition (Saunders). Haydon, D. T., S. Cleaveland, L. H. Taylor, and M. K. Laurenson. 2002. 'Identifying reservoirs of infection: a conceptual and practical challenge', Emerging Infectious Diseases, 8: 1468-73. Hill, N. J., M. L. Power, and E. M. Deane. 2009. 'Absence of Ross River virus amongst Common brushtail possums (Trichosurus vulpecula) from metropolitan Sydney, Australia', European Journal of Wildlife Research, 55: 313-16. Hoffmann, A. A., B. L. Montgomery, J. Popovici, I. Iturbe-Ormaetxe, P. H. Johnson, F. Muzzi, M. Greenfield, M. Durkan, Y. S. Leong, and Y. Dong. 2011. 'Successful establishment of Wolbachia in Aedes populations to suppress dengue transmission', Nature, 476: 454- 57. Holzer, G. W., S. Coulibaly, G. Aichinger, H. Savidis-Dacho, J. Mayrhofer, S. Brunner, K. Schmid, O. Kistner, J. G. Aaskov, F. G. Falkner, H. Ehrlich, P. N. Barrett, and T. R. Kreil. 2011. 'Evaluation of an inactivated Ross River virus vaccine in active and passive mouse immunization models and establishment of a correlate of protection', Vaccine, 29: 4132-41. Hu, W., A. Clements, G. Williams, S. Tong, and K. Mengersen. 2010. 'Bayesian Spatiotemporal Analysis of Socio-Ecologic Drivers of Ross River Virus Transmission in Queensland, Australia', American Journal of Tropical Medicine and Hygiene, 83: 722-28. Jacups, S. P., J. Carter, N. Kurucz, J. McDonnell, and P. I. Whelan. 2015. 'Determining meteorological drivers of salt marsh mosquito peaks in tropical northern Australia', Journal of Vector Ecology, 40: 277-81. Jacups, S. P., P. I. Whelan, and D. Harley. 2011. 'Arbovirus Models to Provide Practical Management Tools for Mosquito Control and Disease Prevention in the Northern Territory, Australia', Journal of Medical Entomology, 48: 453-60. Jansen, C. C., N. A. Prow, C. E. Webb, R. A. Hall, A. T. Pyke, B. J. Harrower, I. L. Pritchard, P. Zborowski, S. A. Ritchie, R. C. Russell, and A. F. van den Hurk. 2009. 'Arboviruses isolated from mosquitoes collected from urban and peri-urban areas of Eastern Australia', Journal of the American Mosquito Control Association, 25: 272-78. Jansen, C. C., M. Shivas, F. May, A. Pyke, M. Onn, K. Lodo, S. Hall-Mendelin, J. L. McMahon, J. Darbro, S. Doggett, and A. van den Hurk. 2019. 'Epidemiologic, entomologic, and virologic factors of the 2014–15 Ross River virus outbreak, Queensland, Australia', Emerging Infectious Diseases, 25(5). Johnston, S. . 2019. "Detection of IgG antibodies in individual horse 12 years following Ross River virus infection." In, edited by Eloise Stephenson. Jones, K. E., N. G. Patel, M. A. Levy, A. Storeygard, D. Balk, J. L. Gittleman, and P. Daszak. 2008. 'Global trends in emerging infectious diseases', Nature, 451: 990. Kay, B. H., and J. G. Aaskov. 1989. 'Ross River virus (epidemic polyarthritis)', The arboviruses: epidemiology and ecology, 4: 93-112.

Kay, B. H., A M Boyd, P. A. Ryan, and R. A. Hall. 2007. 'Mosquito feeding patterns and natural infection of vertebrates with Ross River and Barmah Forest viruses in Brisbane, Australia', American Journal of Tropical Medicine and Hygiene, 76: 417-23. Kay, B. H., R. A. Hall, I. D. Fanning, P. Mottram, P. L. Young, and C. C. Pollitt. 1986. "Experimental infection of vertebrates with Murray Valley encephalitis and Ross River viruses." In Arbovirus Research in Australia, 71-75. Kay, B. H., C. C. Pollitt, I. D. Fanning, and R. A. Hall. 1987. 'The experimental infection of horses with Murray Valley encephalitis and Ross River viruses', Australian Veterinary Journal, 64: 52-55. Keesing, F., R. D. Holt, and R. S. Ostfeld. 2006. 'Effects of species diversity on disease risk', Ecology Letters, 9: 485-98. Kelly-Hope, L. A., D. M. Purdie, and B. H. Kay. 2004. 'Ross River virus disease in Australia, 1886-1998, with analysis of risk factors associated with outbreaks', Journal of Medical Entomology, 41: 133-50. Klapsing, P., J. D. MacLean, S. Glaze, K. L. McClean, M. A. Drebot, R. S. Lanciotti, and G. L. Campbell. 2005. 'Ross River virus disease reemergence, Fiji, 2003-2004', Emerging Infectious Diseases, 11: 613-15. Klowden, M. J. 2007. 'Making generalizations about vectors: Is there a physiology of “the mosquito”?', Entomological Research, 37: 1-13. Koolhof, I. S., and S. Carver. 2017. 'Epidemic host community contribution to mosquito-borne disease transmission: Ross River virus', Epidemiology & Infection, 145: 656-66. Kuno, G., and G. J. Chang. 2005. 'Biological transmission of arboviruses: reexamination of and new insights into components, mechanisms, and unique traits as well as their evolutionary trends', Clinical Microbiology Reviews, 18: 608-37. LaBeaud, D. a., F. Bashir, and C. H. King. 2011. 'Measuring the burden of arboviral diseases: the spectrum of morbidity and mortality from four prevalent infections', Population Health Metrics, 9: 1. Lanciotti, R. S., J. T. Roehrig, V. Deubel, J. Smith, M. Parker, K. Steele, B. Crise, K. E. Volpe, M. B. Crabtree, and J. H. Scherret. 1999. 'Origin of the West Nile virus responsible for an outbreak of encephalitis in the northeastern United States', science, 286: 2333-37. Lau, A. L. D., I. Chi, R. A. Cummins, T. M. C. Lee, K-L. Chou, and L. W. M. Chung. 2008. 'The SARS (Severe Acute Respiratory Syndrome) pandemic in Hong Kong: Effects on the subjective wellbeing of elderly and younger people', Aging and mental health, 12: 746-60. Lau, C., P. Weinstein, and D. Slaney. 2012. 'Imported cases of Ross River virus disease in New Zealand - A travel medicine perspective', Travel Medicine and Infectious Disease, 10: 129-34. Lee, W-S., J. A. Webster, E. T. Madzokere, E. B. Stephenson, and L. J. Herrero. 2019. 'Mosquito antiviral defense mechanisms: a delicate balance between innate immunity and persistent viral infection', Parasites & Vectors, 12: 165. Levi, L. I., and M. Vignuzzi. 2019. 'Arthritogenic Alphaviruses: A Worldwide Emerging Threat?', Microorganisms, 7: 133. Liang, G., X. Gao, and E. A. Gould. 2015. 'Factors responsible for the emergence of arboviruses; strategies, challenges and limitations for their control', Emerging microbes & infections, 4: e18. Lindsay, M. D. A. 1995. Ecology and epidemiology of Ross River virus in Western Australia (University of Western Australia). Lindsay, M. D. A., A. K. Broom, A. E. T. Wright, C. A. Johansen, and J. S. Mackenzie. 1993. 'Ross RIver virus isolation from mosquiotes in arid regions of Western Australia -

implications of vertical transmission as a means of persistence of the virus', American Journal of Tropical Medicine and Hygiene, 49: 686-96. Lucey, D. R., and L. O. Gostin. 2016. 'The emerging Zika pandemic: enhancing preparedness', Jama, 315: 865-66. Luis, A. D., A. J. Kuenzi, and J. N. Mills. 2018. 'Species diversity concurrently dilutes and amplifies transmission in a zoonotic host–pathogen system through competing mechanisms', Proceedings of the National Academy of Sciences, 115: 7979-84. Lwande, O. W., V. Obanda, G. Bucht, G. Mosomtai, V. Otieno, C. Ahlm, and M. Evander. 2015. 'Global emergence of Alphaviruses that cause arthritis in humans', Infection ecology & epidemiology, 5: 29853-53. Marshall, I. D., G. M. Woodroofe, and S. Hirsch. 1982. 'Viruses recovered from mosquitoes and wildlife serum collected in the Murray Valley of Southeastern Australia February 1974 during an epidemic of encephalitis', Australian Journal of Experimental Biology and Medical Science, 60: 456-70. McManus, T. J., and I. D. Marshall. 1986. 'The epidemiology of Ross River virus in Tasmania', Arbovirus Research in Australia, 1986: 127-31. Moreira, L. A., I. Iturbe-Ormaetxe, J. A. Jeffery, G. Lu, A. T. Pyke, L. M. Hedges, B. C. Rocha, S. Hall-Mendelin, A. Day, and M. Riegler. 2009. 'A Wolbachia symbiont in Aedes aegypti limits infection with dengue, Chikungunya, and Plasmodium', Cell, 139: 1268- 78. Mudge, P. R. 1977. 'Survey of epidemic polyarthritis in Riverland Area, 1976', Medical Journal of Australia, 1: 649-51. Mylonas, A. D., A. M. Brown, T. L. Carthew, B. McGrath, D. M. Purdie, N. Pandeya, P. C. Vecchio, L. G. Collins, I. D. Gardner, F. J. de Looze, E. J. Reymond, and A. Suhrbier. 2002. 'Natural history of Ross River virus-induced epidemic polyarthritis', Medical Journal of Australia, 177: 356-60. Ng, V., K. Dear, D. Harley, and A. McMichael. 2014. 'Analysis and Prediction of Ross River Virus Transmission in New South Wales, Australia', Vector-Borne and Zoonotic Diseases, 14: 422-38. O'Callaghan, M. G., and E. Moore. 1986. 'Parasites and serological survey of the common brushtail possum (Trichosurus vulpecula) from Kangaroo Island, South Australia', Journal of Wildlife Diseases, 22: 589-91. Ostfeld, R. S., and F. Keesing. 2012. 'Effects of host diversity on infectious disease', Annual Review of Ecology, Evolution, and Systematics, 43: 157. Pascoe, RRR, TD George, and DH Cybinski. 1978. 'The isolation of a Ross River virus from a horse', Australian Veterinary Journal, 54: 600-00. Patz, J. A., and S. H. Olson. 2006. 'Climate change and health: global to local influences on disease risk', Annals of Tropical Medicine & Parasitology, 100: 535-49. Plowright, R. K., C. R. Parrish, H. McCallum, P. J. Hudson, A. I. Ko, A. L. Graham, and J. O. Lloyd-Smith. 2017. 'Pathways to zoonotic spillover', Nature Reviews Microbiology. Potter, A., C. A. Johansen, S. Fenwick, S. A. Reid, and M. D. A. Lindsay. 2014. 'The seroprevalence and factors associated with Ross River virus infection in Western grey kangaroos (Macropus fuliginosus) in Western Australia', Vector-Borne and Zoonotic Diseases, 14: 740-45. Ross, A. G. P., R. M. Olveda, and L. Yuesheng. 2014. 'Are we ready for a global pandemic of Ebola virus?', International Journal of Infectious Diseases, 28: 217-18. Russell, R. C. 1995. 'Arboviruses and their vectors in Australia: an update on the ecology and epidemiology of some mosquito-borne arboviruses', Review of Medical and Veterinary Entomology, 83: 141-58.

Russell, R. C. 2002. 'Ross River virus: Ecology and distribution', Annual Review of Entomology, 47: 1-31. Russell, R. C. 2009. 'Mosquito-borne disease and climate change in Australia: time for a reality check', Australian Journal of Entomology, 48: 1-7. Russell, R. C., S. E. Cope, A. J. Yund, and L. Hueston. 1998. 'Combatting the enemy— mosquitoes and Ross River virus in a joint military exercise in tropical Australia', American Journal of Tropical Medicine and Hygiene, 59: S307. Ryan, P. A., K. A. Do, and B. H. Kay. 2000. 'Definition of Ross River virus vectors at Maroochy Shire, Australia', Journal of Medical Entomology, 37: 146-52. Shanks, D. G. . 2019. 'Could Ross River Virus be the next Zika?', Journal of Travel Medicine. Silva, E. S., C. M. F. Gontijo, and M. N. Melo. 2005. 'Contribution of molecular techniques to the epidemiology of neotropical Leishmania species', Trends in parasitology, 21: 550-52. Stephenson, E. B., A. K. Murphy, C. C. Jansen, A. J. Peel, and H. McCallum. 2019. 'Interpreting mosquito feeding patterns in Australia through an ecological lens: an analysis of blood meal studies', Parasites & Vectors, 12: 156. Stephenson, E. B., A. J. Peel, S. A. Reid, C. C. Jansen, and H. McCallum. 2018. 'The non- human reservoirs of Ross River virus: a systematic review of the evidence', Parasites & Vectors, 11: 188. Studdert, M. J., J. K. Azoulas, J. R. Vasey, R. A. Hall, N. Ficorilli, and J. A. Huang. 2003. 'Polymerase chain reaction tests for the identification of Ross River, Kunjin and Murray Valley encephalitis virus infections in horses', Australian Veterinary Journal, 81: 76- 80. Takken, W., and N. O. Verhulst. 2013. 'Host preferences of blood-feeding mosquitoes', Annual Review Entomology, 58: 433-53. Taylor, L. H., S. M. Latham, and M. E. J. Woolhouse. 2001. 'Risk factors for human disease emergence', Philosophical Transactions of the Royal Society of London. Series B: Biological Sciences, 356: 983-89. Teixeira, M. D., M. L. Barreto, M. D. N. Costa, L. D. A. Ferreira, P. F. C. Vasconcelos, and S. Cairncross. 2002. 'Dynamics of dengue virus circulation: a silent epidemic in a complex urban area', Tropical Medicine & International Health, 7: 757-62. van den Hurk, A. F., and C. C. Jansen. 2016. 'Arboviruses of Oceania.' in, Neglected Tropical Diseases-Oceania (Springer). Viana, M., R. Mancy, R. Biek, S. Cleaveland, P. C. Cross, J. O. Lloyd-Smith, and D. T. Haydon. 2014. 'Assembling evidence for identifying reservoirs of infection', Trends Ecology Evolution, 29: 270-9. Villar, L., G. H. Dayan, J. L. Arredondo-García, D. M. Rivera, R. Cunha, C. Deseda, H. Reynales, M. S. Costa, J. O. Morales-Ramírez, and G. Carrasquilla. 2015. 'Efficacy of a tetravalent dengue vaccine in children in Latin America', New England Journal of Medicine, 372: 113-23. Watson, T. M., and B. H. Kay. 1999. 'Vector competence of Aedes notoscriptus (Diptera : Culicidae) for Barmah Forest virus and of this species and Aedes aegypti (Diptera : Culicidae) for dengue 1-4 viruses in Queensland, Australia', Journal of Medical Entomology, 36: 508-14. Weaver, S.C., and A. D. Barrett. 2004. 'Transmission cycles, host range, evolution and emergence of arboviral disease', Nature Review Microbiology, 2: 789-801. Webster, J. P., D. H. Molyneux, P. J. Hotez, and A. Fenwick. 2014. 'The contribution of mass drug administration to global health: past, present and future', Philosophical Transactions of the Royal Society B: Biological Sciences, 369: 20130434.

Whitehead, R. 1969. 'Experimental infection of vertebrates with Ross River and Sindbis viruses 2 Group A arboviruses isolated in Australia', Australian Journal of Experimental Biology and Medical Science, 47: 11-&. Whitehead, R. H., R. L. Doherty, R. Domrow, H. A. Standfast, and E. J. Wetters. 1968. 'Studies of the epidemiology of arthropod-borne virus infections at Mitchell River Mission, Cape York Peninsula, North Queensland: III. Virus studies of wild birds, 1964–1967', Transactions of the Royal Society of Tropical Medicine and Hygiene, 62: 439-45. Williams, C. R. 2015. 'Vector-borne disease in South Australia's future climate', Transactions of the Royal Society of South Australia, 139: 121-29. Williams, C. R., S. R. Fricker, and M. J. Kokkinn. 2009. 'Environmental and entomological factors determining Ross River virus activity in the River Murray Valley of South Australia', Australian and New Zealand Journal of Public Health, 33: 284-88. Williams, C. R., M. J. Kokkinn, and B. P. Smith. 2003. 'Intraspecific variation in odor-mediated host preference of the mosquito Culex annulirostris', Journal of chemical ecology, 29: 1889-903. Woolhouse, M. E. J., L. H. Taylor, and D. T. Haydon. 2001. 'Population biology of multihost pathogens', science, 292: 1109-12. Wressnigg, N., M. V. W. van der Velden, D. Portsmouth, W. Draxler, M. O'Rourke, P. Richmond, S. Hall, W. J. H. McBride, A. Redfern, J. Aaskov, N. P. Barrett, and G. Aichinger. 2015. 'An Inactivated Ross River Virus Vaccine Is Well Tolerated and Immunogenic in an Adult Population in a Randomized Phase 3 Trial', Clinical and Vaccine Immunology, 22: 267-73. Yu, W., P. Dale, L. Turner, and S. Tong. 2014. 'Projecting the impact of climate change on the transmission of Ross River virus: methodological challenges and research needs', Epidemiology and Infection, 142: 2013-23.

22 Chapter 2

The non-human reservoirs of Ross River virus: a systematic literature review of the evidence

Statement of contribution to co-authored published paper

The following chapter includes a co-authored paper:

Stephenson, E. B., Peel, A., J., Reid, S. A., Jansen, C. C., & McCallum, H. (2019) The non-human reservoirs of Ross River virus: a systematic literature review of the evidence. Parasites & Vectors. 11:188. doi:10.1186/s13071-018-2733-8

My contribution to the paper involved data collection, planning and performing all analyses, and writing the paper.

Signed: ______Date: ______Corresponding author: Eloise Stephenson

Countersigned: ______Date:______Supervisor: Hamish McCallum

23 Stephenson et al. Parasites & Vectors (2018) 11:188 https://doi.org/10.1186/s13071-018-2733-8

REVIEW Open Access The non-human reservoirs of Ross River virus: a systematic review of the evidence Eloise B. Stephenson1*, Alison J. Peel1, Simon A. Reid2, Cassie C. Jansen3,4 and Hamish McCallum1

Abstract: Understanding the non-human reservoirs of zoonotic pathogens is critical for effective disease control, but identifying the relative contributions of the various reservoirs of multi-host pathogens is challenging. For Ross River virus (RRV), knowledge of the transmission dynamics, in particular the role of non-human species, is important. In Australia, RRV accounts for the highest number of human mosquito-borne virus infections. The long held dogma that marsupials are better reservoirs than placental mammals, which are better reservoirs than birds, deserves critical review. We present a review of 50 years of evidence on non-human reservoirs of RRV, which includes experimental infection studies, virus isolation studies and serosurveys. We find that whilst marsupials are competent reservoirs of RRV, there is potential for placental mammals and birds to contribute to transmission dynamics. However, the role of these animals as reservoirs of RRV remains unclear due to fragmented evidence and sampling bias. Future investigations of RRV reservoirs should focus on quantifying complex transmission dynamics across environments. Keywords: Amplifier, Experimental infection, Serology, Virus isolation, Host, Vector-borne disease, Arbovirus

Background transmission dynamics among arboviruses has resulted in Vertebrate reservoir hosts multiple definitions for the key term “reservoir” [9]. Given Globally, most pathogens of medical and veterinary im- the diversity of virus-vector-vertebrate host interactions, portance can infect multiple host species [1]. Indeed, an there is unlikely to be a single definition suitable for all estimated 60–75% of emerging infectious diseases are systems or for all applications [9]. Here, we are concerned multi-host zoonoses [2]. Zoonotic arboviruses, such as with identifying which vertebrate hosts contribute most to Rift Valley fever virus, West Nile virus, and Japanese en- the pathogen pressure on humans (via infected vectors). cephalitis virus, have complex transmission cycles that in- We therefore adopt the notion that an arbovirus “reser- clude multiple host and vector species in maintenance voir” is a vertebrate host species which, if present in suffi- and spillover [2–4]. Identifying optimal approaches to cient abundance, will contribute to the pathogen pressure mitigate spillover of multi-host pathogens requires an un- on humans. This will require that it has frequent contact derstanding of how the transmission cycles of zoonotic vi- with vector populations, is attractive to a vector as a blood ruses and non-human hosts contribute to spatiotemporal meal source, is susceptible to infection, and can produce changes in the patterns of human disease [1, 5, 6]. The sufficient viraemia to infect another vector [9–11]. Kuno challenge of understanding the complex population biol- and Chang [3] identified three commonly used criteria for ogy of multi-host pathogens comes not only from identify- classifying vertebrate reservoirs of arboviruses: (i) virus ing potential reservoir host species, but in disentangling isolation from suspected animals; (ii) relatively high anti- which species contribute most to transmission and patho- body prevalence in the animals captured in the field; and gen pressure, and whether any species are crucial to per- (iii) demonstration of viraemia (of high virus titre and dur- sistence within the reservoir community [7, 8]. ation) in the suspected animals typically obtained under The definition of a “reservoir” in infectious disease lab conditions. Methods commonly adopted to address epidemiology is not straightforward [7]. This is especially these criteria include virus isolation, serosurveys and ex- the case for arboviruses, where complex and novel perimental infection studies, respectively.

* Correspondence: [email protected] Ross River virus 1Environmental Futures Research Institute, Griffith University, Brisbane, Queensland 4111, Australia Ross River virus (RRV) is a zoonotic alphavirus and Full list of author information is available at the end of the article human infection is nationally notifiable in Australia. It

© The Author(s). 2018 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.

24 Stephenson et al. Parasites & Vectors (2018) 11:188 Page 2 of 13

is responsible for the greatest number of mosquito- that RRV circulates in countries in the Pacific, where borne infections in humans across every state and ter- marsupials are absent [29–31]. ritory of Australia [12]. On average there are 4800 This review aims to: (i) critically review the evidence cases of RRV notified per year Australia-wide, with the supporting the hypothesis ‘marsupials are better reser- majority from Queensland [12]. There are occasional voirs of RRV than placental mammals, which in turn are large outbreaks of RRV involving a significantly higher better reservoirs than birds’; (ii) characterise the limita- number of human cases. In 2015, there were 9800 notifi- tions of that evidence; and (iii) identify research gaps cations of RRV - almost double the national average, 6192 with regards to RRV transmission cycles. of which were reported from Queensland. In 2017, an outbreak occurred in Victoria, with some 1200 notifi- Methods cations reported in January and February alone, ex- We systematically identified original research papers on ceeding the state counts for the previous four years RRV reservoir as follows. First, we searched electronic da- combined [12]. tabases (Web of Science, ProQuest, Science Direct, Ross River Virus is not usually fatal [13]. However, pa- PubMed and Google Scholar) for articles published be- tients with disease caused by RRV infection present with tween 1950 and May 2016 using combinations (Additional symptoms that include polyarthritis, myalgia, and fever file 1: Table S1) of the following keywords: ‘Ross River and chronic joint pain which may last several weeks and, virus,’ ‘Ross River fever,’ ‘endemic polyarthritis,’ ‘host’, ‘reser- in some cases, months [14–16]. The economic costs of voir,’ ‘wild*’,‘captive,’ ‘population,’ ‘serolog*,’ ‘serosurvey*,’ ‘anti- illness were estimated to be A$1070 per case in 2002, bod*,’ ‘virus,’ ‘viral,’ ‘viraemia,’ ‘viremia,’ ‘PCR,’ ‘patholog*,’ averaging more than A$5 million each year [17]. This is ‘serum,’ ‘RNA’, ‘vector*’. The asterisk (*) operator was used likely a conservative estimate of RRV cost as it does not as a wildcard to search for all the possible variations of include broader implications of infection, such as the in- keywords. We then manually searched bibliographies for ability for an individual to work or care for children additional references. Review papers, studies involving [13]. Presently, there is no treatment or commercially only humans, and studies not reporting original data were available vaccine for RRV and the best means to reduce excluded. A flow chart showing the article selection the risk of infection is through mosquito management process is presented in Additional file 2: Figure S1. A list and avoidance of mosquito bites [14, 18]. of the publications included is provided in Additional file More than 40 species of mosquitoes have yielded iso- 1: Table S2. One person (EBS) was responsible for deter- lates of RRV (summarised in [19]), although many are mining if a paper was included and extracting data. By fol- likely to only have a minor role in transmission. Species lowing the inclusion and exclusion criteria there were no most commonly associated with transmission include discrepancies for selecting papers. saltmarsh mosquitoes Aedes camptorhychus, presenting For each article, we recorded the following information: in southern Australia and is replaced by Ae. vigilax north year of publication, location of study, type of study (experi- of its range, and the freshwater mosquito (Culex annu- mental infection, serosurvey, virus isolation/detection), lirostris) that is present throughout Australia, excluding method (e.g. for experimental infection studies: the dose, Tasmania [14]. infection technique, strain of RRV used and post infection A definitive description of the host-vector relationships analysis), species investigated, sample size and results. in the transmission cycle of RRV is currently not available. Species examined in each study were assigned to a spe- Non-human reservoirs of RRV are thought to play a cies group (marsupial, placental mammal, bird) for significant role in RRV endemicity [20–22]. While sev- interpretation. eral authors have suggested that human-mosquito-human transmission of RRV may occur during epidemics [23–25], Statistical analysis such transmission is not believed to be sufficient to ac- A meta-analysis of results across experimental infection count for the total number of reported cases each year in studies was not possible as methods of infection and Australia [19], nor to be responsible for the long-term per- viral detection were highly variable. Instead, we conducted sistence of RRV. two one-way analysis of variance (ANOVA) for the one Marsupials are generally considered better reservoirs experimental infection study that assessed the greatest of RRV than placental mammals, which in turn are bet- number of species (n = 10 species, [27]) to test the hy- ter reservoirs than birds [13, 19, 26, 27]. This hypothesis pothesis that the duration of viraemia and peak viral titre first appeared in the literature in 1971 following epi- differs between species groups (marsupial, placental mam- demiological studies in northern Queensland where high mal and bird). rates of RRV seropositivity were detected in macropods For serosurveys, the seroprevalence range was calcu- (kangaroos and wallabies) [28]. However, the hypothesis lated for each species group (marsupial, placental mammal deserves critical re-evaluation because there is evidence or bird), and plotted as a boxplot. An ANOVA was used

25 Stephenson et al. Parasites & Vectors (2018) 11:188 Page 3 of 13

to test for differences in seroprevalence between species Within each study there was substantial variability in the groups across different studies. viraemic response reported for different species of animal within each species group and study (i.e. marsupials, pla- Results cental mammals and birds; Table 3). Whitehead [37]re- We identified a total of 38 research papers that met our ported the highest peak titres in Antichinus spp. of 8 LD50 criteria. Of these studies, seven described experimental lasting 144 hours, in contrast to 4.75 LD50, lasting 48 infections, five described virus isolation and 29 utilised hours in rabbits (Oryctolagus cuniculus). Kay et al. [27]re- serosurveys (Table 1) (three studies used multiple methods, ported the highest viremia in horses (Equus caballus)at Additional file 1: Table S2). All experimental infection 6.3 SMIC, compared to black ducks (Anas rubripes) devel- studies were undertaken in Queensland. Virus isolation oping a peak titre of 1.8 SMIC. Pigeons (Columba livia studies were undertaken in Queensland and Victoria. We domestica), cats (Felis catus) and dogs (Canis lupus famil- identified a single article that performed molecular iaris) were the only animals that did not develop a detect- identification of virus from horses [32], but this was ex- able viraemia. cluded from the analysis as the paper was methodo- Statistical analysis of the results from Kay et al. [27], logical, describing the novel test method. Serosurveys were the experimental infection study with the greatest num- performed in every state in Australia and the Northern ber of species (n = 10) and largest number of individuals Territory, as well as other countries including Fiji (n =1), tested (n = 92), showed that although viraemia in grey New Guinea (n = 1) and New Zealand (n =2)(Table1). kangaroos (Macropus giganteus) attained moderately The earliest studies of RRV reservoirs included serosurveys high levels and lasted the longest duration (Fig. 1), there in 1966 and virus isolation in 1968. was no significant difference (between species groups (marsupial (n = 11 individuals, 2 species), placental Experimental infection studies mammals (n = 45 individuals, 5 species) and birds (n =35 The seven experimental infection studies included infec- individuals, 3 species) for both duration of viraemia tion of 18 vertebrate species with RRV (summarised in (F(2, 9) = 2.312, P = 0.169) and peak titre level (F(2, 9) = Table 2). At least two strains of RRV were used: the proto- 3.177, P = 0.104). type T48, isolated from a human Townsville in 1959 [13], For horses (Equus caballus) and little corellas (Caca- and B94/20, isolated from a human during an epidemic in tua sanguinea), Kay et al. [27] also used susceptible Cx Queensland in 1994 [33]. Two studies did not state which annulirostris vectors to feed on infected hosts to deter- strain was used [27, 34]. The most common route of in- mine the percentage of mosquitoes that became infected fection was via infected mosquito (n =5),although with RRV. Despite low titre and short duration viraemias subcutaneous (n = 2) and intravenous (n =1)routes (2.3 SMIC, 50 hours; Fig. 1), little corellas infected 14% were also used. All studies assessed the titre of viraemia of an unknown number of recipient vectors. Horses de- in blood, but methods and metrics differed. Four of the veloped the highest titre viraemeia (6.3SMIC), of one of seven studies subsequently exposed infected animals to the longest durations (112 hours), and infected a com- susceptible vectors to determine infectiousness of po- parable 11% of recipient vectors. tential reservoirs. Across all studies, the median sample size was 9 individuals, Kay et al. [27]usingthelargest Virus isolation studies sample size of 20 chickens (Gallus gallus domesticus). Isolation of RRV from non-human vertebrate species has More than half (4 of 7) of the experimental infection been reported in 20 instances in published studies. The studies undertaken for RRV simultaneously co-infected majority (n = 15) of isolates were recovered from horses the same animal with other viruses in addition to RRV, (Equus caballus), whilst two isolates were recovered including Barmah Forest virus, Murray Valley encephal- from agile wallabies (Macropus agilis) and three isolates itis or Sindbis virus [27, 35–37]. came from birds (Table 4). Virus was isolated from the Comparison of viral titres across experimental infection heart tissue of birds [38] and from serum of horses and studies is hampered by different measures of viraemia. wallabies [28, 39]. Virus isolation was achieved using

Table 1 Summary of study types included in the literature review of Ross River virus reservoir studies comprising the number of studies of each type, location and dates of publications Study type Total no. of studies Location of studies Date range of studies Experimental infection 7 Queensland 1969–2001 Virus isolation 5 Queensland, Victoria 1968–2003 Serosurvey 29 Queensland, New South Wales, Western Australia, Victoria, South Australia, 1966–2015 Northern Territory, Tasmania, New Guinea, Fiji, New Zealand

26 Stephenson ta.Prsts&Vectors & Parasites al. et

Table 2 Summary of Ross River virus experimental infection studies included in the review with description of methods and vertebrate species used (2018)11:188 Reference Infection method Virus detection methods Vertebrate species tested (sample size) Virus co-infection RRV strain Infection route Vector used Boyd et al. [35] B94/20 Infected vector bite Yes Infection of a vector; magnitude and Brushtail possum Trichosurus vulpecula (10) Simultaneous infection with duration of viraemia (SMIC) Barmah Forest virus Boyd & Kay [58] B94/20 Infected vector bite Yes Infection of a vector; magnitude and Cat Felis catus (10); dog Canis lupus familiaris (10) duration of viraemia (CCID50) Ryan et al. [33] B94/20 Infected vector bite Yes Infection of a vector; magnitude and Grey-headed flying fox Pteropus poliocephalus (10) duration of viraemia (TCID) Kay et al. [34] Not stated Intravenous injection and Yes Infection of a vector; magnitude and Horse Equus ferus (11) Simultaneous infection with

27 infected vector bite duration of viraemia (SMIC) Murray Valley encephalitis Kay et al. [27] Not stated Infected vector bite Yes Infection of a vector; magnitude and Agile wallaby Macropus agilis (9); grey kangaroo Simultaneous infection with duration of viraemia (SMIC) Macropus giganteus (3); rabbit Oryctolagus cuniculus Murray Valley encephalitis (9); sheep Ovis aries (8); pig Sus scrofa (11); horse Equus ferus (11); cattle Bos taurus (6); chicken Gallus gallus domesticus (20); black duck Anas rubripes (3); little corella Cacatua sanguinea (12) Spradbrow [55] T48 Subcutaneous infection No Magnitude and duration of viraemia Sheep Ovis aries (3); pig Sus scrofa domesticus (3) (LD50) Whitehead [38] T48 Subcutaneous infection No Magnitude and duration of viraemia Rabbit Oryctolagus cuniculus (4); rat Rattus spp. (4); Simultaneous infection (LD50) bandicoot Isoodon macrourus (4); marsupial mouse with Sindbis virus Antechinus spp. (6); chicken Gallus gallus domesticus (16); pigeon Columba livia domestica (5)

Abbreviations: SMIC, suckling mouse intracerebral injection; CCID50, cell culture infectious dose; TCID, tissue culture infectious dose; LD50, lethal dose per gm of whole blood ae4o 13 of 4 Page Stephenson ta.Prsts&Vectors & Parasites al. et Table 3 Summary of species, sample size, viraemia and antibody response to experimental infection of vertebrate species with Ross River virus Species Reference Sample size Proportion with Peak titre level Viraemia duration (h) Infected recipient vectors Proportion with antibody viraemic response response Marsupial Brushtail possum Trichosurus vulpecula Boyd & Kay [37] 10 0.33 7.5 CCID 48 Yes 0.8 Agile wallaby Macropus agilis Kay et al. [27] 9 0.78 5.6 SMIC 81.6 Not reported 1 Grey kangaroo Macropus giganteus Kay et al. [27] 3 1 4.6 SMIC 144 Not reported 1 (2018)11:188

Bandicoot Isoodon macrourus Whitehead [38] 4 Not reported 7.2 LD50 144 Not reported 1

Marsupial mouse Antechinus spp. Whitehead [38] 6 Not reported 8 LD50 144 Not reported Not reported Placental mammal Horse Equus caballus Kay et al. [27, 34] 11 0.1 6.3 SMIC 96 Yes 0.6 Sheep Ovis aries Kay et al. [27] 8 1 3.8 SMIC 57.6 Not reported 1 Spradbrow [55] 14 0.64 Not reported 120 Not reported 1 Pig Sus scrofa domesticus Kay et al. [27] 11 0.91 3.0 SMIC 81.6 Not reported 0.45 28 Spradbrow [55] 3 1 Not reported 48 Not reported 1 Cow Bos taurus Kay et al. [27] 6 0.16 2.3 SMIC 48 Not reported 0.16 Cat Felis catus Boyd & Kay [58] 10 0 0 0 No 0.1 Dog Canis lupus familiaris Boyd & Kay [58] 10 0 0 0 No 0.1 Grey-headed flying fox Pteropus poliocephalus Ryan et al. [33] 10 0.25 2.2 TCID Not reported Yes 0.33

Rat Rattus spp. Whitehead [38] 4 Not reported 7.4 LD50 72 Not reported 1 Rabbit Oryctolagus cuniculus Kay et al. [27] 9 0.67 3.1 SMIC 55.2 Not reported Not reported

Whitehead [38] 4 Not reported 4.7 LD50 48 Not reported 1 Bird Chicken Gallus gallus domesticus Kay et al. [27] 20 0.95 2.8 SMIC 69.6 Not reported 0.55

Whitehead [38] 16 Not reported 5.0 LD50 120 Not reported 0.75 Black duck Anas rubripes Kay et al. [27] 3 0.67 1.8 SMIC 96 Not reported 1 Little corella Cacatua sanguinea Kay et al. [27] 12 0.5 2.3 SMIC 50.4 Yes 0 Pigeon Columba livia domestica Whitehead [38] 5 0 0 0 Not reported 0.6

Fig. 1 Mean peak titre and duration of viraemia measured in different animals experimentally infected with Ross River virus, data extracted from Kay et al. [27]. Squares represent marsupials, circles represent mammals and triangles represent birds. Species in order of number: 1, Cow; 2, Little corella; 3, Rabbit; 4, Sheep; 5, Chicken; 6, Pig; 7, Black duck; 8, Agile wallaby; 9, Horse; 10, Grey kangaroo intracerebral inoculation into infant mice [28, 38] or in- specific [41] and have been used throughout the decades. cubation of tissue culture plates with serum followed by These methods are generally considered more labour identification with antiserum raised in rabbits [40]. intensive, require trained personnel and a minimum of five days to perform. More recent serosurveys have Serosurvey studies used enzyme linked immunosorbent assay (ELISA) We identified a total of 30 serosurveys studies (Additional which can be purchased in commercial kits and are file 1: Table S2) that tested more than 17,000 serum sam- more commonly used in human diagnostic labs (Fig. 2a). ples from 77 host species. The majority of these studies In the first decade (1966–1975) of seroprevalence stud- were undertaken in Australia, with a small number in ies, 80% of all species sampled were birds (Fig. 2b). In New Zealand, Fiji and Papua New Guinea (Table 1). Sero- the following decade (1976–1985) more than 82% of surveys for RRV in non-human species have spanned al- serosurveys were performed on placental mammal spe- most 50 years and, as such, the methods within these cies. In the subsequent two decades (1986–2005) mar- studies vary substantially. Studies were grouped by decade supial species were sampled most frequently (between of publication to accommodate the different serological 50–60% of serosurveys), followed by placental mammals methods and species groups tested (Fig. 2). Earlier studies (between 32–40% of serosurveys) and birds (between favoured haemoglobin inhibition. This technique has now 0–17% of serosurveys). largely been superseded in favour of assays with better Figure 3 shows the mean seropositivity in each of the sensitivity and specificity. Virus neutralisation, either three species groups. Half of all species sampled were through Plaque Reduction Neutralisation (PRNT) (the marsupials (n = 39 species), followed by placental mam- gold standard) or serum microneutralisation are highly mals (n = 27 species) and birds (n = 13 species).

Table 4 The number, study and study sample size for isolates of Ross River virus collected from non-human vertebrates Species Reference Sample size Number of RRV isolations Marsupial Agile wallaby Macropus agilis Doherty et al. [28]17 2 Mammal Horse Equus callabus Azuolas et al. [40] 750 13 Pascoe et al. [39]8 1 Campbell et al. [76] Not reported 1 Bird Magpie lark Grallina cyanoleuca Whitehead et al. [38] 775 (104 species) 1 Flycatcher Myiagra rubecula 1 Masked finch Poephila personata 1

29 Stephenson et al. Parasites & Vectors (2018) 11:188 Page 7 of 13

Fig. 2 a The number and types of method used to test Ross River virus seroprevalence by decade. b The percentage of different vertebrate groups sampled in each decade for Ross River virus seroprevalence. The numbers in each bar represent the number of species tested for each species group Fig. 3 Boxplot of serosurvey results for each vertebrate group with the number of sera sampled in brackets. Minimum, median and maximum values are represented with the box and whiskers, and outliers are represented by circles Placental mammals comprised the largest number of sera (n = 10,126), more than double that of marsupial sera (n = 4304) and quadruple that of bird sera (n = (Oryctolagus cuniculus,6of10positive).Allofthefol- 2621). Within placental mammals, cattle have been sam- lowing marsupial species have tested positive to RRV: pled most frequently (28.9% of samples), closely followed the eastern barred bandicoot Perameles gunnii (n =2/ by horses (28.8% of samples). For marsupials, 46% of ser- 2), the eastern bettong Bettongia gaimardi (n =1/1), osurveys were from one study with a focus on western the long-nosed potoroo Potorous tridactylus (n =2/2), grey kangaroos, Macropus fuliginosus [42]. Within birds, the northern nail-tail wallaby Onychogalea unguifera (n chickens were the most sampled species, contributing = 1/1), the Tasmanian devil Sarcophilus harrisii (n =4/ 38% of all samples in this species group. 4) and the tiger quoll Dasyurus maculatus (n =1/1). Overall, there was a significant difference in seroprevalence between species groups (F(2, 76) = 7.091, P = 0.001). Across studies, the median seroprevalence in Discussion marsupials was greater when compared with placental Identifying reservoirs of multi-host viruses is challenging mammals and birds (44%, 16% and 0%, respectively; Fig. due to the complex interactions that sustain and promote 3). The interquartile range of seroprevalence was great- pathogens and spillover events. For arboviruses, three est in the marsupial group (5–75%) and smallest in the commonly used criteria for classifying vertebrate reser- bird group (0–6%) (Fig. 3). Outliers in the bird sero- voirs include: viraemia, virus isolation and relatively prevalence results included black ducks (Anas superci- high antibody prevalence [3]. In light of these criteria, liosa, 2 of 3 positive) and little corellas (Cacatua this study aimed to review the evidence for non-human sanguinea, 6 of 12 positive). For placental mammals, the reservoirs of RRV against the hypothesis: marsupials are highest seroprevalence was observed in red foxes better reservoirs of RRV than placental mammals, which (Vulpes vulpes, 3 of 4 positive) followed by rabbits are better reservoirs than birds.

30 Stephenson et al. Parasites & Vectors (2018) 11:188 Page 8 of 13

The role of marsupials as reservoirs of RRV from horses and passerine birds and the majority of Results from experimental infection, virus isolation and RRV cases in humans do not overlap with macropod serosurvey studies on 31 marsupial species support the home ranges [13]. hypothesis that marsupials are competent reservoirs and Across all studies, marsupials had the highest RRV sero- likely contribute significantly to RRV transmission. How- prevalence (44.3%), compared with placental mammals ever, the evidence is fragmentary and subject to sampling (22.7%) and birds (11.1%). Although informative, these bias, which limits our ability to extrapolate across species, data must be interpreted with caution because it is evident broad geographical areas, habitat and land use types. that marsupials were more likely to be targeted during Across experimental infection studies, marsupials gen- sampling efforts in the decade 1986–1995 (Fig. 2b). This erally developed high and long-lasting viraemia. This has shift in targeted species group followed the results of ex- previously been interpreted as evidence that marsupials perimental infection studies demonstrating marsupials as are better reservoirs than other species groups, yet we competent amplifiers of RRV in 1986. Further, without in- found no significant difference between the mean duration formation on the age of individuals, seroprevalence data of viraemia or peak viraemia of marsupials, placental should be compared between studies with caution. mammals or birds. At least two factors must be consid- Sampling biases are likely to have arisen from the fre- ered when interpreting results of experimental infection quent use of convenience sampling or ‘active surveillance’ studies. First, experimental infection studies are often con- methods (where investigator-driven data collection is de- strained by small sample sizes both in the number of spe- signed to meet specific information needs [45]). The focus cies and the number of studies that can be compared. on marsupials as hosts for RRV to the exclusion of other Although we statistically analysed results from the RRV host species is premature, and is unlikely to be uniform experimental infection study with the greatest number across all marsupial species. For example, brushtail pos- and diversity of species [27], sample sizes were still limited sums were hypothesised to be the urban reservoir of RRV, and likely influenced the statistical power of the results. In being both marsupials and living in close proximity to particular, the diversity of methods used limits compari- humans [35], resulting in a focus on this species. However, sons and the effect of simultaneous co-infection with targeted surveillance of this species between February and other viruses cannot be discounted. Secondly, while vir- December 2005 in Sydney failed to identify any seroposi- aemia plays an important role in the maintenance and tive individuals [35, 46] of the 10 possums sampled. This transmission of arboviruses, using this measure alone to number of animals is insufficient to draw strong conclu- identify potential reservoirs has limited value. For example, sions about the host status, and further studies are re- in experimental infection studies of West Nile virus, an- quired [46]. Furthermore, it is interesting to note that other zoonotic arbovirus, viraemia alone did not defini- whilst brushtail possums are an abundant urban marsupial, tively identify vertebrate reservoirs: blue jays (Cyanocitta ringtail possums are more commonly reported in major cristata), house finches (Haemorhous mexicanus)and metropolitan areas including Brisbane, Sydney, Perth, house sparrows (Passer domesticus)wereidentifiedasthe Adelaide and Hobart [47] but only two studies (testing four most competent reservoirs on the basis of viraemia profile individuals in total) have been undertaken serological as- [43]. Yet subsequent field investigations identified that sessments of the species (50% seropositivity) [48, 49]. American robins (Turdus migratorius),alessviraemicand relatively uncommon avian species, were responsible for The role of placental mammals as reservoirs of RRV the majority of WNV vector infections due to host feeding Placental mammals comprise the greatest diversity of spe- preferences [44]. cies tested, including ungulates, carnivorous and small The isolation of virus from naturally infected hosts is urban species. While placental mammals meet the three interpreted as evidence that the species can infect vector criteria for arboviral reservoirs as a species group, there mosquitoes and, thus, infect humans. For marsupials, are significant differences among species. the isolation of RRV from two free-living agile wallabies (M. agilis) (from a total of 17 tested) demonstrates a Ungulates vector-host relationship under natural conditions and Ungulate species, including pigs, horses, sheep and cattle, suggests that this species is capable of infecting suscep- are recognised as reservoirs for other zoonotic arboviruses tible vectors, thereby supporting the argument for the [4]. Interestingly for RRV, horses are the only ungulate species as reservoirs of RRV. Together with Kay et al.’s likely to amplify the disease and act as reservoirs. High, [27] observation of viraemia in grey kangaroos, this has long-lasting viral titres, the ability to infect susceptible mos- led to the hypothesis that macropods are important RRV quitoes, frequent virus isolations and high seroprevalences reservoirs within their range. However, the relative im- suggest that horses could contribute significantly to on- portance of this group of species as a reservoir is not going RRV transmission, particularly during epidemic pe- clear, given that RRV has been isolated more frequently riods [50], although it is unclear whether they play a role in

31 Stephenson et al. Parasites & Vectors (2018) 11:188 Page 9 of 13

ongoing endemic circulation of RRV. A possible explan- Ae. funereus vectors in close proximity to a mixed-species ation for the high number of RRV isolates from horses is flying fox colony in Brisbane found that all of the 16 mos- that they are both a domestic species and one of the only quitoes analysed had fed on black flying foxes and none known species that develops clinical symptoms to RRV [51] on grey-headed flying foxes [33]. When considering the and therefore, are more likely to be sampled, particularly if possibility of flying foxes as reservoirs of RRV, it is import- they were infected and symptomatic. The horse population ant to consider the height at which different vectors feed in Australia may exceed 1.2 million individuals [52], and and move. Known vectors of RRV, including Ae. vigilax whilst they are rare in highly urbanised environments, they and Ae. camptorhyncus are likely to feed close to the are abundant in peri-urban areas, where some of the high- ground, potentially avoiding roosting flying foxes [61]. est prevalence of human RRV infection exists [53]. Further blood meal analysis studies are needed to de- In contrast, RRV has not been isolated from cattle, termine this. pigs and sheep [54–56], and these species have demon- Given their small body size, rats, rodents and flying strated low viraemic responses in experimental infection foxes may be considered less desirable as blood-meals studies [27] and in serosurveys [55]. Large numbers of for vectors [62]. However, they may exist in high dens- cattle sera are tested for antibodies to RRV due to the ities close to human populations. A blood-meal analysis use of cattle as sentinel species in the National Arbovirus of RRV vectors found rabbits and rats comprised up to Monitoring Program, which is designed to detect incur- 33% of Cx annulostris blood-meals in urban areas [63]. sions of exotic arboviral infection, such as bluetongue Serological data supporting the hypothesis that small viruses [57]. mammals may be playing a role in the transmission of RRV is currently lacking due to limited numbers tested. Cats and dogs Cats and dogs are the only carnivores that have been The role of birds as reservoirs of RRV assessed as potential reservoir hosts of RRV. Viraemias Birds are the most common arboviral reservoir for zoo- were not detected following experimental infection and notic flaviviruses and alphaviruses globally [4]; however, only 10% of these cats and dogs developed neutralising their contribution as reservoirs of RRV has been largely antibodies to RRV [58]. Seroprevalence studies of do- overlooked. On the basis of experimental infection vir- mestic cats not experimentally infected, found they have aemia data alone, birds appear to be poor amplifiers of a relatively low antibody prevalence (12.1%). The poor RRV. Four species of birds (chickens, pigeons, little cor- amplifier capacity and low seroprevalence suggest these ellas and black ducks) have been experimentally infected domestic species are unlikely to be significant reservoirs with RRV. Across experimental infection studies, birds of RRV. had the lowest peak titre and the shortest duration of viraemia in comparison to marsupials and placental Small mammals (< 2 kg) mammals (Table 3). Furthermore, pigeons were one of The potential role of small placental mammals, such as the only species that did not develop a detectable vir- rodents, rabbits and flying foxes, as reservoir hosts of aemia. However, little corellas (Cactua sanguinea) were RRV is ambiguous. Under experimental infection condi- capable of infecting 14% of susceptible Cx annuilostris tions Whitehead [37] found rodents were capable of de- mosquitoes, despite having a low and short viraemia. veloping viraemia higher than bandicoots, a marsupial, yet This is important because in the same study, horses the viraemia was short lived compared to marsupials. Rab- developed the highest titre but only infected 11% of bits developed mid-range titre peaks of short duration. susceptible vectors. Possible reasons for this were not In experimental infection, grey-headed flying foxes discussed in the original paper, but we suggest the cap- (Pteropus poliocephalus) did not develop a detectable ability of a vertebrate species to infect susceptible mos- viraemia, but were capable of infecting 3% of recipient quito vectors with RRV may be a more relevant Ae. vigilax vectors [33]. Flying foxes are a unique species measure of reservoir capacity than viraemia. group because they have been shown to be the reservoir The isolation of RRV from birds further supports their host for several zoonotic pathogens including, henipa- capacity as amplifiers. More than 750 virus isolation at- viruses lyssaviruses and filoviruses, often without detectable tempts, across 104 species, yielded the first 3 isolates of viraemia [6, 59]. Similar observations have been made for RRV from the heart muscle of passerine birds in North- arboviruses. In an experimental infection of black flying ern Queensland: a magpie lark, a flycatcher and a foxes (Pt. alecto) with Japanese encephalitis virus, all 15 in- masked finch (Table 4). Passerine birds are recognised as dividuals had a low viraemic response; however, two were important amplifiers of other arboviruses including flavi- capable of infecting susceptible mosquitoes [60]. Only the viruses such as West Nile virus [43], tick-borne patho- grey-headed flying fox has been investigated as a potential gens such as Borrelia burgdorferi - the causative agent reservoir host of RRV, yet a blood meal analysis of 20 for Lyme disease [64] and an arthritic alphavirus closely

32 Stephenson et al. Parasites & Vectors (2018) 11:188 Page 10 of 13

related to RRV, Sindbis [65]. Indeed, the isolation of the kangaroos and brushtail possums. Overall these studies alphavirus Sindbis from passerine birds, in combination found that one host alone was insufficient to maintain virus with genetic studies and antibody prevalence investiga- in vector populations. Glass [72] concluded that although tions has implicated birds as the reservoir host of Sind- marsupials such as kangaroos and wallabies are generally bis [62]. assumed to be the most important reservoir hosts, the virus Serological surveys have found low seroprevalence of survived longed under all models when the marsupial host RRV in birds. However almost 40% of bird sera tested was replaced with one with a shorter infectious period and has been from sentinel chickens. Chickens are consid- higher birth rate. Further, the same study reported that very ered appropriate sentinels for flaviviruses such as Mur- large host populations (> 100,000 individuals) were re- ray valley encephalitis because they display a strong quired for the virus to survive for four years. Choi et. al. antibody response [66], however experimental infections [70] similarly found that a kangaroo reservoir did not im- suggest this is not the case for RRV [27, 37]. Notably, pact the number of human infections due to a small popu- birds with positive serology for RRV were free-living na- lation size in the region. Carver et al. [20]reporteda tive species: a Tawny frogmouth owl in NSW [48] and significant negative relationship between the abundance of an Australasian gannet (Morus serrator) sampled in New a marsupial reservoir and RRV transmission. These find- Zealand [67]. Thus, the tendency towards sampling ings are in contrast to the putative reservoir hypothesis. chickens in RRV serosurveys may underestimate the Tompkins & Slaney [73] noted that different species may rates for birds as a whole, and future serosurveys would be reservoirs in different environments, such as high- benefit from inclusion of greater bird species diversity. density urban areas and protected environmental habitats, which can result in different transmission cycles. These Alternative evidence for non-human reservoirs modelling studies highlight the importance of investigat- This review has focused on the intrinsic host variables ing alternative species as potential reservoirs of RRV. important to reservoir capacity. There are other lines of Ideally, the system should be explicitly modelled as a evidence that can be important for investigating poten- multihost system, but obtaining the necessary data to tial reservoirs such as blood meal analysis and modelling parameterise such models is challenging [74]. studies. Determining vector preferences, may indicate a higher feeding frequency, and thus if a capable reservoir, Ross River virus: a multi-host pathogen higher transmission rate. Blood meal analysis studies in- Despite the evidence supporting marsupials as reservoirs vestigate the relationship between the vector and the of RRV, questions remain. Recent studies have found a host. Vector-host choice is a complicated matter, with high seroprevalence of RRV in the Pacific Islands in the factors such as host body size, carbon dioxide emission, absence of marsupial populations [29, 31], suggesting olfaction, availability, abundance and vector genetics that marsupials are not the only species group capable impacting feeding preferences [62, 68]. Blood meal stud- of increasing the community infection for RRV. Studies ies are further complicated as they are easily confounded modelling RRV reservoirs have suggested that the patho- by the environment in which study took place, and as gen has a multi-host system [23, 72]. However, none of such these studies are best when accompanied with ani- the studies reviewed in this paper specifically examined mal abundance and diversity measures. Of the 12 blood this hypothesis. Multi-host systems are not uncommon meal analysis papers in Australia, only one [69] has done for arboviruses but quantifying these systems is challen- this by asking the human residents to estimate numbers ging, requiring coordinated data collection over tem- of animals. Further research is needed to investigate poral and geographical scales for multiple species [3]. vector-host preferences in Australian urban, peri-urban To understand RRV as a multi-host pathogen, two issues and rural environments and determine the influence this must be considered. First, as RRV has an international distri- may have on a species capacity to act as a reservoir. bution spanning different environmental and social bounds Mathematical models are valuable way of describing it is important to define the ecological transmission of RRV and understanding complex disease systems such as across different ecosystems. Expansion of Claflin & Webb’s RRV. Models can test assumptions of a disease system [14] categorisation of potential RRV vectors, habitats and and generate predictions which can be used for manage- reservoirs across inland, metropolitan and coastal regions to ment decisions. For RRV, five studies [20, 70–73] have include transmission cycles is warranted. Secondly, to better utilised mechanistic modelling techniques (e.g. Susceptible- understand the reservoir capacity between different host Infectious-Recovered models) to better understand the communities, identification of amplifying or diluting reser- transmission dynamics underpinning the maintenance of voir hosts is required. Given that humans are not considered the pathogen. Although the models differ in parameters, lo- significant maintenance reservoirs of RRV outside of epi- cation and methods, all include a marsupial reservoir. Spe- demic periods in Australia, this may provide a benchmark cies that have been modelled as reservoirs are western grey for relative comparison of seroprevalence. For example,

33 Stephenson et al. Parasites & Vectors (2018) 11:188 Page 11 of 13

RRV seroprevalence in blood donors shows that the human Acknowledgments IgG seroprevalence ranges from 8.38% in Australia in 2011 We would like to acknowledge Wildlife Health Australia for their assistance with gathering sources for this review. [75] to 34.4% in French Polynesia between 2011 and 2013 [29]. Whilst this only gives an indication of the Funding number of people exposed and does not consider other ES was supported by Australian Government Research Training Program Scholarship. AJP was supported by a Queensland Government Accelerate contributing factors such as duration of antibody re- Postdoctoral Research Fellowship. sponse, these data may be compared with animal serosurvey data to identify species with higher infection Availability of data and materials rates than humans. Vector-host preference may be key The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request. to understanding reservoir and transmission dynamics in other zoonotic arboviruses [3]. Overall, consideration of Authors' contributions RRV as a multi-host pathogen may disentangle the com- ES collected, analysed, interpreted papers and was a major contributor in the writing of the manuscript. AP, SR, CC and HM contributed to the interpretation plex ecological dynamics that may be taking place. of results and writing of the manuscript. All authors read and approved the final manuscript.

Conclusion Ethics approval and consent to participate This study set out to: (i) critically review the evidence for the Not applicable. hypothesis that marsupials are better reservoirs of RRV than Consent for publication mammals, which in turn are better than birds; (ii) identify Not applicable. limitations of this evidence; and (iii) identify research gaps allowing for better assessments of RRV reservoirs in the fu- Competing interests ture. The evidence reviewed in this paper is limited by a sam- The authors declare that they have no competing interests. pling bias in favour of particular species and species groups, ’ cross-sectional serosurveys and a diversity of methods Publisher sNote Springer Nature remains neutral with regard to jurisdictional claims in employed, which reduces the statistical strength for meta- published maps and institutional affiliations. analysis. Notwithstanding these limitations, this review highlights that evidence to support the stated hypothesis, that Author details 1Environmental Futures Research Institute, Griffith University, Brisbane, marsupials are better reservoirs than placental mammals Queensland 4111, Australia. 2The University of Queensland, School of Public which in turn are better reservoirs than birds, is variable. Un- Health, Herston, Brisbane, Queensland 4006, Australia. 3Metro North Public derstanding the non-human reservoirs of RRV has broader Health Unit, Metro North Hospital and Health Service, Windsor, Brisbane, Queensland 4030, Australia. 4Communicable Diseases Branch, Department of applications to other zoonotic arboviruses and, importantly, Health, Queensland Government, Herston, Brisbane, Queensland 4006, can contribute to the management of current and emerging Australia. arboviruses through mitigating infection between host and Received: 18 October 2017 Accepted: 20 February 2018 vector populations. Future research on the non-human reservoirs of RRV should focus on investigating non-marsupial species, including passerine birds and small placental mam- References 1. KareshWB,DobsonA,Lloyd-SmithJO,LubrothJ,DixonMA,BennettM, mals. Ideally this would be done through ecological assess- et al. Ecology of zoonoses: natural and unnatural histories. Lancet. mentsofvector,virusandhostabundanceinareasofhigh 2012;380(9857):1936–45. and low disease in humans. For Australia, reducing the 2. Webster JP, Borlase A, Rudge JW. Who acquires infection from whom and how? Disentangling multi-host and multi-mode transmission dynamics in burden of RRV, the most common arbovirus, would the ‘elimination’era. Phil Trans R Soc B. 2017;372(1719):20160019. have substantial economic and social benefits. 3. Kuno G, Chang GJ. Biological transmission of arboviruses: reexamination of and new insights into components, mechanisms, and unique traits as well as their evolutionary trends. Clin Microbiol Rev. 2005;18(4):608–37. Additional files 4. Weaver SC, Barrett AD. Transmission cycles, host range, evolution and emergence of arboviral disease. Nat Rev Microbiol. 2004;2(10):789–801. 5. Plowright RK, Parrish CR, McCallum H, Hudson PJ, Ko AI, Graham AL, et al. Additional file 1: Table S1. Combinations of search terms used to Pathways to zoonotic spillover. Nat Rev Microbiol. 2017;15:502–10. collect papers for review. Table S2. Detailed summary of included Ross 6. Plowright RK, Peel AJ, Streicker DG, Gilbert AT, McCallum H, Wood J, et al. River virus reservoir studies, including the reference, location, study type Transmission or within-host dynamics driving pulses of zoonotic viruses in and species group assessed in each study. (DOCX 22 kb) reservoir-host populations. PLoS Negl Trop Dis. 2016;10(8):e0004796. Additional file 2: Figure S1. Flowchart outlining the process followed 7. Haydon DT, Cleaveland S, Taylor LH, Laurenson MK. Identifying reservoirs and actions taken to compile the systematic literature review. The box in of infection: a conceptual and practical challenge. Emerg Infect Dis. yellow highlights the total number of studies used in this review. The 2002;8(12):1468–73. total number n is the number of original research papers. (PNG 29 kb) 8. Viana M, Mancy R, Biek R, Cleaveland S, Cross PC, Lloyd-Smith JO, et al. Assembling evidence for identifying reservoirs of infection. Trends Ecol Evol. 2014;29(5):270–9. Abbreviations 9. Kuno G, Mackenzie JS, Junglen S, Hubálek Z, Plyusnin A, Gubler DJ. ANOVA: Analysis of Variance; ELISA: Enzyme linked immunosorbent assay; Vertebrate reservoirs of arboviruses: Myth, synonym of amplifier, or reality? PRNT: Plaque reduction neutralisation; RRV: Ross River virus Viruses. 2017;9(7):185.

34 Stephenson et al. Parasites & Vectors (2018) 11:188 Page 12 of 13

10. World Health Organization. Arthropod-borne and rodent-borne viral 34. Kay BH, Pollitt CC, Fanning ID, Hall RA. The experimental infection of diseases: report of a WHO scientific group meeting held in Geneva from 28 horses with Murray Valley encephalitis and Ross River viruses. Aust Vet J. February to 4 March 1983. 1985. http://apps.who.int/iris/handle/10665/39922. 1987;64(2):52–5. Accessed 31 Aug 2017. 35. Boyd AM, Hall RA, Gemmell RT, Kay BH. Experimental infection of Australian 11. Rodhain F. The idea of natural reservoir in arbovirology. Bull Soci Pathol brushtail possums, Trichosurus vulpecula (Phalangeridae : Marsupialia), with Exot. 1997;91(4):279–82. Ross River and Barmah Forest viruses by use of a natural mosquito vector 12. National Notifiable Diseases Surveillance System. Australian Government, system. Am J Trop Med Hyg. 2001;65(6):777–82. Department of Health. 2017. http://www9.health.gov.au/cda/source/cda- 36. Kay BH, Standfast HA. Ecology of arboviruses and their vectors in Australia. index.cfm. Accessed 12 Sep 2017. Curr Top Vect Res. 1987;3:1–36. 13. Harley D, Sleigh A, Ritchie S. Ross River virus transmission, infection, and 37. Whitehead R. Experimental infection of vertebrates with Ross River and disease: a cross-disciplinary review. Clin Microbio Rev. 2001;14(4):909–32. Sindbis viruses 2 Group A arboviruses isolated in Australia. Aust J Exp Biol 14. Claflin SB, Webb CE. Ross River virus: many vectors and unusual hosts make Med Sci. 1969;47:11. for an unpredictable pathogen. PLoS Pathogens. 2015;11(9):e1005070. 38. Whitehead R, Doherty R, Domrow R, Standfast H, Wetters E. Studies of the 15. Condon RJ, Rouse IL. Acute symptoms and sequelae of Ross River virus epidemiology of arthropod-borne virus infections at Mitchell River Mission, infection in South Western Australia. A follow up study. Clin Diag Virol. Cape York Peninsula, North Queensland: III. Virus studies of wild birds, 1964– 1995;3(3):273–84. 1967. Trans R Soc Trop Med Hyg. 1968;62(3):439–45. 16. Schleenvoigt BT, Baier M, Hagel S, Forstner C, Koetsche R, Pletz MW. Ross 39. Pascoe R, George T, Cybinski D. The isolation of a Ross River virus from a river virus infection in a Thuringian traveller returning from south-east horse. Aust Vet J. 1978;54(12):600. Australia. Infection. 2015;43(2):229–30. 40. Azuolas JK, Wishart E, Bibby S, Ainsworth C. Isolation of Ross River virus from 17. Ratnayake JTB. The valuation of social and economic costs of mosquito- mosquitoes and from horses with signs of musculo-skeletal disease. Aust transmitted Ross River virus: PhD Thesis. Brisbane: Griffith University; 2006. Vet J. 2003;81(6):344–7. 18. Tomerini DM, Dale PE, Sipe N. Does mosquito control have an effect on 41. Hobson-Peters J. Approaches for the development of rapid serological assays for mosquito-borne disease? The case of Ross River virus disease and mosquito surveillance and diagnosis of infections caused by zoonotic flaviviruses of the management in Queensland, Australia. J Am Mosq Cont Assoc. 2011;27(1):39–44. Japanese encephalitis virus serocomplex. BioMed Res Int. 2012;2012:379738. 19. Russell RC. Ross River virus: Ecology and distribution. Annu Rev Entomol. 42. Potter A, Johansen CA, Fenwick S, Reid SA, Lindsay M. The seroprevalence 2002;47:1–31. and factors associated with Ross River virus infection in western grey 20. Carver S, Bestall A, Jardine A, Ostfeld RS. Influence of hosts on the ecology kangaroos Macropus fuliginosus in Western Australia. Vector Borne Zoonotic of arboviral transmission: potential mechanisms influencing dengue, Murray Dis. 2014;14(10):740–5. Valley encephalitis, and Ross River virus in Australia. Vector Borne Zoonotic 43. Komar N, Langevin S, Hinten S, Nemeth N, Edwards E, Hettler D, et al. Dis. 2009;9(1):51–64. Experimental infection of North American birds with the New York 1999 21. Jacups SP, Whelan PI, Currie BJ. Ross River virus and Barmah Forest virus strain of West Nile virus. Emerg Infect Dis. 2003;9(3):311. infections: A review of history, ecology, and predictive models, with implications 44. Kilpatrick AM, Daszak P, Jones MJ, Marra PP, Kramer LD. Host heterogeneity for tropical northern Australia. Vector Borne Zoonotic Dis. 2008;8(2):283–97. dominates West Nile virus transmission. Proc R Soc Lond B Biol Sci. 22. Ng V, Dear K, Harley D, McMichael A. Analysis and prediction of Ross River 2006;273(1599):2327–33. virus transmission in New South Wales, Australia. Vector Borne Zoonotic Dis. 45. Stallknecht D. Impediments to wildlife disease surveillance, research, and 2014;14(6):422–38. diagnostics. In: Childs JE, Mackenzie JS, Richt JA, editors. Wildlife and 23. Koolhof I, Carver S. Epidemic host community contribution to mosquito-borne emerging zoonotic diseases: The biology, circumstances and consequences disease transmission: Ross River virus. Epidemiol Infect. 2017;145(4):656–66. of cross-species transmission. Berlin: Springer; 2007. p. 445–61. 24. Marshall I, Miles J. Ross River virus and epidemic polyarthritis. Curr Top 46. Hill NJ, Power ML, Deane EM. Absence of Ross River virus amongst Vector Res. 1984;2:31–56. Common brushtail possums Trichosurus vulpecula from metropolitan 25. Rosen L, Gubler DJ, Bennett PH. Epidemic polyarthritis (Ross River) virus - Sydney, Australia. Eur J Wildl Res. 2009;55(3):313–6. infection in the Cook Islands. Am J Trop Med Hyg. 1981;30(6):1294–302. 47. Atlas of Living Australia. 2017. http://www.ala.org.au. Accessed 15 Dec 2017. 26. Jacups SP, Whelan PI, Markey PG, Cleland SJ, Williamson GJ, Currie BJ. 48. Gard G, Marshall ID, Woodroof GM. Annually recurrent epidemic polyarthritis Predictive indicators for Ross River virus infection in the Darwin area of and Ross River virus activity in a coastal area of New South Wales. II. tropical northern Australia, using long-term mosquito trapping data. Trop Mosquitos, viruses and wildlife. Am J Trop Med Hyg. 1973;22(4):551–60. Med Int Health. 2008;13(7):943–52. 49. McManus TJ, Marshall ID. The epidemiology of Ross River virus in Tasmania. 27. Kay BH, Hall RA, Fanning ID, Mottram P, Young PL, Pollitt CC. Experimental Arbov Res Aust. 1986:127–31. infection of vertebrates with Murray Valley encephalitis and Ross River 50. Roche SE, Wicks R, Garner MG, East IJ, Paskin R, Moloney BJ, et al. viruses. In: Arboviral Research of Australia: Proceedings Fourth Symposium, Descriptive overview of the 2011 epidemic of arboviral disease in horses in May 6–9, 1986. Brisbane; 1986. p. 71–5. Australia. Aust Vet J. 2013;91(1):5–13. 28. Doherty RL, Standfast HA, Domrow R, Wetters EJ, Whitehead RH, Carley JG. 51. Azuolas JK. Ross River virus disease of horses. Aust Equin Vet. 1998;16(2):56–8. Epidemiology of arthropod-borne virus infections at Mitchell River Mission, 52. McGowan TW, Pinchbeck G, Phillips CJC, Perkins N, Hodgson DR, McGowan Cape York Peninsula, North Queensland. IV. Arbovirus infections of CM. A survey of aged horses in Queensland, Australia. Part 1: Management mosquitoes and mammals, 1967–1969. Trans R Soc Trop Med Hyg. 1971; and preventive health care. Aust Vet J. 2010;8811:420–7. 65(4):504. 53. Hu W, Tong S, Mengersen K, Oldenburg B. Exploratory spatial analysis of 29. Aubry M, Finke J, Teissier A, Roche C, Broult J, Paulous S, et al. Silent circulation social and environmental factors associated with the incidence of Ross River of Ross River virus in French Polynesia. Int J Infect Dis. 2015;37:19–24. virus in Brisbane. Australia. Am J Trop Med Hyg. 2007;76(5):814–9. 30. Aubry M, Finke J, Teissier A, Roche C, Broult J, Paulous S, et al. Seroprevalence 54. Gard G, Shorthose J, Weir R, Walsh S, Melville L. Arboviruses recovered from of arboviruses among blood donors in French Polynesia, 2011–2013. Int J sentinel livestock in northern Australia. Vet Microbio. 1988;18(2):109–18. Infect Dis. 2015;41:11–2. 55. Spradbrow PB. Experimenatl infection of sheep and pigs with Ross River 31. Lau C, Aubry M, Musso D, Teissier A, Paulous S, Desprès P, et al. New virus. Aust Vet J. 1973;49(8):403–4. evidence for endemic circulation of Ross River virus in the Pacific Islands 56. Vale TG, Spratt DM, Cloonan MJ. Serological evidence of arbovirus infection and the potential for emergence. Int J Infect Dis. 2017;57:73–6. in native and domesticated mammals on the south coast of New South 32. Studdert MJ, Azoulas JK, Vasey JR, Hall RA, Ficorilli N, Huang JA. Wales, Australia. Aust J Zoolo. 1991;39(1):1–8. Polymerase chain reaction tests for the identification of Ross River, 57. Gard G, Weir R, Walsh S. Arboviruses recovered from sentinel cattle using Kunjin and Murray Valley encephalitis virus infections in horses. Aust several virus isolation methods. Vet Microbio. 1988;18(2):119–25. Vet J. 2003;8(1):76–80. 58. Boyd AM, Kay BH. Assessment of the potential of dogs and cats as 33. Ryan PA, Martin L, Mackenzie JS, Kay BH. Investigation of gray-headed flying urban reservoirs of Ross River and Barmah Forest viruses. Aust Vet J. foxes Pteropus poliocephalus (Megachiroptera: Pteropodidae) and 2002;80(1):83–6. mosquitoes in the ecology of Ross River virus in Australia. Am J Trop Med 59. Brook CE, Dobson AP. Bats as ‘special’reservoirs for emerging zoonotic Hyg. 1997;57(4):476–82. pathogens. Trends Microbio. 2015;23(3):172–80.

35 Stephenson et al. Parasites & Vectors (2018) 11:188 Page 13 of 13

60. Van den Hurk AF, Smith CS, Field HE, Smith IL, Northill JA, Taylor CT, et al. Transmission of Japanese encephalitis virus from the black flying fox, Pteropus alecto,toCulex annulirostris mosquitoes, despite the absence of detectable viremia. Am J Trop Med Hyg. 2009;81(3):457–62. 61. Johnston E, Weinstein P, Slaney D, Flies AS, Fricker S, Williams C. Mosquito communities with trap height and urban-rural gradient in Adelaide, South Australia: implications for disease vector surveillance. J Vect Ecol. 2014;39(1):48–55. 62. Lyimo IN, Ferguson HM. Ecological and evolutionary determinants of host species choice in mosquito vectors. Trends Parasitol. 2009;25(4):189–96. 63. Jansen CC, Webb CE, Graham GC, Craig SB, Zborowski P, Ritchie SA, et al. Blood sources of mosquitoes collected from urban and peri-urban environments in eastern Australia with species-specific molecular analysis of avian blood meals. Am J Trop Med Hyg. 2009;81(5):849–57. 64. Comstedt P, Bergstrom S, Olsen B, Garpmo U, Marjavaara L, Mejlon H, et al. Migratory passerine birds as reservoirs of Lyme borreliosis in Europe. Emerg Infect Dis. 2006;12(7):1087–95. 65. Lundström JO, Lindström KM, Olsen B, Dufva R, Krakower DS. Prevalence of Sindbis virus neutralizing antibodies among Swedish passerines indicates that thrushes are the main amplifying hosts. J Med Entomol. 2001;38(2):289–97. 66. Chung YS, Spradbrow PB. Survey for antibodies to arboviruses in domestic- fowls in Queensland. Aust Vet J. 1973;49(12):564. 67. Tompkins D, Johansen C, Jakob-Hoff R, Pulford D, Castro I, Mackereth G. Surveillance for arboviral zoonoses in New Zealand birds. West Pac Surveill Resp J. 2013;4(4):16–23. 68. Takken W, Verhulst NO. Host preferences of blood-feeding mosquitoes. Annu Rev Entomol. 2013;58:433–53. 69. Kay BH, Boyd AM, Ryan PA, Hall RA. Mosquito feeding patterns and natural infection of vertebrates with Ross River and Barmah Forest viruses in Brisbane, Australia. Am J Trop Med Hyg. 2007;76(3):417–23. 70. Choi YH, Comiskey C, Lindsay MDA, Cross JA, Anderson M. Modelling the transmission dynamics of Ross River virus in southwestern Australia. Ima J Math Appl Med Biol. 2002;19(1):61–74. 71. Denholm L, Beeton NJ, Forbes LK, Carver S. A model for the dynamics of Ross River Virus in the Australian environment. Letters Biomath. 2017;4(1):187–206. 72. Glass K. Ecological mechanisms that promote arbovirus survival: a mathematical modelofRossRivervirustransmission.TransRSocTropMedHyg. 2005;99(4):252–60. 73. Tompkins DM, Slaney D. Exploring the potential for Ross River virus emergence in New Zealand. Vector Borne Zoonotic Dis. 2014;14(2):141–8. 74. Buhnerkempe MG, Roberts MP, Dobson AP, Heesterbeek H, Hudson PJ, Lloyd-Smith JO. Eight challenges in modelling disease ecology in multi-host, multi-agent systems. Epidemics. 2015;10:26–30. 75. Faddy H, Dunford M, Seed C, Olds A, Harley D, Dean M, et al. Seroprevalence of antibodies to Ross River and Barmah Forest viruses: Possible implications for blood transfusion safety after extreme weather events. EcoHealth. 2015;12(2):347–53. 76. Campbell J, Aldred J, Davis G. Some aspects of the natural history of Ross River virus in south-east Gippsland, Victoria. In: Arbovirus research in Australia Proceedings Fifth Symposium, August 28-September 1, 1989. Brisbane; 1989. p. 24–8.

Submit your next manuscript to BioMed Central and we will help you at every step:

• We accept pre-submission inquiries • Our selector tool helps you to ﬁnd the most relevant journal • We provide round the clock customer support • Convenient online submission • Thorough peer review • Inclusion in PubMed and all major indexing services • Maximum visibility for your research

Submit your manuscript at www.biomedcentral.com/submit

36 2.8 Supplementary Information Additional file 1: Table S1 Combinations of search terms used to collect papers for review.

Search term 1 (virus related) Combined with the following text Combined with the following text relating to animals related to testing methods Ross River virus AND serolog* OR serosurvey* OR antibod*OR virus OR viral OR Ross River fever AND; host OR reservoir OR wild* OR viraemia OR viremia OR PCR OR captive OR population Epidemic polyarthritis patholog* OR serum OR RNA OR vector* The asterisk (*) operator was used as a wildcard to search for all the possible variations of keywords.

Additional file 1: Table S2 Detailed summary of included Ross River virus reservoir studies, including the reference, location, study type and species group assessed in each study.

Reference Location Study type Species group 1 Doherty, R. L., Gorman, B. M., Queensland Serosurvey Marsupial/Placental Whitehead, R. H., & Carley, J. G. mammal/Bird (1966). Studies of arthropod-borne virus infections in Queensland .V. Survey of antibodies to Group A arboviruses in man and other animals Australian Journal of Experimental Biology and Medical Science, 44, 365-&. doi:10.1038/icb.1966.35 2 Whitehead, R., Doherty, R., Queensland Virus Bird Domrow, R., Standfast, H., & isolation/Serosurvey Wetters, E. (1968). Studies of the epidemiology of arthropod-borne virus infections at Mitchell River Mission, Cape York Peninsula, North Queensland: III. Virus studies of wild birds, 1964–1967. Transactions of the Royal Society of Tropical Medicine and Hygiene, 62(3), 439-445. 3 Whitehead, R. (1969). Queensland Experimental Marsupial/Placental Experimental infection of infection mammal/Bird vertebrates with Ross River and Sindbis viruses 2 Group A arboviruses isolated in Australia. Australian Journal of Experimental Biology and Medical Science, 47, 11-&. doi:10.1038/icb.1969.2 4 Sanderson, C. (1969). A serologic Queensland Serosurvey Placental mammal survey of Queensland cattle for evidence of arbovirus infections. The American Journal Of Tropical Medicine And Hygiene, 18(3), 433. 5 Doherty, R. L., Standfast, H. A., Queensland Virus Marsupial/Placental Domrow, R., Wetters, E. J., isolation/Serosurvey mammal/Bird Whitehead, R. H., & Carley, J. G. (1971). Epidemiology of arthropod-borne virus infections at

37 Mitchell River Mission, Cape York Peninsula, North Queensland. IV. Arbovirus infections of mosquitoes and mammals, 1967-1969. Transactions of the Royal Society of Tropical Medicine and Hygiene, 65(4), 504-&. doi:10.1016/0035- 9203(71)90161-1 6 Spradbrow, P. B. (1972). A survey Queensland Serosurvey Placental mammal for arbovirus antibodies in pigs and sheep in Queensland. Australian Veterinary Journal, 48(7), 402- 407. 7 Chung, Y. S., & Spradbrow, P. B. Queensland Serosurvey Placental mammal (1973). Survey for antibodies to arboviruses in domestic-fowls in Queensland. Australian Veterinary Journal, 49(12), 564-&. doi:10.1111/j.1751- 0813.1973.tb06734.x 8 Doherty, R. L., George, T. D. S., & Queensland, Serosurvey Placental mammal Carley, J. G. (1973). Arbovirus Western Australia, infections of sentinel cattlel in Northern Territory, Australia and New-Guinea. New South Wales, Australian Veterinary Journal, New Guinea 49(12), 574-579. doi:10.1111/j.1751- 0813.1973.tb06737.x 9 Gard, G., Marshall, I. D., & New South Wales Serosurvey Marsupial/Placental Woodroof, G. M. (1973). Annually mammal/Bird recurrent epidemic polyarthritis and Ross River virus activity in a coastal area of New South Wales. II. Mosquitos, viruses and wildlife. American Journal of Tropical Medicine and Hygiene, 22(4), 551- 560. 10 Spradbrow, P. B. (1973). Queensland Experimental Placental mammal Experimenatl infection of sheep infection and pigs with Ross River virus. Australian Veterinary Journal, 49(8), 403-404. doi:10.1111/j.1751- 0813.1973.tb09357.x 11 Gard, G. P., Giles, J. R., New South Wales Serosurvey Marsupial/Placental Dwyergray, R. J., & Woodroofe, mammal/Bird G. M. (1976). Serological evidence of inter-epidemic infection of feral pigs in New South Wales with Murray Valley encephalitis virus. Australian Journal of Experimental Biology and Medical Science, 54(JUN), 297-302. doi:10.1038/icb.1976.30 12 Pascoe, R., George, T., & Queensland Virus isolation Placental mammal Cybinski, D. (1978). The isolation of a Ross River virus from a horse. Australian Veterinary Journal, 54(12), 600-600.

38 13 Marshall, I. D., Woodroofe, G. M., New South Wales Sersosurvey Marsupial/Placental & Gard, G. P. (1980). Arboviruses mammal of coastal Southeastern Australia. Australian Journal of Experimental Biology and Medical Science, 58(Pt1), 91-102. doi:10.1038/icb.1980.9 14 Cloonan, M. J., O'Neill, B. J., New South Wales Serosurvey Placental mammal Vale, T. G., Carter, I. W., & Williams, J. E. (1982). Ross River virus activity along the south coast of New South Wales. Australian Journal of Experimental Biology and Medical Science, 60(DEC), 701-706. doi:10.1038/icb.1982.71 15 Marshall, I. D., Brown, B. K., Western Australia Serosurvey Placental mammal Keith, K., Gard, G. P., & Thibos, E. (1982). Variation in arbovirus infection rates in species of birds sampled in a serological survey during an encephalitis epidemic in the Murray-Valley of Southeastern Australia, February 1974. Australian Journal of Experimental Biology and Medical Science, 60(Pt 5), 471-478. doi:10.1038/icb.1982.52 16 Marshall, I., & Miles, J. (1984). Fiji Serosurvey Placental mammal Ross River virus and epidemic polyarthritis. Curr Top Vector Res, 2, 31-56. 17 Kay, B. H., Hall, R. A., Fanning, I. Queensland Experimental Marsupial/Placental D., Mottram, P., Young, P. L., & infection mammal /Bird Pollitt, C. C. (1986). Experimental infection of vertebrates with Murray Valley encephalitis and Ross River viruses. Arbovirus Research in Australia, 1986, 71- 75. 18 McManus, T. J., & Marshall, I. D. Tasmaina Serosurvey Marsupial/Placental (1986). The epidemiology of Ross mammal River virus in Tasmania. Arbovirus Research in Australia, 1986, 127- 131. 19 O'Callaghan, M. G., & Moore, E. South Australia Serosurvey Marsupial/Placental (1986). Parasites and serological mammal survey of the common brushtail possum (Trichosurus vulpecula) from Kangaroo Island, South Australia. Journal of Wildlife Diseases, 22(4), 589-591. 20 Kay, B. H., Pollitt, C. C., Fanning, Queensland Experimental Placental mammal I. D., & Hall, R. A. (1987). The infection experimental infection of horses with Murray Valley encephalitis and Ross River viruses. Australian Veterinary Journal, 64(2), 52-55 21 Campbell, J., Aldred, J., & Davis, Victoria Serosurvey/Virus Placental G. (1989). Some aspects of the isolation mammal/Marsupial natural history of Ross River virus

39 in south east Gippsland, Victoria. Paper presented at the Arbovirus research in Australia. Proceedings Fifth Symposium, August 28- September 1, 1989, Brisbane, Australia. 22 Aldred, J., Campbell, J., Mitchell, Victoria Serosurvey Marsupial/Placental P., Davis, G., & Elliott, J. (1991). mammal Involvement of wildlife in the natural cycles of Ross River and Barmah Forest viruses. Paper presented at the Proceedings of the Wildlife Disease Association (Australasian Section) Annual Conference, Malacoota. 23 Humphery-Smith, I., Cybinski, D. Queensland Serosurvey Birds H., Byrnes, K. A., & George, T. D. S. (1991). Seroepidemiology of arboviruses among seabirds and island residents of the Great Barrier Reef and Coral Sea. Epidemiology and Infection, 107(2), 435-440. 24 Vale, T. G., Spratt, D. M., & New South Wales Serosurvey Marsupial/Placental Cloonan, M. J. (1991). Serological mammal evidence of arbovirus infection in native and domesticated mammals on the south coast of New South Wales Australia. Australian Journal of Zoology, 39(1), 1-8. 25 Azuolas, J. (1997). Arboviral Victoria Serosurvey Marsupial/Placental diseases of horses and possums. mammal Arbovirus Res. Aust, 7, 5-7. 26 Ryan, P. A., Martin, L., Queensland Experimental Placental mammal Mackenzie, J. S., & Kay, B. H. infection (1997). Investigation of gray- headed flying foxes (Pteropus poliocephalus) (Megachiroptera: Pteropodidae) and mosquitoes in the ecology of Ross river virus in Australia. American Journal of Tropical Medicine and Hygiene, 57(4), 476-482. 27 Azuolas, J. K. (1998). Ross River Victoria Serosurvey Placental mammal virus disease of horses. Australian Equine Veterinarian, 16(2), 56-58. 28 Boyd, A. M., Hall, R. A., Queensland Experimental Marsupial Gemmell, R. T., & Kay, B. H. infection (2001). Experimental infection of Australian brushtail possums, Trichosurus vulpecula (Phalangeridae : Marsupialia), with Ross River and Barmah Forest viruses by use of a natural mosquito vector system. American Journal of Tropical Medicine and Hygiene, 65(6), 777-782. 29 Boyd, A. M., & Kay, B. H. (2002). Queensland Experimental Placental mammal Assessment of the potential of infection dogs and cats as urban reservoirs

40 of Ross River and Barmah Forest viruses. Australian Veterinary Journal, 80(1-2), 83-86. 30 Azuolas, J. K., Wishart, E., Bibby, Victoria Virus isolation Placental mammal S., & Ainsworth, C. (2003). Isolation of Ross River virus from mosquitoes and from horses with signs of musculo-skeletal disease. Australian Veterinary Journal, 81(6), 344-347. doi:10.1111/j.1751- 0813.2003.tb11511.x 31 Old, J. M., & Deane, E. M. (2005). New South Wales Serosurvey Marsupial Antibodies to the Ross River virus in captive marsupials in urban areas of eastern New South Wales, Australia. Journal of Wildlife Diseases, 41(3), 611-614. 32 Kay, B. H., Boyd, A. M., Ryan, P. Queensland Serosurvey Marsupial/Placental A., & Hall, R. A. (2007). Mosquito mammal feeding patterns and natural infection of vertebrates with Ross River and Barmah Forest viruses in Brisbane, Australia. American Journal of Tropical Medicine and Hygiene, 76(3), 417-423. 33 Hill, N. J., Power, M. L., & Deane, New South Wales Serosurvey Marsupial E. M. (2009). Absence of Ross River virus amongst Common brushtail possums (Trichosurus vulpecula) from metropolitan Sydney, Australia. European Journal of Wildlife Research, 55(3), 313-316. doi:10.1007/s10344-008-0238-z 34 McFadden, A. M. J., McFadden, B. New Zealand Serosurvey Placental mammal D., Mackereth, G. F., Clough, R. R., Hueston, L., Gradwell, B., & Dymond, M. (2009). A serological survey of cattle in the Thames- Coromandel district of New Zealand for antibodies to Ross River virus. New Zealand Veterinary Journal, 57(2), 116- 120. doi:10.1080/00480169.2009.36888 35 Pacioni, C., Johansen, C. A., Western Australia Serosurvey Marsupial Mahony, T. J., O'Dea, M. A., Robertson, I. D., Wayne, A. F., & Ellis, T. (2013). A virological investigation into declining woylie populations. Australian Journal of Zoology, 61(6), 446-453. doi:10.1071/zo13077 36 Tompkins, D., Johansen, C., New Zealand Serosurvey Bird Jakob-Hoff, R., Pulford, D., Castro, I., & Mackereth, G. (2013). Surveillance for arboviral zoonoses in New Zealand birds. Western Pacific surveillance and response

41 journal : WPSAR, 4(4), 16-23. doi:10.5365/wpsar.2013.4.3.002 37 Potter, A., Johansen, C. A., Western Australia Serosurvey Placental mammal Fenwick, S., Reid, S. A., & Lindsay, M. D. A. (2014). The seroprevalence and factors associated with Ross River virus infection in Western grey kangaroos (Macropus fuliginosus) in Western Australia. Vector-Borne and Zoonotic Diseases, 14(10), 740-745. doi:10.1089/vbz.2014.1617 38 Reiss, A., Jackson, B., Gillespie, Northern Territory Serosurvey Marsupial/Placental G., Stokeld, D., & Warren, K. mammal (2015). Investigation of Potential Diseases Associated with Northern Territory Mammal Declines.

42 Chapter 3

Interpreting mosquito feeding patterns in Australia through an ecological lens: an analysis of blood meal studies

Statement of contribution to co-authored published paper

The following chapter includes a co-authored paper:

Stephenson, E. B., Murphy, A. K., Jansen, C. C., Peel, A. J., & McCallum, H. (2019) Interpreting mosquito feeding patterns in Australia through an ecological lens: an analysis of blood meal studies. Parasites & Vectors. 12:156. doi:10.1186/s13071-019-3405-z

My contribution to the paper involved data collection, planning and performing all analyses, and writing the paper.

Signed: ______Date: ______Corresponding author: Eloise Stephenson

Countersigned: ______Date:______Supervisor: Hamish McCallum

43 Stephenson et al. Parasites Vectors (2019) 12:156 https://doi.org/10.1186/s13071-019-3405-z Parasites & Vectors

RESEARCH Open Access Interpreting mosquito feeding patterns in Australia through an ecological lens: an analysis of blood meal studies Eloise B. Stephenson1* , Amanda K. Murphy2, Cassie C. Jansen3, Alison J. Peel1 and Hamish McCallum1

Abstract Background: Mosquito-borne pathogens contribute signifcantly to the global burden of disease, infecting millions of people each year. Mosquito feeding is critical to the transmission dynamics of pathogens, and thus it is important to understanding and interpreting mosquito feeding patterns. In this paper we explore mosquito feeding patterns and their implications for disease ecology through a meta-analysis of published blood meal results collected across Australia from more than 12,000 blood meals from 22 species. To assess mosquito-vertebrate associations and identify mosquitoes on a spectrum of generalist or specialist feeders, we analysed blood meal data in two ways; frst using a novel odds ratio analysis, and secondly by calculating Shannon’s diversity scores. Results: We fnd that each mosquito species had a unique feeding association with diferent vertebrates, suggesting species-specifc feeding patterns. Broadly, mosquito species could be grouped broadly into those that were primarily ornithophilic and those that fed more often on livestock. Aggregated feeding patterns observed across Australia were not explained by intrinsic variables such as mosquito genetics or larval habitats. We discuss the implications for disease transmission by vector mosquito species classifed as generalist-feeders (such as Aedes vigilax and Culex annulirostris), or specialists (such as Aedes aegypti) in light of potential infuences on mosquito host choice. Conclusions: Overall, we fnd that whilst existing blood meal studies in Australia are useful for investigating mosquito feeding patterns, standardisation of blood meal study methodologies and analyses, including the incorporation of vertebrate surveys, would improve predictions of the impact of vector-host interactions on disease ecology. Our analysis can also be used as a framework to explore mosquito-vertebrate associations, in which host availability data is unavailable, in other global systems. Keywords: Blood meal, Associations, Vector, Vertebrate, Disease risk, Blood-feeding

Background to take a blood meal from a source host and then to sub- Mosquitoes are the most important disease vector glob- sequently feed on a recipient host. Understanding the ally, responsible for infecting millions of people and feeding patterns of mosquitoes can inform disease man- animals annually with pathogens that infuence human agement strategies (such as targeted vector control to health, livestock and economic trade and wildlife bio- reduce vector-host contact) and can contribute to mod- diversity [1]. Mosquitoes comprise a broad taxonomic els forecasting future disease risk in human and animal group with more than 3000 species recognised across 40 populations [3]. genera [2], but not all species are involved in pathogen Mosquito host choice is complex; both intrinsic and transmission. Pathogen transmission requires a mosquito extrinsic factors can infuence feeding preference [3, 4]. Intrinsic variables can include genetics, whereby individ- *Correspondence: [email protected] uals are more likely to feed on the same host as previous 1 Environmental Futures Research Institute, Grifth University, Brisbane, generations [5, 6], and the nutritional state of the mos- QLD 4111, Australia quito, with nutrition-poor individuals being more likely Full list of author information is available at the end of the article

© The Author(s) 2019. This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/ publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.

44 Stephenson et al. Parasites Vectors (2019) 12:156 Page 2 of 11

to feed on non-preferred hosts [4]. Extrinsic host-seeking this review if they were original peer-reviewed research behaviour is predominantly guided by detection of heat articles, undertaken on mainland Australia (e.g. [17] was and carbon dioxide (CO2), and is also afected by host undertaken in Badu Island in the Torres Strait and thus abundance, biomass, various odorants and chemicals excluded), analysed feld-collected mosquitoes that had that are released by hosts, and host defensive behaviour fed under natural conditions on free-living vertebrates [7–13]. Other extrinsic factors may include climatic vari- (e.g. [18] used tethered animal baits and was excluded), ables such as relative humidity, along with habitat charac- and assessed at least 3 potential vertebrate species (e.g. teristics that determine availability and diversity of hosts [19] only tested for a single fying fox species, and was [12, 14]. In addition to these broad intrinsic and extrinsic excluded). variables, evidence suggests that mosquitoes may adjust Te following information was extracted from identi- their host-seeking behaviours based on positive and neg- fed articles: the geographical area in which the study ative experiences, in essence, adapting feeding choices took place, including site location, bioregion of each site according to their individual circumstances [4]. (as defned by Tackway & Cresswell [20]), the mosquito Mosquito-host relationships in Australia are largely collection method used (including the year, month and understudied. Te island biogeography of Australia collection method/trap type), and the methods used to and its varied climatic zones and bioregions promote a determine the vertebrate origin of blood meals (includ- unique endemic biodiversity for mosquito and vertebrate ing the vertebrate species investigated, the source of host species. Along with a high diversity of native marsu- vertebrate reference samples and laboratory technique). pials, placental mammals and birds in Australia, there are Additional notes were made on stated limitations (if any) more than 300 species of mosquitoes described, many of of each paper and whether data on vertebrate abundance which are unique to the continent [15]. Te interactions and diversity was included. A database of blood meal between these populations of mosquitoes and vertebrate results was populated and is reported in Additional fle 1: hosts across diferent climatic zones provide opportuni- Table S1. ties for maintenance and emergence of mosquito-borne pathogens. Te transmission of numerous medically- Data analysis important arboviruses has been documented in Australia Mosquito‑vertebrate associations to date, including dengue (DENV), Ross River (RRV), Odds ratios were used to calculate the direction (positive Murray valley encephalitis (MVEV), Barmah Forest or negative) and strength of associations in the database (BFV) and Kunjin [16] viruses. between each mosquito species and vertebrate taxon. For Taking into account the complexity of contributing fac- this analysis, the blood meal origin data compiled from tors, critical analysis of the feeding patterns of mosqui- the literature were aggregated such that each vertebrate toes may represent an important approach to explore species was grouped into the broader taxonomic groups disease risks for both human and animal populations. of humans, Carnivora (cats, dogs and foxes), Aves (all Tis study aims to synthesise existing literature describ- birds), Diprotodontia (all possum and kangaroo species), ing blood meal studies in Australia, specifcally assess- Artiodactyla (cows, sheep, pigs and goats) and Perisso- ing the most likely mosquito-host associations, and the dactyla (horses). Flying fox [21], rodent [22, 23] and rab- diversity of feeding patterns for common mosquito spe- bit [22, 24] species were excluded from this analysis, as cies. In light of these feeding patterns, we discuss broad the sample size of blood meals from these species was too implications for disease ecology. small (either one or two studies, with these species comprising less than 9% of blood-meal origins within each). Methods To be included in the analysis, mosquito species needed Data collection to meet all minimum data criteria of (i) their blood meal Original research articles were systematically searched origins being reported more than twice in the literature; by using the following search terms in diferent combi- and (ii) having an arbitrary minimum of 35 blood meals nation across fve search engines (Web of Science, Pro- identifed. Quest, Science Direct, PubMed and Google Scholar): Log odds ratios were calculated between each mos- ‘bloodmeal*’, ‘blood meal’, blood-meal’, ‘feeding’, ‘habit’, quito and vertebrate taxon using two by two feeding ‘pattern*’, ‘preference*’, ‘interaction*’, ‘mosquito*’, ‘vec- frequency tables, derived from the raw data (Additional tor*’, ‘vector-host’, ‘host*’, ‘vertebrate*’, ‘animal*’ and ‘Aus- fle 1: Table S2). Positive log odds ratios indicate a posi- tralia’. Te asterisk (*) operator was used as a wildcard to tive feeding association between a given mosquito species search for all possible variations of keywords. We then and vertebrate host, whereby there is a higher likelihood manually searched the reference lists of papers to iden- than random chance that a blood meal of that mosquito tify additional relevant articles. Papers were included in species would originate from the given vertebrate taxon.

45 Stephenson et al. Parasites Vectors (2019) 12:156 Page 3 of 11

Te greater the log odds ratio, the stronger the feeding To collect blood-fed mosquitoes, most studies (n = association. Conversely, a negative log odds ratio suggests 8) used Centers for Disease Control (CDC) CO2-baited a negative feeding association, whereby there is a lower miniature light traps [28], supplemented with 1-octen- likelihood that a blood meal from the mosquito species 3-ol in some cases (Table 1). Other methods included would originate from the given vertebrate taxon. Log unbaited BioGent® (BG) sentinel traps [22] and aspira- odds ratios close to 0 indicate no association between the tion of resting sites [21, 29, 30]. One study [29] also used mosquito species and vertebrate taxon. vehicle-mounted traps. To analyse blood meals, early Te log odds ratios were plotted in a heatmap chart studies employed precipitin tests [29–31] and serological and sorted using hierarchical clustering. Te cluster- gel difusion techniques [32, 33] (Table 1). More recent ing grouped mosquitoes with similar feeding patterns studies adopted enzyme-linked immunosorbent assay together by similarity in log odds ratio across all verte- (ELISA) and various molecular techniques including pol- brate taxa. All calculations and graphs were generated ymerase chain reaction (PCR) and gene sequencing [22– using R software, with packages gplots and RColorBrewer 24, 34]. Vertebrate reference sources most commonly [25] with modifed script from Raschka [26]. employed in immunoassays and gel difusion techniques were commercially-available anti-sera and included Mosquito feeding diversity horse, rabbit, rat, dog, chicken, cat, bird, kangaroo, cow We used the Shannonʼs diversity index to place mosquito and pig. Molecular studies which included wildlife used species on a spectrum between generalist or specialist vertebrate references provided through wildlife hospi- feeders. Te inclusion criteria for this analysis were that tals, zoos and roadkill. Tese studies also included DNA each mosquito species needed to have fed on greater sequence data available on GenBank. than three vertebrate species and had to have a minimum number of 10 blood meals analysed. Vertebrates were Mosquito‑vertebrate associations not aggregated by taxonomic group in this analysis but Of the 10 studies of blood meal origins, data on 41 mos- remained at the level reported in the literature (mostly as quito species were reported, and data on 12 of these met species but, for the case of birds, several studies reported the criteria to be included in analysis: Aedes norman- as class). A total of 15 vertebrate species were included ensis, Ae. notoscriptus, Ae. procax, Ae. vigilax, Anoph- in this analysis as blood-meal origins, and 13,934 blood eles annulipies, An. bancroftii, Coquillettidia linealis, meals from 21/41 mosquito species met the criteria Cq. xanthogaster, Culex annulirostris, Cx. sitiens, Cx. (Additional fle 1: Table S1). quinquefasciatus and Mansonia uniformis (Fig. 2). All Shannonʼs diversity index was calculated for each mos- species, except Ae. procax, showed signifcant positive quito species [27] and expressed as an h-index. A higher associations with at least one vertebrate host. Te strong- h-index is associated with a greater feeding diversity, as est positive log odds ratio was between Cx. annulirostris it suggests a mosquito species has fed on a greater num- and the Diprotodontia taxa (possums and kangaroos; log ber of vertebrate species and/or feeds evenly across ver- odds ratio (LOR) = 2.77), followed by Ma. uniformis and tebrates. Conversely, a lower h-index suggests mosquito humans (LOR = 2.2). All mosquito species except Ae. species have a low feeding diversity, and are associated vigilax and Cq. xanthogaster had strong negative asso- with feeding on fewer vertebrate species and/or a greater ciation with at least one vertebrate taxon. Te strongest number of feeds on a small number of vertebrates. negative log odds ratio was between Cx. quinquefasciatus Within this dataset, we categorised an h-index in the top and Diprotodontia (LOR = -3.8), followed by Ae. noto- quartile as ‘high feeding diversity’, whilst an h-index in scriptus and Artiodactyla (cows, sheep, pigs and goats; the lowest quartile was considered a ‘low feeding diver- LOR = -2.8). sity’. Shannon’s diversity index was calculated in Excel. Te mosquito species clustered together in two broad groups. Te frst cluster group consisted of seven mos- Results quito species (Cx. quinquefasciatus, Ae. vigilax, Cq. xan- Characteristics of the selected studies thogaster, Ae. procax, Cq. linealis, Cx. sitiens and Ae. We identifed ten papers that met the search criteria, notoscriptus), of which most shared a negative associa- comprising 14,044 mosquito blood meals across 48 mos- tion with the Artiodactyla (6 of the 7 species) and Dipro- quito species. Study characteristics and methodologies todontia vertebrates (5/7), and a positive association with are summarised in Table 1. Tese studies took place at humans (7/7) and Aves (6/7). Within this cluster, Ae. 32 sites across 14 bioregions, in all mainland states and vigilax and Cq. xanthogaster were in the same clade and territories in Australia (Fig. 1). Te selected studies were shared a strong positive association with Perissodactyla. undertaken over a 62 year-period, from 1954 to 2016. Coquillettidia linealis and Cx. sitiens also shared a clade

46 Stephenson et al. Parasites Vectors (2019) 12:156 Page 4 of 11 cat, bird, kangaroo, kangaroo, cat, bird, pig cow, dog, chicken, cat, bird, chicken, cat, bird, dog, 50 pig, cow, kangaroo, species wild bird rabbit, horse, donkey, donkey, rabbit, horse, dog, human, cat, fox, brushtail possum, quokka, Western grey mouse, kangaroo, chicken, duck pig, chicken, brushtail pig, and ringtail possum, koala, kangaroo, red lorikeet,rainbow galah, magpie Australian cow horse, dog, dog, chicken, cat, bird, chicken, cat, bird, dog, pig cow, kangaroo, horse, human, brushtail horse, possum, fying fox species marsupial, chicken marsupial, pig, dog, cat, man, dog, pig, bat, rodent, horse, cow, carnivore marsupial, Horse, rat, human, dog, rat, human, dog, Horse, Vertebrate species tested Vertebrate Horse, rabbit, rat, human, Horse, Cow, sheep, goat, pig, goat, pig, sheep, Cow, Dog, cat, cow, sheep, sheep, cat, cow, Dog, bird, Human, kangaroo, Horse, rabbit, rat, human, Horse, Bird, kangaroo, cat, dog, cat, dog, kangaroo, Bird, Cow, horse, dog, human, dog, horse, Cow, Reptile, amphibian, bird, amphibian, bird, Reptile, - - - using avian- and using avian- mammalian-specifc gene primers, sequencing using vertebrate, using vertebrate, and mamma - avian gene lian primers, sequencing assay assay assay Double-antibody ELISA Blood-meal analysis method Indirect ELISA; PCR Double-antibody ELISA Cytochrome b PCR Cytochrome Gel difusion immuno Gel difusion immuno Gel difusion immuno Precipitin test Precipitin Precipitin test Precipitin 199 (8) 763 (15) 865 (10) 1128 (1) 1180 (15) 2606 (29) 2582 (15) 1628 (18) 5431 (16) No. of blood meals No. of species) (no. analysed - 2 CO + / − octe + / − OCT + / − OCT 2 2 2 2 2 2 CO CO CO CO CO CO nol traps specifed) and vehicle specifed) and vehicle trap mounted sites EVS + Mosquito collection method EVS + Unbaited BG sentinel Unbaited CDC + CDC, EVS + CDC + CDC + Light trap (not EVS + Aspiration of resting of resting Aspiration - (1993–2004) 2015) (2000–2001) (2005–2008) ber (2002–2006) 2001) (1974–1976) (1998–2000) All year round round All year Collection months (years) August–May (2006– August–May September–April September–April All year round round All year February, March, Octo March, February, January–May (1995– All year round round All year All year round round All year February (1976) February Urban and rural Site habitat typeSite Urban and rural Urban Urban Rural Rural Rural Rural Rural Australia) City, Murray River val - City, ley (South Australia) Brisbane (Queensland), (NewNewcastle South (New Sydney Wales), South Wales) Territory) (Queensland) Territory) (Queensland) 32 sites (Western32 sites Site name (state) Site Adelaide Hills, Adelaide Hills, Adelaide Adelaide Brisbane (Queensland) Cairns (Queensland), Cairns (Queensland), (Queensland, Northern(Queensland, Gulf Plains York, Cape Beatrice Hill (Northern Shoalwater Bay Bay Shoalwater Charleville (Queensland) Location and methods for mosquito collection and blood meal analysis for the studies included in this review collection mosquito and blood meal analysis for and methods for Location 24 ] [ Johansen et al. 1 Table Reference al. [ 23 ] et al. Flies Kay [ 21 ] et al. al. [ 22 ] Jansen et al. al. [ 34 ] Hall-Mendelin et al. [ 33 ] van den Hurk et al. al. [ 29 ] Muller et al. al. [ 32 ] et al. Frances Kay [ 30 ] et al.

47 Stephenson et al. Parasites Vectors (2019) 12:156 Page 5 of 11 - dog, rabbit, horse, mar rabbit, horse, dog, supial (unspecifed) Vertebrate species tested Vertebrate Human, chicken, cow, Human, chicken, cow, Blood-meal analysis method Precipitin test Precipitin 1231 (15) 14,044 (48) No. of blood meals No. of species) (no. analysed sites Mosquito collection method Aspiration of resting of resting Aspiration (1951–1952) Collection months (years) December–January December–January Site habitat typeSite Mostly rural , carbon dioxide; CDC, centers for disease control; BG, BioGent; PCR, polymerase chain reaction; ELISA, enzyme-linked BG, BioGent; PCR, polymerase disease control; for octane CDC, centers immunosorbent assay;, carbon dioxide; OCT, 2 CO South Wales), Texas Texas South Wales), Golburn(Queensland), (Victoria),Valley Canberra (Australian Territory) Capital Site name (state) Site Moree, Hornsby (New Moree, (continued) 1 Table survey;: EVS, vector Abbreviations encephalitis Reference al. [ 31 ] et al. Lee Total

48 Stephenson et al. Parasites Vectors (2019) 12:156 Page 6 of 11

Fig. 1 Bioregions in which blood meal studies took place (indicated in red) across Australia (derived from Google Map Data ©) and strong negative association with both Artiodactyla was reported in fve mosquito species; of these, Ae. vigi- and Carnivora. lax had the highest diversity (h-index = 2.17). Te second major cluster group consisted of fve mosquito species (Cx. annulirostris, Ae. normanesis, An. Discussion annulipies, An. bancroftii and Ma. uniformis). Tese spe- Our analysis revealed that each mosquito species had a cies all shared a positive association with Artiodactyla unique feeding association with diferent vertebrates, and a negative association with Aves. Culex annulirostris suggesting species-specifc feeding patterns. Te hierar- and Ae. normanesis were on the same clade and shared chical clustering from the odds ratio analysis sorted mos- a strong positive association with Perissodactyla and quitoes into two broad groups: mosquitoes that either Diprotodontia. Although they both also had a negative had a positive association with birds (Aves) and negative association with Aves, Carnivora and humans, this was association with livestock (Artiodactyla), or vice-versa. strongest only for Ae. normanensis. Anopheles annulipies Interpreting the feeding patterns of these particular mos- and Ma. uniformis were on a single clade and were both quito species is important, given that at least half of these had high associations with humans. mosquitoes have been found to be competent vectors for notifable arboviruses in Australia [35–40], whilst the Mosquito feeding diversity other half have been demonstrated to carry some viruses, Twenty-two mosquito species met the criteria for inclu- although their ability to transmit them has not been fully sion in the diversity analysis based on Shannonʼs index investigated [41–43]. (Fig. 3), comprising 12,424 individual blood meals in Intrinsic drivers of mosquito host choices (such as total. Te median h-index reported across all species genetics, larval ecology and dispersal) did not explain was 1.40, and the mean was 1.34. Low feeding diver- feeding patterns in this analysis. Specifcally, mosquito sity (h-index = < 0.99) was observed in fve mosquito species did not group together by taxonomic related- species; of which Ae. aegypti had the lowest diveristy ness (e.g. genus). Studies examining the efect of genet- (h-index = 0.72). High feeding diversity (h-index = > 1.64) ics on mosquito host choices have found that ofspring

49 Stephenson et al. Parasites Vectors (2019) 12:156 Page 7 of 11

Fig. 2 Feeding associations between Australian mosquito species and vertebrate taxa. Log odds ratio for mosquito species (right hand side) indicate feeding likelihood on vertebrate taxa (bottom). Mosquito species are sorted in the chart using a hierarchal cluster (left) according to how similar their vertebrate feeding patterns are

Fig. 3 Shannon’s diversity (h-index) of blood meal origins for Australian mosquitoes, error bars represent the standard error for all measures

50 Stephenson et al. Parasites Vectors (2019) 12:156 Page 8 of 11

are more likely to feed on the same host as previous despite being exposed to the same vertebrates in a single generations [5, 6]. However, this has only been demon- location. strated within-species and is unlikely to be important In the odds ratio analysis, Cx. annulirostris exhibited between species belonging to same genus, particularly an unexpected feeding pattern. Tis species is consid- since potential for rapid evolution (due to short genera- ered an important vector for medically-important arbo- tion span) likely reduces the infuence that taxonomic viruses [36, 49, 50]; however, the meta-analyses which relatedness may have on mosquito feeding host behav- included more than 5700 blood meals, found that Cx. iour. Another intrinsic factor, mosquito larval ecology, annulirostris had only a weak feeding association with may partially explain some clustering. For example, Ae. humans (LOR = -0.74). Tis is consistent with an early normanensis and Cx. annulirostris grouped together and feld study assessing mosquito feeding preferences using larvae of both typically inhabit inland freshwater; simi- live baits, in which Cx. annulirostris preferred cows, pigs larly, the larval habitat of both An. bancroftii and Ma. and dogs more than humans [37]. Te analysis based on uniformis is freshwater swamps. Although this pattern Shannonʼs diversity, along with other studies, have iden- did not explain all clusters, it implies that local environ- tifed Cx. annulirostris as a generalist feeder with plastic mental infuences, at least partially, drive mosquito host feeding patterns that may shift temporally or spatially choice. Tis is perhaps not surprising when considering [51, 52]. Tis knowledge, in combination with the wide- potential limitations on dispersal from larval habitats for spread distribution of Cx. annulirostris across Australia, various mosquito species. For example, Ae. vigilax is rec- suggests that localised studies of Cx annulirostris feeding ognised as having large dispersal capability, being found are required to assess the role the species plays in disease more than 50 km from potential saltwater larval habitats, transmission for which it is theoretically an important albeit likely wind-assisted in some cases [44]. Tis high vector. dispersal potential suggests that Ae. vigilax can move In addition to Cx. annulirostris, Aedes vigilax and Ae. readily between locations, allowing feeding on a diversity notoscriptus were identifed as generalists due to their of vertebrate taxa, as refected in our feeding diversity high diversity scores in the Shannon’s diversity analy- analysis. As such, whilst genetics and larval habitats may sis. International studies [53–55] suggest that generalist be important within species on a local scale, they do not feeders are capable of playing a role as bridge vectors due explain the aggregated feeding patterns observed across to their ability to acquire pathogens from animal hosts, Australia. and subsequently transmitting the pathogen to humans. Extrinsic variables, such as species abundance and Bridge vectors are particularly important for enzootic diversity, explain in part some of the feeding associa- amplifcation of arboviruses and are often associated with tions in this analysis, but not all. Mosquitoes have com- outbreaks [53]. For the species identifed in this analy- plex interactions with their environment. Tus, factors sis as generalists, they have been demonstrated to be broader than vertebrate abundance alone are important competent vectors of zoonotic arboviruses in Australia to consider for mosquito feeding patterns. For example, [45, 56–58], and as such should be closely monitored to mosquito fying/resting height has been linked to host reduce transmission between vectors and humans. feeding patterns [45–48]. In two Australian studies, more Te disease ecology associated with specialist feeders is Cx. sitiens and Cx. quiquefasciatus were caught in traps also important to consider. Here we identifed Ae. aegypti set at least 8 m of the ground, whilst a higher abundance as having the lowest feeding diversity, indicating the spe- of Ae. vigilax were found in traps 1.5 m of the ground, cies is a specialist feeder. Indeed, more than 70% of the for the same locations [47, 48]. In our meta-analysis, Cx. blood meals originated from humans. Te anthropophilic sitiens and Cx. quinquefasciatus had strong positive asso- feeding observed in Ae. aegypti is similar to that reported ciations with blood meals originating from tree dwelling in international studies, where 80–99% of all blood meals bird species (i.e. Australasian fgbirds Sphecotheres vieil- are human in origin [59, 60]. Tis feeding pattern for Ae. loti, common myna Sturnus tristis and helmeted friar- aegypti is consistent with its role as an important vector birds Philemon buceroides [22]), whilst Ae. vigilax had the of several arboviruses which are transmitted between strongest positive associations with ground-dwelling spe- humans without an animal reservoir, including dengue, cies (horses and humans). Tis could suggest that whilst Zika and chikungunya viruses. Interestingly, although the overall vertebrate abundance within a given environment importance of Ae. aegypti is recognised, under laboratory can infuence the availability of a particular host, mosqui- conditions the species has been observed to demonstrate toes are highly mobile and may seek a blood meal across relatively poor transmission rates for DENV, when com- ecological niches within a given habitat. Tus, diferent pared to other mosquito species [61, 62]. In this case, mosquito species can exhibit diferent feeding patterns being a specialist feeder, preferring mainly humans, is

51 Stephenson et al. Parasites Vectors (2019) 12:156 Page 9 of 11

what determines the status of Ae. aegypti as an important likely infuenced by a suite of intrinsic and extrinsic vari- disease vector, rather than its competence [63]. ables. Broader ecological considerations alongside these feeding patterns could be useful for the interpretation of Future directions these complex biological systems, but at present data avail- An absence of data on host availability in the regions able to do this are limited. Future studies should utilise where mosquitoes were collected limits inferences on multidisciplinary approaches to collect data on vertebrate host preference specifcally. Of the blood meal studies communities in parallel with mosquito communities. More reviewed here, only one considered host abundance [21]. data from both top-down (broad assessments of blood Tat study assessed abundance through a local resident meals) and bottom-up approaches (specialised host choice survey on the number of pets, people and estimated experiments) are needed in conjunction with modelling number of possums in the vicinity, adding confdence to techniques to bring these data together for meaningful the interpretation of vector-feeding patterns. Such col- interpretation of arbovirus transmission risk in Australia. lection of host ecology data in conjunction with blood- fed mosquitoes can be considerably labour-intensive; Additional fle however, it provides a more thorough assessment of how host abundance and biomass may infuence observed mosquito feeding patterns and informs the selection of Additional fle 1: Table S1. Reported blood-meal results for Australian mosquito species. Numbers in the columns representing number of blood appropriate reference samples against which to com- meals for each vertebrate. Table S2. Derivation of 2 2 contingency pare blood meals in the laboratory. Although limited in tables for mosquito feeding preferences. × their application, other blood-meal studies [64, 65] have utilised databases, such as the Atlas of Living Australia Acknowledgements (ALA) or the Global Biodiversity Information Facility Not applicable.

(GBIF), to identify potential available vertebrates in the Funding absence of formal vertebrate surveys. Whilst they are no ES was supported by Australian Government Research Training Program substitute for vertebrate surveys, these datasets could be Scholarship. benefcial for noting the presence of common hosts in Availability of data and materials future blood meal studies but are limited in estimating The datasets used and/or analysed during the present study are available from host density or true absence of a given species. the corresponding author upon reasonable request.

Although a range of reference vertebrates were often Authors’ contributions included in Australian blood-meal studies, they were ES collected, analysed, interpreted papers and was a major contributor in the rarely a true representation of the vertebrates available to writing of the manuscript. AM, CC, AP and HM contributed to the interpretation of results and writing of the manuscript. All authors read and approved mosquitoes for feeding. At present there are large gaps in the fnal manuscript. understanding the role of cryptic, migratory or smaller mammalian species in mosquito feeding patterns. For Ethics approval and consent to participate example, only two studies included rabbits [22, 24] and Not applicable. rodents [22, 23] in their analysis. Despite their small Consent for publication size, rabbits and rodents were identifed to be the origin Not applicable. of blood meals for Cx. sitiens and Cq. linealis [22]. Mos- Competing interests quito-rodent associations have also been identifed in the The authors declare that they have no competing interests. literature, where by at least 27% of mice were seropositive to RRV [58, 66]. It is therefore important that, despite Publisher’s Note small body size, rats and rodents are included in future Springer Nature remains neutral with regard to jurisdictional claims in pub- investigations of mosquito blood meals. lished maps and institutional afliations. Author details Conclusions 1 Environmental Futures Research Institute, Grifth University, Brisbane, QLD 4111, Australia. 2 QIMR, Herston, QLD 4006, Australia. 3 Communicable Improved understanding of mosquito feeding patterns Diseases Branch, Department of Health, Queensland Government, Herston, can lead to better management and risk predictions for QLD 4006, Australia. medically important arboviruses. Here we fnd that of the Received: 15 November 2018 Accepted: 20 March 2019 Australian mosquito species tested, each had a unique feeding pattern; however, the particular specialist or generalist feeding patterns of mosquito species could be a key determinant of the risk they pose for human disease. Tese patterns, and the resulting human disease risk, are

52 Stephenson et al. Parasites Vectors (2019) 12:156 Page 10 of 11

References 26. Raschka S. Instant heat maps in R: how-to. Birmingham UK: Packt 1. Gubler DJ. Human arbovirus infections worldwide. Ann N Y Acad Sci. Publishing Ltd.; 2013. 2001;951:13–24. 27. Morris EK, Caruso T, Buscot F, Fischer M, Hancock C, Maier TS, 2. Wongkamchai S, Khongtak P, Leemingsawat S, Komalamisra N, Junsong et al. Choosing and using diversity indices: insights for ecological N, Kulthanan K, et al. Comparative identifcation of protein profles and applications from the German Biodiversity Exploratories. Ecol Evol. major allergens of saliva, salivary gland and whole body extracts of mos- 2014;4:3514–24. quito species in Thailand. Asian Pac J Allergy Immunol. 2010;28:162–9. 28. Sudia W, Chamberlain R. Battery-operated light trap, an improved model. 3. Lyimo IN, Ferguson HM. Ecological and evolutionary determinants of Mosq News. 1962;22:126–9. host species choice in mosquito vectors. Trends Parasitol. 2009;25:189–96. 29. Muller M, Murray M, Edwards J. Blood-sucking midges and mosquitoes 4. Takken W, Verhulst NO. Host preferences of blood-feeding mosquitoes. feeding on mammals at Beatrice Hill, NT. Aust J Zool. 1981;29:573–88. Annu Rev Entomol. 2013;58:433–53. 30. Kay B, Boreham P, Williams G. Host preferences and feeding patterns of 5. Mukwaya L. Genetic control of feeding preferences in the mosquitoes mosquitoes (Diptera: Culicidae) at Kowanyama, Cape York Peninsula, Aedes (Stegomyia) simpsoni and aegypti. Physiol Entomol. 1977;2:133–45. northern Queensland. Bull Entomol Res. 1979;69:441–57. 6. Ulloa García A, Arredondo Jiménez JI, Fernández Salas I, Rodríguez MH, 31. Lee DJ, Clinton K, O’gower A. The blood sources of some Australian González Cerón L. Innate host selection in Anopheles vestitipennis from mosquitoes. Aust J Biol Sci. 1954;7:282–301. southern Mexico. J Am Mosq Control Assoc. 2004;20:337–41. 32. Frances SP, Cooper RD, Rowclife KL, Chen N, Cheng Q. Occurrence of 7. Edman JD, Taylor DJ. Culex nigripalpus: seasonal shift in the bird-mammal Ross River virus and Barmah Forest virus in mosquitoes at Shoalwa- feeding ratio in a mosquito vector of human encephalitis. Science. ter Bay military training area, Queensland, Australia. J Med Entomol. 1968;161:67–8. 2004;41:115–20. 8. Balenghien T, Fouque F, Sabatier P, Bicout DJ. Horse-, bird-, and human- 33. van den Hurk AF, Johansen CA, Zborowski P, Paru R, Foley PN, Beebe seeking behavior and seasonal abundance of mosquitoes in a West Nile NW, et al. Mosquito host-feeding patterns and implications for Japanese virus focus of southern France. J Med Entomol. 2006;43:936–46. encephalitis virus transmission in northern Australia and Papua New 9. Kilpatrick AM, Kramer LD, Jones MJ, Marra PP, Daszak P. West Nile virus Guinea. Med Vet Entomol. 2003;17:403–11. epidemics in North America are driven by shifts in mosquito feeding 34. Hall-Mendelin S, Jansen CC, Cheah WY, Montgomery BL, Hall RA, Ritchie behavior. PLoS Biol. 2006;4:e82. SA, et al. Culex annulirostris (Diptera: Culicidae) host feeding patterns and 10. Eiras AE, Jepson P. Responses of female Aedes aegypti (Diptera: Culicidae) Japanese encephalitis virus ecology in northern Australia. J Med Entomol. to host odours and convection currents using an olfactometer bioassay. 2012;49:371–7. Bull Entomol Res. 1994;84:207–11. 35. Watson TM, Kay BH. Vector competence of Aedes notoscriptus (Diptera: 11. Bowen M. The sensory physiology of host-seeking behavior in mosqui- Culicidae) for Barmah Forest virus and of this species and Aedes aegypti toes. Ann Rev Entomol. 1991;36:139–58. (Diptera: Culicidae) for dengue 1–4 viruses in Queensland. Australia. J 12. Clements A. The biology of mosquitoes. London: Chapman & Hall; 1992. Med Entomol. 1999;36:508–14. 13. Day JF, Ebert KM, Edman JD. Feeding patterns of mosquitoes (Diptera: 36. Boyd AM, Kay BH. Vector competence of Aedes aegypti, Culex sitiens, Culex Culicidae) simultaneously exposed to malarious and healthy mice, includ- annulirostris, and Culex quinquefasciatus (Diptera: Culiciclae) for Barmah ing a method for separating blood meals from conspecifc hosts. J Med Forest virus. J MedEntomol. 2000;37:660–3. Entomol. 1983;20:120–7. 37. Ryan PA, Kay BH. Vector competence of mosquitoes (Diptera: Culicidae) 14. Olanga EA, Okal MN, Mbadi PA, Kokwaro ED, Mukabana WR. Attraction of from Maroochy Shire, Australia, for Barmah Forest virus. J Med Entomol. Anopheles gambiae to odour baits augmented with heat and moisture. 1999;36:856–60. Malar J. 2010;9:6. 38. Wells PJ, Russell RC, Cloonan MJ. Investigating vector competence of 15. Webb C, Russell R, Doggett S. A guide to mosquitoes of Australia. New Culex annulirostris and Aedes vigilax for Ross River virus and other alpha- York: CSIRO Publishing; 2016. and bunyaviruses. Arbov Res Aust. 1993;6:10–4. 16. van den Hurk AF, Jansen CC. Arboviruses of Oceania. In: Loukas A, editor. 39. Kay BH, Fanning ID, Carley JG. Vector competence of Culex pipiens Neglected tropical diseases - Oceania. Berlin: Springer; 2016. p. 193–235. quinquefasciatus for Murray Valley encephalitis, Kunjin, and Ross River 17. van den Hurk AF, Nisbet DJ, Johansen CA, Foley PN, Ritchie SA, Mackenzie viruses from Australia. Am J Trop Med Hyg. 1982;31:844–8. JS. Japanese encephalitis on Badu Island, Australia: the frst isolation of 40. Kay BH, Carley JG, Fanning ID, Filippich C. Quantitative studies of the vec- Japanese encephalitis virus from Culex gelidus in the Australasian region tor competence of Aedes aegypti, Culex annulirostris and other mosquitos and the role of mosquito host-feeding patterns in virus transmission (Diptera, Culicidae) with Murray Valley encephalitis and other Queensland cycles. Trans R Soc Trop Med Hyg. 2001;95:595–600. arboviruses. J Med Entomol. 1979;16:59–66. 18. Muller M, Murray M. Blood-sucking fies feeding on sheep in eastern 41. Jefery JA, Ryan PA, Lyons SA, Kay BH. Vector competence of Coquillet- Australia. Aust J Zool. 1977;25:75–85. tidia linealis (Skuse) (Diptera: Culicidae) for Ross River and Barmah Forest 19. Ryan PA, Martin L, Mackenzie JS, Kay BH. Investigation of gray-headed viruses. Aust J Entomol. 2002;41:339–44. fying foxes (Pteropus poliocephalus) (Megachiroptera: Pteropodidae) and 42. van den Hurk AF, Nisbet DJ, Foley PN, Ritchie SA, Mackenzie JS, Beebe mosquitoes in the ecology of Ross River virus in Australia. Am J Trop Med NW. Isolation of arboviruses from mosquitoes (Diptera: Culicidae) col- Hyg. 1997;57:476–82. lected from the Gulf Plains region of Northwest Queensland, Australia. J 20. Thackway R, Cresswell I. A bioregional framework for planning Med Entomol. 2002;39:786–92. the national system of protected areas in Australia. Nat Areas J. 43. Harley D, Phillips D, Ritchie S, Atkin C, van den Hurk A. Isolation of Ross 1997;17:241–7. River virus from mosquitoes trapped in North Queensland. Arbov Res 21. Kay BH, Boyd AM, Ryan PA, Hall RA. Mosquito feeding patterns and Aust. 1997;7:92–4. natural infection of vertebrates with Ross River and Barmah Forest viruses 44. Chapman HF, Hughes JM, Jennings C, Kay BH, Ritchie SA. Population in Brisbane, Australia. Am J Trop Med Hyg. 2007;76:417–23. structure and dispersal of the saltmarsh mosquito Aedes vigilax in 22. Jansen CC, Webb CE, Graham GC, Craig SB, Zborowski P, Ritchie SA, et al. Queensland, Australia. Med Vet Entomol. 1999;13:423–30. Blood sources of mosquitoes collected from urban and peri-urban envi- 45. Lee HI, Seo BY, Shin EH, Burkett DA, Lee WJ, Shin YH. Study of fying height ronments in eastern Australia with species-specifc molecular analysis of of culicid species in the northern part of the Republic of Korea. J Am avian blood meals. Am J Trop Med Hyg. 2009;81:849–57. Mosq Cont Assoc. 2006;22:239–45. 23. Flies EJ, Flies AS, Fricker SR, Weinstein P, Williams CR. Regional compari- 46. Swanson D, Adler P. Vertical distribution of haematophagous Dip- son of mosquito bloodmeals in South Australia: implications for Ross tera in temperate forests of the southeastern USA. Med Vet Entomol. River virus ecology. J Med Entomol. 2016;53:902–10. 2010;24:182–8. 24. Johansen CA, Power SL, Broom AK. Determination of mosquito (Dip- 47. Jansen CC, Zborowski P, Ritchie SA, Van Den Hurk AF. Efcacy of bird- tera: Culicidae) bloodmeal sources in Western Australia: implications baited traps placed at diferent heights for collecting ornithophilic mos- for arbovirus transmission. J Med Entomol. 2009;46:1167–75. quitoes in eastern Queensland, Australia. Aust J Entomol. 2009;48:53–9. 25. Neuwirth E. Package ‘RColorBrewer’. 2015. http://www.colorbrewe 48. Johnston E, Weinstein P, Slaney D, Flies AS, Fricker S, Williams C. Mosquito r2.org. communities with trap height and urban-rural gradient in Adelaide,

53 Stephenson et al. Parasites Vectors (2019) 12:156 Page 11 of 11

South Australia: implications for disease vector surveillance. J Vect Ecol. 59. Scott TW, Amerasinghe PH, Morrison AC, Lorenz LH, Clark GG, Strick- 2014;39:48–55. man D, et al. Longitudinal studies of Aedes aegypti (Diptera: Culicidae) 49. Jansen CC, Webb CE, Northill JA, Ritchie SA, Russell RC, van den Hurk AF. in Thailand and Puerto Rico: blood feeding frequency. J Med Entomol. Vector competence of Australian mosquito species for a North American 2000;37:89–101. strain of West Nile virus. Vector Borne Zoonotic Dis. 2008;8:805–11. 60. Scott TW, Chow E, Strickman D, Kittayapong P, Wirtz RA, Lorenz LH, 50. Kay BH, Hearnden MN, Oliveira NM, Sellner IN, Hall RA. Alphavirus Edman JD. Blood-feeding patterns of Aedes aegypti (Diptera: Culicidae) infection in mosquitoes at the Ross River reservoir, north Queensland, collected in a rural Thai village. J Med Entomol. 1993;30:922–7. 1990–1993. J Am Mosq Cont Assoc. 1996;12:421–8. 61. Calvez E, Guillaumot L, Girault D, Richard V, O’Connor O, Paoaafaite T, 51. Williams CR, Kokkinn MJ, Smith BP. Intraspecifc variation in odor-medi- et al. Dengue-1 virus and vector competence of Aedes aegypti (Diptera: ated host preference of the mosquito Culex annulirostris. J Chem Ecol. Culicidae) populations from New Caledonia. Parasit Vectors. 2017;10:381. 2003;29:1889–903. 62. Brady OJ, Golding N, Pigott DM, Kraemer MU, Messina JP, Reiner RC Jr, 52. Standfast H, Barrow G. Studies of the epidemiology of arthropod-borne et al. Global temperature constraints on Aedes aegypti and Ae. albopictus virus infections at Mitchell River Mission, Cape York Peninsula, North persistence and competence for dengue virus transmission. Parasit Vec- Queensland. Trans R Soc Trop Med Hyg. 1968;62:418–29. tors. 2014;7:338. 53. Kilpatrick AM, Kramer LD, Campbell SR, Alleyne EO, Dobson AP, Daszak P. 63. Jansen CC, Williams CR, van den Hurk AF. The usual suspects: compari- West Nile virus risk assessment and the bridge vector paradigm. Emerg son of the relative roles of potential urban chikungunya virus vectors in Infect Dis. 2005;11:425. Australia. PLoS One. 2015;10:e0134975. 54. Hongoh V, Berrang-Ford L, Ogden NH, Lindsay R, Scott ME, Artsob H. A 64. González-Salazar C, Stephens CR, Sánchez-Cordero V. Predicting the review of environmental determinants and risk factors for avian-associ- potential role of non-human hosts in Zika virus maintenance. EcoHealth. ated mosquito arboviruses in Canada. Biodiversity. 2009;10:83–91. 2017;14:171–7. 55. Andreadis TG. The contribution of Culex pipiens complex mosquitoes to 65. Chahad-Ehlers S, Fushita AT, Lacorte GA, de Assis PC, Del Lama SN. transmission and persistence of West Nile virus in North America. J Am Efects of habitat suitability for vectors, environmental factors and host Mosq Cont Assoc. 2012;28:137–51. characteristics on the spatial distribution of the diversity and prevalence 56. Jacups SP, Carter J, Kurucz N, McDonnell J, Whelan PI. Determining of haemosporidians in waterbirds from three Brazilian wetlands. Parasit meteorological drivers of salt marsh mosquito peaks in tropical northern Vectors. 2018;11:276. Australia. J Vect Ecol. 2015;40:277–81. 66. Gard G, Marshall ID, Woodroof GM. Annually recurrent epidemic polyar- 57. Hu W, Mengersen K, Dale P, Tong S. Diference in mosquito species thritis and Ross River virus activity in a coastal area of New South Wales. II. (Diptera: Culicidae) and the transmission of Ross River Virus between Mosquitoes, viruses and wildlife. Am J Trop Med Hyg. 1973;22:551–60. coastline and Inland Areas in Brisbane, Australia. Environ Entomol. 2010;39:88–97. 58. Stephenson EB, Peel AJ, Reid SA, Jansen CC, McCallum H. The non-human reservoirs of Ross River virus: a systematic review of the evidence. Parasit Vectors. 2018;11:188.

Ready to submit your research ? Choose BMC and benefit from:

• fast, convenient online submission • thorough peer review by experienced researchers in your ﬁeld • rapid publication on acceptance • support for research data, including large and complex data types • gold Open Access which fosters wider collaboration and increased citations • maximum visibility for your research: over 100M website views per year

At BMC, research is always in progress.

Learn more biomedcentral.com/submissions

54 3.9 Supplementary Information

Supplementary Table 1: Derivation of 2x2 contingency tables for mosquito feeding preferences Focal host taxon All other hosts Totals

Focal mosquito species a A-a A

All other mosquito species B-a N-A-B+a N-A

Totals B N-B N

a Number of records of the focal mosquito species feeding on the focal host taxon A Total number of records for the focal mosquito species B Total number of records for the focal host taxon N Overall number of records Values in grey shaded cells are obtained by subtraction

55 Supplementary Table 2: Reported blood meal results for Australian mosquito species. Numbers in the columns representing number of blood meals for each vertebrate.

Reference Human Dog Cat Horse Bird Possum Marsupial Flying fox Cattle Pig Rabbit Sheep Goat Fox Rat Total Mosquito species studies Ae aegypti [22] 131 23 2 0 10 - 0 - 0 - 1 - - - 0 174 Ae alboannualtus [31] 2 - - - - - 2 - - - 1 - - - - 5 Ae alternans [31] ------1 ------1 Ae camptorhynchus [23] 5 0 0 - 7 2 - - 83 - - 21 2 2 1 125 [22] Ae lineatopennis [29] 0 1 0 0 0 - 3 - 5 - 0 - - - 0 9 Ae multiplex [22] 0 0 0 0 0 - 1 - 0 - 0 - - - 0 1 [30] Ae normanensis [29] 1 3 0 17 5 - 54 - 21 2 - - - - - 104 [33] [23] [22] Ae notoscriptus [30] 38 63 4 0 32 44 16 4 2 0 0 0 0 0 2 207 [21] [22] Ae procax [21] 3 17 0 1 8 0 5 0 2 - 0 - - - 0 36 Ae queenslandis [31] 5 - - - - - 24 ------29 Ae reesi [29] 0 0 - 0 - - 1 - 1 ------2 Ae subauridorsum [31] 1 ------1 Ae theobaldi [31] 1 - - - 1 - - - - - 2 - - - - 4 [32] [22] [30] Ae vigilax [21] 37 47 3 25 45 12 30 1 37 2 30 - - - 6 278 [31] [29] [33] [21] Ae vittiger [31] 6 2 0 3 1 1 1 0 1 0 - - - - - 15 [33] [30] An amictus [29] 0 3 0 0 0 - 0 - 2 0 - - - - - 5 [23] [22] An annulipes [30] 41 187 18 29 62 0 104 - 276 3 579 0 0 0 0 1332 [31] [29]

56 [30] An bancroftii [29] 28 26 3 2 2 - 12 - 18 1 - - - - - 93 [30] An farauti [29] 0 1 0 0 0 - 2 - 6 0 - - - - - 9 An hilli [30] 1 2 0 0 0 - 0 - 0 0 - - - - - 3 An meraukensis [29] 0 0 - 0 - - 0 - 2 ------2 An novaguinensis [29] 0 0 - 0 - - 0 - 6 ------6 An stigmaticus [31] ------2 ------2 [23] Cq linealis [22] 4 3 0 4 7 1 8 0 1 - 10 0 0 0 2 41 [21] [22] Cq xanthogaster [21] 0 15 1 5 17 0 10 0 7 - 0 - - - 0 55 [29] [23] [32] [34] [22] [24] Cx annulirostris [30] 183 673 42 258 503 57 2884 1 808 347 143 11 0 0 44 6089 [21] [31] [29] [33] [22] Cx australicus [21] 0 1 0 5 17 2 0 0 0 - 0 - - - 0 25 [30] Cx bitaeniorhynchus [29] 0 3 0 0 2 - 6 - 0 0 - - - - - 11 [33] Cx fatigans [31] 54 12 - 15 197 - - - 2 ------280 Cx globocoxitus [23] 2 0 2 - 9 0 - - 1 - - 0 0 0 0 14 Cx halifaxii [22] 0 0 0 0 1 - 0 - 0 - 0 - - - 0 1 Cx hilli [22] 3 1 0 0 0 - 0 - 0 - 0 - - - 0 4 [30] Cx Lophoceramyia sp [29] 10 6 3 0 5 - 0 - 3 0 - - - - - 27 [23] Cx molestus [22] 3 1 0 0 10 0 0 - 3 - 0 0 0 0 0 17 Cx orbostiensis [22] 1 0 0 0 1 - 0 - 1 - 0 - - - 0 4 Cx palpalia [33] 0 2 0 0 1 - 12 - 1 0 - - - - - 16 [22] Cx pullus [29] 1 0 0 0 1 - 2 - 2 - 0 - - - 0 6 [23] Cx quinquefasciatus [22] 398 2417 73 93 1514 2 35 0 49 58 24 0 - 0 0 4746

57 [30] [21] [31] [29] [33] [32] [22] Cx sitiens 6 8 0 3 39 0 13 0 1 2 2 - - - 1 [21] 77 [33] Cx squamosus [30] 0 13 2 2 62 - 2 - 1 0 - - - - - 84 Cx starckeae [30] 0 3 1 0 4 - 2 - 1 0 - - - - - 11 Cx whitmorei [33] 0 0 0 0 0 - 1 - 0 1 - - - - - 2 Ma linealis [31] - - - - 1 ------1 Ma septempunctata [33] 0 0 0 0 1 - 3 - 0 0 - - - - - 4 [22] [30] Ma uniformis [29] 25 4 1 0 2 - 5 - 22 1 0 - - - 0 60 [33] U albescens [30] 0 0 0 0 0 - 1 - 0 0 - - - - - 1 Ve carmenti [22] 1 1 0 0 0 - 4 - 0 - 0 - - - 0 6 [22] Ve funerea [33] 3 0 0 0 0 - 0 - 0 0 0 - - - 0 3 Grand Total 994 3538 155 462 2567 121 3245 22 1366 417 792 32 2 2 56 14044

0 1 2

58 Chapter 4

A systematic review of human seroprevalence for Dengue, Ross

River, and Barmah Forest viruses in Australia and the Pacific

Statement of contribution to co-authored published paper

This chapter is formatted to the specifications of PLoS Neglected Tropical Diseases. The following chapter includes a co-authored paper:

Stephenson, E. B., Madzokere, E. T., Webster, J. A., Harley, D. & Herrero, L. J. (2019) A systematic review of human seroprevalence for Dengue, Ross River, and Barmah Forest viruses in Australia and the Pacific. PLoS Neglected Tropical Diseases. In preparation.

My contribution to the paper involved data collection and analysis collaboratively with Eugene Madzokere and Julie Webster, planning and performing all analyses, and writing the paper.

Signed: ______Date: ______Corresponding author: Eloise Stephenson

Countersigned: ______Date: ______Supervisor: Lara Herrero

59 Thesis chapter statement:

The primary focus of this PhD is on the transmission dynamics of RRV. Rarely, however, do viruses circulate between organisms independently. The primary vectors of RRV are also capable of transmitting dengue and Barmah Forest virus, both of which cause clinical disease in humans. This chapter compiles published seroprevalence reports or RRV, dengue and BFV in humans throughout the distribution of RRV. In doing so, the results inform Chapter 5 through the assessment of serological methods to be used later in this thesis.

60 A systematic review of human seroprevalence for Dengue, Ross River, and Barmah Forest viruses in Australia and the Pacific

Author List Eloise B. Stephensonª1,2, Eugene T. Madzokereª1, Julie A. Webster1,3, David Harley4, and Lara J. Herrero1,5,*

Affiliations

1Institute for Glycomics, Griffith University, Gold Coast Campus, Southport, QLD 4215, Australia; [email protected] (E.T.M); [email protected] (L.J.H). 2Environmental Futures Research Institute, Griffith University, Brisbane, Queensland, 4111, Australia; [email protected] (E.B.S). 3QIMR Berghofer Medical Research Institute, 300 Herston Road, Herston, QLD 4006, Australia; [email protected] (J.A.W). 4Queensland Centre for Intellectual and Developmental Disability, Mater Research Institute— University of Queensland, Brisbane, QLD 4101, Australia; [email protected] (D.H). 5School of Medicine, Griffith University, Gold Coast Campus, Southport, QLD 4215, Australia; ªThese authors have contributed equally to this study and should be considered joint first author *Correspondence: [email protected] (L.J.H); Tel.: +61-7-5552-7858

4.1 Abstract

4.1.1 Introduction

Dengue (DENV), Ross River (RRV) and Barmah Forest viruses (BFV) are the commonest human arbovirus infections in Australia and the Pacific Island Countries and Territories (PICTs), and are associated with debilitating symptoms. All are nationally notifiable in Australia, but routine surveillance is limited to a few areas in the PICTs. Understanding the level of human exposure to these viruses would inform disease management and mitigation strategies. So, we conducted a systematic literature review of all quantitative serosurveys published for these viruses in Australia and the PICTs over a 52-year period (1966-2018).

4.1.2 Methods

61 The Preferred Reporting of Items for Systematic Reviews and Meta-Analyses (Checklist S1: PRISMA Checklist) procedures were adopted to produce a protocol to systematically search for published studies reporting the seroprevalence of DENV, RRV and BFV in Australia and the PICTs. We identified 34 papers. Data for year, location, study population, serosurvey methods, results and study limitations were extracted.

4.1.3 Results

The 34 papers reported 74 serosurveys of DENV, RRV and BFV including 109,440 samples. Seroprevalence varied depending on the assay used and location of the study population. Dengue seroprevalence ranged from 0.16% to 95.59% seropositivity, while RRV and BFV ranged from 0.00% to 73.97%, and 0.28% to 6.47%, respectively. We discuss some of the causes of variation including serological methods used (such as sensitivity and specificity of assays), study populations (including age), and study site.

4.1.4 Conclusion

Human serosurveys provide important information on the extent of human exposure to arboviruses across: 1) time, 2) location, and 3) person (e.g. age and gender). Interpreting results obtained at these scales has the potential to inform us about transmission cycles, improve diagnostic surveillance, and mitigate future outbreaks. Future research should streamline methods and reduce bias to allow a better understanding of the burden of these diseases and the factors associated with seroprevalence.

4.2 Author summary:

Mosquito-borne viruses contribute significantly to the global burden of diseases. Understanding infection rates in human populations is important for informing management. Serosurveys measure population immunity consequent upon arbovirus infections. Here we systematically review serosurveys conducted for three viruses associated with large public health burdens in Australia and the PICTs; dengue virus (DENV), Ross River virus (RRV), and Barmah Forest virus (BFV). We identified 34 studies reporting 74 serosurveys between 1966 and 2018. For DENV we found the highest seroprevalence was reported in the Pacific with surveys in American Samoa, French Polynesia and Papua New Guinea reporting seropositivity greater than 80%. For RRV we found early evidence (before 1975) of circulation outside Australia. There were fewer studies of BFV and these reported low seroprevalence (<6%) and were conducted in Australia. Researchers used different serological methods so

62 study comparisons must be nuanced and it is often not appropriate to pool results from different years and locations. This limits ability to identify risk groups within populations.

Keywords: Exposure, epidemiology, risk factor, public health, transmission dynamics

4.3 INTRODUCTION Arboviruses are considered a global public health priority with more than four billion people at risk (1). Efforts to manage the global spread of arboviruses are hampered by asymptomatic infections, complicated vector-host dynamics, and environmental change (2, 3). This has been most evident through the globally increasing number and severity of arboviral epidemics, including Zika, West Nile, and dengue viruses, over the past 50 years (2, 4). Human serosurveys are an epidemiological tool regularly adopted to better understand and inform the management of arboviruses. When serosurveys are conducted over wide temporal and spatial scales, they can be used to: 1) identify exposure rates and viruses currently circulating within a population, 2) identify susceptible populations and thus potential outbreaks, 3) and distinguish symptomatic from asymptomatic infections. Serosurveys measure population immunity and this is useful for assessing population risk and building predictive transmission models.

Several arboviruses have been reported to circulate in Australia and the Pacific Island Countries (PICTs), including dengue virus (DENV), Ross River virus (RRV) and Barmah Forest virus (BFV). Of these, DENV is associated with the greatest public health threat. Dengue is a single stranded, positive-sense RNA flavivirus. It comprises four serotypes (DENV-1, DENV-2, DENV-3 and DENV-4) with the genome 65% identical, near identical clinical syndromes, and the occupation of the same ecological niche (5). Infection ranges from asymptomatic, to classic dengue fever or more severe forms of disease, such as haemorrhagic fever and dengue shock syndrome (6). Severe DENV cases are associated with high human morbidity and around 20,000 deaths per year globally (7). The cost of DENV disease management in Australia is around AU$17 million annually (8). Recently, there has been an increase in number of dengue cases reported in some of the PICTs, including the Solomon Islands, Vanuatu, Fiji and Palau (9). Dengue transmission is primarily via human-mosquito-human transmission involving Aedes aegypti. Regional variation in epidemiology arises because of host (e.g. behaviour, population age structure), agent (e.g. co-circulation of serotypes, extrinsic incubation period) and environmental (e.g. Aedes aegypti behaviour, built environment, global climate change) factors (10).

63 Both RRV and BFV are single stranded positive-sensed RNA alphaviruses (11). Ross River virus is the most common and widespread arbovirus in Australia, accounting for more than 64% of all human vector-borne infection notifications since 1991 (12). Although RRV and BFV have a nation-wide distribution in Australia, the number of notifications for BFV is approximately half those of RRV, averaging around 2,500 notifications annually (12). Ross River and BFV are zoonotic, maintained in multiple vector and host species (13, 17). Patients infected with RRV or BFV may be asymptomatic; present with mild symptoms such as a fever and rash; or succumb to debilitating polyarthralgia or arthritis (11, 13). Reports of RRV in the PICTs date back to the 1970’s, however until recently, RRV was thought to be endemic only in Australia (14). There is evidence for transmission of RRV outside Australia, including French Polynesia and American Samoa (15, 16). No evidence exists for the circulation of BFV outside Australia.

The significant public health and economic burden strongly justifies assessment of past and current circulation of DENV, RRV and BFV. Consequently, we systematically reviewed serosurveys for these three viruses in Australia and the PICTS. Thorough synthesising these studies, we aim to identify trends in seroprevalence and critique study designs.

4.4 METHODS

4.4.1 Search strategy and selection criteria

We searched Web of Knowledge TM Thomson Reuters, PubMed, Google Scholar and Science Direct for original research articles on the seroprevalence of Barmah Forest (BFV), Dengue (DENV) and Ross River virus (RRV) among humans in Australia and the Pacific Island Countries and Territories (PICTs). Broad search terms were used to maximise the chances of capturing all relevant research articles. The following combinations of keywords were used to search the literature: ‘Ross River virus’, ‘Ross River fever’, ‘Barmah Forest virus’, ‘Dengue virus’, ‘sero-prevalence’, ‘seroprev*’, ‘serosurv*’, ‘serolog*’, ‘sero-epidemiology’, ‘seroepidemiology’, ‘serum’, ‘antigen’, ‘antibod*’, ‘hemagglutination’, ‘hemagglutination inhibition’, ‘IgM’, ‘IgG’, ‘immunoglobulin’, ‘neutralization’, ‘neutralisation’, ‘assay’, ‘enzyme-linked immunosorbent assay’, ‘ELISA’, ‘patholog*’. We used the asterisk * operator as a wildcard to search for all possible variations of a keyword. We also searched reference lists of selected papers to identify additional articles. All the retrieved articles were screened by three authors (EBS, ETM and JAW) based on inclusion and exclusion criteria (Table 1).

64 Table 1. Article selection criteria.

Inclusion criteria Exclusion criteria Quantitative serological survey conducted Case Reports using the following assays: 1) ELISA (IgM or IgG), 2) Neutralisation tests (NT) and, 3) Hemagglutination Inhibition test (HI) Study conducted on humans Study conducted on non-human species

Serological testing for DENV, RRV or BFV The following studies: 1) Modelling studies 2) Genetic analysis of known cases 3) Retrospective analysis of an outbreak 4) Studies focusing only on optimising methods Study undertaken within Australia or the Studies not conducted in geographical PICTs region of interest (i.e. Australia or the PICTs) Studies written in English Studies not conducted on DENV, RRV or BFV

4.4.2 Data extraction For each reference we extracted the study year(s), site, risk factors (such as age or sex), virus(es) investigated, serological method(s) used, number of samples tested, and the number of serologically positive samples. Data were extracted by a single author for a single virus and once complete, all extracted data were cross-checked by a different author. Table S2 shows all the data extracted from each study.

4.4.3 Data analysis

Forest plots were made incorporating serosurvey results (Figures 2a, 2b and 2c) using the method developed by Bailey (18). Results were ordered by the serological method (Hemagglutination inhibition (HI), neutralisation test (NT), enzyme-linked immunosorbent assay (ELISA) IgM and ELISA IgG) used. Studies using multiple methods therefore produced more than one result. Statistical heterogeneity between included studies reporting seroprevalence using different serological methods was determined by calculation of Cochrane’s Q and the I2 score as described by Neyeloff et al (S3 Table) (19). We then used ArcGIS (version 10.4.1), to visualise seroprevalence during 1960 to 1979, 1980 to 1999, and 2000 to 2018.

65 4.5 RESULTS

4.5.1 Reported seroprevalence for DENV, RRV and BFV

Of 1684 studies identified, 34 met our inclusion criteria (Figure 1) and these reported 74 serosurveys. Serosurveys spanned 52 years (1966 – 2018) and included 109,440 human samples. The majority of samples were tested for RRV (n = 52,201), followed by DENV (n = 30,134) and BFV (n = 27,105). Dengue virus had the largest seropositivity range (between 0.16% to 95.59%, Figure 2a). Dengue virus also had the highest median seroprevalence of 23.05% positivity, compared to 9.10% and 1.60% in RRV and BFV, respectively. Studies assessing RRV also reported a large range of seropositivity (0.00% to 73.97%, Figure 2b). However, in contrast to DENV and RRV, the reported seroprevalence range for BFV was small (0.28% to 6.47%, Figure 2c).

Figure 1. Flow diagram of the systematic search strategy used to identify and include studies into our systematic review of reported DENV, RRV and BFV seroprevalence among humans in Australia and the PICTs.

67 There was high variability of reported seroprevalence using the HI, NT, ELISA IgM and ELISA IgG methods for DENV and RRV but low variability for BFV serosurveys across all serological methods (Figure’s 2a-c). For each virus, Cochrane’s Q was greater than the number of degrees of freedom and the I2 score exceeded 75% indicating high variability among studies (Figures 2a-c).

Figure 2a. Forest plot for DENV. Reported seroprevalence within each study (square brackets), categorised by serological method (HI, NT, ELISA IgM and ELISA IgG). For each study; the size of the square represents the total number of samples tested, while the horizontal line indicates the range with the midpoint of each square being the total reported seroprevalence rate. The diamond represents the pooled (“Overall meta”) estimate, whereas the dashed line signifies the overall mean of the pooled estimate and the tips of the diamond on the left and right side represent the lower and upper range limits.

Figure 2b. Forest plot for RRV. Reported seroprevalence within each study (square brackets), categorised by serological method (HI, NT, ELISA IgM and ELISA IgG). For each study; the size of the square represents the total number of samples tested, while the horizontal line indicates the range with the midpoint of each square being the total reported seroprevalence rate. The diamond represents the pooled (“Overall meta”) estimate, whereas the dashed line signifies the overall mean of the pooled estimate and the tips of the diamond on the left and right side represent the lower and upper range limits.

Figure 2c. Forest plot for BFV. Reported seroprevalence within each study (square brackets), categorised by serological method (HI, NT, ELISA IgM and ELISA IgG). For each study; the size of the square represents the total number of samples tested, while the horizontal line indicates the range with the midpoint of each square being the total reported seroprevalence rate. The diamond represents the pooled (“Overall meta”) estimate, whereas the dashed line signifies the overall mean of the pooled estimate and the tips of the diamond on the left and right side represent the lower and upper range limits.

4.5.2 Serosurvey methods adopted

The majority of DENV (n = 7/14) and BFV (n = 7/15) surveys utilised ELISA methods for IgG antibodies (Table 2). For RRV, the HI assay was the most common (15 of 45 studies and 41% of samples) (Table 2). Neutralisation assays were used only in RRV serosurveys (n = 14/45). ELISA (IgM) was used in fewer studies but more individuals were analysed than any other single method (9/74 = 12% and 52129/109440 = 47.63%, respectively).

70 Table 2. Serological methods used to test samples for each virus

Virus ELISA (IgG) ELISA (IgM) HI NT Total

No. of No. of No. of No. of No. of No. of No. of No. of No. of No. of studies samples studies samples studies samples studies samples studies samples

DENV 7 10,233 3 16,256 2 2,723 2 924 14 30,134

RRV 13 8,942 3 18,050 15 21,314 14 3,895 45 52,201

BFV 7 2,839 3 17,825 2 5,550 3 891 15 27,105

Total 27 22,014 9 52,129 19 29,587 19 5,710 74 109,440

Studies that used ELISA IgM produced the smallest variability in seroprevalence (median values = 1.00%, 0.90%, and 1.90% for DENV, RRV and BFV, respectively). A significant number of studies (DENV = 54.50%; RRV = 53.80%; and BFV = 75.00%) reported a seroprevalence below the pooled estimate (“Overall meta” in Figures 2a, 2b and 2c).

4.5.3 Spatiotemporal trends in reported seroprevalence

Dengue virus serosurveys were performed in eight countries plus two Australian states (Figure 3). The majority of human samples tested for DENV originated from Queensland (66.95%, n = 20177/30134). The highest DENV seroprevalence was reported in American Samoa (95.59%, n = 759/794), followed by French Polynesia (76.89%, n = 1055/1372) and Indonesia (69.29%, n = 2216/3198). The lowest DENV seroprevalence was reported in Victoria (4.78%, n = 59/1232), Queensland (6.56%, n = 1324/20177) and the Philippines (9.01%, n=77/854).

Figure 3: The spatiotemporal trends in reported DENV, RRV and BFV seroprevalence in Australia and the PICTs between 1960 and 2018. The figure shows variation in reported seroprevalence within Australia and in individual PICTs (coloured) across different time periods, and notes areas within Australia and the PICTs which were not tested for each virus across these time periods.

72 Ross River virus serosurveys were undertaken in 13 countries, and additionally within Australia each state and territory (with the exception of the Australian Capital Territory (ACT) (Figure 3). One third (n = 17932/52201) of the samples were collected in Queensland. The highest RRV seroprevalence was reported in American Samoa (64.2%, n = 145/226), followed by Papua New Guinea (33.8%, n = 204/604) and New South Wales (29%, n = 3439/11827). All samples originating from Malaysia (n = 219), New Caledonia (n = 72), the Philippines (n = 344), Taiwan (n = 36) and Vanuatu (n = 104) were seronegative for RRV.

Serological surveys for BFV have only been undertaken in Australia, but were reported in every state and territory except the ACT (Figure 3). The majority of samples were collected in Queensland (n = 11686/27105). The Northern Territory reported the highest BFV seroprevalence (3.7%, n = 13/354), followed by New South Wales (2.4%, n = 103/4368). Tasmania and Western Australia had the lowest BFV seroprevalence with only 0.8% (n = 3/355) and 1.2% (n = 8/680) samples testing positive for BFV, respectively.

4.5.4 Reported risk factors

Of the 14 DENV studies, four reported age-specific seroprevalence. Within these, three tested for statistical associations between seroprevalence and age, and two reported a significant positive relationship (Table 3). For RRV, more studies reported age-specific seroprevalence (n =7). However only two studies statistically tested associations with one reporting an increasing seroprevalence with increasing age. Three Barmah Forest virus studies reported age-specific seroprevalence and two statistically tested the relationship, with one finding a positive association (Table 3).

Most studies included age, gender, Indigenous status and occupation. Most studies in which age was measured showed higher seropositivity with increasing age in both males and females. Very few studies collected occupational data.

73 Table 3. Studies reporting risk factors Study (Year) Risk Statistical test Major findings [Reference] factor(s) measured Dengue virus Prayitno et al (2017) Age Odds ratio Seroprevalence increased with age [31] (p<0.0001). Faddy et al (2013) [32] Age Test not specified Seroprevalence increased with age (p<0.05). Ratnam et al (2012) [33] Age Univariate regression Age was not significantly associated with analysis seropositivity. Aubry et al (2018) [35] Age Fishers exact test Seroprevalence among schoolchildren was significantly lower than in the general population. McBride et al (1998) Age Not assessed Seroprevalence increased with age. [34] Ross River virus Faddy et al (2015) [20] Age Odds ratio, Age group 45-54 and >65 had a multivariate logistic significant relationship (p<0.001) with regression analysis IgG seropositivity. Sex Odds ratio, Males were significantly more likely to multivariate logistic be IgG positive (p<0.001) regression analysis Aubry et al (2017) [29] Age Fishers exact test Seroprevalence among schoolchildren was significantly lower than in the general population. Phillips et al (1990) [27] Age Not assessed Seroprevalence increased with age Sex Not assessed Seroprevalence was greater in males Fraser et al (1986) [23] Age Logistic regression Seroprevalence increased with age coefficients (p<0.001) Sex Not assessed Not reported Liehne et al (1976) [21] Age Not assessed Not reported Tesh et al (1975) [22] Age Not assessed Not reported Doherty et al (1968)[24] Age Not assessed Not reported Stallman et al (1976) Age Not assessed Not reported [25] Sex Not assessed Not reported Dodsley et al (2001)[26] Age Not assessed Not reported Doherty (1971) [28] Age Not assessed Not reported Sex Not assessed Not reported Barmah Forest virus Faddy et al (2015) [20] Age Odds ratio, Age group 45-54 and >65 had a multivariate logistic significant relationship (p<0.001) with regression analysis IgG seropositivity. Sex Odds ratio, Males were significantly more likely to multivariate logistic be IgG positive (p<0.001). regression analysis Hawkes et al (1987) Age Test not specified Seroprevalence increased with age [30] (p<0.05). Phillips et al (1990) [27] Age Not assessed Seroprevalence increased with age Sex Not assessed Seroprevalence was greater in males

74 4.6 DISCUSSION Our review of DENV, RRV and BFV human seroprevalence in Australia and the PICTs reveals variability in serological methods, study populations and potential risk factors assessed. Study heterogeneity makes it challenging to estimate human exposure. The findings from this research have practical implications for management of these arboviruses in Australia and the Pacific. Specifically, they can inform public health authorities on the historical and spatial transmission dynamics of DENV, RRV and BFV. Policies that develop from these findings could target areas of high risk, or better inform local and migrating populations of potential transmission risks.

4.6.1 Dengue virus seroprevalence

Globally the highest burden of DENV is in low- and middle-income countries (37, 38). This was also true in the studies reviewed here. Several countries in the PICTs, including Papua New Guinea, Fiji and the Samoan Islands are low income countries. These countries have limited resources and infrastructure, and consequently do not undertake routine active surveillance for DENV (39). This may be why there are no published DENV serosurveys between 1966 and 1979, and between 1980 and 1999 serosurveys were only undertaken in Australia and New Zealand (34, 40). Serosurveys conducted in more recent decades (2000– 2018) reported a higher DENV seroprevalence in the PICTs than in Australia, with as much as 95% and 77% of all samples in American Samoa and the French Polynesia, respectively, testing seropositive (35, 41). Several factors cause the high burden of DENV in the Pacific. Firstly, the abundance of competent anthropophilic vectors (including Aedes albopictus and Aedes aegypti) in the PICTs maintain on-going transmission in human populations. Secondly, resource availability to manage vector populations and diagnose DENV cases is limited. In most of these countries, diagnostic testing is reliant on healthcare practitioners and laboratory staff and the requisite expertise and equipment are not readily available (41). Efforts for surveillance of DENV are on-going in these areas, and while serological surveys can provide a measure for exposure rates of DENV in humans to identify future at risk populations, more surveys are needed in the PICTs.

Geographical aggregation of DENV serosurvey results may be misleading. The time and place at which serosurveys were conducted reflect the epidemiology and ecology of the virus, but are also determined by the research interests, funding and infrastructure. Dengue seroprevalences in Victoria and New Zealand (4.40% and 10.00%, respectively) do not result

75 from authochthonous transmission (33, 40). Study samples comprised returning travellers who had acquired infection elsewhere, most commonly Southeast Asia and the Pacific, where DENV is endemic. Infected travellers potentially pose a threat to local populations. This is particularly important in Northern Queensland where the primary vector of DENV, Ae. aegypti, is well established.

4.6.2 Ross River virus seroprevalence Ross River virus was first isolated in 1954 in Townsville and is considered endemic in Australia (11) consequent upon the presence of competent vectors and hosts (42). However, RRV may have been circulating outside Australia at least since 1975 when human seropositivity between 20 and 34% was reported in Indonesia and Papua New Guinea (22). An outbreak of RRV involved New Caledonia, Fiji, Samoa and the Cook Islands between 1979 and 1980 (43, 44). Rates of infection up to 90% inferred from serosurveys (14) suggested the people were serologically naïve. However, it is possible that RRV was transmitted before 1979 but was only studied and reported during the outbreak.

Our review indicates that transmission is increasing in some countries in this region. Seroprevalence in the Solomon Islands was 0% in 1972 (22, 44) but 43.80% in 1980 (22), and 74% in 2017 (16). Ross River virus can be transmitted by several mosquito vectors and amplified by a number of vertebrate hosts (17). Consequently, it is geographically mobile and has been reported to move into serologically naïve human populations. The clinical signs can appear similar to that of chikungunya (CHIKV) or DENV, and so may have been misdiagnosed. Testing should be performed in areas where CHIKV and DENV occur but RRV is thought not to, such as the Philippines, Taiwan and Malaysia.

4.6.3 Barmah Forest seroprevalence

Barmah Forest virus was first isolated in 1974 (45) and there were no serosurveys until 1986 (46). Cases may have been misdiagnosed as RRV as these viruses share the same geographical distribution, clinical symptoms, vectors, and possibly reservoir hosts (13). The highest seroprevalence was 6%. The causes for rarity relative to RRV are unknown. Infection with one of these alphaviruses may protect against the other (47) so high RRV population immunity may limit the number of BFV infections. Barmah Forest virus serosurveys have been conducted in all states and territories in Australia but not in the PICTs (20, 27, 30, 46, 48-50). Future serological surveys conducted in the PICTs should also include BFV.

4.6.4 Serological assays

76 The gold standard for serological testing is the plaque reduction neutralisation test (PRNT) (51). Although highly specific this is labour intensive, requires skilled personnel, takes several days for a result and requires the handling of live virus within biosafety facilities. Consequently, in lower to middle income countries this assay is rarely available. Alternative methods such as ELISA assays with high sensitivity and lower specificity are more readily available. Neutralisation assays were used in earlier studies but ELISA is more common in more recent years. This may reflect greater accessibility of ELISA assays in low- and middle- income countries.

Generally IgM antibodies are produced immediately after exposure to a pathogen and can only rarely be detected several years post-exposure (52). Immunoglobulin G (IgG) levels rise a few weeks post-infection and can sometimes be detected several years following exposure (52). More than half of the samples in this review were tested for IgM antibodies. This explains the consistently low seroprevalence reported using ELISA IgM assays because outside outbreak events, only a small proportion of the population is likely to be currently or to have been recently infected with DENV, RRV or BFV. The primary aim of studies investigating the IgM seroprevalence of these arboviruses was to quantify the current risk of infecting people through blood transfusions from donor blood supplies (20, 32, 53).

4.6.5 Risk factors

A number of social and demographic variables are associated with arboviral infections. Social- economic variables, including housing and income, are also important. People in developed countries are more likely than others to have piped water, insect screening and air-conditioning, which reduce breeding of and contact with disease vectors (64). These factors, however, are highly dependent on the study area and the disease system. We note a higher reported seroprevalence of DENV in low- or middle-income countries, which may be attributed to housing conditions or finite medical resources [36]. For RRV and BFV, however, this pattern is not observed. Across all studies reviewed here, none collected data on socio-economic variables, and thus no conclusions can be drawn on a finer, local scale.

Despite age being importantly associated with arboviral disease, few studies reviewed statistically tested the relationship between age and seroprevalence. Contradictory significant

77 positive and negative associations were reported for each virus. This might be because the reported serosurveys include biased samples. Several of the studies here used opportunistic sampling (such as blood donors or volunteers) [20, 32], which are not representative of age groups and sexes within the population. More representative sampling for serosurveys would ameliorate this. Limitations This review aimed to summarise all published research documenting human seroprevalence of RRV, DENV and BFV in Australia and the PICTs. One key limitation of this is that a lack of published seroprevalence reports does not indicate that one of these viruses does not circulate in each region or at a given time period. As such the aggregated findings of this literature review are limited to identify long-term spatial and temporal trends for the circulation of these viruses, but valuably highlights the spatial and temporal trends in diagnostic testing.

4.7 CONCLUSION Serosurveys are an important tool to measure human exposure to arboviruses. They are more powerful when standardised and conducted using highly sensitive and specific diagnostic methods. This review aimed to identify epidemiologically relevant trends, potential biases and risk factors associated with DENV, RRV and BFV seropositivity in Australia and the PICTs. There is variability in reported seroprevalence, study designs, and methods adopted for DENV, RRV and BFV seropositivity. Most studies did not conduct statistical analyses to determine the significance of risk factors, or the accuracy of the method used to detect seroconversion. To be able to rationally compare studies and therefore reach statistically significant results in a meta- analysis, we believe future serosurveys conducted in Australia and the PICTs should be standardised. The emergence of medically important arboviruses in new geographical areas (e.g Japanese Encephalitis virus in India, West Nile virus in North America, and Zika virus in Brazil [65]) emphasises the importance of building knowledge of the transmission and maintenance of these viruses. Serosurveys provide important data to this end. Knowledge of DENV, RRV and BFV transmission is important for understanding these viruses in Australia and the PICTs and will translate to a better understanding of other arboviruses.

Acknowledgements

We would like to broadly acknowledge the LJH’s lab members for their guidance and revisions of this manuscript. We would also like to acknowledge Hamish McCallum, Alison Peel, Simon

78 Reid and Cassie Jansen for being part of a supervisory team, for which this research contributes to. Funding LJH is the recipient of the Australian National Health and Medical Research Council Career Development Award (ID: 105760). EBS is the recipient of an Australian Government Research Training Program Author Contributions Conceptualisation: LJH, ETM, EBS, JAW, DH. Data curation: ETM, EBS, and JAW. Formal analysis: ETM and EBS. Funding acquisition: LJH. Investigation: ETM, and EBS. Methodology: ETM and EBS. Project administration: LJH and DH. Supervision: LJH. Visualisation: ETM and EBS. Writing – original draft: ETM and EBS. Writing review and editing: ETM, EBS, JAW, DH, and LJH.

4.8 REFERENCES

1. Shaw WR and Catteruccia F. Vector biology meets disease control: using basic research to Fight vector-borne diseases. Nature Microbiology, 2018; p. 1. 2. Weaver SC. Prediction and prevention of urban arbovirus epidemics: A challenge for the global virology community. Antiviral Research, 2018; 156: p. 80 - 84. 3. Wilder-Smith A, Gubler DJ, Weaver SC, Monath TP, Heymann DL, Scott TW. Epidemic arboviral diseases: priorities for research and public health. The Lancet infectious diseases. The Lancet Infectious Diseases, 2017; 17(3): p. e101 - e106. 4. Calvez E, Mousson L, Vazeille M, O'Connor O, Cao-Lormeau VM, Mathieu-Daude F, et al. Zika virus outbreak in the Pacific: Vector competence of regional vectors. PLoS Neglected Tropical Diseases, 2018; 12(7): p. e0006637. 5. Halstead SB. Dengue virus–mosquito interactions. Annual Review of Entomology, 2008; 53: p. 273 - 291. 6. Bandyopadhyay S, Lum LC, Kroeger A. Classifying dengue: a review of the difficulties in using the WHO case classification for dengue haemorrhagic fever. Tropical Medicine & International Health, 2006; 11(8): p. 1238 - 1255. 7. Fitzpatrick C, Haines A, Bangert M, Farlow A, Hemingway J, Velayudhan R. An economic evaluation of vector control in the age of a dengue vaccine. PLoS Neglected Tropical Diseases, 2017; 11(8): p. e0005785.

79 8. Canyon DV. Historical analysis of the economic cost of dengue in Australia. Journal of Vector Borne Diseases, 2008; 45: p. 245 - 248. 9. International Federation of Red Cross and Red Crescent Societies. Solomon Islands: Dengue outbreak Emergency Plan of Action (EPoA) DREF Operation MDRSB005, 2016. 10. Akter R, Naish S, Hu W, Tong S. Socio-demographic, ecological factors and dengue infection trends in Australia. PLoS One, 2017; 12(10): p. e0185551. 11. Harley D, Sleigh A, Ritchie S., Ross River virus transmission, infection, and disease: a cross-disciplinary review. Clinical Microbiology Reviews, 2001; 14(4): p. 909 - 932. 12. Government of Australia. National Notifiable Disease Surveillance System. 2018. 13. Jacups SP, Whelan PI, Currie B. Ross River virus and Barmah Forest virus infections: A review of history, ecology, and predictive models, with implications for tropical northern Australia. Vector-Borne and Zoonotic Diseases, 2008; 8(2): p. 283 - 297. 14. Aaskov JG, Mataika JU, Lawrence GW, Rabukawaga V, Tucker MM, Miles JA et al. An epidemic of Ross River virus infection in Fiji. American Journal of Tropical Medicine and Hygiene, 1981; 30(5): p. 1053 - 1059. 15. Aubry M, Finke J, Teissier A, Roche C, Broult J, Paulous S et al. Silent Circulation of Ross River Virus in French Polynesia. International Journal of Infectious Diseases, 2015; 37: p. 19 - 24. 16. Lau C, Aubry M, Musso D, Teissier A, Paulous S, Despres P et al. New evidence for endemic circulation of Ross River virus in the Pacific Islands and the potential for emergence. International Journal of Infectious Diseases, 2017; 57: p. 73 - 76. 17. Stephenson EB, Peel AJ, Reid SA, Jansen CC, McCallum H. The non-human reservoirs of Ross River virus: a systematic review of the evidence. Parasites & Vectors, 2018; 11(1): p. 188. 18. Bailey TM. ForestPlot Tool version 504 (MS Excel workbook). 2009. 19. Neyeloff JL, Fuchs S, Moreira LB. Meta-analyses and Forest plots using a Microsoft excel spreadsheet: a step-by-step guide focusing on descriptive data analysis. BioMedCentral Research Notes, 2012; 5(52): p. 1 - 6. 20. Faddy H, Dunford M, Seed C, Olds A, Harley D, Dean M et al. Seroprevalence of Antibodies to Ross River and Barmah Forest Viruses: Possible Implications for Blood Transfusion Safety After Extreme Weather Events. Ecohealth, 2015; 12(2): p. 347 - 353. 21. Liehne CG, Stanley NF, Alpers MP, Paul S, Leihne PF and CHAN KH. Ord River Arboviruses - Serological Epidemiological. Australian Journal of Experimental Biology and Medical Science, 1976; 54(5): p. 505 - 512. 22. Tesh RB, Gajdusek DC, Garruto RM, Cross JH, Rosen L. The distribution and prevalence of group A arbovirus neutralizing antibodies among human populations in Southeast Asia and the Pacific islands. American Journal of Tropical Medicine and Hygiene, 1975; 24(4): p. 664 - 675. 23. Fraser JR, Christie DG, Gust ID, White J, Leach R, Macaulay ED et al. Arbovirus Infection in a Murray valley Comunity. Australian and New Zealand Journal of Medicine, 1986; 16: p. 52 - 57. 24. Doherty RL, Standfast HA, Wetters EJ, Whitehead RH, Barrow GJ, Gorman BM. Virus iolation and Serological Studies of Arthropod-borne virus infectios in a high rainfall area of North Queensland. Transactions of the Royal Society of Tropical Medicine and Hygiene, 1968; 62(6): p. 862 - 867. 25. Stallman ND, Wiemers MA, Bourke ATC. Serology in diagnosis and surveys of man. 1976. 26. Dodsley NA, Broom AK, Smith DW, Plant AJ, Lindsay MD. Ross River virus: Determining the Prevelance in the South West of Western Australia. Arbovirus Research in Australia, 2001; 8: p. 122 - 125.

80 27. Phillips DA, Murray JR, Aaskov JG, Wiemers M. Clinical and subclinical Barmah Forest virus infection in Queensland. Medical Journal of Australia, 1990; 152: p. 463 - 466. 28. Doherty RL. Surveys of Haemagglutination-Inhibiting Antibody to Arboviruses in Aborigines and Other Population Groups in Northern and Eastern Australia, 1966–1971. Transactions of the Royal Society of Tropical Medicine and Hygiene, 1973; 67(2): p. 197 - 205. 29. Aubry M, Teissier A, Huart M, Merceron S, Vanhomwegen J, Roche C et al. Ross River Virus Seroprevalence, French Polynesia, 2014 - 2015. Emerging Infectious Diseases, 2017; 23(10): p. 1751 - 1753. 30. Hawkes RA, Boughton CR, Naim HM, Myrick BA, Ramsay LG. Barmah Forest virus infections in humans in New South Wales. The Medical Journal of Australia, 1987; 146: p. 569 - 573. 31. Prayitno A, Taurel A-F, Nealon J, Satari HI, Karyanti MR, Soedjatmiko S et al. Dengue seoprevalence and force of primary infection in a representative population of urban dwelling Indonesian children. PLoS Neglected Tropical Diseases, 2017; 11(6): p. e0005621. 32. Faddy HM, Seed CR, Fryk JJ, Hyland CA, Ritchie SA, Taylor CT et al.,, Implications of Dengue Outbreaks for Blood Supply, Australia. Emerging Infectious Diseases, 2013; 19(5): p. 787 - 789. 33. Ratnam I, Black J, Leder K, Biggs B-A, Matchett E, Padiglione A et al.,, Incidence and seroprevalence of dengue virus infections in Australian travellers to Asia. European Journal of Clinical Microbiology and Infectious Disease, 2012; 31: p. 1203 - 1210. 34. McBride WJH, Mullner H, LaBrooy JT, Wronski I. The 1993 Dengue 2 Epidemic In North Queensland: A Serosurvey and Comparison of Hemagglutination Inhibition with an ELISA. American Journal of tropical Medicine and Hygiene, 1998; 59(3): p. 457 - 461. 35. Aubry M, Teissier A, Huart M, Merceron S, Vanhomwegen J, Mapotoeke M et al. Seroprevalence of Dengue and Chikungunya Virus Antibodies, French Polynesia, 2014-2015. Emerging Infectious Diseases, 2018; 24(3): p. 558 - 561. 36. World Health Organization. Dengue and severe dengue. 2014. 37. Bhatt S, Gething PW, Brady OJ, Messina JP, Farlow AW, Moyes CL et al. The global distribution and burden of dengue. Nature, 2013; 496(7446): p. 504. 38. Shepard DS, Undurraga EA, Halasa YA, Stanaway JD. The global economic burden of dengue: a systematic analysis. The Lancet Infectious Diseases, 2016; 16(8): p. 935 - 941. 39. Beatty ME, Stone A, Fitzsimons DW, Hanna JN, Lam SK, Vong S et al. Best practices in dengue surveillance: a report from the Asia-Pacific and Americas Dengue Prevention Boards. PLoS Neglected Tropical Diseases, 2010; 4(11): p. e890. 40. Maguire T. Do Ross River and dengue viruses pose a threat to New Zealand? The New Zealand Medical Journal 1994; 107(989): p. 448 - 450. 41. Duncombe J, Lau C, Weinstein P, Aaskov J, Rourke M, Grant R et al. Seroprevalence of Dengue in American Samoa, 2010. Emerging Infectious Diseases, 2013; 19(2): p. 324 - 326. 42. Claflin SB and Webb C. Ross River Virus: Many Vectors and Unusual Hosts Make for an Unpredictable Pathogen. PLoS Pathogens, 2015; 11(9): p. e1005070. 43. Rosen L, Gubler DJ, Bennett PH.,, Epidemic polyarthritis (Ross River) virus - infection in the Cook Islands. American Journal of Tropical Medicine and Hygiene, 1981; 30(6): p. 1294 - 1302. 44. Tesh RB, McLean RG, Shroyer DA, Calisher CH, Rosen L. Ross River virus (Togaviridae, Alphavirus) infection (epidemic polyarthritis) in American Samoa. Transactions of the Royal Society of Tropical Medicine and Hygiene, 1981; 75(3): p. 426 - 431. 45. Marshall ID, Woodroofe GM, Hirsch S.,, Viruses recovered from mosquitoes and wildlife serum collected in the Murray Valley of Southeastern Australia February 1974 during

81 an epidemic of encephalitis. Australian Journal of Experimental Biology and Medical Science, 1982; 60(5): p. 456 - 470. 46. Vale TG, Carter IW, McPhie KA, James GS, Cloonan MJ.,, Human Arbovirus Infections along the Eastern South Coast of New South Wales. Australian Journal of Experimental Biology and Medical Science, 1986; 64(3): p. 307 - 309. 47. Partidos CD, Paykel J, Weger J, Borland EM, Powers AM, Seymour R et al.,, Cross- protective immunity against o'nyong-nyong virus afforded by a novel recombinant chikungunya vaccine. Vaccine, 2012; 30: p. 4638 - 4643. 48. Humphery-Smith I, Cybinksi DH, Byrnes KA, St George TD.,, Seroepidemiology of arboviruses among seabirds and island residents of the Great barrier Reef and Coral Sea. Epidemiology and Infection, 1991; 107: p. 435 - 440. 49. Dean MM, Dunford M, Flower RL, Faddy HM.,, Biological markers of infection may assist in the identification of early stage viral infection and serve as a surrogate biomarker to identify asymptomatic Ross River or Barmah Forest virus infection. The Journal of Immunology, 2016; 196(1): p. 124 - 136. 50. Johansen CA, Mackenzie JS, Smith DW, Lindsay MDA.,, Prevalence of neutralising antibodies to Barmah Forest, Sindbis, and Trubanaman viruses in animals and humans in the south-west of Western Australia. Australian Journal of Zoology, 2005; 53: p. 51 - 58. 51. Hobson-Peters J. Approaches for the development of rapid serological assays for surveillance and diagnosis of infections caused by zoonotic flaviviruses of the Japanese encephalitis virus serocomplex. Journal of Biomedicine and Biotechnology, 2012; 2012: p. 379738. 52. Smith DW, Speers DJ, Mackenzie JS. The viruses of Australia and the risk to tourists. Travel Medicine and Infectious Disease, 2011; 9(3): p. 113 - 125. 53. Faddy Tran TV, Hoad VC, Seed CR, Viennet E, Chan HT et al. Ross River virus in Australian blood donors: possible implications for blood transfusion safety. Transfusion, 2018; 58(2): p. 485 - 492. 54. Alera MT, Srikiantkhachorn A, Velasco JM, Tac-An IA, Lago CB, Clapham HE et al. Incidence of Dengue Virus Infection in Adults and Children in a Prospective Longitudinal Cohort in the Philippines. PLoS Neglected Tropical Diseases, 2016; 10(2). 55. Luang-Suarkia D, Ernst T, Alpers MP, Garruto R, Smith D, Imrie A. Serological evidence for transmission of multiple dengue virus serotypes in papua New Guinea and West Papua prior to 1963. PLoS Neglected Tropical Diseases, 2017; 11(4): p. e0005488. 56. Darcy A, Clothier H, Phillips D, Bakote'e B, Stewart T. Solomon Islands dengue seroprevalence study - previous circulation of dengue confirmed. PNG Mdeical Journal, 2001; 44(1): p. 43 - 47. 57. Clarke JA, Marshall ID, Gard G. Annually Recurrent Epidemic Polyarthritis and Ross River virus Activity in A Coastal Area of New South Wales I. Occurrence of the Disease*. The American Journal of tropical Medicine and Hygiene 1973. 22(4): p. 543 - 550. 58. Doherty RL, Gorman BM, Whitehead RH, Carley JG. Studies of Arthropod-borne Infections in Queensland. Australian Journal of Experimental Biology and Medical Science, 1966; 44: p. 365 - 378. 59. Hawkes RA, Pamplin J, Boughton CR, Naim HM. Arbovirus infections of humans in high-risk areas of south-eastern Australia: a continuing study. Medical Journal of Australia, 1993; 159(3): p. 159 - 162. 60. Aubry M, Finke J, Teissier A, Roche C, Broult J, Paulous S et al. Seroprevalence of arboviruses among blood donors in French Polynesia, 2011-2013. International Journal of Infectious Diseases, 2015; 41: p. 11 - 12.

82 61. Campbell J, Aldred J, Davis G. Some aspects of the natural history of Ross River virus in South East Gippsland, Victoria. Arbovirus Research in Australia - Proceedings 5th Symposium, 1989; p. 24 - 28. 62. Cloonan MJ, O'Neill BJ, Vale TG, Carter IW, Williams JE. Ross River Virus Activity Along the South Coast of New South Wales. Australian Journal of Experimental Biology and Medical Science, 1982; 60(6): p. 701 - 706. 63. Weinstein P, Worswick D, McIntyre A, Cameron S. Human sentinels for arbovirus surveillance and regional risk classification in South Australia. Medical Journal of Australia, 1994; 160(8): p. 494 - 499. 64. World Health Organisation. A global brief on vector-borne diseases. No. WHO/DCO/WHD/2014.1). World Health Organization, 2014. 65. Gubler DJ. The global emergence/resurgence of arboviral diseases as public health problems. Archives of medical research. 2002; 33(4):330-42.

83 4.9 Supplementary Information Supplementary Table 1. Summary of data extracted from publications

Author Year Virus Location Method Event Outcom Low_rate High_rate Sample size Age_categor (Year) s e (need y to divide by 100) Aaskov 1981 RRV QLD NT 139 45 45 45 308 N (1981) 1981 RRV QLD HI 76 22 22 22 345 N Alera (2016) 2016 DENV PHILIPP HI 77 9.01639 4.39 15.76 854 Y Aubry (2018) 2018 DENV FR_POLY ELISA(IgG 1055 76.895 60 83 1372 Y ) Aubry (2017) 2017 RRV FR_POLY ELISA(IgG 197 14.3586 1.26 48.98 1372 N ) Aubry (2015) 2015 RRV FR_POLY ELISA(IgG 214 36.0877 21.43 42.42 593 N ) Campbell 1989 RRV VIC ELISA(IgG 130 24.8566 24.856596 24.85659656 523 N (1989) ) 6 Clarke (1973) 1973 RRV NSW HI 215 16.8232 15.24 24.5 1278 N Cloonan 1982 RRV NSW ELISA(IgG 31 6.62393 5 14.28 468 N (1982) ) Darcy (2001) 2001 DENV SOLO ELISA(IgG 186 36.1165 36.116504 36.11650485 515 ) 9 Dean (2016) 2016 BFV UNK ELISA(IgM 168 2.4 2.4 2.4 7001 N ) 2016 RRV UNK ELISA(IgM 161 2.3 2.3 2.3 7001 N ) Dodsley 2001 RRV WA HI 58 7.2 3.2 8.4 806 Y (2001) Doherty 1968 RRV QLD HI 128 45.0704 10.5 76.47 284 Y (1968) 1966 RRV QLD HI 317 49.921 19.23 63.95 636 Y

84 Doherty 1966 RRV QLD NT 139 45.1299 16.95 58.82 308 Y (1966) Doherty 1973 RRV QLD HI 614 41.9686 7.05 64.7 1463 Y (1973) 1973 RRV NT HI 98 55.0562 40 90 178 Y Duncombe 2013 DENV AMERICA_SAMO ELISA(IgG 759 95.5919 89.1 99.5 794 Y (2013) A ) Faddy (2013) 2013 DENV QLD ELISA(IgM 25 0.16151 0.08 0.33 15479 N ) 2013 DENV QLD ELISA(IgG 292 9.43152 7.18 11.48 3096 Y ) 2013 DENV VIC ELISA(IgG 31 6.78337 4.4 7.44 457 Y ) Faddy (2015) 2015 BFV QLD ELISA(IgM 93 1.04718 0.21 2.88 8881 Y ) 2015 BFV VIC ELISA(IgM 36 1.8528 1.81 1.9 1943 Y ) 2015 BFV TAS ELISA(IgG 3 0.84507 0 1.8 355 Y ) 2015 BFV SA ELISA(IgG 5 1.39276 0.18 2.61 359 Y ) 2015 BFV QLD ELISA(IgG 9 1.29683 1.13 1.47 694 Y ) 2015 BFV NT ELISA(IgG 13 3.67232 1.71 5.63 354 Y ) 2015 BFV VIC ELISA(IgG 1 0.27855 0 0.82 359 Y ) 2015 BFV WA ELISA(IgG 5 1.39665 0.18 2.61 358 Y ) 2015 BFV NSW ELISA(IgG 11 3.05556 1.28 4.83 360 Y ) 2015 RRV QLD ELISA(IgM 74 0.81265 0.21 1.66 9106 Y ) 2015 RRV VIC ELISA(IgM 18 0.9264 0.9 0.95 1943 Y )

85 2015 RRV TAS ELISA(IgG 19 5.35211 3.01 7.69 355 Y ) 2015 RRV SA ELISA(IgG 13 3.62117 1.69 5.55 359 Y ) 2015 RRV QLD ELISA(IgG 123 17.7233 13.8 21.83 694 Y ) 2015 RRV NT ELISA(IgG 54 15.2542 11.51 19 354 Y ) 2015 RRV VIC ELISA(IgG 3 0.83565 0 1.78 359 Y ) 2015 RRV WA ELISA(IgG 10 2.7933 1.09 4.5 358 Y ) 2015 RRV NSW ELISA(IgG 16 4.44444 2.32 6.57 360 Y ) Fraser (1986) 1986 RRV VIC HI 123 16.6441 3 29.5 739 Y Hawkes 1987 BFV NSW HI 70 1.9774 0.3 6 3540 Y (1987) Hawkes 1993 RRV NSW HI 1028 39 39 39 2635 N (1993) Humphrey- 1991 BFV QLD NT 2 1.9802 1.9801980 1.98019802 101 N Smith (1991) 2 1991 RRV QLD NT 9 8.91089 8.9108910 8.910891089 101 N 9 Johansen 2005 BFV WA NT 3 0.93168 0.9316770 0.931677019 322 N (2005) 2 Lau (2017) 2017 RRV AMERICA_SAMO ELISA(IgG 145 73.9796 45 80 196 Y A ) Liehne (1976) 1976 RRV WA HI 181 41.0431 20.6 77.9 441 Y Luang- 2013 DENV PNG NT 59 47.9675 40 80 123 N Suarkia (2013) Maguire 1994 DENV NZ HI 187 10 10 10 1869 N (1994) 1994 RRV NZ HI 79 43 43 43 183 N

86 McBride 1998 DENV QLD ELISA(IgG 617 77.0287 77.028714 77.02871411 801 Y (1998) ) 1 1998 DENV QLD NT 390 48.6891 48.689138 48.68913858 801 Y 6 Phillips 1990 BFV QLD HI 130 6.46766 0 18.6 2010 Y (1990) 1990 RRV QLD HI 636 31.6418 3.2 62.5 2010 Y Prayitno 2013 DENV INDO ELISA(IgG 2216 69.2933 69.293308 69.29330832 3198 Y (2013) ) 3 Ratnam 2011 DENV VIC ELISA(IgM 24 6.20155 6.2015503 6.201550388 387 (2011) ) 9 2011 DENV VIC ELISA(IgM 4 1.03093 1.0309278 1.030927835 388 ) 4 Stallman 1976 RRV QLD HI 206 7.69518 4.26 12.5 2677 Y (1976) 1976 RRV NT HI 8 1.44665 4.55 22.58 553 Y Tesh (1975) 1975 RRV TAIWAN NT 0 0 0 0 36 Y 1975 RRV SOUTH_VIETNA NT 1 0.76923 0 0.769230769 130 Y M 1975 RRV MALAYSIA NT 0 0 0 0 219 Y 1975 RRV PHILIPP NT 0 0 0 0 240 Y 1975 RRV INDO NT 225 21.7391 0 95 1035 Y 1975 RRV PNG NT 204 33.7748 0 64 604 Y 1975 RRV SOLO NT 40 6.62252 0 17 604 Y 1975 RRV VANUATU NT 0 0 0 0 104 Y 1975 RRV NCA NT 0 0 0 0 72 Y 1975 RRV AMERICA_SAMO NT 0 0 0 0 30 Y A 1975 RRV PHILIPP NT 0 0 0 0 104 Y

Vale (1986) 1986 BFV NSW NT 22 4.70085 2 8 468 Weinstein 1994 RRV SA ELISA(IgG 236 8 8 8 2951 N (1994) )

Chapter 5

Species traits and hotspots associated with Ross River virus infection in non-human vertebrates in South-East Queensland

Statement of contribution to co-authored published paper

This chapter is formatted to the specifications of Transactions of the Royal Society of Tropical Medicine and Hygiene. This chapter has been submitted as is currently under review. The following chapter includes a co-authored paper:

Stephenson, E. B., Rudd, P. A., Peel, A. J., McCallum, H., Reid, S. A. & Herrero, L. J. (2019) Species traits and hotspots associated with Ross River virus infection in non-human vertebrates in South-East Queensland. Transactions of the Royal Society of Tropical Medicine and Hygiene. Under Review.

My contribution to the paper involved undertaking serological analyses collaboratively with Penny Rudd, planning and performing all analyses, and writing the paper.

Signed: ______Date: ______Corresponding author: Eloise Stephenson

Countersigned: ______Date: ______

Supervisor: Lara Herrero

Species traits and hotspots associated with Ross River virus infection in non-human vertebrates in South-East Queensland

Eloise B. Stephensona,b*, Penny A. Ruddb, Alison J. Peela, Hamish McCalluma, Simon A.

Reidc, and Lara J. Herrerob,d*

Affiliations aEnvironmental Futures Research Institute, Griffith University, Brisbane, Queensland, 4111, Australia bInstitute for Glycomics, Griffith University, Gold Coast Campus, Southport, QLD 4215, cThe University of Queensland, School of Public Health, Herston, Brisbane, Queensland 4006,

Australia dRedlands Hospital, QLD Health, Weippin Street, Cleveland 4163, Australia

*corresponding authors – Tel: +61405162000, E-mail: [email protected]; [email protected].

5.1 Abstract:

Background: Ross River virus (RRV) is a mosquito-borne zoonotic arbovirus associated with high public health and economic burdens across Australia, but particularly in South-East

Queensland. Despite this high burden, humans are considered incidental hosts. Transmission of RRV is maintained among mosquitoes and many non-human vertebrate reservoir hosts, though the relative contributions of each of these hosts is unclear.

Methods: To clarify the importance of a range of vertebrates in RRV transmission in South-

East Queensland, a total of 595 serum samples from 31 species were examined for RRV exposure using a gold-standard plaque reduction neutralisation test. Data was analysed statistically using a random forest analysis and a coefficient inference tree, and spatially.

Results: RRV exposure was highly variable between and within species groups. Critically, species group (‘placental mammal’, ‘marsupial’ and ‘bird’) which has previously been used as a proxy for reservoir hosts, was a poor correlate for exposure. Instead, we found that ‘diet’,

‘body mass’ and ‘longevity’ species traits were most important in determining seropositivity.

We also identified significant differences in seropositivity between the two major possum species (ringtail possums and brushtail possums), which are ecologically and taxonomically different. Lastly, we identified distinct hotspots and coldspots of seropositivity in non-human vertebrates which correlated with human notification data.

Conclusion: This is the largest diversity of species tested for RRV in a single study to date. The analysis methods within this study provide a framework for analysing serological data in combination with species traits for other zoonotic disease, but more specifically for RRV highlights areas to target further public health research and surveillance effort.

Keywords: arbovirus; exposure; immunology; random forest regression; reservoir; seroprevalence.

5.2 Introduction Ross River virus (RRV) accounts for the greatest number of human arboviral infections in

Australia (1). It is an alphavirus of the Togaviridae family, and is associated with clinical symptoms including polyarthritis, rash, fever and myalgia (2). At present there is no vaccine or treatment available for RRV and management of individual human cases focuses on symptomatic care (2). As RRV is a zoonotic arbovirus that is maintained in complex transmission dynamics between multiple vectors and non-human vertebrate hosts, current vector-centric efforts to reduce RRV transmission would be enhanced by also considering the roles of environmental drivers and reservoir communities in transmission (3).

The vectors of RRV are diverse: the virus has been isolated from more than 40 species of mosquitoes to date (2, 4). These species are predominantly of the Aedes and Culex genera and persist in a range of climates and habitats, including freshwater and saltwater environments.

The different vertebrate feeding patterns exhibited by these mosquitoes may also alter their importance as vectors (5). A recent review of the non-human reservoirs of RRV found that a broad range of vertebrate species may play a role in the maintenance and transmission of the virus, including marsupials, birds, horses and flying foxes (6). Along with reports of RRV transmission in humans in the Pacific Island Countries, in the absence of marsupial hosts (7-

9), these studies challenge the long held dogma that marsupials are more important reservoirs than placental mammals, which in turn are more important than birds.

For other zoonotic arboviruses, including West Nile virus (WNV), Sindbis virus and Murray

Valley encephalitis virus, serosurveys have been informative in understanding transmission among non-human vertebrate species (10-13). Serosurveys provide evidence of past exposure by measuring antibodies in blood, which are easier to detect and typically longer lasting than the infectious agent itself (14). Ross River virus serosurveys in non-human vertebrates have previously targeted marsupial populations due to them being considered the primary vertebrate

reservoir. To date, other potentially important reservoir hosts, such as birds and flying foxes have been greatly underrepresented in the literature (6).

Serosurveys are most informative when combined with additional data, such as spatial, temporal or species trait information. This can be challenging in free-living populations as it requires repeated sampling from the same individuals or populations (14). A recent meta- analysis took a novel approach to interpreting RRV serology by assessing life-history traits for any mammalian species that had been reported in the literature to have previous exposure to

RRV (15). The findings of that study were that diet range, population density and length of gestation positively correlated with exposure to RRV. The addition of life-history traits to serological assessments provides informative data which can be used to in management and prevention strategies.

This study aims to identify putative non-human vertebrate hosts of RRV in South-East

Queensland (SEQ), and ecological and spatial determinates of seroprevalence. The majority of human notifications for RRV originate in this region (1). This, in combination with its subtropical climate (promoting a diversity of habitats and species) makes it an ideal location to study the disease ecology of a multi-host vector borne disease.

5.3 Materials and methods

5.3.1 Sample collection and storage

The cross-sectional study included wildlife and domestic horse populations in SEQ from

February 2017 to June 2018 and was performed in accordance with Griffith University Animal

Ethics Committee standards (approval number: ENV/04/17/AEC). Clinics were largely recruited through the Australian Veterinary Association and Wildlife Health Australia. Serum samples were collected from horses of consenting owners and wildlife admitted for veterinary care. This study excluded horses with clinical symptoms of RRV (pyrexia, ataxia, stiffness,

lameness, swollen joints, reluctance to move, lethargy, inappetence, mild colic and ‘poor performance’) up to one year prior to admission. Each sample was accompanied with the following information: species, sex, age (either number of years for horses, or age class

(juvenile, sub-adult, adult, mature-adult) for wildlife), location animal was found/housed, reason for veterinary attention and any additional relevant notes (such as body condition score).

For horses, samples were collected using a 19 gauge needle from the jugular vein, into a plain tube and allowed to coagulate at ambient temperature. For wildlife, samples were collected in serum separator tubes using a 22-27 gauge needle from either the jugular vein, caudal vein or medial metatarsal vein, depending on the species and the individual animal. A minimum of 400 uL of whole blood was collected from each individual, unless this exceeded 10 % of their body weight, in which case no more than 10 % of whole blood was collected. Whole blood samples were labelled, centrifuged at 5514 g for 6 minutes and stored at -20 C until analysis.

5.3.2 Sample analysis

Vero cells (ATTC # CCL-81) were seeded at 4 x 105 cells/mL in with 1 mL added in each well for 12-well plates and cultured at 37°C for 24 hours. Serum samples were inactivated for 30 minutes in a 56°C water-bath and diluted 1:10 in serum-free Opti-MEM media (Gibco). Ross

River virus (T48) was diluted to 350 PFU in 500uL and mixed with an equal volume of diluted serum and incubated for 1 hour at 37°C. A volume of 400 ul virus/serum mixture (giving a final virus concentration as 140 PFU per well) was then inoculated onto the vero cells after removing the culture media and incubated for an additional 2 hours at 37°C with gentle shaking every 10-15 minutes. The mixture was removed and 1ml of a 1 % agarose overlay containing

1 % penicillin/streptomycin (Gibco) and 2 % foetal bovine serum (FBS) was added to each well. Plates were incubated for 48-60 hours at 37°C after which, the agarose overlay was removed, and cells were stained with 0.1 % crystal violet solution (in 20 % EtOH). Plates were

washed with tap water, dried, and the plaques were counted. Samples were considered to have neutralising antibodies if the test serum sample had at least 50 % less plaques than the virus- only control wells.

5.3.3 Species trait data

A dataset of ecological and life-history traits potentially relevant for RRV exposure was compiled for each of the vertebrates (Supplementary Table 1). Traits included: longevity; mean number of offspring per season; mean body mass diet rank based on increasing protein and energy content [1 = grass/leaves, 2 = seeds, forbs, grass, roots and fungi, 3 = nectar, gum and insects (insects <50% of the diet), 4 = insects of vertebrates (>50% of the diet)]; habitat number

(number of categories of vegetation structure in which the species occurs); sex of the individual; estimated age class of the individual; latitude and longitude of the individual; season the sample was collected in and associated taxonomic information (order, family, genus and species group). For mammals this data was obtained from the PANTHERIA dataset (16) and for birds it was obtained from Garnett, Duursma (17).

5.3.4 Statistical analysis

To test for associations between the each species’ seroprevalence and the ecological and life- history traits a random forest regression tree was constructed using all species, following the methodology described by Cutler et al. (18). This classification method attempts to develop rules to assign individuals into known classes (here, seropositive or seronegative) according to potential predictor or explanatory variables. The method develops rules by using explanatory variables to recursively partition data into groups that are increasingly homogenous with respect to the known classes (19). This method was chosen as it can identify interactions in which the same variable repeatedly enters a model at different levels (20). To visualise the results of this analysis, the variables identified as most important in the random forest analysis

were used to construct a conditional inference tree using the function "ctree” in the R package

“party” (21).

To analyse spatial trends for RRV serology in non-human vertebrates, a two step-approach was used. First, point maps demonstrating positive and negative results were constructed for the most commonly assessed species and taxa, namely, brushtail possums, ringtail possums, koalas

(Phascolarctos cinereus), horses (Equus caballus), flying foxes (Pteropus alecto and P. poliocephalus) and all bird species. Second, to identify spatial clusters of positive or negative vertebrates, a hotspot analysis was undertaken in ArcGIS, which uses the Getis-Ord Gi* statistic (22). This analysis works by assessing each point in the context of surrounding points.

For example, if a single sample is positive, but the samples surrounding it are negative, it would not be statistically significant as a hot spot of seropositivity. Instead, hotspots of seropositivity must be positive and also in close proximity to other positive samples.

5.4 Results

5.4.1 Collected samples

Serum samples from a total of 595 individuals from 31 species were tested for the presence of

RRV antibodies. Of these, 327 individuals or 54.96% (95% CI: 40.17-69.1) from 23 species were seropositive (Supplementary Table 2). Antibodies to RRV were detected across all species groups (marsupials, placental mammals and birds) (Figure 1), with the highest average seroprevalence reported in placental mammals (70.83%; 95% CI: 61.84-78.78) and lowest in birds (28.87%; 95% CI: 20.11-38.95), though high heterogeneity was observed between species and within species groups. Results for individual species are reported in the Supporting

Information (Supplementary Table 2).

Figure 1: The prevalence of RRV antibodies, expressed as a proportion on the y-axis, for species (x-axis) with a minimum of nine samples, grouped by a) birds, b) marsupials and, c) placental mammals. Numbers above columns indicate the total number of samples tested, and error bars are exact 95% binomial confidence intervals.

5.4.2 Species traits

The random forest model was used to determine the relative importance of all species traits for seropositivity. The importance of a species trait was assessed using the ‘varImpPlot’ and

‘importance’ functions on the output of the model (Figure 2). From this, log body mass, maximum longevity, number of offspring per season, social group size, diet and latitude of the individual were identified as the variables most likely to be accurately predicting seropositivity

(known as the Mean Decrease Accuracy), and thus selected for the conditional inference tree analysis.

* * * * * *

Figure 2. Variable importance as determined by the Random Forest model. The length of the horizontal bar indicates the variable importance for explaining seropositivity, based on the Mean Decrease Accuracy. Variables with an * at the end were identified as most important and used in the coefficient inference tree.

The conditional inference tree (Figure 3) indicated that diet, log body mass, number of offspring and longevity drive key splits in the dataset between seropositive and seronegative individuals, and that the crucial (first) dichotomous split was based on diet. Species with the lowest protein intake (diet ≤1: namely grazers including horses and macropod species) had higher seropositivity than all other species (Figure 1). The second node split species based on their log mass, with heavier animals (in this case horses) having higher seropositivity. Log mass also separated pigeons (0% seropositive) from ringtail possums (69.7% seropositive) in node 10. Amongst animals with higher protein diets, carnivores (diet >3) had lower seropositivity than the combined omnivore group (diet ≤ 3), which then split according to number of offspring per season, effectively separating ringtail possums from the other possum species, as well as separating birds from flying foxes. Finally, the longevity split in node 7 largely separated flying foxes from brushtail possums and short-eared possums.

Figure 3: Conditional inference tree based on the six traits most strongly associated with seropositivity (as determined by the random forest model). The numbered ellipses identify the nodes and variables that best split the data for seropositivity at each point in the tree, with Bonferroni adjusted P values shown together with the cut-off values for the split along the branches. At the bottom of the tree, shading in the boxes representing the terminal nodes shows the proportion of seropositive individuals (darkest shading), and n= the number of individuals in each of the final groups. Although social group size and latitude were included in the analysis, they did not generate a significant split between seropositive or seronegative individuals.

There was a highly significant difference in seropositivity between ringtail possums

Pseudocheirus peregrinus and brushtail possums Trichosurus vulpecula (x²=26.276, p=

<0.001), with the odds of a ringtail possum being seropositive 3.2 times higher (CI 2.02-5.34) than for a brushtail possum being seropositive.

5.4.3 Spatial analysis

Point maps showing seropositive and negative results were derived for ringtail possums, brushtail possums, koalas, horses, flying foxes and birds, separately (Figure 4a). No strong

spatial patterns were observed for the three marsupial species (ringtail possums, brushtail possums and koalas). Conversely, horses, flying foxes and birds had a greater aggregation of negative samples in northern SEQ and a greater aggregation of positive samples in southern

SEQ (Figure 4a). This trend was more apparent in the hotspot analysis, which identified a significant cold spot (99% confidence) of RRV seropositivity in the northern SEQ locations, and a significant hotspot (99% confidence) formed around southern SEQ (Figure 4b).

100

Figure 4. Maps of South-East Queensland representing locations and results of samples tested for RRV. Specifically, the: a) spatial distribution of positive (red) and negative (blue) results for ringtail possums, brushtail possums, koalas, horses, flying foxes and birds; and b) a hotspot analysis across all samples and reported significance of these spots in the right-hand legend.

5.5 Discussion Ross River virus is Australia’s most epidemiologically important arbovirus and effectively managing transmission requires an understanding of the natural exposure rates of RRV in non- human vertebrates. This study tested a greater diversity of vertebrate species for RRV seroconversion than previous studies, using a plaque reduction neutralisation assay (PRNT50), which is recognised as the gold standard for arboviruses (23). In contrast to the dogma that that marsupials are the most important reservoirs of RRV, we found that placental mammals as a group had higher seropositivity than marsupials or birds. However, these aggregated results may be misleading as there is high heterogeneity of RRV seropositivity within each taxonomic group indicating that these groupings may not be useful predictors of RRV seropositivity.

Rather than taxonomic relatedness, ecological and life-history traits could be driving some heterogeneity in seroprevalence results. In particular, ringtail possums had a much higher seroprevalence against RRV than brushtail possums, despite both being marsupials. This is interesting because brushtail possums have long been implicated as an important urban reservoir of RRV (24). Experimental infection studies have demonstrated that brushtail possums are competent amplifiers of RRV, capable of infecting 53% of susceptible Ae. vigilax vectors (24). However, a recent serosurvey of 72 brushtail possums in urban Sydney failed to detect any RRV antibodies in the surveyed population (25). By comparison, ringtail possums have been greatly understudied as reservoirs of RRV, with no experimental infection studies undertaken and only four free-living individuals being previously tested for RRV exposure (26,

27). Brushtail possums and ringtail possums are both abundant in Brisbane and other Australian urban areas. They differ somewhat ecologically: brushtail possums primarily inhabit roofs in urban areas whereas ringtail possums build nests in trees (28, 29). These differences could

101

bring ringtail possums into contact with mosquitoes more frequently than brushtail possums, but more studies, including mosquito bloodmeal analyses, would be needed to test this hypothesis.

Vector-host interactions are critical for the transmission of arboviruses and are influenced by a number of factors. Mosquito host-seeking behaviour is predominantly guided by detection of heat and carbon dioxide (CO2), but can also be influenced by various odorants and chemicals that are released by hosts, host abundance and host defensive behaviour (5, 30). In the conditional inference tree analysis ‘diet’ was the first characteristic influencing an individual’s likelihood of being seropositive or seronegative. Diet is closely linked with CO2 production

(31). In general, large-bodied herbivores produce larger quantities of CO2, in part due to the rumination and fermentation process required to digest their meal, than carnivorous or omnivorous species (32). In this study, species with herbivore diets (category 1) were split from other species and had a higher than average seroprevalence. Within this group the larger bodied vertebrates (horse) had the highest seroprevalence, which could be due to a larger body surface area and heat for mosquitoes to detect and feed on (33, 34). By comparison, species with the greatest protein intake (in this case insectivorous and carnivorous birds), had a lower than average seroprevalence, which may be a result of smaller body sizes or exhibiting host behaviours which may include feeding on mosquitoes (35).

Although horses are assumed to be the only species other than humans to have clinical signs to

RRV (36), this has been based largely on prospective case studies, as opposed to experimental infection results (37, 38). Regardless, the high seroprevalence in horses in our study echo that of other published seroprevalence studies (39, 40) and is of interest for several reasons. Firstly, the individuals in this study had no previous signs of RRV infection within the past 12 months, suggesting either that infections were asymptomatic (or sufficiently so to be unobserved by their owners and veterinarian) or that infections occurred more than 12 months prior to sample

102

collection, indicating that RRV antibodies may be long-lasting in horses. Long-term immunity for RRV in horses requires further investigation given that early experimental studies demonstrate the hemagglutination-inhibition response lasting ~100 days (Kay et al. 1987), whereas anecdotal reports from horse racing in Australia has found detectable IgG more than

12 years post original infection (41). Secondly, whilst naïve horses have a demonstrated ability to transmit RRV to mosquitoes when experimentally infected (Kay et al. 1986), the high herd immunity reported here and in other studies may suggest that they could reduce ongoing transmission via a dilution effect if infected mosquitoes bite previously exposed, and immune, horses. Having a large body size, and limited host defensive strategies, horses are an ideal target for mosquitoes, which is reflected in their frequent detection in blood feeding studies (5). The duration of RRV antibodies in horses, and the association between antibody levels and immunity against reinfection, are therefore critical areas for further research to assess the role of horses in maintaining RRV in peri-urban environments.

Differences in immune responses and abilities of producing neutralising antibodies among vertebrate species might influence serological findings across the species studied here. For example, no experimentally infected little corellas seroconverted, despite 50% developing detectable viraemias and being capable of infecting susceptible mosquitoes (42). Similarly, experimentally infected grey-headed flying foxes showed highly variable immune responses, and were capable of transmitting RRV to susceptible mosquitoes despite undetectable viraemias (43). As such, the immune response and assay sensitivity needs to be assessed on a species by species basis to fully evaluate whether seroprevalences observed in this study are a true reflection of RRV exposure under natural conditions. To date, no long-term studies for the persistence of RRV antibodies in non-human vertebrates have been done. Such studies would be critical in quantifying the variability and detectability of RRV in non-human vertebrates.

103

Human notification patterns of RRV share both similarities and inconsistencies with the results of the hotspot analysis. Consistent with our results, the south-eastern hotspot in Brisbane has previously been identified as a hotspot of human infections over 1991-1996 time period (44), and again in 2001 (45). This could suggest the area experiences frequent spill-over from reservoir species and could be of interest for management interventions. In contrast, more recent data on human notifications in SEQ (from 2018) identify the Sunshine Coast (a cold spot in our analysis) as consistently reporting the highest notification rates (number of notifications/human population) compared to the rest of SEQ (1). Human notification data differs from the seroprevalence data presented here, in that notifications occur closer to the timing of an infection, and only represent those infections that were associated with clinical signs of RRV in humans (as opposed to asymptomatic infections). This, in combination with the complicated disease ecology of RRV (involving multiple vectors and vertebrate hosts) suggest that transmission patterns between humans and non-human vertebrates may vary and/or be highly localised (46, 47).

5.5.1 Ecological applications

The random forest model, in combination with the conditional inference tree, has practical applications for targeted management of RRV. For example, if there are high levels of exposure identified in a given species that is known to amplify RRV, mosquito control measures can be put in place to reduce mosquito-host contact rates. The output of the results in a decision tree makes it easy to visualise which populations are best to target. A limitation with this method is that the results are only applicable to the species and life history traits that are included in the random forest analysis. Whilst our study has assessed the greatest diversity of vertebrate species tested for RRV, it does not reflect the full biodiversity of species in SEQ and there may be other important life history traits that we have not yet considered. However, the framework

104

will be a valuable addition to future studies incorporating more species and to assess the outcomes of management efforts when used over a temporal scale.

5.5.2 Future directions

Serosurveys provide an informative tool to understand patterns of transmission, however for vector-borne diseases they are limited in identifying reservoir hosts. An individual that has a detectable immune response to RRV has previously been infected, but this does not correlate to an acute infection capable of amplification. Few experimental infection studies have been undertaken to determine which vertebrates can infect susceptible mosquitoes with RRV, and without this information it is unclear whether or not the seropositive species identified here are amplifying or diluting RRV in this complex transmission system. Experimental infection studies which utilise multiple vector species to infect and measure infectivity of vertebrates with RRV are needed to fill this gap.

Equally important is long-term serological sampling of a community. For WNV, the use of localised studies which incorporated i) vector availability, ii) host abundance, iii) host exposure rates and, iv) human infections were successful in identifying migratory bird species as a key reservoir (10). For RRV, there are no published studies which take repeated samples from a community. Repeated sampling from a free-living community in combination with climatic and vector variables would provide novel insights into the local transmission dynamics of

RRV.

5.6 Conclusion

Seroprevalence data provides critical information about the transmission and distribution of pathogens in non-human species. Here we interpret the RRV seroprevalence data across the greatest diversity of species in a single study, using a novel modelling methodology. We confirm that non-human vertebrates are frequently infected with RRV across SEQ. The highest

105

seroprevalence was reported in horses, followed by koalas and ringtail possums. Species traits, in particular diet and body mass, may play a role in vector-host contact rates, ultimately influencing exposure and ongoing transmission. Ringtail possums seroprevalence was significantly higher than brushtail possums, which is particularly interesting as only the latter have been investigated as potential reservoirs for RRV. Further research is needed encompassing experimental infection studies and long-term serosurvey studies to identify the transmission dynamics between vectors and hosts. The results from this study provide a framework to examine serological results and ultimately inform management of pathogens more broadly.

Authors contributions

EBS collected all samples; EBS, PAR, LJH undertook all laboratory analyses; EBS, AJP,

HM conducted modelling analyses. All authors contributed to the conceptualisation of the study, interpretation of results and writing and editing of the manuscript.

Acknowledgements

We would like to acknowledge the following veterinary clinics for contributing to this research:

RSPCA Wildlife Hospital, Wacol; Currumbin Wildlife Sanctuary, Currumbin; Australia Zoo

Wildlife Hospital, Beerwah; Brisbane Bird Vets, Chermside; WestVets, Anstead; Gold Coast

Equine Clinic, Bundall.

Cassie Jansen is a supervisor of Eloise Stephenson and contributed to the project design. We would also like to acknowledge Narayan Gyawali from QIMR for providing the serology protocol. We would like to broadly acknowledge the LJH’s and HM’s lab members for their guidance and revisions of this manuscript.

Funding

106

LJH is the recipient of the Australian National Health and Medical Research Council Career

Development Award (ID: 105760). AJP was supported by a Queensland Government

Accelerate Postdoctoral Research Fellowship. EBS is the recipient of an Australian

Government Research Training Program. Sample collection was supported by a Student

Research Award from the Ecological Society of Australia, and a Future Fellow award from

Graduate Women Gold Coast.

Competing interests: None declared.

Ethical approval

This work was performed in accordance with Griffith University Animal Ethics Committee standards (approval number: ENV/04/17/AEC).

Data accessibility

Data is currently accessible though the CloudStor Repository via the link below, once published, data will be uploaded to figshare: https://protect- au.mimecast.com/s/JEHjCYWLoQTQ8mG4F0vuY9?domain=cloudstor.aarnet.edu.au

5.7. References

1. Australian Government. National Notifiable Diseases Surveillance System 2018 [Available from: http://www9.health.gov.au/cda/source/cda-index.cfm. 2. Harley D, Sleigh A, Ritchie S. Ross River virus transmission, infection, and disease: a cross- disciplinary review. Clinical Microbiology Reviews. 2001;14(4):909-32. 3. Garchitorena A, Sokolow S, Roche B, Ngonghala C, Jocque M, Lund A, et al. Disease ecology, health and the environment: a framework to account for ecological and socio-economic drivers in the control of neglected tropical diseases. Philosophical Transactions of the Royal Society B: Biological Sciences. 2017;372(1722):20160128. 4. Russell RC. Ross River virus: Ecology and distribution. Annual Review of Entomology. 2002;47:1-31. 5. Stephenson EB, Murphy AK, Jansen CC, Peel AJ, McCallum H. Interpreting mosquito feeding patterns in Australia through an ecological lens: an analysis of blood meal studies. Parasites & vectors. 2019;12(1):156. 6. Stephenson EB, Peel AJ, Reid SA, Jansen CC, McCallum H. The non-human reservoirs of Ross River virus: a systematic review of the evidence. Parasites & vectors. 2018;11(1):188. 7. Lau C, Aubry M, Musso D, Teissier A, Paulous S, Desprès P, et al. New evidence for endemic circulation of Ross River virus in the Pacific Islands and the potential for emergence. International Journal of Infectious Diseases. 2017;57:73-6.

107

8. Aubry M, Teissier A, Huart M, Merceron S, Vanhomwegen J, Roche C, et al. Ross River Virus Seroprevalence, French Polynesia, 2014-2015. Emerging Infectious Diseases. 2017;23(10):1751-3. 9. Aubry M, Finke J, Teissier A, Roche C, Broult J, Paulous S, et al. Silent Circulation of Ross River Virus in French Polynesia. International Journal of Infectious Diseases. 2015;37:19-24. 10. Kilpatrick AM, Daszak P, Jones MJ, Marra PP, Kramer LD. Host heterogeneity dominates West Nile virus transmission. Proceedings of the Royal Society of London B: Biological Sciences. 2006;273(1599):2327-33. 11. Carver S, Bestall A, Jardine A, Ostfeld RS. Influence of Hosts on the Ecology of Arboviral Transmission: Potential Mechanisms Influencing Dengue, Murray Valley Encephalitis, and Ross River Virus in Australia. Vector-Borne and Zoonotic Diseases. 2009;9(1):51-64. 12. Marshall ID, Brown BK, Keith K, Gard GP, Thibos E. Variation in arbovirus infection rates in species of birds sampled in a serological survey during an encephalitis epidemic in the Murray- Valley of Southeastern Australia, February 1974. Australian Journal of Experimental Biology and Medical Science. 1982;60(Pt 5):471-8. 13. Lundström JO, Lindström KM, Olsen B, Dufva R, Krakower DS. Prevalence of Sindbis virus neutralizing antibodies among Swedish passerines indicates that thrushes are the main amplifying hosts. Journal of medical entomology. 2001;38(2):289-97. 14. Gilbert AT, Fooks AR, Hayman DT, Horton DL, Müller T, Plowright R, et al. Deciphering serology to understand the ecology of infectious diseases in wildlife. EcoHealth. 2013;10(3):298-313. 15. Walsh MG. Ecological and life history traits are associated with Ross River virus infection among sylvatic mammals in Australia. BMC ecology. 2019;19(1):2. 16. Jones KE, Bielby J, Cardillo M, Fritz SA, O'Dell J, Orme CDL, et al. PanTHERIA: a species‐ level database of life history, ecology, and geography of extant and recently extinct mammals. Ecology. 2009;90(9):2648-. 17. Garnett ST, Duursma DE, Ehmke G, Guay P-J, Stewart A, Szabo JK, et al. Biological, ecological, conservation and legal information for all species and subspecies of Australian bird. Scientific data. 2015;2:150061. 18. Cutler DR, Edwards TC, Beard KH, Cutler A, Hess KT, Gibson J, et al. Random forests for classification in ecology. Ecology. 2007;88(11):2783-92. 19. Bielby J, Cardillo M, Cooper N, Purvis A. Modelling extinction risk in multispecies data sets: phylogenetically independent contrasts versus decision trees. Biodiversity and Conservation. 2010;19(1):113-27. 20. Murray KA, Rosauer D, McCallum H, Skerratt LF. Integrating species traits with extrinsic threats: closing the gap between predicting and preventing species declines. Proceedings of the Royal Society B: Biological Sciences. 2010;278(1711):1515-23. 21. Hothorn T, Hornik K, Zeileis A. Unbiased recursive partitioning: A conditional inference framework. Journal of Computational and Graphical statistics. 2006;15(3):651-74. 22. Getis A, Ord JK. The analysis of spatial association by use of distance statistics. Perspectives on Spatial Data Analysis: Springer; 2010. p. 127-45. 23. Hobson-Peters J. Approaches for the development of rapid serological assays for surveillance and diagnosis of infections caused by zoonotic flaviviruses of the Japanese encephalitis virus serocomplex. BioMed Research International. 2012;2012. 24. Boyd AM, Hall RA, Gemmell RT, Kay BH. Experimental infection of Australian brushtail possums, Trichosurus vulpecula (Phalangeridae : Marsupialia), with Ross River and Barmah Forest viruses by use of a natural mosquito vector system. American Journal of Tropical Medicine and Hygiene. 2001;65(6):777-82. 25. Hill NJ, Power ML, Deane EM. Absence of Ross River virus amongst Common brushtail possums (Trichosurus vulpecula) from metropolitan Sydney, Australia. European Journal of Wildlife Research. 2009;55(3):313-6. 26. Gard G, Marshall ID, Woodroof GM. Annually recurrent epidemic polyarthritis and Ross River virus activity in a coastal area of New South Wales. II. Mosquitos, viruses and wildlife. American Journal of Tropical Medicine and Hygiene. 1973;22(4):551-60. 27. McManus TJ, Marshall ID. The epidemiology of Ross River virus in Tasmania. Arbovirus Research in Australia. 1986;1986:127-31.

108

28. Statham M, Statham H. Movements and habits of brushtail possums (Trichosurus vulpecula Kerr) in an urban area. Wildlife Research. 1997;24(6):715-26. 29. Inions G, Tanton M, Davey S. Effect of fire on the availability of hollows in trees used by the common brushtail possum, Trichosurus vulpecula Kerr, 1792, and the ringtail possum, Pseudocheirus peregrinus Boddaerts, 1785. Wildlife Research. 1989;16(4):449-58. 30. Takken W, Verhulst NO. Host preferences of blood-feeding mosquitoes. Annu Rev Entomol. 2013;58:433-53. 31. Gillies M, Wilkes T. The range of attraction of animal baits and carbon dioxide for mosquitoes. Studies in a freshwater area of West Africa. Bulletin of Entomological Research. 1972;61(3):389-404. 32. Passey BH, Robinson TF, Ayliffe LK, Cerling TE, Sponheimer M, Dearing MD, et al. Carbon isotope fractionation between diet, breath CO2, and bioapatite in different mammals. Journal of Archaeological Science. 2005;32(10):1459-70. 33. Kleiber M. Body size and metabolic rate. Physiological reviews. 1947;27(4):511-41. 34. Franz R, Soliva CR, Kreuzer M, Steuer P, Hummel J, Clauss M. Methane production in relation to body mass of ruminants and equids. Evolutionary Ecology Research. 2010;12(6):727-38. 35. Torr SJ, Mangwiro T, Hall DR. The effects of host physiology on the attraction of tsetse (Diptera: Glossinidae) and Stomoxys (Diptera: Muscidae) to cattle. Bulletin of Entomological Research. 2006;96(1):71-84. 36. Azuolas JK. Ross River virus disease of horses. Australian Equine Veterinarian. 1998;16(2):56-8. 37. Kay BH, Pollitt CC, Fanning ID, Hall RA. The experimental infection of horses with Murray Valley encephalitis and Ross River viruses. Australian Veterinary Journal. 1987;64(2):52-5. 38. Barton AJ, Bielefeldt-Ohmann H. Clinical presentation, progression, and management of five cases of Ross River virus infection in performance horses located in southeast Queensland: a longitudinal case series. Journal of equine veterinary science. 2017;51:34-40. 39. Gummow B, Tan R, Joice R, Burgess G, Picard J. Seroprevalence and associated risk factors of mosquito‐borne alphaviruses in horses in northern Queensland. Australian veterinary journal. 2018;96(7):243-51. 40. Azuolas J. Arboviral diseases of horses and possums. Arbovirus Res Aust. 1997;7:5-7. 41. Johnston S. Detection of IgG antibodies in individual horse 12 years following Ross River virus infection. In: Stephenson E, editor. 2019. 42. Kay BH, Hall RA, Fanning ID, Mottram P, Young PL, Pollitt CC, editors. Experimental infection of vertebrates with Murray Valley encephalitis and Ross River viruses. Arbovirus Research in Australia; 1986 1986. 43. Ryan PA, Martin L, Mackenzie JS, Kay BH. Investigation of gray-headed flying foxes (Pteropus poliocephalus) (Megachiroptera: Pteropodidae) and mosquitoes in the ecology of Ross river virus in Australia. American Journal of Tropical Medicine and Hygiene. 1997;57(4):476-82. 44. Muhar A, Dale PE, Thalib L, Arito E. The spatial distribution of Ross River virus infections in Brisbane: significance of residential location and relationships with vegetation types. Environmental health and preventive medicine. 2000;4(4):184-9. 45. Hu W, Tong S, Mengersen K, Oldenburg B. Exploratory spatial analysis of social and environmental factors associated with the incidence of Ross River virus in Brisbane, Australia. American Journal of Tropical Medicine and Hygiene. 2007;76(5):814-9. 46. Flies EJ, Weinstein P, Anderson SJ, Koolhof I, Foufopoulos J, Williams CR. Ross River virus and the necessity of multi-scale, eco-epidemiological analyses. The Journal of infectious diseases. 2017. 47. Claflin SB, Webb CE. Ross River Virus: Many Vectors and Unusual Hosts Make for an Unpredictable Pathogen. Plos Pathogens. 2015;11(9).

109

5.8 Supplementary Information

Supplementary Table 1: Species and their associated traits assessed in the Random Forest model

Species Order Family Genus Longevity Mean Log Mean Social Diet Number (months) body body number group rank habitats mass mass offspring category occupied (g) Birds Australian Passeriformes Oriolidae Sphecotheres 203 123 2.09 2.7 5 3 3 figbird Sphecotheres vieilloti Australian Passeriformes Artamidae Gymnorhina 289.5 280 2.45 3.3 6 4 2 magpie Cracticus tibicen Australian Pelecaniformes Threskiornithidae Threskiornis 316.8 1810 3.26 2.8 5 4 2 white ibis Threskiornis moluccus Common Columbiformes Columbidae Phaps 73.23 331 2.52 1.8 3 2 2 bronzewing Phaps chalcoptera Common koel Cuculiformes Cuculidae Eudynamys 121 243 2.39 1 3 3 3 Eudynamys orientalis Common Passeriformes Sturnidae Acridotheres 148.23 113 2.05 4 6 3 3 myna Sturnus tristis

110

Crested pigeon Columbiformes Columbidae Ocyphaps 210.9 192 2.28 2 4 2 2 Ocyphaps lophotes Grey Passeriformes Artamidae Cracticus 245.37 90.8 1.96 3.2 6 4 3 butcherbird Cracticus torquatus Laughing Coraciiformes Alcedinidae Dacelo 206.33 312 2.49 3 4 4 3 kookabuura Dacelo novaeguineae Magpie lark Passeriformes Monarchidae Grallina 147.73 88 1.94 3.8 2 4 2 Grallina cyanoleuca Noisy friarbird Passeriformes Meliphagidae Philemon 119.43 100 2.00 3 6 3 3 Philemon corniculatus Noisy minor Passeriformes Meliphagidae Manorina 151.43 71.3 1.85 2.9 5 3 3 Manorina melanocephala Pied Passeriformes Artamidae Cracticus 269.57 123 2.08 3 3 4 3 butcherbird Cracticus nigrogularis Pied Passeriformes Artamidae Strepera 293.83 308 2.49 3.1 6 4 3 currawong Strepera graculina Masked Charadriiformes Charadriidae Vanellus 169.33 316 2.50 3.5 4 4 1 lapwing Vanellus miles Rainbow Psittaciformes Psittaculidae Trichoglossus 250.37 126 2.10 2.5 5 3 3 lorikeet

111

Trichoglossus moluccanus Rock dove Columbiformes Columbidae Columba 133.37 306 2.49 2 2 2 2 Columba livia Spotted dove Columbiformes Columbidae Streptopelia 119.97 161 2.21 2 6 2 2 Streptopelia chinensis Tawny Caprimulgiformes Podargidae Podargus 217.9 277 2.44 2.3 6 4 3 frogmouth Podargus strigoides Torresian crow Passeriformes Corvidae Corvus 29.82 582 2.76 4.6 6 4 3 Corvus orru Wedgtail eagle Accipitriformes Accipitridae Aquila 144.5 3470 3.54 2 6 4 4 Aquila audax White headed Columbiformes Columbidae Columba 119.97 420 2.62 1 3 2 2 pigeon Columba leucomela Marsupials Brushtail Diprotodontia Phalangeridae Trichosurus 176.4 2700 3.43 1.03 1 3 3 possum Trichosurus vulpecula Eastern grey Diprotodontia Macropodidae Macropus 288 33600 4.53 1 5 1 2 kangaroo Macropus gigantus Koala Diprotodontia Phascolarctidae Phascolarctos 240 6580 3.82 1 1 1 3 Phascolarctos cinereus Red necked Diprotodontia Macropodidae Macropus 228 16800 4.23 1 2 1 2 wallaby

112

Macropus rufogriseus Ringtail Diprotodontia Pseudocheiridae Pseudocheirus 98 894 2.95 1.94 4 3 3 possum Pseudocheirus peregrinus Short-eared Diprotodontia Phalangeridae Trichosurus 204 3140 3.50 1.03 1 3 3 possum Trichosurus caninus Placental mammals Black flying Chiroptera Pteropodidae Pteropus 85.2 607 2.78 1 5 3 3 fox Pteropus alecto Grey headed Chiroptera Pteropodidae Pteropus 85.2 702 2.85 1 5 3 3 flying fox Pteropus poliocephalus Horse Equus Perissodactyla Equidae Equus 744 400000 5.60 1 4 1 2 caballus

113

Supplementary Table 2: Serology results for each species, including total number tested, apparent prevalence and confidence intervals

Species Total seropositive/Total tested Apparent prevalence (lower- upper confidence interval)

Birds 28/97 28.9 (20.1-38.9) Australian figbird Sphecotheres vieilloti 2/2 100.0 (84.2-100.0) Australian magpie Cracticus tibicen 6/33 18.2 (11.2-35.5) Australian white ibis Threskiornis moluccus 1/1 100.0 (97.5-100.0) Common bronzewing Phaps chalcoptera 1/1 100.0 (97.5-100.0) Common koel Eudynamys orientalis 1/1 100.0 (97.5-100.0) Common myna Sturnus tristis 1/1 100.0 (97.5-100.0) Crested pigeon Ocyphaps lophotes 0/1 0.0 (0.0-97.5) Grey butcherbird Cracticus torquatus 0/3 0.0 (0.0-70.8) Laughing kookabuura Dacelo novaeguineae 1/5 20.0 (19.5-71.6) Magpie lark Grallina cyanoleuca 0/1 0.0 (0.0-97.5) Noisy friarbird Philemon corniculatus 0/7 0.0 (0.0-40.9) Noisy minor Manorina melanocephala 1/1 100.0 (97.5-100.0) Pied butcherbird Cracticus nigrogularis 1/4 25.0 (24.4-80.6) Pied currawong Strepera graculina 0/1 0.0 (0.0-97.5) Masked lapwing Vanellus miles 0/2 0.0 (0.0-84.2) Rainbow lorikeet Trichoglossus moluccanus 2/5 40.0 (34.7-85.3) Rock dove Columba livia 0/2 0.0 (0.0-84.2) Spotted dove Streptopelia chinensis 1/2 50.0 (48.7-98.7) Tawny frogmouth Podargus strigoides 3/4 75.0 (55.6-99.4) Torresian crow Corvus orru 6/18 33.3 (19.9-59.0) Wedgtail eagle Aquila audax 0/1 0.0 (0.0-97.5) White headed pigeon Columba leucomela 1/1 100.0 (97.5-100.0) Marsupials 214/378 56.6 (51.4-61.7) Brushtail possum Trichosurus vulpecula 75/182 41.2 (7.2-48.7) Eastern grey kangaroo Macropus gigantus 5/6 83.3 (47.5-99.6) Koala Phascolarctos cinereus 24/35 68.6 (17.9-83.2)

114

Red necked wallaby Macropus rufogriseus 3/3 100.0 (70.8-100.0) Ringtail possum Pseudocheirus peregrinus 101/145 69.7 (8.2-77.0) Short-eared possum Trichosurus caninus 6/7 85.7 (43.6-99.6) Placental mammals 85/120 70.8 (61.8-78.8) Black flying fox Pteropus alecto 5/29 17.2 (11.4-35.8) Grey headed flying fox Pteropus 3/9 33.3 (25.8-70.1) poliocephalus Horse Equus caballus 77/82 93.9 (7.6-97.9) Total 327/595 54.96

115

Chapter 6 Associations between Ross River virus infection in humans and vector-vertebrate community ecology in Brisbane, Australia

Statement of contribution to co-authored submitted paper

The following chapter includes a co-authored paper:

Stephenson, E. B., Murphy, A., Jansen, C. C., Shivas, M., McCallum, H., Onn, M. B., Reid, S. A. & Peel, A. J. (2019) Associations between Ross River virus infection in humans and vector-vertebrate community ecology in Brisbane, Australia. Vector-Borne and Zoonotic Diseases. Submitted.

My contribution to the paper involved data collection collaboratively with Amanda Murphy, Michael Onn, and Martin Shivas, planning and performing all analyses, and writing the paper.

Signed: ______Date: ______Corresponding author: Eloise Stephenson

Countersigned: ______Date:______Supervisor: Alison Peel

116

Associations between Ross River virus infection in humans and vector-vertebrate community ecology in Brisbane, Australia

Stephenson, E. B.1, Murphy, A.2,3, Jansen, C. C.4, Shivas, M. A.5, McCallum, H.1, Onn, M. B.5, Reid, S. A.6 & Peel, A. J.1

1Environmental Futures Research Institute, Griffith University, Nathan, Queensland, 4111 2Queensland University of Technology, School of Biomedical Sciences, Brisbane, Queensland, 4000 3QIMR Berghofer Medical Research Institute, Mosquito Control Laboratory, Herston, Queensland, 4006 4Communicable Diseases Branch, Queensland Health, Herston, Queensland, 4006 5Brisbane City Council, Field Services, Brisbane CBD, Queensland, 4000 6The University of Queensland, School of Public Health, Herston, , Queensland, 4006

Keywords (3-6): arbovirus ecology; abundance; competence; diversity; mosquito; vertebrate

6.1 Abstract:

Vector-borne disease transmission can vary in complexity from single-vector, single-host systems through to multi-vector, multi-host community and vertebrate systems.

Understanding the dynamics of transmission is important for disease prevention efforts but is dependent on disentangling complex interactions within coupled natural systems. Ross River virus (RRV) is a multi-vector, multi-host pathogen responsible for the greatest number of notified vector-borne pathogen infections in humans in Australia. Yet, there is a limited knowledge of which mosquito vector species and amplifying vertebrate host species are most important for spillover of RRV to humans.

We conducted field surveys of vertebrates and mosquitoes in the RRV-endemic city of Brisbane, Australia, to assess the effect of vector and host community structure on human

RRV notifications. Six suburbs were selected across a gradient of human disease notification rates. Differences in vertebrate and mosquito compositions were observed across all suburbs.

Suburbs with higher RRV notification rates contained significantly a higher vertebrate biomass and mosquito abundances. This study suggests that horse-mosquito interactions should be considered in more detail; and that future RRV studies incorporate vertebrate

117

biomass and mosquito abundance into modelling and public health strategies for RRV management.

6.2 Introduction

Obtaining a detailed understanding of pathogen-host-vector dynamics within a given environment is particularly challenging for multi-vector, multi-host pathogens. The relative role of different vectors and/or vertebrate reservoir species in supporting pathogen circulation is often largely unknown under natural conditions. However, knowledge of these determinants of transmission are essential for the design of effective prevention and control strategies (Gardner et al., 2017; Hoffmann, Ross, & Rašić, 2015; Janousek et al., 2014;

Weinstein, 1997), both for predicting net effects on human populations, and for evaluating management and prevention of zoonotic mosquito-borne diseases.

Mosquito-borne diseases contribute significantly to the global burden of infectious disease in both human and animal populations (Hill et al., 2005). Widely distributed across

Australia, Ross River virus (RRV) is Australia’s most common mosquito-borne disease in humans and is nationally notifiable (Australian Government, 2018). Ross River virus is a multi-host, multi-vector pathogen with a complex ecology: it persists in all Australian bioregions and climates and has been isolated from more than 30 species of mosquito under field conditions (Russell, 2002). Humans are generally not considered important amplifying or maintenance hosts of RRV (Harley, Sleigh, & Ritchie, 2001); however, the relative roles of various non-human vertebrate host species in maintaining transmission of RRV are poorly understood. Strong experimental evidence supports marsupials as competent reservoirs, but amplification and transmission competence has also been demonstrated in other species groups, including placental mammals and birds (Kay et al., 1986; Stephenson et al., 2018). It is unclear how transmission competence of hosts in an experimental setting reflects their

118

relative role in transmission in natural settings (Claflin & Webb, 2015) and many potential reservoir species have not yet been investigated.

Although at least 20 mosquito species have been demonstrated to have some level of vector competence of RRV in the laboratory, it is unclear what the relative contributions of each species are to both enzootic transmission and/or human disease outbreaks (Jansen et al.,

2019; Russell, 2002; van den Hurk et al., 2010). It is likely that the vectors associated with human outbreaks vary across different bioregions of Australia, and this may also be the case with reservoir hosts (Claflin & Webb, 2015). Simply, the key vectors and hosts associated with the maintenance of RRV, and its spillover into humans, in different regions of Australia remain unclear.

Of the approximately 5,000 human RRV notifications reported in Australia each year, the majority are reported from the state of Queensland (Australian Government, 2018).

Residents of Queensland’s capital city, Brisbane (27.4698° S, 153.0251° E), consistently represent the highest proportion of the state’s cases. In 2014-2015, Queensland experienced the largest outbreak of RRV ever recorded in Australia, totalling nearly 3,000 human notifications (Queensland Government, 2019; Jansen et al., 2019). Brisbane has a subtropical climate, which accommodates a diversity of habitats for both vector and reservoir species.

This diversity, in combination with the persistent high human notification rates of RRV, makes the city an ideal location to study RRV disease ecology.

This study aims to assess mosquito and vertebrate host community ecology across suburbs of Brisbane encompassing locations with a range of RRV notification rates.

Specifically, we aim to identify relationships between abundance, diversity and biomass of potential vertebrate reservoirs; mosquito vectors; and human notification rates of RRV.

6.3 Materials and Methods

6.3.1 Human notification data

119

Long-term human notification data for Brisbane suburbs were obtained for 2001 to

2016 (inclusive) from the Queensland Department of Health (Ethics application: QIMRB

HREC ref no. P2238). Notifications were based on laboratory evidence of probable or confirmed RRV infection, determined either via (i) isolation of RRV from patients, (ii) detection of RRV by nucleic acid testing, or (iii) IgG seroconversion or a significant increase in IgG antibody level in previously tested patients (Australian Government, 2019) and the suburb of residence of notified cases was recorded. As human notifications of RRV had high levels of interannual variability (Supplementary Table 1), mean annual notification rates from

2001 to 2016 were calculated to account for this and identify areas with persistently high or low cases of RRV. These were expressed relative to the population size of the suburb.

Suburbs were represented spatially using the ‘State Suburb (SSCs)’ geographical unit.

Notifications for each of Brisbane’s SSCs were matched to population data from the 2016 census, available from the Australian Bureau of Statistics (Australian Bureau of Statistics,

1999).

6.3.2 Site selection

Six SSCs were chosen across Brisbane (Figure 1). SSCs were chosen so that they (i) included a diverse range of human RRV notification rates, and (ii) varied geographical and ecological environments (urban, suburban, coastal, inland), and preferentially (iii) were included in Brisbane City Council (BCC) mosquito monitoring scheme. Study sites were identified within each of these SSCs, as described below.

120

Figure 1: a) Location of Brisbane Local Government Area within Australia, b) Distribution of SSCs in which study sites were located, and from which data was collected in Brisbane. BR = Bracken Ridge, CW = Chermside West, IN = Indooroopilly, CO = Corinda, WI = Wishart, LO = Lota.

Table 1. Ecological, geographical and social characteristics of suburbs as containing study sites. Human population size and density are derived from the 1Australian Bureau of Statistics

121

2016 census (Australian Bureau of Statistics, 2016) and the RRV notification rate is averaged from 2001-2016, from 2Queensland Department of Health data.

Average Suburb (SSC) Geography Human Human annual RRV name in which within Ecosystem(s) pop. Geocordinates pop. Size1 notification study sites Brisbane classification Density1 (2016) rate2 (2001- were located City (2016) 2016) 27.3190° S, North, Metropolitan, 19.92 Bracken Ridge 16,741 43/100,000 153.0340° E coastal wetland people/ha 27.4667° S, South, Metropolitan, 16.85 Lota 3,401 17/100,000 153.1833° E coastal wetland people/ha Metropolitan,

27.3814° S, riparian 17.3 Chermside North 5,883 33/100,000 153.0122° E bushland, dry people/ha West eucalypt forest 27.5500° S, Metropolitan, 23.44 South 11,238 18/100,000 Wishart 153.1000° E parkland people/ha 27.4984° S, South-west, Metropolitan, 17.04 Indooroopilly 12,800 24/100,000 152.9712° E on the river parkland people/ha 27.5443° S, South-west, Metropolitan, 17.63 5,084 50/100,000 Corinda 152.9824° E on the river parkland people/ha

6.3.3. Vertebrate surveys

Within each SSC, three study sites were selected so that they satisfied two additional criteria: (a) Sites were required to be an accessible public park or nature reserve where it was feasible to establish transects of adequate size, and (b) at least one site within each SSC had to include the location of the mosquito trap (selected by Brisbane City Council). Each study site comprised a 100 m x 100 m transect (marked using hand-held GPS devices; Garmin-62S) between 50m-1000m apart from other sites. Dawn and dusk vertebrate surveys were carried out monthly in each site over a 6-month period that encompassed the peak RRV transmission season, between October 2017 and March 2018. Methods were adapted from the ‘Standard

Early Bird Search’ and ‘Arboreal Spotlight Search’ detailed in the Queensland Fauna Survey

Guidelines (Eyre et al., 2014). In brief, dawn surveys commenced 30 minutes before sunrise and ran for 30 minutes in each site, sites within a SSC were sampled sequentially, and the order was mixed at each visit. The following data was recorded for all non-human vertebrate species: species, number of individuals, record type (i.e. seen, heard, seen and heard),

122

microhabitat (i.e. on the ground, in the canopy, above the canopy), and whether the observation was in, near (<50m) or off (>50m) the transect. The same data were also recorded for evening surveys, which commenced after sunset and were carried out in each transect for 30 person minutes. Equipment used to identify vertebrates included binoculars, handheld torches, field guides and the Birds of Australia mobile device application (Knight &

Menkhorst, 2010; Morcombe, 2003).

Mosquito surveys

Mosquitoes were collected across the six SSCs in Brisbane using light traps (Pacific

Biologics, Scarborough, Australia) baited with carbon dioxide (CO2; as approximately 1kg dry ice) and 1-octen-3-ol. Trapping was undertaken by BCC as part of its ongoing mosquito surveillance and control program. Traps were set once per week before dusk (starting the 5th of September 2017, ending the 26th of March) and collected the following morning. A single trap was located within one of the three sites in each SSC used for vertebrate surveys. Traps were set 1.5 m off the ground in proximity to a productive larval mosquito habitat.

Mosquitoes were identified to species level by a single experienced entomologist from BCC familiar with the mosquito fauna of Brisbane, with the aid of morphological keys (Lee et al.,

1989; Webb, Russell, & Doggett, 2016).

6.4 Data analysis

6.4.1 Variable analysis

Community ecology variables, including biomass (vertebrates only), abundance and species diversity, were calculated for vertebrate and mosquito communities within each site.

Species diversity was calculated using the Shannon-Weiner index (Boyle, Smillie, Anderson,

& Beeson, 1990). Biomass was calculated by multiplying the total number of observed individuals for a given species by the average adult body mass (kg) of that species, and log transformed for analysis. Body mass was derived from PANTHERIA dataset for mammals

123

(Jones et al., 2009) and Garnett et al. (2015) was used for birds. Linear regression models were used to test whether human notification rate for RRV was correlated with a given community ecology variable.

6.4.2 Community analysis

Vertebrate and mosquito communities within each SSC were analysed in two ways.

First, stacked column charts were constructed to visually represent community composition.

Secondly, to examine the effect of community compositions on human notification rates, non-metric multidimensional scaling (NDMS) analyses were performed using the vegan R package (Oksanen et al., 2010) and BiodiversityR (Kindt, 2019). These ordinations analysed community composition of vertebrates and mosquitoes (separately) for each month at each

SSC. The data were auto-transformed and scaled using Bray Curtis dissimilarity indices using the ‘metaMDS’ function. The degree of stress for each NDMS plot was calculated, which indicates the reliability of the outcome, i.e. lower stress corresponds with a higher reliability.

The ordination of elements was considered arbitrary for stress values of 0.3 or above. The dissimilarity matrixes were based on abundances for both vertebrates and mosquitoes, and an additional biomass ordination was generated for vertebrates.

6.5 Results

6.5.1 Human cases of RRV within each SSC

The mean annual notification rate across all Brisbane suburbs was 31 cases/100,000 people over year. The six SSCs selected for our study exhibited wide variation in the rate of reported RRV (Figure 2). Of these, Corinda reported the highest mean annual notification rate, followed by Bracken Ridge, Chermside West, Indooroopilly, Wishart and Lota, respectively.

124

Figure 2. Mean annual notification rate (per 100,000 population) for SSCs in Brisbane with a population greater than 1000. SSCs shown in orange are those within which study sites were located, and the dashed black line represents the average notification rate across all SSCs (31/100,000).

6.5.1 Vertebrate and mosquito survey summary statistics

Given the limited number of sites, and complexity of the system, the power to detect statistically significant results overall is limited. However, of the variables tested, log vertebrate biomass had a strong association with human notification rate (GLM: R-squared =

0.11, p=0.03, d.f. = 31). The suburbs reporting the highest human notification rates, Corinda and Bracken Ridge, had the greatest vertebrate biomass, which was followed by a decline in biomass in suburbs with lower human notification rates (Figure 3a). Total mosquito abundance was also significantly correlated with human notification rate (R-squared =

0.2271, p<0.01, d.f. = 31). However, this trend was not linear, with the Bracken Ridge SSC reporting the highest mosquito abundance, and the Wishart SSC reporting the lowest (Figure

3b). There were statistically significant differences in mosquito diversity, placental mammal abundance, marsupial diversity and marsupial abundance across suburbs, but these significant differences did not correlate with human notification rates (Supplementary Figure 1c, d, f, g).

No significant differences between SSCs, nor associations with human notification rates were

125

observed for vertebrate abundance and diversity, bird abundance and diversity, or placental mammal diversity (Supplementary Table 2, Supplementary Figure 1a, b, e, h, i).

Figure 3. Boxplots of (a) vertebrate biomass (on a log scale) and (b) mosquito abundance per suburb. Suburbs are ordered from highest human notification rate on the left to lowest on the

126

right. The boxes and whiskers show the extent of variation between months for the relevant variable per suburb.

6.5.3 Community composition

The following mosquito species were frequently reported across all SSCs: Culex annulirostris, Aedes procax, Ae. notoscriptus, Ae. vigilax, Anopheles annulipes,

Coquillettidia linealis, and Cx. sitiens (Figure 4). SSCs with the highest RRV notification rates were not distinguishable from other SSCs in terms of mosquito species composition. In all SSCs, except Bracken Ridge, both Cx. annulirostris and Ae. procax contributed > 55% of the species composition. Compared with other SSCs, Bracken Ridge had a unique species composition where Cx. annulirostris and Ae. procax comprised less than one third, while Ae. vigilax and Cq. linealis dominated (>62% combined). At least 20% of the species composition in Wishart and Lota (SSCs with the lowest RRV notification rates) were species that have not previously been considered as candidate vectors of RRV and, thus, their vector status is unknown (such as Cx. orbostiensis and Ae. vittiger (Harley et al., 2001; Jansen et al.,

2019)).

The two SSCs with the highest notification rates were distinguishable from all other

SSCs based on the vertebrate biomass composition (Figure 5). Specifically, Corinda and

Bracken Ridge had high total vertebrate biomass, almost entirely dominated by

Perissodactyla (horses), which was driven by the proximity of horses to a single sampling site. Otherwise, the composition patterns were relatively similar across the remaining four

SSCs. Minor differences included a larger proportion of marsupials (Diprotodonts) in

Chermside West and Indooroopilly compared to other SSCs, and a greater proportion of avian species (including Pscittaformes, Passeriformes and Pelicaniformes) in SSCs with the lowest human notification rates. Overall, the relative composition of Carnivores was generally consistent across SSCs with mid-range or low human RRV notification rates

127

(Chermside West, Indooroopilly, Wishart and Lota) ranging from 33.8% in Indooroopilly to

45.8% in Wishart.

100%

90%

80%

70% Other mosquito species 60% Cx sitiens 50% Cq linealis

40% An annulipes

30% Ae vigilax Ae notoscriptus 20%

Mosquitocommunity composition Ae procax 10% Cx annulirostris 0%

14000

12000

Other mosquito species 10000 Cx sitiens 8000 Cq linealis An annulipes 6000 Ae vigilax

4000 Ae notoscriptus Totalnumber mosquitoes

2000 Ae procax Cx annulirostris 0

Figure 4. Ross River virus mosquito community composition within each SSC, represented as (a) proportions and (b) totals. SSCs are ordered from highest (left) to lowest (right) RRV notification rate.

128

a) 100%

90%

80% Psittaciformes Pelecaniformes 70% Passeriformes 60% Diprodontia 50% Coraciiformes

40% Perissodactyla Chiroptera 30% Charadriiformes 20% Carnivora 10%

Vertebrate Vertebrate biomass community composition Anseriformes

0% Corinda Bracken Chermside Indooroopilly Wishart Lota Ridge West

25000

Psittaciformes 20000 Pelecaniformes Passeriformes Diprodontia 15000 Coraciiformes Perissodactyla 10000 Chiroptera

Total biomass Totalbiomass (kg) Charadriiformes Carnivora 5000 Anseriformes

0 Corinda Bracken Chermside Indooroopilly Wishart Lota Ridge West

Figure 5. Vertebrate community biomass (kg) categorised by taxonomic class within each SSC, represented as (a) proportions and (b) totals. SSCs are ordered from highest human notification rate (left) to lowest (right) notification rate.

129

6.5.4 Ordinations

Dissimilarity matrices resulting from NMDS analyses had moderately high stress, and demonstrated no clear groupings of suburbs with higher or lower RRV notification rates based on vertebrate biomass, vertebrate abundance or mosquito abundance. In general, surveys grouped most closely together by the SSC they were undertaken in (i.e. Bracken

Ridge, or Indooroopilly), but showed no trend in grouping between the highest to lowest human notification rate (Figure 6).

130

Figure 6. Results of NMDS analysis for the (a) vertebrate biomass (auto-transformation = square root, stress = 0.1922) and (b) mosquito abundance (auto-transformation = square root, stress = 0.2051). Each point represents a unique monthly survey within each SSC. SSCs are labelled on the right-hand side and coloured from highest (dark red) to lowest (light yellow) notification rate.

6.6 Discussion By integrating data on the abundance, diversity and biomass of potential RRV vertebrate reservoirs and mosquito vectors, we identified associations between high RRV notification rates in humans, total mosquito abundance and total vertebrate biomass

(particularly due to the presence of horses).

While variability existed between study sites, the overall positive relationship observed between total mosquito abundance and long-term human notification rates of RRV is as expected for a mosquito-borne virus. This result is also consistent with previous investigations, which found that a greater number of mosquitoes (Lindsay et al., 1996;

Lindsay et al., 1997), or closer proximity to mosquito habitats (Jardine, Neville, & Lindsay,

2015) corresponded with higher human notification rates (Ryan, Do, & Kay, 1999).

Importantly, however, this measure of overall mosquito abundance does not necessarily take the vector competence or vectorial capacity of specific species into account. Ross River virus is a multi-vector pathogen that has been isolated from more than 30 species of mosquitoes, each differing in general biology, ecology and, most notably, host feeding patterns (Claflin &

Webb, 2015; Stephenson et al., 2019). Thus, it is likely that some species are more important than others as enzootic and/or bridge vectors. Further investigations are required to implicate specific mosquitoes as the most important vectors of RRV in Brisbane, however we observed notable differences in mosquito communities between suburbs. The dominance of Ae. vigilax and Cq. linealis in Bracken Ridge is of interest as these two species have demonstrated relatively high vector competence under laboratory investigations (Jeffery et al., 2002; Ryan,

131

Do, & Kay, 2000) and frequently feed on both humans and horses (which were abundant in this suburb) (Stephenson, Murphy, et al., 2019). For suburbs that reported low RRV notification rates several of the mosquito species collected have not previously been investigated as vectors of RRV (no virus isolations, or transmission studies; e.g. Cx. orbostiensis and Ae. vittiger (Harley et al., 2001)). While these species are present in relatively low proportions in these sites, the lack of data on their competence and feeding patterns makes it difficult to determine their contribution to RRV transmission.

The significant positive relationship between vertebrate biomass and RRV notification rate is novel, and of interest when mosquito host seeking behaviour is considered.

Studies on mosquito bloodmeals have found that mosquitoes can feed preferentially on particular vertebrate species, independent of their relative abundance (Janousek et al., 2014;

Kilpatrick et al., 2006; Lyimo & Ferguson, 2009). When seeking a bloodmeal, mosquitoes detect hosts through CO2 and body heat (Takken & Verhulst, 2013), in addition to numerous olfactory factors. Both CO2 and heat emission increase with vertebrate body size (Franz et al.,

2010). Biomass accounts for the size of vertebrates as well as their total abundance. As such, although marsupials are considered potential reservoirs of RRV, their biomass contributed less than 5% of the community composition in four of the six SSCs. In SSCs with relatively low RRV notification rates, there was a moderate biomass of species that have an unknown or limited reservoir potential, including Carnivores (namely cats and dogs) and Psittaformes

(namely rainbow lorikeets Trichoglossus moluccanus and little corellas Cacatua sanguinea)

(Boyd et al., 2001; Stephenson et al., 2018). By comparison, sites with higher notification rates were dominated by a relatively large biomass of horses, which are competent amplifiers when experimentally infected (Kay et al., 1987; Stephenson et al., 2018), have demonstrated high exposure under natural conditions (Gummow et al., 2018; Stephenson, Rudd, et al.,

2019) and have yielded multiple RRV isolates (Azuolas, 1998). The potential role of horses

132

in the maintenance and transmission of RRV should be explored further, as their high biomass and interactions with vectors may be important for spillover of RRV to human populations. Overall, our findings suggest that vertebrate biomass rather than abundance alone should be assessed in greater detail for RRV ecology, but it is most informative when combined with existing knowledge on the amplification potential of vertebrate species.

The ordinations offered little insight into the disease ecology of RRV. Specifically, no relevant groupings of sites or human notification were ascertained from either the vertebrate or mosquito communities. One possible explanation may be that all sites are situated within the same bioregion (‘South Eastern Queensland’), characterised by similar climatic conditions, with the furthest distance between sites ~30 km. Although it is assumed that most vertebrates and mosquitoes did not disperse widely outside of the sites, flying foxes (Roberts et al., 2012; Tidemann & Nelson, 2004), some bird species (Smith & Smith, 2012) and Ae. vigilax (Chapman et al., 1999; Webb & Russell, 2019) readily travel distances greater than 10 km. In general, while the total community did not vary significantly by SSC within this study, it is unlikely that this would be the case if sites with a greater geographic separation were considered, particularly since RRV has a nationwide distribution which supports diverse mosquito and vertebrate communities.

6.6.1 Limitations and future directions

Confounding factors within this study include the number of replicates, inherent survey method bias, and the choice of sites based on human notifications. First, the limited number and purposeful selection of sites and survey months means that our findings may not be generalisable across Brisbane or to other cities. Nevertheless, this study provides the most comprehensive assessment of vertebrate and mosquito communities in comparison to human notifications to date. The nature of the data collection, both for mosquito and vertebrate communities is time consuming and costly. We found the use of a single mosquito trap in

133

each SSC successful for capturing a diversity of mosquito communities which largely corresponded to the habitat conditions (such as freshwater or saltwater habitats). Future studies, however, would benefit from additional traps. There would be great value for future investigations to expand data collection with additional SSCs across the gradient of human notifications, as well as additional mosquito traps and host surveys.

A caveat of selecting study sites on the basis of human notifications is that it may not accurately represent locations with high or low infection rates of RRV. Not all cases of RRV cause clinical manifestations and/or are notified, and the place of residence of an infected person may not have been the place of infection. We found a long-term notification rate to be most suitable to identify areas with persistent RRV cases, which we assume are more likely to be locally acquired. The sparsity of RRV cases on an annual basis limits that ability to match human notifications to the same sample period as the vertebrate and mosquito data was collected, as these are relatively low numbers.

Future studies on RRV community ecology would benefit from including broader vertebrate survey methods (such as trapping and mosquito blood meal analysis) to incorporate a wider diversity of vertebrates. For example, small and/or cryptic vertebrates, such as murids, were infrequently detected in this study. However, some serological evidence suggests that these species are exposed to RRV (Vale, Spratt, & Cloonan, 1991). Inclusion of mosquito blood meal analyses would further inform the potential role of the vertebrates considered in this study, by demonstrating which mosquito vectors readily interact with which vertebrate hosts. Of three published blood-feeding studies conducted in Brisbane between 1995 and 2001, birds, humans, placental mammals and marsupials were all variously dominant among the vertebrate species fed on by mosquitoes, but this was likely influenced by differing mosquito collection sites between the studies (Jansen et al., 2009; Kay et al.,

2007; Ryan et al., 1997). Only one of these three studies attempted to quantify the relative

134

availability of host species in trapping sites, finding that, overall, dogs were the most dominant blood meal relative to their abundance, but this was based largely on subjective householder survey (rather than vertebrate observation surveys) and was undertaken in SSCs where animals with higher biomass such as horses were not abundant (Kay et al., 2007). This also supports the inclusion of vertebrate and blood meal surveys, plus consideration of the effect of biomass in future studies.

6.7 Conclusion

Improved understanding of both mosquito and vertebrate communities can lead to a better prediction and management of transmission for zoonotic mosquito-borne diseases.

Here, we find that for RRV vertebrate biomass (particularly horses) and mosquito abundance had the strongest association with human notification rates of RRV in selected suburbs of

Brisbane. Further investigations, including ongoing vertebrate surveillance and modelling studies, are needed to quantify the importance of vector competences and mosquito feeding patterns in association with the novel finding of vertebrate biomass from this study. This study adopts a multidisciplinary approach to consider both vertebrate host and vector contributions to RRV ecology, and highlights both the benefits and challenges associated with this study design.

Acknowledgements

The collection of this data would not have been possible without the following field volunteers: Georgia Braun, Cara Parsons, Ian Parsons, Hannah Thomas, Jack Dodd, Scout

Fisher, Stevie Tozer, Stuart MacLeod, Robin Rowland, Jaylan Schabrod, Michael Johnson and Trina Kateifides.

We would like to acknowledge Lara Herrero for supervising the PhD study of which this research is a part, and more broadly HM lab members and Lara Herrero lab members for feedback on data collection and analysis.

135

This project was supported by an Ecological Society of Australia Graduate Grant. EBS was supported by an Australian Government Research Training Program, and a Griffith Graduate

Research School Publication Assistance Scholarship. AJP was supported by a Queensland

Government Accelerate Postdoctoral Research Fellowship.

Author Disclosure Statement

No competing financial interests exist.

6.8 References

Australian Bureau of Statistics. (1999). Australian Standard Geographical Classification (AGSC) Retrieved from https://www.abs.gov.au/ausstats/[email protected]/Lookup/by%20Subject/1270.0.55.001~July%2020 16~Main%20Features~Non%20ABS%20structures~10008 Australian Bureau of Statistics. (2016). 2016 Census QuickStats. Retrieved from https://quickstats.censusdata.abs.gov.au/census_services/getproduct/census/2016/quickstat/3G BRI?opendocument Australian Government. (2018). National Notifiable Diseases Surveillance System. Retrieved from http://www9.health.gov.au/cda/source/cda-index.cfm Australian Government. (2019). Ross River virus infection case definition. Department of Health,. Retrieved from https://www1.health.gov.au/internet/main/publishing.nsf/Content/cda-surveil- nndss-casedefs-cd_rrv.htm Azuolas, J. K. (1998). Ross River virus disease of horses. Australian Equine Veterinarian, 16(2), 56- 58. Boyd, A. M., Hall, R. A., Gemmell, R. T., & Kay, B. H. (2001). Experimental infection of Australian brushtail possums, Trichosurus vulpecula (Phalangeridae : Marsupialia), with Ross River and Barmah Forest viruses by use of a natural mosquito vector system. American Journal of Tropical Medicine and Hygiene, 65(6), 777-782. Boyle, T. P., Smillie, G. M., Anderson, J. C., & Beeson, D. R. (1990). A sensitivity analysis of nine diversity and seven similarity indices. Research Journal of the Water Pollution Control Federation, 749-762. Chapman, H. F., Hughes, J. M., Jennings, C., Kay, B. H., & Ritchie, S. A. (1999). Population structure and dispersal of the saltmarsh mosquito Aedes vigilax in Queensland, Australia. Medical and Veterinary Entomology, 13(4), 423-430. doi:10.1046/j.1365-2915.1999.00195.x Claflin, S. B., & Webb, C. E. (2015). Ross River Virus: Many Vectors and Unusual Hosts Make for an Unpredictable Pathogen. Plos Pathogens, 11(9). doi:10.1371/journal.ppat.1005070 Eyre, T. J., Ferguson, D. J., Hourigan, C. L., Smith, G. C., Mathieson, M. T., Kelly, A. L., . . . Rowland, J. (2014). Terrestrial vertebrate fauna survey assessment guidelines for Queensland. Department of Science. Information Technology, Innovation and the Arts, Queensland Government, Brisbane. Franz, R., Soliva, C. R., Kreuzer, M., Steuer, P., Hummel, J., & Clauss, M. (2010). Methane production in relation to body mass of ruminants and equids. Evolutionary Ecology Research, 12(6), 727-738. Gardner, A. M., Muturi, E. J., Overmier, L. D., & Allan, B. F. (2017). Large-Scale Removal of Invasive Honeysuckle Decreases Mosquito and Avian Host Abundance. Ecohealth, 14(4), 750-761. Garnett, S. T., Duursma, D. E., Ehmke, G., Guay, P.-J., Stewart, A., Szabo, J. K., . . . Drynan, D. (2015). Biological, ecological, conservation and legal information for all species and subspecies of Australian bird. Scientific data, 2, 150061.

136

Gummow, B., Tan, R. H. H., Joice, R. K., Burgess, G., & Picard, J. (2018). Seroprevalence and associated risk factors of mosquito‐borne alphaviruses in horses in northern Queensland. Australian Veterinary Journal, 96(7), 243-251. Harley, D., Sleigh, A., & Ritchie, S. (2001). Ross River virus transmission, infection, and disease: a cross-disciplinary review. Clinical Microbiology Reviews, 14(4), 909-932. doi:10.1128/cmr.14.4.909-932.2001 Hill, C. A., Kafatos, F. C., Stansfield, S. K., & Collins, F. H. (2005). Arthropod-borne diseases: vector control in the genomics era. Nature Reviews Microbiology, 3(3), 262. Hoffmann, A. A., Ross, P. A., & Rašić, G. (2015). Wolbachia strains for disease control: ecological and evolutionary considerations. Evolutionary applications, 8(8), 751-768. Janousek, W. M., Marra, P. P., & Kilpatrick, A. M. (2014). Avian roosting behavior influences vector-host interactions for West Nile virus hosts. Parasites & Vectors, 7(1), 399. Jansen, C. C., Shivas, M., May, F., Pyke, A., Onn, M., Lodo, K., . . . van den Hurk, A. (2019). Epidemiologic, Entomologic, and Virologic Factors of the 2014–15 Ross River Virus Outbreak, Queensland, Australia. Emerging Infectious Diseases, 25(12). Jansen, C. C., Webb, C. E., Graham, G. C., Craig, S. B., Zborowski, P., Ritchie, S. A., . . . van den Hurk, A. F. (2009). Blood sources of mosquitoes collected from urban and peri-urban environments in eastern Australia with species-specific molecular analysis of avian blood meals. American Journal of Tropical Medicine and Hygiene, 81(5), 849-857. doi:10.4269/ajtmh.2009.09-0008 Jardine, A., Neville, P. J., & Lindsay, M. (2015). Proximity to Mosquito Breeding Habitat and Ross River Virus Risk in the Peel Region of Western Australia. Vector-Borne and Zoonotic Diseases, 15(2), 141-146. doi:10.1089/vbz.2014.1693 Jeffery, J. A., Ryan, P. A., Lyons, S. A., & Kay, B. H. (2002). Vector competence of Coquillettidia linealis (Skuse) (Diptera : Culicidae) for Ross River and Barmah Forest viruses. Australian Journal of Entomology, 41, 339-344. doi:10.1046/j.1440-6055.2002.00316.x Jones, K. E., Bielby, J., Cardillo, M., Fritz, S. A., O'Dell, J., Orme, D. L., . . . Carbone, C. (2009). PanTHERIA: a species‐level database of life history, ecology, and geography of extant and recently extinct mammals. Ecology, 90(9), 2648-2648. Kay, B. H., Boyd, A. M., Ryan, P. A., & Hall, R. A. (2007). Mosquito feeding patterns and natural infection of vertebrates with Ross River and Barmah Forest viruses in Brisbane, Australia. American Journal of Tropical Medicine and Hygiene, 76(3), 417-423. Kay, B. H., Hall, R. A., Fanning, I. D., Mottram, P., Young, P. L., & Pollitt, C. C. (1986, 1986). Experimental infection of vertebrates with Murray Valley encephalitis and Ross River viruses. Paper presented at the Arbovirus Research in Australia. Kay, B. H., Pollitt, C. C., Fanning, I. D., & Hall, R. A. (1987). The experimental infection of horses with Murray Valley encephalitis and Ross River viruses. Australian Veterinary Journal, 64(2), 52-55. Kilpatrick, A. M., Daszak, P., Jones, M. J., Marra, P. P., & Kramer, L. D. (2006). Host heterogeneity dominates West Nile virus transmission. Proceedings of the Royal Society of London B: Biological Sciences, 273(1599), 2327-2333. Kindt, R. (2019). Package ‘BiodiversityR’: Obtenido de http://vps. fmvz. usp. br/CRAN/web/packages/BiodiversityR …. Knight, F., & Menkhorst, P. (2010). Field guide to mammals of Australia: Oxford University Press: Melbourne. Lee, D. J., Hicks, M. M., Griffiths, M., Russell, R. C., Bryan, J. H., & Marks, E. N. (1989). The Culicidae of the Australian region. Canberra: Australian Government Publishing Service, 7. Lindsay, M., Oliveira, N., Jasinska, E., Johansen, C., Harrington, S., Wright, A. E., & Smith, D. (1996). An outbreak of ross river virus disease in southwestern Australia. Emerging Infectious Diseases, 2(2), 117-120. Lindsay, M., Oliveira, N., Jasinska, E., Johansen, C., Harrington, S., Wright, T., . . . Shellam, G. (1997). Western Australia' s largest recorded outbreak of Ross River virus disease. Arbovirus Research in Australia, 7, 147-152.

137

Lyimo, I. N., & Ferguson, H. M. (2009). Ecological and evolutionary determinants of host species choice in mosquito vectors. Trends Parasitology, 25(4), 189-196. doi:10.1016/j.pt.2009.01.005 Morcombe, M. K. (2003). Field guide to Australian birds: Steve Parish Publishing. Oksanen, J., Blanchet, G. F., Kindt, R., Legendre, P., O’hara, R. B., Simpson, G. L., . . . Wagner, H. (2010). Vegan: community ecology package. R package version 1.17-4. http://cran. r-project. org>. Acesso em, 23, 2010. Queensland Government. (2019). Notifiable conditions annual reporting. Retrieved from https://www.health.qld.gov.au/clinical-practice/guidelines-procedures/diseases- infection/surveillance/reports/notifiable/annual Roberts, B. J., Catterall, C. P., Eby, P., & Kanowski, J. (2012). Long-distance and frequent movements of the flying-fox Pteropus poliocephalus: implications for management. Plos One, 7(8), e42532. Russell, R. C. (2002). Ross River virus: Ecology and distribution. Annual Review of Entomology, 47, 1-31. doi:10.1146/annurev.ento.47.091201.145100 Ryan, P. A., Do, K. A., & Kay, B. H. (1999). Spatial and temporal analysis of Ross River virus disease patterns at Maroochy Shire, Australia: Association between human morbidity and mosquito (Diptera : Culicidae) abundance. Journal of Medical Entomology, 36(4), 515-521. Ryan, P. A., Do, K. A., & Kay, B. H. (2000). Definition of Ross River virus vectors at Maroochy Shire, Australia. Journal of Medical Entomology, 37(1), 146-152. doi:10.1603/0022-2585- 37.1.146 Ryan, P. A., Martin, L., Mackenzie, J. S., & Kay, B. H. (1997). Investigation of gray-headed flying foxes (Pteropus poliocephalus) (Megachiroptera: Pteropodidae) and mosquitoes in the ecology of Ross river virus in Australia. American Journal of Tropical Medicine and Hygiene, 57(4), 476-482. Smith, P., & Smith, J. (2012). Climate change and bird migration in south-eastern Australia. Emu- Austral Ornithology, 112(4), 333-342. Stephenson, E. B., Murphy, A. K., Jansen, C. C., Peel, A. J., & McCallum, H. (2019). Interpreting mosquito feeding patterns in Australia through an ecological lens: an analysis of blood meal studies. Parasites & Vectors, 12(1), 156. Stephenson, E. B., Peel, A. J., Reid, S. A., Jansen, C. C., & McCallum, H. (2018). The non-human reservoirs of Ross River virus: a systematic review of the evidence. Parasites & Vectors, 11(1), 188. Stephenson, E. B., Rudd, P. A., Peel, A. J., McCallum, H., Reid, S. A., & Herrero, L. J. (2019). Species traits and hotspots associated with Ross River virus infection in non-human vertebrates in South-East Queensland. Transactions of the Royal Society of Tropical Medicine and Hygiene, Submitted 26th July 2019. Takken, W., & Verhulst, N. O. (2013). Host preferences of blood-feeding mosquitoes. Annual Review Entomology, 58, 433-453. doi:10.1146/annurev-ento-120811-153618 Tidemann, C. R., & Nelson, J. E. (2004). Long‐distance movements of the grey‐headed flying fox (Pteropus poliocephalus). Journal of Zoology, 263(2), 141-146. Vale, T. G., Spratt, D. M., & Cloonan, M. J. (1991). Serological evidence of arbovirus infection in native and domesticated mammals on the south coast of New South Wales Australia. Australian Journal of Zoology, 39(1), 1-8. van den Hurk, A. F., Craig, S. B., Tulsiani, S. M., & Jansen, C. C. (2010). Emerging tropical diseases in Australia. Part 4. Mosquitoborne diseases. Annals of Tropical Medicine and Parasitology, 104(8), 623-640. doi:10.1179/136485910x12851868779984 Webb, C. E., Russell, R., & Doggett, S. (2016). A guide to mosquitoes of Australia: Csiro Publishing. Webb, C. E., & Russell, R. C. (2019). Dispersal of the Mosquito Aedes vigilax (Diptera: Culicidae) From Urban Estuarine Wetlands in Sydney, Australia. Journal of Medical Entomology. Weinstein, P. (1997). An ecological approach to public health intervention: Ross River virus in Australia. Environmental Health Perspectives, 105(4), 364-366. doi:10.2307/3433322

138

6.9 Supplementary Information

Supplementary Table 1. Reported RRV notifications between 2001 and 2016 for SSC’s included in this study, and the average annual notification rate within each SSC (calculated per 100,000 people).

Year Bracken Chermside Corinda Indooroopilly Lota Wishart Ridge West notifications notifications notifications notifications notifications notifications 2001 3 2 1 4 0 2 2002 1 0 0 0 0 0 2003 13 3 5 2 0 3 2004 8 1 7 0 3 2 2005 1 0 0 0 0 1 2006 13 2 3 2 0 0 2007 5 2 0 4 0 1 2008 6 3 3 8 0 4 2009 5 1 0 2 0 4 2010 4 2 3 0 0 2 2011 4 1 1 0 0 1 2012 8 1 1 10 0 1 2013 4 2 2 3 0 3 2014 12 3 6 1 3 2 2015 26 11 8 10 3 4 2016 4 0 0 1 0 0 Average 43/100,000 33/100,000 50/100,000 24/100,000 17/100,000 18/100,000 annual notification rate

Supplementary Table 2. Reported significance for notification rates across variables

lm(Notification.Rate ~ Variable) Variable Significance (p-value) R-squared Mosquito abundance <0.01 0.227 Mosquito diversity 0.830 -0.031 Vertebrate abundance 0.108 0.052 Vertebrate diversity 0.374 -0.006 Vertebrate biomass <0.001 0.254 Mammal abundance 0.883 -0.032 Mammal diversity 0.388 -0.007 Marsupial abundance 0.690 -0.027 Marsupial diversity 0.520 -0.018 Bird abundance 0.091 0.060 Bird diversity 0.561 -0.021

139

140

141

142

143

Supplementary Figure 1. Variance within and between monthly surveys across SSCs for vertebrate abundance (a), diversity (b), mosquito diversity (c), placental mammal abundance (d), placental mammal diversity (e), marsupial abundance (f), marsupial diversity (g), bird abundance (h) and bird diversity (i). Sites are ordered from highest human notification rate on the left to lowest on the right.

144 Chapter 7

Discussion and conclusions

7.1 Research summary

The research presented in this dissertation addressed the limited understanding of the reservoirs of Ross River virus (RRV) using a multifaceted approach. The approach consisted of three processes: interpretation of the current and historic evidence for non-human reservoirs and their interacting vectors, assessment of contemporary serological exposure of RRV in vertebrates, and conducting ecological investigations to quantify the availability of both vertebrates and vectors across a spatial gradient of human infection rates. Given that the virus persists across a broad geographic range, climates and seasons (in which multiple vectors and vertebrates involved in transmission frequently shift), it was clear that identifying the specific reservoir species for RRV would not be possible within a single dissertation. Regardless, the outcomes from the work presented in this dissertation have advanced the understanding of RRV transmission under the following themes: i) the identification of likely and unlikely RRV reservoir species for future investigations, ii) the application of a multidisciplinary approach for investigating arbovirus transmission communities, iii) providing the foundations of a multi- host, multi-vector model for RRV, and finally iv) the identification of challenges and future research on RRV transmission ecology. Rather than enumerate results from previous discussion sections in this chapter, below is a general discussion on these major themes and the implications of work within this dissertation.

7.2 Non-human reservoirs of RRV

7.2.1 Horses

This dissertation provides further evidence that horses play an important role in RRV transmission, highlighting that critical attention should be placed on this species in future investigations. Historically, horses have been identified as a reservoir species of interest due to their competence under experimental infection studies (Kay et al. 1986) and their high seroprevalence under natural conditions (Gummow et al. 2018; Azuolas 1998) (reviewed in Chapter 2 (Stephenson et al. 2018)). Despite this knowledge, there has been a paucity of studies investigating the importance of the role of horses in RRV transmission to humans. In this

145 dissertation, new evidence that horses may be important for RRV transmission came from the ecological study (Chapter 6), in combination with findings from the mosquito blood meal study (Chapter 3). In the ecological study, horses were highly ‘available’ in suburbs with the highest rates of human disease, contributing more than 90% of the available biomass. Aedes vigilax and Cq. linealis were among the most commonly caught mosquito species in the same areas. These species demonstrated the highest feeding correlation with horses and humans (Chapter 3), potentially highlighting the importance of their role as bridge hosts and the combination of these species contributing to ongoing RRV transmission. A key determining factor in assessing the role of horses as reservoirs of RRV is interpreting the high seroprevalence reported in Chapter 5. Horses had the highest seroprevalence with more than 90% of individuals testing positive for RRV exposure, despite the horses enrolled in this study having displayed no symptoms in the 12 months prior to collection. As a long-lived species, it is critical to know whether RRV antibodies in horses are protective against reinfection and the duration of that immunity. Current knowledge on the immune response of horses to RRV infection is limited. If horses are highly attractive to mosquito vectors of RRV (resulting in high exposure rates) but develop a long-lived immune response to RRV, this could result in high herd immunity in horse populations. Re-exposure of a previously infected individual would not contribute to ongoing transmission and thus, a high herd immunity would implicate this species as a dead-end host. This conclusion would be contradictory to the hypothesis that horses are important as reservoirs for RRV and a new finding from this dissertation for RRV reservoirs. Recommendation for future research – Given the nature and longevity of the immune response in horses is such a critical determinant of their role as potential reservoirs, the findings from this dissertation suggests that future investigations (including modelling and long-term immunity studies) should prioritise this knowledge gap. Furthermore, as horses are a domestic species and are of interest from an economic perspective, studies investigating the acute and long-term immune responses of horses to RRV infection is more feasible and accessible than studies of wildlife.

7.2.2 Ringtail possums

Among the marsupial species investigated here, ringtail possums should be prioritised in future studies of RRV reservoirs. Ringtail possums did not feature in historical assessments (virus isolations, serology or experimental infections) of RRV reservoirs (Chapter 2), nor has the

146 species been assessed in previous blood meal studies (Chapter 3). However, the serological investigation undertaken in Chapter 5 indicated that they were frequently exposed to RRV, and significantly more so than brushtail possums. Brushtail possums have previously been demonstrated to have a high capacity as an amplifier under experimental infection studies (Boyd et al. 2001), but may not be frequently exposed to the virus (Kay et al. 2007; Hill, Power, and Deane 2009) under natural conditions, limiting their capacity as reservoirs. The high seroprevalence in ringtail possums identified here, their abundance in Australian suburbs, and a complete lack of knowledge on their amplification capabilities, or the mosquito feeding patterns on them, make the species a priority for future investigations. Recommendations for future research – The greatest advance in this knowledge gap would be to undertake experimental infection studies using multiple vector species to identify whether ringtail possums have the capacity to amplify RRV. Experimental infection studies are limited in size due to cost and welfare considerations, so should be designed with mathematical models in mind. One alternative would be to utilise the new population and exposure knowledge identified in this dissertation and integrate this data in future models. A mathematical model incorporating factors of ringtail possum biology may inform whether or not this species has the potential to increase or decrease the transmission in a given community.

7.2.3 Flying foxes and birds

In comparison to other species, there was equivocal evidence to support birds and flying foxes as important reservoirs of RRV. Chapter 2 highlighted that multiple species within these taxa can infect susceptible mosquitoes with RRV, suggesting they could contribute to transmission. Further supporting this, comparable free living populations are exposed to RRV, with Chapter 5 reporting seropositivity in flying foxes as 25.3% and seropositivity across birds (13 out of 22 species) as 28.9%. In comparison to other species in my study, these are among the lowest exposure rates. Given that these species were identified to have high abundance counts in the ecological study area (Chapter 6), low exposure rates may still contribute to circulation of the virus in the environment, but this is dependent on contact rates with vectors. In Chapter 3, flying foxes were rarely reported in mosquito blood meal studies in Australia, but this is likely a bias as they were rarely tested. For birds, two mosquito species predominantly fed on birds; Cx. sitiens and Cx. quiquefasciatus. These are among the 40+ species of mosquito from which RRV has been isolated, but are not recognised as the most important vectors for RRV transmission to humans (van den Hurk et al. 2010). As with all species within this dissertation, little is known about the immune response to RRV in birds and flying foxes, in particular if

147 they develop on-going immunity following exposure. Most other alphaviruses are commonly maintained in a mammalian-mosquito cycles (Weaver and Barrett 2004), though Sindbis virus is an exception (Modlmaier et al. 2002), for which passerine birds are considered important reservoir hosts. Further investigations of birds and flying foxes as potential reservoirs would advance knowledge on RRV reservoirs, however, if resources are finite, these are not species of the highest priority.

7.2.4 Humans

Broadly, humans are not considered important in the maintenance of RRV. However, understanding exposure in human population provides insight on RRV transmission cycles. When comparing RRV transmission to that of other arboviruses (Chapter 4), RRV had a wide geographic exposure, but was most prevalence in Australia. Human exposure may be increasing in the Pacific Islands (Duncombe et al. 2013; Tesh et al. 1975), but data is still lacking on the availability and relative importance of vectors and hosts in these areas. A recommendation for future research would be to test for RRV in humans in South-East Asia (and more broadly) to determine its full geographical extent; and in turn the potential global threat RRV may pose (Flies et al. 2018).

7.3 A multidisciplinary approach for arboviral disease ecology

Table 7.1 summarises the key findings within each chapter of this dissertation, and importantly, highlights the multidisciplinary methods that have been adopted to address the key research aims. For spillover of zoonotic arboviruses to occur in human populations, a number of intraspecific and extrinsic factors must align, including host immunity, vector availability and contact between vectors and hosts (Zinsstag et al. 2011; Plowright et al. 2017; Tompkins and Slaney 2014; Craft et al. 2009). To understand any one of these factors, different research methods must be applied. This body of work brings together methods and analyses across multiple disciplines—including veterinary medicine, virology, public health and ecology—in an attempt to do just this. This approach is critical for investigation of multi-host vector borne zoonotic pathogens as disease patterns in humans are inextricably linked to the ecology of and interactions among reservoir hosts, vectors and pathogens.

A multidisciplinary approach has been advantageous in this dissertation because it provided a more comprehensive understanding when compared with results that are typically discipline-

148 specific. The best example of this is in Chapter 5, which required undertaking, interpreting and comparing serological assays across numerous non-model species (Table 7.1). Given that many methods are traditionally optimised for humans and were not appropriate for the diversity of species tested within this study, multiple virological and serological methods were compared to identify an optimal method across species. The real strength to this chapter however, was combining the serological results with ecological and life-history traits for each individual. This extended beyond the virology discipline and identified important traits and locations for seroprevalence in non-human vertebrates across South-East Queensland.

Multidisciplinary approaches have been adopted successfully for other zoonotic arboviruses, but a challenge for multidisciplinary approaches is that they are dependent on existing infrastructure within the scientific community to support collaboration. Within Australia this is less of an issue compared with in lower income countries because infrastructure (including veterinary networks, nationally notifiable diseases surveillance for human health, local and state government surveillance and control programs for vectors and pest vertebrate species) are available for multidisciplinary studies. The same resources, however, are not as well established elsewhere, and, as such, limits the advancement of understanding RRV transmission throughout its distribution.

149 Table 7.1. Summary of the aims, methods and main findings from this dissertation. Chapter Aim Method(s) Key finding(s) 2 – Literature review 1. Critically review the evidence Systematic literature 1. Evidence to support the hypothesis was weaker than - Non-human supporting the hypothesis review of published expected. reservoirs ‘marsupials are better reservoirs papers. 2. Future research on the non-human reservoirs of RRV Published - of RRV than placental should focus on investigating non-marsupial species, (Stephenson et al. mammals, which in turn are including passerine birds and small placental mammals. 2018) better reservoirs than birds’; Ideally this would be done through ecological assessments of 2. Characterise the limitations of vector, virus and host abundance in areas of high and low that evidence; disease in humans. 3. Identify research gaps with regards to RRV transmission cycles. 3 – Data chapter – 1. Synthesised existing literature Systematic literature 1. Of the Australian mosquito species tested, each had a Blood meal meta- describing blood meal studies in review + unique feeding pattern; however, the particular specialist or analysis Australia; log odds ratio analysis. generalist feeding patterns of mosquito species could be a Published - 2. Assess the most likely key determinant of the risk they pose for human disease. (Stephenson, Murphy, mosquito-host associations; 2. Broader ecological considerations alongside these feeding et al. 2019) 3. Distinguish between generalist patterns could be useful for the interpretation of these and specialist mosquito species complex biological systems, but at present data available to in Australia. do this are limited. Future studies should utilise multidisciplinary approaches to collect data on vertebrate communities in parallel with mosquito communities. 4 – Literature review 1. Identify epidemiologically Systematic literature 1. RRV may have been circulating outside Australia at least – relevant trends, potential biases review + since 1975 when human seropositivity between 20 and 34% Human and risk factors associated with meta-analysis using was reported in Indonesia and Papua New Guinea. seroprevalence DENV, RRV and BFV forest plots. Transmission may also be increasing in some countries in Manuscript seropositivity in humans in the Pacific. Australia and the PICTs. 2. Locally in Queensland, average seroprevalence in humans (since 1966-2015) has been reported as 19.7% seropositivity.

150 5 – Data chapter – 1. Characterise the natural Serological assay 1. Species traits, in particular diet and body mass, may play a Vertebrate exposure of RRV in non-human (plaque reduction role in vector-host contact rates, ultimately influencing seroprevalence vertebrates; neutralisation test) + exposure. Submitted - 2. Classify species and/or life- random forest 2. The highest seroprevalence was reported in horses, (Stephenson, Rudd, et history traits associated with regression analysis + followed by koalas and ringtail possums. Ringtail possums al. 2019) seropositivity; spatial hotspot analysis. were significantly more seropositive than brushtail possums, 3. Identify any potential hotspots which is particularly interesting as only the latter have been or cold spots of seropositivity. investigated as potential reservoirs for RRV. 3. The results from this study provide a novel method to examine serological results and ultimately inform management of RRV in South-East Queensland. 6 – Data chapter – 1. Characterise vector and host Vertebrate surveys + 1. Abundance of marsupials and mosquitoes was the most Field ecology communities across different vector surveys + important ecological value correlating with human incidence investigations suburbs of Brisbane, with ordination analysis + rates. Submitted – varying human incidence rates linear models. 2. When assessing vertebrates by their community (Stephenson, Murphy, of RRV; composition, biomass makes a significant difference to the et al. 2019) 2. Determine if variables biomass, ‘availability’ of potential hosts. abundance or diversity are most important in areas of high or low disease for vectors or vertebrates .

151 7.4 Towards a multi-host, multi-vector model for RRV

Modelling plays a central role in disease ecology. In principle, mathematical models are an attempt to generalize or understand an ecological process. Modelling can be particularly useful to generate testable hypotheses to guide the collection of data or management scenarios for disease control (Restif et al. 2012; McCallum 2016). Mosquito-borne diseases have a long history of use in mathematical modelling, with the earliest model for malaria developed more than a century ago (Reiner et al. 2013; Smith et al. 2012; Ross 1915). Since then, global interest in mosquito-borne diseases has expanded and in response, models have been developed to include different components important to transmission including; immune response, infection dynamics, seasonality, pathogen evolution, host and mosquito behaviour, and multiple host or mosquito species.

In this dissertation, the primary interest is to identify non-human reservoirs most important for the transmission and maintenance of RRV. Further insights have previously been gained through modelling approaches (Carver et al. 2009; Denholm et al. 2017; Koolhof and Carver 2017), and future progress in RRV modelling requires consideration of the dynamics of multiple mosquito species and multiple hosts, accounting for their differing availability, and their differing capacity to transmit RRV. One such model would be a next-generation matrix (NGM) model, which calculates R0 across different categories of individuals, where R0 is defined as the average number of secondary cases caused by one infectious individual in a population of susceptible individuals (Webster, Borlase, and Rudge 2017; Diekmann, Heesterbeek, and Roberts 2009; Diekmann, Heesterbeek, and Metz 1990; van den Driessche and Watmough 2002). NGM models have been adopted for several mosquito-borne diseases, including Zika virus, dengue virus and West Nile virus (WNV) (Brauer et al. 2016; Ding, Tao, and Zhu 2016; Hartemink et al. 2007). Most relevant to this dissertation is a NGM model for WNV which has been used to determine the relative importance of different vector-host combinations for increasing or decreasing WNV R0 within a community (Wonham, de- Camino-Beck, and Lewis 2004).

There are several steps to building a NGM model (Figure 7.1). The collective findings from this dissertation contribute to each of these steps, but ultimately more research is needed to parametrise a NGM model for RRV. The first step of a NGM model is to define the categories of individuals that will be assessed. For RRV this should represent a mixture of vertebrate hosts and mosquitoes. Ultimately, a NGM model for RRV would include a diversity of species with

152 different capacities to transmit RRV. Horses and ringtail possums have already been discussed as key interest species based on the findings within this dissertation. Additional species of interest to model would include; humans (given they suffer clinical disease, and their role in transmission, particularly in the Pacific, is vague); some bird species (magpie-larks Grallina cyanoleuca could be important to include given their nation-wide distribution in Australia, and being one of the only species of birds which have yielded an isolate of RRV), and finally; black flying foxes (which are known to be fed on by vectors of RRV, and are recognised as reservoirs for other viral pathogens; Figure 7.1). To select vectors, it is important that they reflect a diversity of feeding patterns, larval ecology and vector competence. They must also persist in the same geographic range, and temporally, as the suggested vertebrates otherwise the interactions assessed within the model would not be not representative of transmission potential. From these criteria, suggested vectors would include; Ae. vigilax, Ae. notoscriptus, Cq. linealis and Cx. annulirostris (Figure 7.1).

The second stage of a NGM model is to build two matrices (Figure 7.1). These are known as ‘whom acquires infection from whom’ (WAIF) matrices and represent the contact, availability, infectivity and susceptibility of a given vector and a given host (Webster, Borlase, and Rudge 2017). The elements within each matrix characterise the rates of infection between all possible combinations of species. The elements of the first matrix represent transmission from a given mosquito species to a given host species, and the elements of the second matrix represents transmission from a given host species to a mosquito species. R0 is then derived from the largest eigenvalue of the multiplication of these matrices (Diekmann, Heesterbeek, and Roberts 2009). For WNV, differential equations incorporating Susceptible-Infectious-Recovered populations of vertebrates and Susceptible-Infectious populations of vectors have been used in NGM modelling. Wonham, de-Camino-Beck, and Lewis (2004) collected parameters for mosquito and vertebrate species which represented; i) the relative abundance of each species; ii) duration of infectiousness; and iii) contact rates between the two species. Table 7.2 derives these parameters in the context of current knowledge for RRV vertebrate hosts and mosquito vectors.

153

Figure 7.1 Steps necessary to build a NGM model (derived from (Diekmann, Heesterbeek, and Roberts 2009; Wonham, de-Camino-Beck, and Lewis 2004; Webster, Borlase, and Rudge 2017) (left) and the data appropriate for RRV (right)

154 Table 7.2. Host and vector parameters needed for a NGM model for RRV, derived from Wonham, de-Camino-Beck, and Lewis (2004). Cells coloured in green represent those with readily available information, cells in yellow represent partial information, and cells in brown represent missing information.

Study population Data availability for RRV Vectors Hosts Vectors Source Hosts Source State variables Susceptible SV SH All vectors Calculated from: Horses, ringtail possums, humans available. Calculated from: available AV - IV Assumptions would need to be made for NH-RH-IH magpie-larks and black flying foxes on the basis of taxonomic order. Infectious IV IH All vectors Calculated from: Horses and humans available. Assumptions Ideally calculated from available SV( βH αH ωV) would need to be made for ringtail virus isolation from hosts possums, magpie-larks and black flying (not currently available). foxes on the basis of taxonomic order. Potentially estimated from: SH( βH αV ωH) Recovered RH Horses, ringtail possums, humans available. Derived from Assumptions would need to be made for seroprevalence rate in magpie-larks and black flying foxes on the Chapter 4 and 5 basis of taxonomic order. Larval mosquitoes LV All vectors Calculated from: available AV (bV mV dL) Adult mosquitoes AV All vectors Derived from available count data in Chapter 6 Total population NH All hosts available. Assumptions would Derived from count data in need to be made for total population size Chapter 6 from count data. Parameters Birth rate bV bH Data limited. Data limited, but assumptions can be made for each species Maturation rate mV Data limited.

155 Study population Data availability for RRV Vectors Hosts Vectors Source Hosts Source Parameters (cont) Death rate dL, dA dH All adult Russell (1987) Data limited, but assumptions can be made vectors for each species available. Missing data on larval survival. Vector-host contact rate βH All vectors Partially derivable Horses and humans available. Assumptions Partially derivable from available from odds ratio in would need to be made for ringtail odds ratio in Chapter 3 Chapter 3 possums, magpie-larks and black flying foxes on the basis of taxonomy. Virus transmission rate (from) αV αH All vectors (Jeffery et al. Horses and humans available. Other species Kay et al. (1986) available 2002; Ryan, Do, unavailable. and Kay 2000) Incubation period ωV ωH All vectors (Jeffery et al. Horses and humans available. Other species Kay et al. (1986) available 2002; Ryan, Do, unavailable. and Kay 2000) Recovery rate γH Horses and humans available. Other species Kay et al. (1986) unavailable.

156 From this dissertation, new data are available which contribute towards construction of a NGM model, to calculate RRV R0, and to test hypotheses about which combinations of hosts and vectors are most important for increasing R0 in a community. Specifically, from Chapter 3, contact rates between mosquitoes and hosts could be derived from the log odds ratio analysis. For abundance and availability, population estimates could be parameterised from the field findings collected in Chapter 6. The proportion of the host population that is susceptible would have to be calculated, but could be derived from removing the proportion of ‘recovered’ individuals, estimated from seroprevalence results in Chapter 4 and 5, this could be further refined by calculating the proportion of ‘infected’ individuals as a parameter of infectivity and duration of infection based on the literature reviewed in Chapter 2. However, without additional data, attempting to derive a plausible NGM is premature at this stage. Additional data for the candidate mosquito vectors would need to be acquired: in particular, further detail of the proportion that become infected after feeding on a blood meal with a given titre of RRV, the duration of infection and the life expectancy of the mosquito species under natural conditions.

Once built the NGM model is build, its greatest benefit to understanding RRV transmission ecology would be through hypothesis testing by modifying the parameters for each host within the matrices. For example, if contact between vectors and horses were reduced to zero, would this result in an increase or decrease in the community R0, compared to the original model?

Further if it did result in a decreased R0, would this be a viable control strategy for RRV, where vector elimination is focussed on areas where mosquitoes and horses could come into contact?

Despite this approach being an advance in the understanding of RRV transmission ecology, the model would not be without its limitations. Any model is just a representation of the real world and an attempt to simplify complex ecology. Although it would incorporate the latest knowledge on non-human reservoirs, host immune responses would have to be broadly derived. The development of a NGM model, however, would still provide a foundation for future investigations to update the model as new data is generated for host immunity for RRV.

7.5 Final conclusions and future directions

Understanding the complex transmission ecology of multi-host pathogens has been declared as one of the major challenges to biomedical science in the 21st century (Woolhouse, Taylor, and Haydon 2001; Webster, Borlase, and Rudge 2017). Although population biologists have been successful in understanding the evolution and dynamics of single-host pathogens, most infectious disease globally are capable of infecting more than one host species (Cleaveland,

157 Laurenson, and Taylor 2001). To accurately target interventions and manage outbreaks of such multi-host pathogens, knowledge of transmission dynamics must be integrated (Cleaveland, Haydon, and Taylor 2007). This dissertation took a systematic approach to better understand the reservoir hosts for RRV, but has also led to new insights pertaining to each of the three components. Specifically, this research highlighted ecological (including life-history traits), epidemiological (including exposure rates) and behavioural (such as vector feeding patterns) insights that are important for RRV transmission. Arboviruses have varied transmission ecology. On one end of the spectrum, there are viruses such as dengue viruses which are specialist viruses; mostly maintained in transmission with a single vertebrate host (humans), and a single vector (Ae. aegypti) (Morin, Comrie, and Ernst 2013). On the opposite end of the spectrum, there are highly complex generalist viruses which persist in multiple hosts and multiple vectors (such as WNV (Hayes et al. 2005)) and RRV. Identifying the level of transmission complexity is a critical component for targeted interventions against an arbovirus. For example, successful dengue virus strategies are highly vector-centric with most recent efforts focussed on reducing the reproduction and vector competence of Ae. aegypti (Hoffmann et al. 2011; Hoffmann, Ross, and Rašić 2015). By comparison, management strategies for WNV have in recent years been host-centric, either through sampling of culled birds in Italy to inform the annual timing of blood donor screening (ultimately preventing transfusion transmission events)(Pupella et al. 2013), or increased public education surrounding avian migration events in North America (Swetnam et al. 2018). As RRV is a zoonotic arbovirus with one of the most complex transmission dynamics, guidance can be derived from the complex WNV systems, however further research is necessary before implementation of effective host-centric management strategies.

To mitigate and manage ongoing circulation in the complex RRV system, models are required to fully explore RRV dynamics. This dissertation has provided critical information to structure and parameterise those models, and has identified critical species of interest for future RRV research.

158 7.6 References

Azuolas, J. K. 1998. 'Ross River virus disease of horses', Australian Equine Veterinarian, 16: 56-58. Boyd, A. M., R. A. Hall, R. T. Gemmell, and B. H. Kay. 2001. 'Experimental infection of Australian brushtail possums, Trichosurus vulpecula (Phalangeridae : Marsupialia), with Ross River and Barmah Forest viruses by use of a natural mosquito vector system', American Journal of Tropical Medicine and Hygiene, 65: 777-82. Brauer, F., C. Castillo-Chavez, A. Mubayi, and S. Towers. 2016. 'Some models for epidemics of vector- transmitted diseases', Infectious Disease Modelling, 1: 79-87. Carver, S., H. Spafford, A. Storey, and P. Weinstein. 2009. 'Dryland Salinity and the Ecology of Ross River Virus: The Ecological Underpinnings of the Potential for Transmission', Vector-Borne and Zoonotic Diseases, 9: 611-22. Cleaveland, S., D. T. Haydon, and L. H. Taylor. 2007. 'Overviews of pathogen emergence: which pathogens emerge, when and why?' in, Wildlife and Emerging Zoonotic Diseases: The Biology, Circumstances and Consequences of Cross-Species Transmission (Springer). Cleaveland, S., M. K. Laurenson, and L. H. Taylor. 2001. 'Diseases of humans and their domestic mammals: pathogen characteristics, host range and the risk of emergence', Philosophical Transactions of the Royal Society of London. Series B: Biological Sciences, 356: 991-99. Craft, M. E., E. Volz, C. Packer, and L. A. Meyers. 2009. 'Distinguishing epidemic waves from disease spillover in a wildlife population', Proceedings Biological Sciences, 276: 1777-85. Denholm, L., N. J. Beeton, L. K. Forbes, and S. Carver. 2017. 'A model for the dynamics of Ross River Virus in the Australian environment', Letters in Biomathematics, 4: 187-206. Diekmann, O., J. A. P. Heesterbeek, and J. A. J. Metz. 1990. 'On the definition and the computation of the basic reproduction ratio R 0 in models for infectious diseases in heterogeneous populations', Journal of mathematical biology, 28: 365-82. Diekmann, O., J. A. P. Heesterbeek, and M. G. Roberts. 2009. 'The construction of next-generation matrices for compartmental epidemic models', Journal of The Royal Society Interface, 7: 873- 85. Ding, C., N. Tao, and Y. Zhu. 2016. "A mathematical model of Zika virus and its optimal control." In 2016 35th Chinese control conference (CCC), 2642-45. IEEE. Duncombe, J., C. Lau, P. Weinstein, J. Aaskov, M. Rourke, R. Grant, and A. Clements. 2013. 'Seroprevalence of Dengue in American Samoa, 2010', Emerging Infectious Diseases, 19: 324-26. Flies, E. J., C. L. Lau, S. Carver, and P. Weinstein. 2018. 'Another emerging Mosquito-Borne disease? Endemic Ross River virus transmission in the absence of marsupial reservoirs', BioScience, 68: 288-93. Gummow, B., R. H.H. Tan, R. K. Joice, G. Burgess, and J. Picard. 2018. 'Seroprevalence and associated risk factors of mosquito‐borne alphaviruses in horses in northern Queensland', Australian Veterinary Journal, 96: 243-51. Hartemink, N. A., S. A. Davis, P. Reiter, Z. Hubálek, and J. A. P. Heesterbeek. 2007. 'Importance of bird-to-bird transmission for the establishment of West Nile virus', Vector-Borne and Zoonotic Diseases, 7: 575-84. Hayes, E. B., N. Komar, R. S. Nasci, S. P. Montgomery, D. R. O'Leary, and G. L. Campbell. 2005. 'Epidemiology and transmission dynamics of West Nile virus disease', Emerging Infectious Diseases, 11: 1167. Hill, N. J., M. L. Power, and E. M. Deane. 2009. 'Absence of Ross River virus amongst Common brushtail possums (Trichosurus vulpecula) from metropolitan Sydney, Australia', European Journal of Wildlife Research, 55: 313-16.

159 Hoffmann, A. A., B. L. Montgomery, J. Popovici, I. Iturbe-Ormaetxe, P. H. Johnson, F. Muzzi, M. Greenfield, M. Durkan, Y. S. Leong, and Y. Dong. 2011. 'Successful establishment of Wolbachia in Aedes populations to suppress dengue transmission', Nature, 476: 454-57. Hoffmann, A. A., P. A. Ross, and G. Rašić. 2015. 'Wolbachia strains for disease control: ecological and evolutionary considerations', Evolutionary applications, 8: 751-68. Jeffery, J. A., P. A. Ryan, S. A. Lyons, and B. H. Kay. 2002. 'Vector competence of Coquillettidia linealis (Skuse) (Diptera : Culicidae) for Ross River and Barmah Forest viruses', Australian Journal of Entomology, 41: 339-44. Kay, B. H., A M Boyd, P. A. Ryan, and R. A. Hall. 2007. 'Mosquito feeding patterns and natural infection of vertebrates with Ross River and Barmah Forest viruses in Brisbane, Australia', American Journal of Tropical Medicine and Hygiene, 76: 417-23. Kay, B. H., R. A. Hall, I. D. Fanning, P. Mottram, P. L. Young, and C. C. Pollitt. 1986. "Experimental infection of vertebrates with Murray Valley encephalitis and Ross River viruses." In Arbovirus Research in Australia, 71-75. Koolhof, I. S., and S. Carver. 2017. 'Epidemic host community contribution to mosquito-borne disease transmission: Ross River virus', Epidemiology & Infection, 145: 656-66. McCallum, H. 2016. 'Models for managing wildlife disease', Parasitology, 143: 805-20. Modlmaier, M., R. Kuhn, O. R. Kaaden, and M. Pfeffer. 2002. 'Transmission studies of a European Sindbis virus in the floodwater mosquito Aedes vexans (Diptera : Culicidae)', International Journal of Medical Microbiology, 291: 164-70. Morin, C. W., A. C. Comrie, and K. Ernst. 2013. 'Climate and dengue transmission: evidence and implications', Environmental Health Perspectives, 121: 1264-72. Plowright, R. K., C. R. Parrish, H. McCallum, P. J. Hudson, A. I. Ko, A. L. Graham, and J. O. Lloyd-Smith. 2017. 'Pathways to zoonotic spillover', Nature Reviews Microbiology. Pupella, S., G. Pisani, K. Cristiano, L. Catalano, and G. Grazzini. 2013. 'West Nile virus in the transfusion setting with a special focus on Italian preventive measures adopted in 2008– 2012 and their impact on blood safety', Blood Transfusion, 11: 563. Reiner, Jr. R. C., A. T. Perkins, C. M. Barker, T. Niu, L. F. Chaves, A. M. Ellis, D. B. George, A. Le Menach, J. R. C. Pulliam, and D. Bisanzio. 2013. 'A systematic review of mathematical models of mosquito-borne pathogen transmission: 1970–2010', Journal of The Royal Society Interface, 10: 20120921. Restif, O., D. T. Hayman, J. R. Pulliam, R. K. Plowright, D. B. George, A. D. Luis, A. A. Cunningham, R. A. Bowen, A. R. Fooks, T. J. O'Shea, J. L. Wood, and C. T. Webb. 2012. 'Model-guided fieldwork: practical guidelines for multidisciplinary research on wildlife ecological and epidemiological dynamics', Ecology Letters, 15: 1083-94. Ross, R. 1915. 'Some a priori pathometric equations', British Medical Journal, 1: 546. Russell, R. C. 1987. 'The mosquito fauna of Conjola State Forest on the south coast of New South Wales. Part 3. Female survival and population age composition', General and Applied Entomology: The Journal of the Entomological Society of New South Wales, 19: 25. Ryan, P. A., K. A. Do, and B. H. Kay. 2000. 'Definition of Ross River virus vectors at Maroochy Shire, Australia', Journal of Medical Entomology, 37: 146-52. Smith, D. L., K. E. Battle, S. I. Hay, C. M. Barker, T. W. Scott, and F. E. McKenzie. 2012. 'Ross, Macdonald, and a theory for the dynamics and control of mosquito-transmitted pathogens', Plos Pathogens, 8: e1002588. Stephenson, E. B., A. K. Murphy, C. C. Jansen, A. J. Peel, and H. McCallum. 2019. 'Interpreting mosquito feeding patterns in Australia through an ecological lens: an analysis of blood meal studies', Parasites & Vectors, 12: 156. Stephenson, E. B., A. J. Peel, S. A. Reid, C. C. Jansen, and H. McCallum. 2018. 'The non-human reservoirs of Ross River virus: a systematic review of the evidence', Parasites & Vectors, 11: 188.

160 Stephenson, E. B., P. A. Rudd, A. J. Peel, H. McCallum, S. A. Reid, and L. J. Herrero. 2019. 'Species traits and hotspots associated with Ross River virus infection in non-human vertebrates in South-East Queensland', Transactions of the Royal Society of Tropical Medicine and Hygiene, Submitted 26th July 2019. Swetnam, D., S. G. Widen, T. G. Wood, M. Reyna, L. Wilkerson, M. Debboun, D. A. Symonds, D. G. Mead, B. J. Beaty, and H. Guzman. 2018. 'Terrestrial Bird Migration and West Nile Virus Circulation, United States', Emerging Infectious Diseases, 24: 2184. Tesh, R. B., D. C. Gajdusek, R. M. Garruto, J. H. Cross, and L. Rosen. 1975. 'Distribution and prevalence of Group-A arbovirus neutralizing antibodies among human populations in South-East Asia and Pacific Islands. ', American Journal of Tropical Medicine and Hygiene, 24: 664-75. Tompkins, D. M., and D. Slaney. 2014. 'Exploring the Potential for Ross River Virus Emergence in New Zealand', Vector-Borne and Zoonotic Diseases, 14: 141-48. van den Driessche, P., and J. Watmough. 2002. 'Reproduction numbers and sub-threshold endemic equilibria for compartmental models of disease transmission', Mathematical biosciences, 180: 29-48. van den Hurk, A. F., S. B. Craig, S. M. Tulsiani, and C. C. Jansen. 2010. 'Emerging tropical diseases in Australia. Part 4. Mosquitoborne diseases', Annals of Tropical Medicine and Parasitology, 104: 623-40. Weaver, S. C., and A. D. Barrett. 2004. 'Transmission cycles, host range, evolution and emergence of arboviral disease', Nature Review Microbiology, 2: 789-801. Webster, J. P., A. Borlase, and J. W. Rudge. 2017. 'Who acquires infection from whom and how? Disentangling multi-host and multi-mode transmission dynamics in the ‘elimination’era', Philosophical Transactions of the Royal Society B: Biological Sciences, 372: 20160091. Wonham, M. J., T. de-Camino-Beck, and M. A. Lewis. 2004. 'An epidemiological model for West Nile virus: invasion analysis and control applications', Proceedings of the Royal Society of London B: Biological Sciences, 271: 501-07. Woolhouse, M. E. J., L. H. Taylor, and D. T. Haydon. 2001. 'Population biology of multihost pathogens', science, 292: 1109-12. Zinsstag, J., E. Schelling, D. Waltner-Toews, and M. Tanner. 2011. 'From “one medicine” to “one health” and systemic approaches to health and well-being', Preventive veterinary medicine, 101: 148-56.

161