ABSTRACT
SPENCE BEAULIEU, MEREDITH RUTH. The Ecology of Dog Heartworm Disease. (Under the direction of Drs. Michael Reiskind and Robert Dunn).
Suburban development is rapidly transforming the global landscape. This anthropogenically induced land-use change affects the species around us, including mosquitoes and the pathogens that they transmit. Although the effect of urbanization has been extensively studied for key mosquito species of interest, relatively little research has addressed the effect of urbanization on mosquito species assemblages as a whole. Knowledge of the effects of urbanization on entire mosquito communities is critical, as many vector-borne diseases are transmitted by an assembly of vectors and therefore could be sensitive to community-level alterations after land-use change. Additionally, the effects of vector diversity on the transmission of multi-vectored pathogens has scarcely been explored either empirically or theoretically. In this dissertation, I investigate the effects of suburban development on mosquito assemblages and the subsequent consequences for vector-borne disease transmission using the dog heartworm, Dirofilaria immitis, as a model system. In the first experiment, I used a field study to determine the effects of suburbanization on mosquito diversity and community composition. I found that species diversity declined with neighborhood age, with the oldest suburban areas having species assemblages that are characteristically less diverse than those of undeveloped areas. The community composition differed between habitat types, with the suburban mosquito assemblage being dominated by the invasive Aedes albopictus, the Asian tiger mosquito. In the second experiment, I performed an epidemiological study comparing the within-mosquito heartworm and within-host heartworm prevalence by land-use type. I found that suburban areas had lower levels of heartworm within both the host and the vector than did the undeveloped field sites, with undeveloped wooded sites showing an intermediate pathogen presence. Since the suburban areas were overwhelmingly dominated by Ae. albopictus and had significantly lower heartworm prevalence, I then investigated the vector competence of the local population of Ae. albopictus with the hypothesis that it was a poor vector locally. In the third study, I performed a laboratory experiment testing the D. immitis developmental times, mosquito longevity and fecundity, and the vector efficiency index for Ae. albopictus and a suspected highly competent local vector, Aedes triseriatus. I found that Ae. albopictus is indeed a poor vector of heartworm in Wake County, as evidenced by significant infection-induced mortality and a vector efficiency index more than 4x lower than that of Ae. triseriatus. In the fourth study, I created a single host and multi-vector SIR-type mathematical model of dog heartworm disease. Utilizing the laboratory generated parameters for Ae. albopictus and Ae. triseriatus as well as the field data on the diversity and community composition of suburban and undeveloped areas in Wake County, I compared the model’s basic reproduction number, !", for these empirically driven scenarios. Using the critical value of !" > 1 as a threshold, I validated that Ae. triseriatus is a competent vector of D. immitis, while Ae. albopictus is insufficient to support disease transmission as the sole vector in a natural setting. I also demonstrated that the suburban mosquito assemblage had a notably lower !" than the natural area assemblage, supporting my previous observations on heartworm prevalence by habitat type. To test theoretical predictions about the effect of vector diversity on disease transmission, I then simulated many iterations of models with one, three, five, and ten vector species. While changes in diversity did not significantly affect the average !", variation in !" values decreased as the number of vectors increased. This demonstrates the stabilizing effect of vector diversity for multi-vectored pathogens while also highlighting the importance of species identity and community composition for determining the magnitude of disease transmission. In a final brief chapter, I consider the parasite manipulation hypothesis and its potential implications for vector- borne diseases.
© Copyright 2019 by Meredith Ruth Spence Beaulieu
All Rights Reserved The Ecology of Dog Heartworm Disease
by Meredith Ruth Spence Beaulieu
A dissertation submitted to the Graduate Faculty of North Carolina State University in partial fulfillment of the requirements for the degree of Doctor of Philosophy
Entomology
Raleigh, North Carolina 2019
APPROVED BY:
______Dr. Michael Reiskind Dr. Robert Dunn Committee Co-Chair Committee Co-Chair
______Dr. David Watson Dr. Cristina Lanzas
______Dr. Kevin Gross ii BIOGRAPHY
A North Carolina native, Meredith was born and raised in Elizabeth City. She attended NC State as an undergraduate, where she was first exposed to the insect world as an undergraduate research assistant investigating ant-mediated seed dispersal. She graduated valedictorian and summa cum laude, receiving a BS in Zoology with a minor in Mathematics in 2011. She worked as a veterinary technician for three years before deciding to combine her passions for veterinary medicine, mathematics, and entomology by pursuing a PhD in Entomology at NC State. Her PhD research focuses on how human-driven land-use change affects mosquito assemblages, and how that in turn affects pathogen transmission utilizing dog heartworm disease as a model system. She was awarded a National Science Foundation Graduate Research Fellowship in 2016 to support her PhD research. She is interested broadly in vector- borne disease ecology, public health, and science policy. Outside of research, she enjoys drinking inordinate amounts of coffee and cuddling her horde of furry children: Wren the lab- poodle-shar pei mix, Logan and Cora the boxers, Grim the cat, and Ratticus Finch and Ratt Skiba the dumbo fancy rats.
iii ACKNOWLEDGMENTS
First and foremost I would like to thank my advisors, Michael Reiskind and Rob Dunn, for their guidance, support, and encouragement. I was incredibly lucky to have people that so deeply love science as mentors and sources of inspiration. I am indebted to the entirety of my committee, including Michael, Rob, Wes Watson, Cristina Lanzas, and Kevin Gross, for their feedback on my ideas and work. Your insights greatly improved my learning and my research. All of the Reiskind lab members have graciously shared their expertise, time, and advice, but none more so than Kristen Hopperstad. Special thanks also to Paul Labadie for assisting with my molecular work and troubleshooting when things inevitably went wrong, and to Shawn Janairo for being an encyclopedia of mosquito and heartworm rearing knowledge. My research was funded by the National Science Foundation Graduate Research Fellowship Program under Grant No. DGE-1746939 and NC State’s Department of Entomology and Plant Pathology. I would also like to acknowledge the financial support that I received from the American Mosquito Control Association, the Society for Vector Ecology, the Entomological Society of America, Central Life Sciences, Bayer, the Mid-Atlantic Mosquito Control Association, and the North Carolina Mosquito and Vector Control Association for travel to various conferences to communicate my work and enhance my professional development. I had the privilege of working with many fantastic people through a variety of extracurricular activities, including Linnaean Games, the Entomology Graduate Student Association, the university-level Graduate Student Association, and Student Government. Thank you for inspiring me to be a better team member and leader, and for helping me to leave a small mark on NC State. This also seems an appropriate place to acknowledge Michael Reiskind again, for not panicking (too much) when I insisted on being so involved in so many things. I am grateful for your flexibility and trust while I pursued my non-research passions. To the many wonderful friends that have stood by me on this journey and that I’ve met in the process: thank you, thank you, thank you. For the trivia nights, for the dancing, for the gin and tequila drinking, for the long distance phone calls, for everything. You have brought joy and reminded me that I’m still a person outside of my research when I needed reminding. My most heartfelt gratitude to my parents, John and Debbie Spence, and my sister, Leah Creed, for their constant encouragement. Finally, a million thank yous to my patient and supportive husband, Matt Beaulieu. All my love, for now, forever, for on and on and on.
iv TABLE OF CONTENTS
LIST OF TABLES ...... vi LIST OF FIGURES ...... vii
Chapter 1: Introduction ...... 1 References ...... 4
Chapter 2: Simplification of Vector Communities During Suburban Succession...... 6 Abstract ...... 6 Introduction ...... 7 Methods ...... 9 Study Overview ...... 9 Site Selection and Owner Permission ...... 10 Trapping ...... 12 Specimen Identification ...... 12 Land-use Classification ...... 12 Statistical Analyses ...... 13 Results ...... 14 Discussion ...... 21 Acknowledgements ...... 26 Supporting Information ...... 26 References ...... 30
Chapter 3: Mosquito Diversity and Dog Heartworm Prevalence in Suburban Areas ...... 35 Abstract ...... 35 Background ...... 36 Methods ...... 38 Site Selection and Owner Permission ...... 38 Trapping ...... 39 Molecular Analysis ...... 40 Within-host Heartworm Data Acquisition ...... 41 Statistical Analyses ...... 41 Results ...... 43 Discussion ...... 48 Conclusions ...... 52 Acknowledgements ...... 53 Additional Files ...... 53 References ...... 55
Chapter 4: Comparative Vector Efficiency of Two Prevalent Mosquito Species for Dog Heartworm in North Carolina...... 59 Abstract ...... 59 Introduction ...... 59 Materials and Methods ...... 63 Mosquito Rearing ...... 63
v D. immitis Developmental Times ...... 63 Longevity and Fecundity ...... 64 Vector efficiency index ...... 64 Results ...... 65 Discussion ...... 68 Acknowledgements ...... 70 References ...... 71
Chapter 5: The Effects of Vector Diversity and Community Composition on Dog Heartworm Transmission ...... 75 Abstract ...... 75 Introduction ...... 76 Methods ...... 79 Model Formulation ...... 79 Efficient Vector vs. Poor Vector ...... 84 Suburban Mosquito Assemblage vs. Natural Mosquito Assemblage ...... 85 Vector Diversity ...... 87 Results ...... 87 Discussion ...... 89 Acknowledgements ...... 92 References ...... 93
Chapter 6: The Role of Parasite Manipulation in Vector-Borne Diseases ...... 96 Definition and Background ...... 96 Examples in Public Health ...... 96 Evolutionary Perspectives and Implications ...... 97 References ...... 98
Chapter 7: Conclusions...... 99 References ...... 102
vi LIST OF TABLES
Chapter 2
Table 1 Effects of neighborhood age, housing density, home price, and household income on mosquito diversity measures...... 18
S1 Table Relationship between neighborhood age and mosquito species abundance...... 29
Chapter 3
Table 1 Heartworm prevalence within mosquito species...... 43
Chapter 4
Table 1 Dirofilaria immitis developmental times within Aedes triseriatus and Aedes albopictus...... 66
Chapter 5
Table 1 Parameters of the dog heartworm model...... 81
Table 2 Vector parameters for Aedes triseriatus and Aedes albopictus...... 85
Table 3 Vector parameters for species in the suburban assemblage and the natural assemblage...... 86
vii LIST OF FIGURES
Chapter 2
Figure 1 Trapping sites within Wake County, North Carolina...... 11
Figure 2 Comparison of diversity measures between control sites and suburban sites by age category in 2015...... 16
Figure 3 Comparison of diversity measures between control sites and suburban sites by age category in 2016...... 17
Figure 4 Log abundance of Aedes albopictus increases with log neighborhood age...... 19
Figure 5 Partial Canonical Correspondence Analysis of mosquito assemblages and habitat classifications at each trapping site...... 20
S1 Figure Trends in research on mosquito diversity changes during urbanization...... 26
S2 Figure Species accumulation curves based on random sampling...... 27
S3 Figure Log abundance of Culex erraticus decreases with log neighborhood age...... 27
S4 Figure Comparison of suburban site CCA values versus neighborhood age...... 28
Chapter 3
Figure 1 Species-level heartworm prevalence by land-use type...... 44
Figure 2 Relationship between mosquito parity and within-mosquito heartworm prevalence...... 45
Figure 3 Comparison of within-mosquito heartworm prevalence by land-use type...... 46
Figure 4 Visualization of heartworm prevalence by zip code in Wake County, North Carolina...... 46
Figure 5 Within-host heartworm prevalence increases with mosquito diversity measures. ...47
Addl 1 Modified DNA extraction protocols...... 53
Addl 2 Within-mosquito heartworm prevalence throughout the trapping season...... 54
viii Chapter 4
Figure 1 Comparison of longevity with and without Dirofilaria immitis infection...... 67
Figure 2 Effects of Dirofilaria immitis infection on fecundity...... 67
Chapter 5
Figure 1 Compartmental model of dog heartworm transmission with a single host and multiple vectors...... 79
Figure 2 Average !" for systems with varying vector diversity and a 2:1 mosquito to dog ratio...... 88
1 CHAPTER 1
Introduction
Mosquitoes (Diptera: Culicidae) are a diverse family of flies that include many species of public health and veterinary significance. Globally, mosquitoes are the arthropods of greatest concern to human health, transmitting a plethora of pathogens that have plagued us for centuries, including the malaria parasite, dengue virus, and yellow fever virus. Mosquitoes are also responsible for transmitting pathogens that have recently emerged as global concerns, such as Zika virus, which became a focus of research and control efforts after a large outbreak correlated with microcephaly was noted in Brazil in 2015 (Faria et al. 2016). While these previously mentioned diseases are generally considered tropical diseases and autochthonous transmission is rare in the United States, travel-related cases do occur and the threat of emerging pathogens is ever present. Additionally, the US harbors a variety of endemic mosquito-borne pathogens, including West Nile virus, La Crosse virus and other encephalitic viruses, and filarial parasites such as the dog heartworm. No single mosquito species is of greatest public health importance in the US, as each pathogen is transmitted by its own primary vector or assemblage of vectors. Given that different mosquitoes transmit different pathogens, it is necessary to understand the habitats that individual species utilize so that potential disease foci can be identified, reflecting the concept of natural nidality for vector-borne disease (Pavlovsky 1966). Land-use is an important determinant of the species assemblage found in a given area, with mosquitoes showing fidelity between field, wooded, and edge habitats at a scale of less than 20m (Reiskind et al. 2017). While mosquito assemblages have been well studied within natural habitats (O'Brien and Reiskind 2013, Gardner et al. 2014, Steiger et al. 2016a), less research has investigated the effects of land-use on mosquitoes in human dominated areas, including urban and suburban landscapes. Certain species of interest, such as container-breeding Aedes that vector many viruses of human concern, have been extensively studied in urbanized areas (Braks et al. 2003, Reiskind and Lounibos 2013, Hopperstad and Reiskind 2016). However, the few investigations into the effects of anthropogenic land-use change on entire mosquito assemblages (Kun and Kremsner 2000, Steiger et al. 2016b) have taken place in settings very different from the
2 dominant human land-use type in the US: suburbia. Suburban development is the fastest growing anthropogenic land-use in the US and will continue to grow rapidly (Brown et al. 2005, Terando et al. 2014), making our understanding of the response of mosquito communities to suburbanization a pressing issue. In addition to the gap in knowledge with regard to the effects of suburbanization on mosquito communities, we also know very little about how changes in mosquito diversity affect disease transmission. While many vector-borne diseases are transmitted by one or a few primary vectors, there are numerous examples of diseases where an assembly of vectors is responsible for pathogen transmission, including malaria, Rift Valley fever, West Nile encephalitis, Zika fever, lymphatic filariasis, and dog heartworm disease (Bockarie et al. 2009, Sinka et al. 2010, Ledesma and Harrington 2011, Roche et al. 2012, Linthicum et al. 2016, Evans et al. 2017). For diseases that are transmitted by multiple vectors, changes in vector diversity will likely alter disease risk. Investigations into the importance of diversity for multi-vectored diseases is rare, but the empirical and theoretical studies that exist suggest that increases in diversity can also lead to increases in disease transmission (Roche et al. 2012, Fuller et al. 2016). As anthropogenic land-use change is occurring globally and is likely to alter the diversity of mosquito assemblages, research into the association between vector diversity and pathogen transmission for multi- vectored diseases is critically needed. This dissertation examines how suburban development affects mosquito species assemblages and subsequent disease transmission. I approach these questions using the dog heartworm, Dirofilaria immitis, as a study system. Dog heartworm disease is likely the most prevalent vector-borne disease in the US, with a nationwide average of 1% to 12.5% of dogs infected (Lee et al. 2010). The disease has devastating pathology for both domestic and wild canines, causing respiratory distress and eventually congestive heart failure if the infection is left untreated. It has been diagnosed in all fifty states and is considered endemic in the contiguous US (American Heartworm Society 2018). Additionally, this filarial parasite is competently vectored by at least 25 mosquito species in the US with varying vectorial capacities (Ledesma and Harrington 2011), making it an ideal pathosystem with which to study how changes in mosquito assemblage due to suburbanization impact disease transmission. I addressed this question using a combination of field experiments, epidemiological studies, laboratory
3 experiments, and mathematical modeling of dog heartworm disease in Wake County, North Carolina. In the first study, I assessed the effects of suburbanization on mosquito species assemblages. I sampled mosquitoes in neighborhoods of different ages and in undeveloped wooded and field areas as controls, then compared various diversity metrics and the community composition between land-use types. In the second study, I analyzed the previously sampled mosquitoes for the presence of D. immitis DNA and acquired data on heartworm positive rates within domestic dogs entering a local animal shelter. I used these data to investigate landscape level trends in heartworm prevalence within both the vector and the host in suburbanized and undeveloped habitats, explicitly linking anthropogenic-induced mosquito diversity changes to altered vector-borne disease risk. In the third study, I tested the vector competence of two prominent NC mosquito species for D. immitis. Recognizing that vector competence can vary between geographically distinct populations of the same species (Nayar and Knight 1999), this study contributes the first known vector efficiency estimate for NC populations of the eastern treehole mosquito, Aedes triseriatus, as well as an updated estimate of the vector efficiency of the local population of Asian tiger mosquito, Aedes albopictus, which was last investigated in 1989 shortly after its introduction to NC (Apperson et al. 1989). By investigating D. immitis developmental times, longevity, and fecundity of these two species when infected with heartworm in a laboratory setting, I also generated parameter values that could be utilized in a mathematical disease transmission model. In the last study, I constructed a system of ordinary differential equations to describe dog heartworm transmission for a single host and multiple vectors. This study synthesizes and quantifies information from the aforementioned laboratory and field experiments, while also testing theoretical predictions about the role of vector diversity in the multi-vectored dog heartworm disease system. In a brief final chapter, I consider the parasite manipulation hypothesis with regard to vector-borne diseases. I discuss how the evolution of parasite adaptations, or host or vector adaptations in response to infection by a parasite, can impact vector-borne disease transmission.
4 REFERENCES
American Heartworm Society. 2018. Current Canine Guidelines for the Prevention, Diagnosis, and Management of Heartworm (Dirofilaria immitis) Infection in Dogs.
Apperson, C. S., B. Engber, and J. F. Levine. 1989. Relative suitability of Aedes albopictus and Aedes aegypti in North Carolina to support development of Dirofilaria immitis. J. Am. Mosq. Control Assoc. 5:377-382.
Bockarie, M. J., E. M. Pedersen, G. B. White, and E. Michael. 2009. Role of vector control in the global program to eliminate lymphatic filariasis. Annu. Rev. Entomol. 54:469-487.
Braks, M. A., N. A. Honório, R. Lourenço-De-Oliveira, S. A. Juliano, and L. P. Lounibos. 2003. Convergent habitat segregation of Aedes aegypti and Aedes albopictus (Diptera: Culicidae) in southeastern Brazil and Florida. J. Med. Entomol. 40:785-794.
Brown, D. G., K. M. Johnson, T. R. Loveland, and D. M. Theobald. 2005. Rural land-use trends in the conterminous United States, 1950-2000. Ecol. Appl. 15:1851-1863.
Evans, M. V., T. A. Dallas, B. A. Han, C. C. Murdock, and J. M. Drake. 2017. Data-driven identification of potential Zika virus vectors. Elife 6:e22053.
Faria, N. R., da Silva Azevedo, Raimunda do Socorro, M. U. Kraemer, R. Souza, M. S. Cunha, S. C. Hill, J. Thézé, M. B. Bonsall, T. A. Bowden, and I. Rissanen. 2016. Zika virus in the Americas: early epidemiological and genetic findings. Science 352:345-349.
Fuller, D. O., T. Alimi, S. Herrera, J. C. Beier, and M. L. Quiñones. 2016. Spatial association between malaria vector species richness and malaria in Colombia. Acta Trop. 158:197-200.
Gardner, A. M., R. L. Lampman, and E. J. Muturi. 2014. Land use patterns and the risk of West Nile Virus transmission in central Illinois. Vector-Borne and Zoonotic Diseases 14:338-345.
Hopperstad, K. A., and M. H. Reiskind. 2016. Recent Changes in the Local Distribution of Aedes aegypti (Diptera: Culicidae) in South Florida, USA. J. Med. Entomol. 53:836-842.
Kun, J. F., and P. G. Kremsner. 2000. Mosquito distribution and entomological inoculation rates in three malaria-endemic areas in Gabon. Trans. R. Soc. Trop. Med. Hyg. 94:652-656.
Ledesma, N., and L. Harrington. 2011. Mosquito vectors of dog heartworm in the United States: vector status and factors influencing transmission efficiency. Topics in companion animal medicine 26:178-185.
Lee, A. C., S. P. Montgomery, J. H. Theis, B. L. Blagburn, and M. L. Eberhard. 2010. Public health issues concerning the widespread distribution of canine heartworm disease. Trends Parasitol. 26:168-173.
5 Linthicum, K. J., S. C. Britch, and A. Anyamba. 2016. Rift Valley fever: an emerging mosquito-borne disease. Annu. Rev. Entomol. 61:395-415.
Nayar, J. K., and J. W. Knight. 1999. Aedes albopictus (Diptera: Culicidae): an experimental and natural host of Dirofilaria immitis (Filarioidea: Onchocercidae) in Florida, USA. J. Med. Entomol. 36:441-448.
O'Brien, V. A., and M. H. Reiskind. 2013. Host-seeking mosquito distribution in habitat mosaics of southern Great Plains cross-timbers. J. Med. Entomol. 50:1231-1239.
Pavlovsky, E. N. 1966. Natural Nidality of Transmissible Diseases with special reference to the Landscape Epidemiology of Zooanthroponoses.
Reiskind, M. H., R. H. Griffin, M. S. Janairo, and K. A. Hopperstad. 2017. Mosquitoes of field and forest: the scale of habitat segregation in a diverse mosquito assemblage. Med. Vet. Entomol. 31:44-54.
Reiskind, M. H., and L. P. Lounibos. 2013. Spatial and temporal patterns of abundance of Aedes aegypti L.(Stegomyia aegypti) and Aedes albopictus (Skuse)[Stegomyia albopictus (Skuse)] in southern Florida. Med. Vet. Entomol. 27:421-429.
Roche, B., P. Rohani, A. P. Dobson, and J. Guégan. 2012. The impact of community organization on vector-borne pathogens. Am. Nat. 181:1-11.
Sinka, M. E., M. J. Bangs, S. Manguin, M. Coetzee, C. M. Mbogo, J. Hemingway, A. P. Patil, W. H. Temperley, P. W. Gething, and C. W. Kabaria. 2010. The dominant Anopheles vectors of human malaria in Africa, Europe and the Middle East: occurrence data, distribution maps and bionomic précis. Parasites & vectors 3:117.
Steiger, D. B. M., S. A. Ritchie, and S. G. W. Laurance. 2016a. Mosquito communities and disease risk influenced by land use change and seasonality in the Australian tropics. Parasites & Vectors 9:387.
Steiger, D. B. M., S. A. Ritchie, and S. G. Laurance. 2016b. Land use influences mosquito communities and disease risk on remote tropical islands: a case study using a novel sampling technique. Am. J. Trop. Med. Hyg. 94:314-321.
Terando, A. J., J. Costanza, C. Belyea, R. R. Dunn, A. McKerrow, and J. A. Collazo. 2014. The southern megalopolis: using the past to predict the future of urban sprawl in the Southeast US. PLoS One 9:e102261.
6 CHAPTER 2
Simplification of Vector Communities During Suburban Succession
(This work was published in PLoS One: Spence Beaulieu MR, Hopperstad K, Dunn RR, Reiskind MH (2019) Simplification of vector communities during suburban succession. PLoS One 14(5): e0215485. https://doi.org/10.1371/journal.pone.0215485)
ABSTRACT Suburbanization is happening rapidly on a global scale, resulting in changes to the species assemblages present in previously undeveloped areas of land. Community-level changes after anthropogenic land-use change have been studied in a variety of organisms, but the effects on arthropods of medical and veterinary importance remain poorly characterized. Shifts in diversity, abundance, and community composition of such arthropods, like mosquitoes, can significantly impact vector-borne disease dynamics due to varying vectorial capacity between different species. In light of these potential implications for vector-borne diseases, we investigated changes in mosquito species assemblage after suburbanization by sampling mosquitoes in neighborhoods of different ages in Wake County, North Carolina, US. We found that independent of housing density and socioeconomic status, mosquito diversity measures decreased as suburban neighborhoods aged. In the oldest neighborhoods, the mosquito assemblage reached a distinct suburban climax community dominated by the invasive, peridomestic container-breeding Aedes albopictus, the Asian tiger mosquito. Aedes albopictus is a competent vector of many pathogens of human concern, and its dominance in suburban areas places it in close proximity with humans, allowing for heightened potential of host-vector interactions. While further research is necessary to explicitly characterize the effects of mosquito community simplification on vector-borne disease transmission in highly suburbanized areas, the current study demonstrates that suburbanization is disrupting mosquito communities so severely that they do not recover their diversity even 100 years after the initial disturbance. Our understanding of the community-level effects of anthropogenic land-use change on arthropod
7 vectors will become increasingly important as we look to mitigate disease spread in a global landscape that is continually developed and altered by humans.
INTRODUCTION The world is becoming increasingly urban. As recently as 1950, urban areas were occupied by just 30% of the world’s population. Today, 55% of people live in urban areas, with the proportion of people living in cities projected to be as high as 68% by 2050 [1]. In some regions, nearly all urban dwellers live in relatively high-density cities. But in others, a large proportion of human homes are at the margins of cities, in suburbs. Suburban development is the fastest growing anthropogenic land-use in the United States, expanding approximately seven- to ten-fold from 1950–2000 [2]. In line with national trends, this phenomenon is rapidly changing the southeastern United States, and models predict a continued growth in urban and suburban land-use of 101% to 192% over the next 50 years [3]. This land-use change due to anthropogenic development simultaneously disfavors species dependent on wild forests and grasslands, and favors species associated with urban areas, with suburban areas having the potential to act as an ecotone between urban and natural areas or to be completely distinct from natural areas. Among the species suburban and urban development may favor are arthropod species able to vector human pathogens [4]. Vector-borne diseases are serious public health threats, affecting more than one billion people yearly [5] and causing significant health impacts and economic burdens. Globally, mosquitoes are the arthropod disease vectors of greatest importance, transmitting pathogens that cause significant morbidity and mortality such as malaria, dengue virus, the recently emerging Zika virus, and filarial parasites [6]. In recent years, the geographic ranges of many vector-borne diseases have been expanding, increasing their already significant public health and economic impacts. Many factors are potentially contributing to the expansion of vector-borne disease ranges, such as human travel, transport of products, and global climate change, but also anthropogenic land-use changes, including urbanization [7]. Land use changes are of particular interest because of their ability to affect entire species assemblages. Changes in species assemblage can significantly affect the diversity and composition of mosquito species and their relative abundance, all of which have important implications for the spread of vector-borne diseases [8–10].
8 When natural landscapes are converted to suburban development, the accompanying land clearing and construction acts as a large disturbance event in the environment, impacting species assemblages. Some ecological studies suggests that intermediate levels of disturbance will result in increased diversity, both because disturbance introduces habitat heterogeneity and because disturbance can reduce the abundance of dominant species, increasing potential for invasion [11]. Within the urban setting, this equates to a theoretical diversity peak at intermediate levels of development, when the biotic limitations of rural areas and the physical limitations of urban areas are both alleviated [12]. Bird diversity in Twin Cities, Minnesota and Chicago, Illinois, for instance, is highest in moderately disturbed areas and then subsequently declines in highly urban areas [13,14]. Similarly, butterfly diversity in an area of former oak woodlands in Palo Alto, California is lowest in the most urbanized areas, but in this case, even moderate levels of land development are detrimental to the natural species assemblage despite the overall intermediate disturbance diversity peak [12]. It is currently unclear whether the fine-scale heterogeneity in land-use associated with suburban development may lead to increases in mosquito biodiversity compared to that in natural woodlands or grasslands, as is seen with other species [15–17]. While understudied, the issue of mosquito species diversity in disturbed areas is important when considering the spread of vector-borne diseases. Vector competence for a given pathogen varies between species, leading to obvious implications for disease risk after community-level changes. Changes in the diversity or evenness of a community could affect vector-borne disease transmission if the species assemblage is shifted toward one dominated by mosquitoes with greater vectorial capacity. This is particularly relevant in suburban areas, as these disturbed environments place vector mosquitoes in close proximity and routine contact with hosts, including humans and their companion animals, which may increase risk for disease spread [4,18]. Despite the rise in urbanization and its potential effects on mosquito communities, and a rapid increase in the number of studies of the effects of urbanization, few studies have considered the ways in which urbanization in general, and suburbanization in particular, influence the overall composition of mosquito communities (S1 Fig). Where the influence of habitat on mosquito communities has been studied, it is most often in the context of exclusively natural habitats. For example, studies investigating the species assemblages present at wooded sites when compared to pasture or grassland sites have shown, perhaps unsurprisingly, that these
9 habitats tend to sustain different mosquito species [8,19,20]. These links between mosquito species and habitat can arise because of adaptation to certain habitats through specificity in breeding sites or selectivity of use by adult mosquitoes, with fine scale (< 20m) variation in flying, host-seeking, and adult distribution [21]. In contrast to the amount of studies on mosquito assemblages in natural areas, those investigating similar questions in urbanized areas are less common. The studies that have considered primarily anthropogenic landscapes have tended to focus on species-specific distributions rather than communities as a whole (e.g. [22–24]). The few studies that have approached questions of community-level effects have had study sites that are vastly different from suburban areas in the United States (e.g. [9,25]), leaving a gap in knowledge as to the effects of suburbanization on mosquito assemblages. Given that suburban areas are heterogeneous landscapes composed of grass, shrubs, trees, man-made structures, and perhaps surrounded by undeveloped natural areas, it is unclear whether mosquito assemblages present in these areas are more like those in fields, woodlots, or something unique to the suburban landscape. Here we sought to characterize the changes to mosquito assemblages that occur after human driven land-use change in the context of suburban development, and, if such changes exist, the rate at which these changes occur post-development. We hypothesize that 1) suburban species assemblages are distinct from those in either undeveloped fields or undeveloped woodlots, and 2) the mosquito assemblage changes rapidly after the initial disturbance, approaching a suburban climax community through time. We tested these a priori predictions by establishing a chronosequence of suburban developments in which to sample mosquitoes, and comparing these species assemblages to those present in uncultivated field and woodlot areas.
METHODS Study Overview We conducted this research in Wake County, North Carolina, USA. Wake County has a temperate climate and consists of a major urban center, Raleigh, and extensive suburbs, making it North Carolina’s second most populous county with around 1 million residents [26]. To characterize the changes in mosquito assemblage that occur after human driven land-use change, we sampled mosquitoes in neighborhoods that were previously woodlots before development as well as neighborhoods that were fields prior to development, based upon historical aerial photos
10 in Google Earth [27]. To compare the community assemblages to those in natural areas, we also sampled mosquitoes in natural woodlots and natural fields and grasslands as controls. To define the rate at which any changes to the mosquito assemblage occur after suburban development, we established a chronosequence by sampling in neighborhoods of various ages. This approach allowed us to determine the effects of approximate time scales within a two-year study.
Site Selection and Owner Permission We identified candidate suburban neighborhoods using Google Earth current and historical imagery. We defined suburban neighborhoods as those consisting of detached single- family homes not located on a city block. Based on historical images, we categorized candidate neighborhoods in Wake County, NC by age and then classified them as previously fields or previously woodlots before development. To ensure that neighborhoods of various ages were being represented in the study, we created age categories to guide neighborhood selection: developed before 1993, between 1993 and 2002, between 2003 and 2007, between 2008 and 2012, and from 2013 to present. As changes in mosquito assemblage were predicted to happen fairly quickly after the initial disturbance, age category intervals were designed to be shorter when neighborhood development was closer to the present time, and longer when development was further removed. This allowed us to better capture periods of predicted rapid change in our chronosequence, while reducing sampling efforts in periods of less rapid change. Within the age categories, we selected neighborhoods from each previous land use (i.e. fields or woodlot). To ensure even sampling throughout Wake County, we split the county into geographical quadrants and selected neighborhoods across all quadrants in a haphazard manner. Because of the stringent requirements of neighborhood age, previous land use, and the necessity for homeowner permission for trapping, site selection was fairly limited, but each geographical quadrant had at least one site representative of each age category and previous land use. Overall, we selected 30 neighborhoods in 2015 that spanned the age categories and previous uses from across Wake County (Fig 1). Within each neighborhood, we selected a single house for trap placement based on homeowner approval. We verified that the homeowner did not intend to perform any mosquito pesticide applications during the course of the study. Location of the trap within the yard was based entirely on homeowner preference, although the majority of homeowners elected to have the trap placed in their backyard near the property edge.
11
Figure 1. Trapping sites within Wake County, North Carolina. All 2015 sites denoted by a circle were also sampled in 2016, with the addition of 6 older suburban sites that were sampled exclusively in 2016, denoted by a triangle. Map created using public domain data from Wake County Government’s Wake County GIS [28] and the U.S. Census Bureau’s 2017 TIGER/Line Shapefiles [29].
We also sampled undeveloped field and woodlot habitats as controls. We used three natural habitat sites: Schenck Memorial Forest at North Carolina State University (NC State), NC State’s Equine Educational Unit, and NC State’s Lake Wheeler Beef Unit. Each of these natural sites featured both field habitat and woodland habitat, and we placed traps at least 100m away from a habitat edge, consistent with previous findings of mosquito habitat fidelity [21]. As large undisturbed wild areas are not especially common around Raleigh, NC, we also sampled at smaller parcels of land consisting of natural woodlots and natural fields as control sites. Each of these smaller control sites had at least a 100m radius of non-developed natural land around the trap, again consistent with previous findings of mosquito habitat fidelity [21]. We sampled at six of these smaller wooded control sites and five smaller field control sites, giving us an overall total of 9 woodlot control sites and 8 field control sites.
12 We repeated the study in 2016, trapping at all of the same control sites and at different houses within the same neighborhoods that were sampled in 2015. Given that the oldest neighborhood that we sampled in 2015 was 40 years old, we also added six additional older neighborhoods that were developed between 50 and 102 years prior in 2016, resulting in a total of 36 neighborhoods sampled (Fig 1) and giving us an even broader time scale for chronosequence analysis.
Trapping We used CDC light traps (JW Hock Co., Gainesville, FL) to sample mosquitoes throughout the course of the study. We removed the lights from the CDC light traps to reduce by-catch, and baited the traps with a small cooler containing 1kg of dry ice (solid CO2) to attract host-seeking mosquitoes. We set traps at each of the neighborhood and control sites for approximately 16 hours overnight biweekly from June through mid-October in 2015 and June through the end of October 2016.
Specimen Identification We enumerated and identified all collected mosquitoes to species using Burkett-Cadena’s Mosquitoes of the Southeastern United States [30] and Darsie and Ward’s Mosquitoes of North America [31] dichotomous keys. We then preserved the mosquitoes in 95% ethanol for future reference.
Land-use Classification We characterized site habitats by their current land use and vegetative structure within 100m radius of the trap. In the geographic information system (GIS) software ArcMap [32], we hand digitized the landscape using contemporary imagery and classified the resultant polygons as one of 9 categories: deciduous forests and evergreen forests (coarse vegetation), scrub/shrub (medium vegetation), grassland (fine vegetation), barren land (bare soil), buildings and pavement (impervious surfaces), cultivated crops, and water. We considered cultivated crops and grassland as field components and classified deciduous and evergreen forests as woodlot components. In the data analysis, we used field components, woodlot components, scrub/shrub, buildings, and
13 pavement as predictor variables and the remaining land-use categories as covariables that were not of immediate interest.
Statistical Analyses As we sampled at the same neighborhoods and control sites in both years with the addition of six unique older neighborhoods in 2016, we chose to treat the two years separately rather than combining into a single dataset. This allowed us to avoid pseudoreplication of the overlapping neighborhoods between years, verify findings from the first year of trapping, and assess for the effects of adding in neighborhoods greater than 40 years old in the second year of trapping. All statistical analyses were completed using the Vegan package in R 3.2.2 statistical software [33]. Species accumulation curves were performed to validate that there were enough trapping sites to ensure sufficient mosquito sampling. Recognizing that each technique for measuring diversity has limitations, we approached our analysis of mosquito diversity at each site holistically by measuring rarefied species richness, Pielou’s evenness index, and Shannon- Wiener diversity index. We chose the Shannon-Wiener index due to its widespread use in comparable published studies, as well as its recognition as a valid expression of species diversity [34]. We calculated rarefied species richness, Pielou’s evenness, and Shannon-Wiener diversity index and compared these diversity measures between control woodlots, control field sites, and neighborhood age categories via ANOVA to test the hypothesis that natural sites differ from suburban sites with regard to their mosquito diversity. We performed linear regressions modeling the effects of neighborhood age, housing density, home price, and median household income on rarefied species richness, evenness, and Shannon-Wiener diversity index. Although we are primarily interested in the effects of neighborhood age, housing density and socioeconomic factors could also affect mosquito diversity and covary with neighborhood age. For housing density measurements, we simply counted the number of houses within 100m radius of the trapping site. To approach socioeconomic status, we acquired median household income data by census tract from the 2013 American Community Survey via the United States Census Bureau’s Census Explorer tool [35]. Recognizing that census tracts are much larger than a given neighborhood, we also approached socioeconomic status at the neighborhood scale by calculating the average estimated home price of the trapping site and its two nearest neighbors
14 using Zillow [36], then normalized all values to the average home price in Wake County [37]. We calculated variance inflation factors to assess for collinearity between neighborhood age, housing density, and both measures of socioeconomic status. We also performed Partial Canonical Correspondence Analysis (PCCA) in R to investigate the relationship between land-use classification and mosquito assemblages for a given site. PCCA is a method for exploring relationships between two multivariate sets of variables measured on the same individual [38]. Here, the individual on which we are measuring our variables is trapping site, and the two multivariate sets of variables are the mosquito species assemblages and the GIS habitat classifications of the land use and vegetative structure around the trapping site. This approach allows us to summarize the relationships between these factors using a lesser number of statistics while preserving the main facets of the relationship. Utilizing the GIS land use classifications described above, we used field components, wood components, scrub/shrub, buildings, and pavement as predictor variables in the PCCA analysis, given our a priori assumptions of their effects on mosquito habitats. We controlled for the barren land and permanent water components by treating them as covariates, as their effects were presumed to be minimal when compared to the other land use classification categories. We plotted the CCA axis values against neighborhood age for each of the suburban sites to visualize the effect of neighborhood age in conjunction with the site’s land-use structure and mosquito species assemblage. To complement the PCCA, we also performed analysis of similarities (ANOSIM) using Bray-Curtis dissimilarity to compare the community composition between suburban, field control, and wood control sites. Finally, we utilized similarity percentages analysis (SIMPER) to evaluate the contribution of individual species to the overall Bray-Curtis dissimilarity.
RESULTS The species accumulation curves appeared to reach an asymptote for both years, indicating that our trapping sites were adequate to sufficiently sample mosquitoes in this area (S2 Fig). In total, we trapped 22 species of mosquitoes (out of 28 known to be historically present in Wake County, NC [B. Byrd, personal communication, February 8, 2019]) and a total of more than 10,000 individuals. In 2015, we trapped a total of 20 species and 4269 individual mosquitoes, 47.1% of which were the Asian tiger mosquito, Aedes albopictus, an invasive
15 species in the U.S. Similar numbers were seen in 2016, with 19 total species and 5975 total individuals, 38.1% of which were our most commonly trapped species, Ae. albopictus. Two trapping sites in 2015, one suburban site and one field control site, were subjectively different from the other sites. The suburban site’s trap had been placed close to the home’s HVAC unit at the owner’s request, deviating from the typical back-of-yard placement at other suburban sites. The field site was agricultural land as opposed to the other field controls, which were pasture or unused land. In 2016, one field control site suffered from loss of the land owner’s trapping permission after a single trap night. These three sites yielded fewer than 10 total individuals across the entire summer’s trapping efforts. All of the aforementioned factors could have resulted in lower mosquito yields and made the sites sufficiently different for exclusion, so the three sites were excluded from further analysis. Rarefied species richness, evenness, and Shannon-Wiener diversity all differed significantly when comparing the natural sites to the suburban sites grouped according to their age categories in 2015 and in 2016. In 2015, the suburban neighborhood age categories differed only among themselves with regard to all diversity measures, and were not significantly different from either of the natural categories (Fig 2). The overall model for rarefied richness was significant (p = 0.013, df = 6, F = 3.156), with the oldest suburban neighborhoods being significantly different from the two youngest age categories. No significant differences found between suburban neighborhoods and control sites. Despite the overall model being significant (p = 0.032, df = 6, F = 2.614), none of the categories were significantly different in their evenness after multiple comparison correction. The two oldest suburban age categories significantly differed from the youngest suburban category with regard to Shannon-Wiener diversity (overall model p = 0.007, df = 6, F = 3.561). Again, no significant differences were found between suburban neighborhoods and control sites. However, in 2016, the oldest neighborhood age category (pre-1993 development) significantly differed from both the natural field and the natural woodlots with regard to all three diversity measures (Fig 3). Various suburban age categories differed when compared with other suburban age categories with their rarefied richness, but of interest is the significant difference between the oldest suburban age category and both the field and wooded control sites (overall model p < 0.001, df = 6, F = 8.173). As with rarefied richness, evenness significantly differed between the oldest suburban sites and both of the control categories, as well as between various of the suburban age categories (overall
16
Figure 2. Comparison of diversity measures between control sites and suburban sites by age category in 2015. For all boxplots, the median is represented by the thick black line. Lower and upper hinges denote the first and third quartiles respectively, while whiskers extend from the hinge to the furthest value that is within 1.5x the interquartile range. Outliers are plotted individually. Boxplots are presented for (A) rarefied richness (p = 0.013, df = 6, F = 3.156), (B) evenness (p = 0.032, df = 6, F = 2.614), and (C) Shannon-Wiener diversity (p = 0.007, df = 6, F = 3.561), with significant comparisons denoted by letter above the boxplots.
17
Figure 3. Comparison of diversity measures between control sites and suburban sites by age category in 2016. For all boxplots, the median is represented by the thick black line. Lower and upper hinges denote the first and third quartiles respectively, while whiskers extend from the hinge to the furthest value that is within 1.5x the interquartile range. Outliers are plotted individually. Boxplots are presented for (A) rarefied richness (p < 0.001, df = 6, F = 5.25), (B) evenness (p < 0.001, df = 6, F = 5.25), and (C) Shannon-Wiener diversity (p < 0.001, df = 6, F = 8.655), with significant comparisons denoted by letter above the boxplots.
18 model p < 0.001, df = 6, F = 5.25). Similar trends were seen with Shannon-Wiener diversity, with the oldest suburban sites significantly differing from both of the control categories being the comparison of interest (overall model p < 0.001, df = 6, F = 8.655). When considering the suburban sites specifically, rarefied richness, evenness, and Shannon-Wiener diversity index were all well explained by a model including neighborhood age, housing density, home price, and median household income for both years of sampling (p ≤ 0.011 for all years and diversity measures, Table 1). When considering these factors individually, neighborhood age was significant in all models (p ≤ 0.005 for all models) and was the only significant factor across both years for any of the diversity measures (Table 1). As neighborhoods age, they become less species rich, less diverse, and species are less evenly distributed. Since Ae. albopictus made up a significant portion of the collected mosquitoes in both years, we performed a linear regression of Ae. albopictus abundance versus the primary determinant of diversity changes, neighborhood age. Aedes albopictus abundance increased as
Table 1. Effects of neighborhood age, housing density, home price, and household income on mosquito diversity measures. Each diversity measure was modeled with neighborhood age, housing density, home price, and median household income as predictor variables for both years. Overall model p-values, degrees of freedom, F- statistics, and R2 values are reported, as well as individual p-values for each of the predictors. All significant variables were negatively correlated with all diversity metrics. Variance inflation factors were less than 1.5 for all predictors across both years. Significant relationships are denoted in bold with an asterisk.
Year Diversity Model R2 Age Density Price Income Measure p-value, df, F- p-value p-value p-value p-value statistic Rarefied <0.001*, 4 and 0.527 <0.001* 0.099 0.444 0.831 Richness 24, 6.689 0.011*, 4 and Evenness 0.406 0.005* 0.623 0.569 0.457 24, 4.105 2015 Shannon- Wiener <0.001*, 4 and 0.552 <0.001* 0.064 0.941 0.511 Diversity 24, 7.417 Index Rarefied <0.001*, 4 and 0.497 <0.001* 0.064 0.974 0.521 Richness 31, 7.643 0.002*, 4 and Evenness 0.417 0.002* 0.070 0.839 0.777 31, 5.547 2016 Shannon- Wiener <0.001*, 4 and 0.489 <0.001* 0.048* 0.776 0.568 Diversity 31, 7.4 Index
19 neighborhoods got older, with neighborhood age explaining 34.6% of variation in Ae. albopictus abundance in 2015 and 48.4% of variation in 2016 (p < 0.001, Fig 4). The remaining four most abundant species were Ae. vexans, Culex pipiens, Cx. salinarius, and Cx. erraticus in 2015, and Cx. salinarius, Ae. vexans, Ae. atlanticus, and Cx. erraticus in 2016 (S1 Table). Of these species, the only significant relationship consistent across both years was a negative correlation between Cx. erraticus abundance and neighborhood age (p ≤ 0.001, S3 Fig).
Figure 4. Log abundance of Aedes albopictus increases with log neighborhood age. Control field and woodlot sites were excluded from this analysis to focus on the effect of suburban development as these areas become further established. (A) In 2015, a significant positive correlation was noted (p < 0.001, df = 1 and 27, F = 14.3, R2 = 0.346, r = 0.614). (B) With the addition of older neighborhoods in 2016, a similar significant positive correlation was noted (p < 0.001, df = 1 and 34, F = 31.86, R2 = 0.484, r = 0.67).
Given the differences in mosquito diversity and the relative abundance of Ae. albopictus among site types, the overall composition of mosquitoes also differed among site types. The environmental variables explained 35.9% of the total variation in the PCCA in 2015, with 86.4% of that variation explained by the first two CCA axes. Overall, field control sites and wooded control sites had distinct assemblages (Fig 5A), consistent with previous findings [21]. There did not appear to be any legacy effects of previous land use; that is, the suburban sites that were previously fields before development and those that were previously woods before development did not cluster together, but rather were fairly evenly dispersed throughout the range of suburban sites on the PCCA. Because of this lack of a legacy effect, suburban sites are shown in a single category on the PCCA for clarity. The suburban sites span across the field and wood habitat
20 types on the CCA1 axis but separate out well on the CCA2 axis, suggesting that the pavement and buildings habitat components are important determinants of a suburban site’s mosquito assemblages in a manner unique to suburbia relative to grassland or forested sites (Fig 5A). Similar trends were seen in 2016, with the environmental variables explaining 26.6% of the total variation and 88.4% of that variation explained by the first two CCA axes. Again, there did not appear to be any legacy effects of previous land use among the suburban sites. The wooded and field control sites were distinct, and the suburban sites again separated out best on the CCA2 axis, suggesting that the buildings and pavement habitat components were most important in determining the suburban mosquito assemblage (Fig 5B).
Figure 5. Partial Canonical Correspondence Analysis of mosquito assemblages and habitat classifications at each trapping site. For both PCCA figures, field control sites are represented by triangles, wood control sites are represented by squares, and suburban sites are represented by circles. Points are placed on the plot by the site’s mosquito community composition. Arrows for environmental factor of land-use classifications are overlaid, with the length of the arrow representing the strength of the relationship. (A) PCCA of the 2015 mosquito assemblage data. Environmental variables explained 35.9% of the total variation, with the first two axes explaining 86.4% of the environmental variation. (B) PCCA of the 2016 mosquito assemblage data. Environmental variables explained 26.6% of the total variation, with the first two axes explaining 88.4% of the environmental variation.
To explicitly investigate neighborhood age in relation to the mosquito assemblage and land-use classification information in the PCCA, we plotted the CCA values for each site against neighborhood age. After adding a constant to all values so that all CCA values were non- negative, a power curve best fit the data for CCA1 values (residual SE = 0.72) and CCA2 values (residual SE = 0.61) in 2015, as well as for CCA1 values (residual SE = 0.59) and CCA2 values
21 (residual SE = 0.80) in 2016 (S4 Fig). Based on these CCA regressions, the oldest suburban neighborhoods are in the upper right quadrant of the PCCA for 2015 (Fig 5A) and the upper left quadrant of the PCCA for 2016 (Fig 5B). In general, older neighborhoods tended to have mosquito community compositions more similar to those of woods than those of fields. Nonetheless, the suburban components of buildings and pavement are still the most important predictive land-use classification for the oldest neighborhoods, suggesting a unique suburban climax community. ANOSIM confirmed that the mosquito assemblage significantly differed between suburban sites, field control sites, and wooded control sites in 2015 (p = 0.001, R = 0.3061) and in 2016 (p = 0.003, R = 0.2449). Aedes albopictus was the most influential species, contributing 30% to 45.8% of the overall community dissimilarity in 2015 and 27.6% to 34.9% in 2016, depending on the site comparison (suburban vs. wood, suburban vs. field, or wood vs. field).
DISCUSSION In comparing suburban neighborhoods with wooded and field sites, we found that the oldest neighborhoods have mosquito communities that are less species rich, less even, and less diverse when compared with either of the natural habitats. This decline in all diversity metrics as suburban neighborhoods age is concordant with a shift in species assemblage to a seemingly distinct suburban assemblage. The suburban assemblage is well correlated with the presence of impervious surfaces like pavement and buildings. Dominant among this suburban mosquito assemblage is the invasive species Aedes albopictus, whose abundance is positively correlated with neighborhood age. Comparing diversity measures between control sites and suburban sites by their neighborhood age category did not yield significant differences between natural and suburban sites in 2015 (Fig 2). However, when incorporating older neighborhoods with the 2016 data, significant differences were found between the oldest neighborhood age category and both field and woodlot control sites in terms of rarefied richness, evenness, and Shannon-Wiener diversity (Fig 3). This shows that mosquito diversity in suburban neighborhoods becomes fundamentally different from that in undeveloped natural sites over time. Additionally, we performed linear regressions to assess the effect of neighborhood age, housing density, home price, and income on our three diversity measures. Variance inflation factors were less than 1.5 for all predictors
22 across both years, indicating that there is not significant concern for collinearity among neighborhood age, housing density, home price, and median household income in the current study. Analysis of linear regressions revealed that neighborhood age was the only significant variable across both years for any diversity measure (Table 1), indicating that neighborhood age is the primary driver of mosquito diversity changes among suburban sites. However, a more appropriate way to analyze the changes that occur after suburban development is in looking beyond just diversity measures and instead to changes in species composition, which we accomplished with the PCCAs. PCCA of the 2015 mosquito assemblage data shows that the mosquitoes collected at field control sites cluster together and are well defined by the field environment, with similar trends seen with the wooded control sites. In contrast, the mosquitoes collected at suburban control sites are best defined by pavement, buildings, and scrub/shrub environmental variables, suggesting that there is a distinct mosquito assemblage created as a result of common anthropogenic landscape components (Fig 5A). Similar results are seen in the PCCA of the 2016 mosquito assemblage data. The wooded and field sites are similar in their mosquito assemblages and separate out well based on their distinctive environments, while the mosquito assemblages at the suburban sites are best defined by the buildings, pavement, and scrub/ shrub environmental variables on the opposite CCA axis (Fig 5B). When plotting CCA values against neighborhood age to explicitly investigate the effect of neighborhood age on mosquito communities, both CCA1 and CCA2 values for suburban sites in 2015 increased with neighborhood age toward a maximum value in the climax community (S4A and S4B Fig). Taken together, these plots show that the oldest neighborhoods are in the upper right quadrant of the PCCA (Fig 5A), correlating with the woods environmental variable on the CCA1 axis, and the pavement, buildings, and scrub/shrub variables on the CCA2 axis. For 2016, CCA1 values decreased with neighborhood age, while CCA2 values increased with neighborhood age (S4C and S4D Fig). This shows that the oldest neighborhoods in 2016 fall into the upper left quadrant of the PCCA (Fig 5B), again correlating with woods on the CCA1 axis, and suburban environmental factors on the CCA2 axis. Overall, we found that as neighborhoods age, their mosquito assemblages shift somewhat from species associated with fields to those associated with woods. However, despite the potential shift from fine vegetation to coarse vegetation as a possible mechanism for this slight
23 shift in mosquito assemblage as a neighborhood matures, pavement and buildings remain extremely important in determining suburban mosquito species assemblages with increasing neighborhood age. Indeed, the oldest neighborhoods did not recover their mosquito diversity and revert to a wooded species assemblage, but instead became increasingly uneven and less diverse, suggesting the suburban assemblage is different from that present in either natural habitat. A recent study has shown that urbanization caused not only functional diversity loss in bird communities, but also a loss of phylogenetic diversity, where evolutionary distinctiveness of species is lessened in urban areas [39]. It is hard to assess whether this holds for mosquito communities as mosquito phylogenies are currently greatly debated, but future work could investigate the potential for evolutionary impacts created by shifts to a suburban mosquito assemblage. The most commonly encountered species in our study was Ae. albopictus, an invasive peridomestic species that thrives in artificial containers associated with anthropogenic land use [40,41], particularly in suburban habitats. The differences in mosquito assemblages seen when transitioning from natural habitats to suburban habitats are primarily driven by changes in the abundance of Ae. albopictus. While Cx. erraticus abundance was also significantly correlated with neighborhood age across both of the years in this study, Ae. albopictus more profoundly affected the suburban mosquito assemblage, as evidenced by the large difference between the two species in numbers of individuals caught as well as the individual contributions by Ae. albopictus of 27.6% to 45.8% to the overall community dissimilarity between site types detected in the SIMPER analysis. In general, the most conspicuous pattern seen throughout the study is that as neighborhoods age, their mosquito communities become less species rich, less even, less diverse, and more dominated by the introduced species Ae. albopictus. A possible mechanism for the abundance of Ae. albopictus in suburban areas is the increased availability of container larval habitat when compared to rural counterparts. Mosquitoes as a whole utilize a wide variety of larval habitats, with each individual species best adapted to a specific type of larval habitat. These aquatic habitats can include both permanent and temporary bodies of water such as ephemeral rainwater pools in fields and woodlands, tree holes, floodwater pools, ditches, ponds and marshes, sluggish streams and stream overflow pools, and man-made containers [30]. With the exception of man-made containers, most of these larval habitat types are likely decreasing in availability as land-use shifts from undeveloped to
24 highly suburbanized. A recent study [42] found significant increases in 3rd to 4th instar Ae. albopictus larvae in suburban areas, and, given significant differences in container availability between urban and rural areas, the study concluded that urbanization generally increases larval habitats for Ae. albopictus. This could, however, be complicated by socioeconomic factors, as Ae. albopictus prevalence varies by level of abandonment, with poor neighborhoods having more of this species than wealthier neighborhoods even within highly urbanized city centers [43]. While neither home price nor median household income as proxies for socioeconomic status significantly affected mosquito communities in the present study, larval habitat may be a key factor in the abundance of Ae. albopictus that was observed in suburban areas [42]. Further research is warranted to categorize the differences in container larval habitat in developed areas versus natural areas, as this may help guide land development policies or homeowner behavior to benefit vector management. Aside from their ability to easily exploit artificial containers, previous work has also shown that Ae. albopictus is an edge species, flourishing at ecotones between field and wooded habitats in natural areas [21]. Given that suburban areas are fragmented, heterogeneous landscapes where small plots of grasslands (lawns) are interspersed with trees and other medium to coarse vegetation, it is possible that suburban neighborhoods are functioning as large spaces of contiguous edge habitats, allowing edge species such as Ae. albopictus to thrive. This trend for an abundance of edge species in suburban areas has previously been documented in ants [44]. In that study, the authors found an increase in ant species diversity in suburban areas, which they attributed to the presence of forest species that are rare in fully open areas, but likely to be present at ecotones, which the suburban landscape created. Despite similar attribution of suburbia as an edge habitat causing changes in species diversity, results from the ant study [44] do not agree with the current study’s findings of decreased overall diversity, likely because of the presence and dominance of the invasive Ae. albopictus. Previous studies in bird communities have found that as neighborhoods age, species diversity decreases and the community composition shifts from primarily native species to heightened dominance of invasive species [14]. This aligns with the prediction that in areas of anthropogenic disturbance, the intermediate disturbance hypothesis only holds when native species are considered, as invasive species thrive in disturbed areas [45]. Our findings also support this prediction, as demonstrated by the increasing dominance of Ae. albopictus as suburban neighborhoods age.
25 Our study was potentially limited by the choice to sample exclusively with a dry ice baited CDC light trap with the light removed. Each trapping method has its own biases, and one of the most prominent criticisms against the CDC light trap is its underrepresentation of Ae. albopictus [46]. Despite this bias against Ae. albopictus, it was our most abundant collected species. As traps were left overnight for approximately 16 hours, the 1kg of dry ice did not last throughout the entire trapping period. However, since traps were placed in the late afternoon, it is reasonable to assume that day-biting, crepuscular, and nocturnal mosquitoes were all potentially sampled for at least part of their host-seeking period before the dry ice sublimated. Removing the light had the benefit of decreasing bycatch, but it likely decreased the attractive range of the trap, particularly for light-attractive species. Since the attractive range of the CO2 baited CDC light trap is less than 15m even with the light intact [47], we anticipated our attractive power to be at a short range. Active attractive range of the CDC light trap is limited, but common mosquitoes in our study areas have been shown to disperse 300m to 1km [48], with Ae. albopictus having a daily mean distance traveled of 119m [49]. Housing density within 100m of a trap, and therefore within the average daily distance traveled by mosquitoes, ranged from 1 (in an area of active new construction) to 59 houses (in an extremely dense area with very small yards), with an average of about 20 houses per 100m across both years of trapping. This indicates that our adult trapping could have sampled from mosquitoes produced in larval habitats in many of the neighboring yards and surrounding areas, despite trapping at a single house per neighborhood. Although the limitations of trap choice and sampling scheme were not prohibitive in this study, future research could use varied mosquito sampling techniques and increased intra-neighborhood replication to further investigate questions of suburban succession of mosquito communities. Previous studies have shown that mosquito assemblages in low diversity urban areas tend to be dominated by vector species of concern for human diseases [50]. Our results suggest that with steadily increasing land-use change and the particular fervor of suburban development, mosquito species assemblages will be altered and Ae. albopictus, a known vector, will become more common. These community level changes can significantly alter vector-borne disease risk for many pathosystems. Aedes albopictus readily bites humans, and is a competent vector of myriad pathogens of global human health concern, including dengue virus, West Nile virus, chikungunya virus, and Zika virus [41,51], as well as pathogens of focal concern in North Carolina, including La Crosse virus [52,53]. The overall community simplification and
26 increasing abundance of Ae. albopictus in highly suburbanized areas make human-vector interactions more likely, heightening the potential for transmission of these concerning vector- borne diseases through both greater vector population size alone, and an increase in host biting rate due to proximity for contact [54]. In light of the overwhelming success of Ae. albopictus in established suburban areas, further research is warranted to elucidate how to mitigate the potentially devastating effects of suburban development favoring competent vector species.
ACKNOWLEDGEMENTS We thank Shawn Janairo, Tommy Pleasant, Hannah Jenkins, Cole Keenan, Dakota Palacio, and Chris Hayes for their work on this project.
SUPPORTING INFORMATION
S1 Figure. Trends in research on mosquito diversity changes during urbanization. Number of peer-reviewed publications by year on topics “urbanization” and “diversity,” “urbanization” and “mosquito,” as well as “urbanization” and “mosquito” and “diversity.” Citation reports accessed from Web of Science Core Collection on September 3, 2018.
27
S2 Figure. Species accumulation curves based on random sampling. Species accumulation curves for (A) the 2015 data and (B) the 2016 data. Vertical bars represent 95% confidence intervals.
S3 Figure. Log abundance of Culex erraticus decreases with log neighborhood age. Control field and woodlot sites were excluded from this analysis to focus on the effect of suburban development as these areas become further established. (A) In 2015, a significant negative correlation was noted (p = 0.001, df = 1 and 27, F = 12.54, R2 = 0.317, r = -0.544). (B) In 2016, a similar significant negative correlation was noted (p < 0.001, df = 1 and 34, F = 19.66, R2 = 0.366, r = -0.613).
28
S4 Figure. Comparison of suburban site CCA values versus neighborhood age. CCA values for the suburban sites were plotted against neighborhood age to explicitly assess the effect of time on the suburban mosquito assemblage. (A) CCA1 values from 2015 were best modeled by a power curve (residual SE = 0.72). (B) CCA2 values from 2015 were also best modeled by a power curve (residual SE = 0.61). (C) CCA1 values from 2016 best fit a power curve with a negative relationship (residual SE = 0.59), while (D) CCA2 values from 2016 best fit a power curve with a positive relationship (residual SE = 0.8).
29 S1 Table. Relationship between neighborhood age and mosquito species abundance. The five most prevalent species during both of the trapping years were assessed for their relationship with neighborhood age. Due to differential numbers of species caught in each trapping year, a total of six mosquito species were assessed. P-values, degrees of freedom, and F-statistics for the linear regressions of log abundance versus log neighborhood age are presented, with significant relationships denoted in bold with an asterisk. R2 and Spearman’s coefficient are given for significant relationships.
30 REFERENCES
1. United Nations. World Urbanization Prospects: The 2018 Revision. 2018.
2. Brown DG, Johnson KM, Loveland TR, Theobald DM. Rural land-use trends in the conterminous United States, 1950-2000. Ecol Appl. 2005; 15: 1851-1863. https://doi.org/10.1890/03-5220
3. Terando AJ, Costanza J, Belyea C, Dunn RR, McKerrow A, Collazo JA. The southern megalopolis: using the past to predict the future of urban sprawl in the Southeast US. PLoS One. 2014; 9: e102261. https://doi.org/10.1371/journal.pone.0102261 PMID: 25054329
4. Allan BF, Keesing F, Ostfeld RS. Effect of forest fragmentation on Lyme disease risk. Conserv Biol. 2003; 17: 267-272.
5. World Health Organization. A global brief on vector-borne diseases. 2014.
6. Foster WA, Walker ED. Mosquitoes (Culicidae). In: Mullen GR, Durden LA, editors. Medical and veterinary entomology. Cambridge, MA: Elsevier; 2019. pp. 261-325.
7. Mayer SV, Tesh RB, Vasilakis N. The emergence of arthropod-borne viral diseases: A global prospective on dengue, chikungunya and Zika fevers. Acta Trop. 2017; 166: 155- 163. https://doi.org/10.1016/j.actatropica.2016.11.020 PMID: 27876643
8. Steiger DBM, Ritchie SA, Laurance SGW. Mosquito communities and disease risk influenced by land use change and seasonality in the Australian tropics. Parasites & Vectors. 2016; 9: 387. https://doi.org/10.1186/s13071-016-1675-2 PMID: 27388293
9. Steiger DBM, Ritchie SA, Laurance SGW. Land Use Influences Mosquito Communities and Disease Risk on Remote Tropical Islands: A Case Study Using a Novel Sampling Technique. Am J Trop Med Hyg. 2016; 94: 314-321. https://doi.org/10.4269/ajtmh.15- 0161 PMID: 26711512
10. Carlson JC, Byrd BD, Omlin FX. Field assessments in western Kenya link malaria vectors to environmentally disturbed habitats during the dry season. BMC Public Health. 2004; 4: 33. https://doi.org/10.1186/1471-2458-4-33 PMID: 15296512
11. Connell JH. Diversity in Tropical Rain Forests and Coral Reefs—High Diversity of Trees and Corals is Maintained Only in a Non-Equilibrium State. Science. 1978; 199: 1302- 1310.
12. Blair RB, Launer AE. Butterfly diversity and human land use: Species assemblages along an urban gradient. Biol Conserv. 1997; 80: 113-125. https://doi.org/10.1016/S0006- 3207(96)0056-0
31 13. Chapman KA, Reich PB. Land use and habitat gradients determine bird community diversity and abundance in suburban, rural and reserve landscapes of Minnesota, USA. Biol Conserv. 2007; 135: 527-541. https://doi.org/10.1016/j.biocon.2006.10.050
14. Loss SR, Ruiz MO, Brawn JD. Relationships between avian diversity, neighborhood age, income, and environmental characteristics of an urban landscape. Biol Conserv. 2009; 142: 2578-2585.
15. Tscharntke T, Steffan-Dewenter I, Kruess A, Thies C. Contribution of small habitat fragments to conservation of insect communities of grassland-cropland landscapes. Ecol Appl. 2002; 12: 354-363.
16. Chaves LF, Hamer GL, Walker ED, Brown WM, Ruiz MO, Kitron UD. Climatic variability and landscape heterogeneity impact urban mosquito diversity and vector abundance and infection. Ecosphere. 2011; 2: UNSP art70. https://doi.org/10.1890/ES11- 00088.1
17. Fahrig L. Effects of habitat fragmentation on biodiversity. Annual review of ecology, evolution, and systematics. 2003; 34: 487-515.
18. Junglen S, Kurth A, Kuehl H, Quan P, Ellerbrok H, Pauli G, et al. Examining landscape factors influencing relative distribution of mosquito genera and frequency of virus infection. Ecohealth. 2009; 6: 239-249. https://doi.org/10.1007/s10393-009-0260-y PMID: 19915916
19. O’Brien VA, Reiskind MH. Host-seeking mosquito distribution in habitat mosaics of southern Great Plains cross-timbers. J Med Entomol. 2013; 50: 1231-1239. PMID: 24843927
20. Gardner AM, Lampman RL, Muturi EJ. Land use patterns and the risk of West Nile Virus transmission in central Illinois. Vector-Borne and Zoonotic Diseases. 2014; 14: 338-345. https://doi.org/10.1089/vbz.2013.1477 PMID: 24746038
21. Reiskind MH, Griffin RH, Janairo MS, Hopperstad KA. Mosquitoes of field and forest: the scale of habitat segregation in a diverse mosquito assemblage. Med Vet Entomol. 2017; 31: 44-54. https://doi.org/10.1111/mve.12193 PMID: 27759165
22. Hopperstad KA, Reiskind MH. Recent Changes in the Local Distribution of Aedes aegypti (Diptera: Culicidae) in South Florida, USA. J Med Entomol. 2016; 53: 836-842. https://doi.org/10.1093/jme/tjw050 PMID: 27113103
23. Kothera L, Godsey M, Mutebi J, Savage HM. A comparison of aboveground and belowground populations of Culex pipiens (Diptera: Culicidae) mosquitoes in Chicago, Illinois, and New York City, New York, using microsatellites. J Med Entomol. 2010; 47: 805-813. PMID: 20939375
32 24. Kristan M, Fleischmann H, Della Torre A, Stich A, Curtis CF. Pyrethroid resistance/susceptibility and differential urban/rural distribution of Anopheles arabiensis and An. gambiae ss malaria vectors in Nigeria and Ghana. Med Vet Entomol. 2003; 17: 326-332. PMID: 12941018
25. Kun JF, Kremsner PG. Mosquito distribution and entomological inoculation rates in three malaria-endemic areas in Gabon. Trans R Soc Trop Med Hyg. 2000; 94: 652-656. PMID: 11198649
26. Wake County Government [Internet]. Census, Demographics and Population Data; [cited 2018 Apr 30]. http://www.wakegov.com/planning/demographic/Pages/default.aspx.
27. Google Inc. Google Earth. 7.1.5.1557.
28. Wake County Government [Internet]. Download Data; [cited 2018 Sep 6]. http://www.wakegov.com/gis/services/Pages/data.aspx.
29. United Status Census Bureau [Internet]. TIGER/Line; [cited 2018 Sep 6]. https://www.census.gov/geo/maps-data/data/tiger-line.html.
30. Burkett-Cadena ND. Mosquitoes of the southeastern United States. University of Alabama Press; 2013.
31. Darsie RF. Identification and geographical distribution of the mosquitoes of North America, north of Mexico. Gainsville: University Press of Florida; 2005.
32. Esri. ArcMap. 2014; 10.2.2.
33. Oksanen J, Blanchet FG, Friendly M, Kindt R, Legendre P, McGlinn D. et al. vegan: Community Ecology Package. 2018.
34. Magurran AE, McGill BJ. Biological diversity: frontiers in measurement and assessment. Oxford University Press; 2011.
35. United States Census Bureau [Internet]. Census Explorer; [cited 2019 Feb 13]. https://www.census.gov/censusexplorer/.
36. Zillow [Internet]. c2006-2018 [cited 2018 Oct 23]. https://www.zillow.com.
37. Zillow [Internet]. Wake County NC Home Prices and Values; c2006-2018 [cited 2018 Oct 23]. https://www.zillow.com/wake-county-nc/home-values/.
38. Ter Braak CJ, Verdonschot PF. Canonical correspondence analysis and related multivariate methods in aquatic ecology. Aquat Sci. 1995; 57: 255-289.
33 39. Morelli F, Benedetti Y, Ibáñez-Álamo JD, Jokimäki J, Mänd R, Tryjanowski P, et al. Evidence of evolutionary homogenization of bird communities in urban environments across Europe. Global Ecol Biogeogr. 2016; 25: 1284-1293.
40. Benedict MQ, Levine RS, Hawley WA, Lounibos LP. Spread of the tiger: Global risk of invasion by the mosquito Aedes albopictus. Vector-Borne and Zoonotic Diseases. 2007; 7: 76-85. https://doi.org/10.1089/vbz.2006.0562 PMID: 17417960
41. Bonizzoni M, Gasperi G, Chen X, James AA. The invasive mosquito species Aedes albopictus: current knowledge and future perspectives. Trends Parasitol. 2013; 29: 460- 468. https://doi.org/10.1016/j.pt.2013.07.003 PMID: 23916878
42. Li Y, Kamara F, Zhou G, Puthiyakunnon S, Li C, Liu Y, et al. Urbanization increases Aedes albopictus larval habitats and accelerates mosquito development and survivorship. PLoS neglected tropical diseases. 2014; 8: e3301. https://doi.org/10.1371/journal.pntd.0003301 PMID: 25393814
43. Little E, Biehler D, Leisnham PT, Jordan R, Wilson S, LaDeau SL. Socio-ecological mechanisms supporting high densities of Aedes albopictus (Diptera: Culicidae) in Baltimore, MD. J Med Entomol. 2017; 54: 1183-1192. https://doi.org/10.1093/jme/tjx103 PMID: 28605549
44. Nuhn TP, Wright CG. An ecological survey of ants (Hymenoptera: Formicidae) in a landscaped suburban habitat. Am Midl Nat. 1979: 353-362.
45. Kowarik I. On the role of alien species in urban flora and vegetation. In: Marzluff JM, Shulenberger E, Alberti M, Bradley G, Ryan C, Simon U, and ZumBrunnen C. editors. Urban ecology. New York, NY: Springer; 2008. pp. 321-338.
46. Farajollahi A, Kesavaraju B, Price DC, Williams GM, Healy SP, Gaugler R, et al. Field efficacy of BG-Sentinel and industry-standard traps for Aedes albopictus (Diptera: Culicidae) and West Nile virus surveillance. J Med Entomol. 2009; 46: 919-925. PMID 19645298
47. Brown HE, Paladini M, Cook RA, Kline D, Barnard D, Fish D. Effectiveness of mosquito traps in measuring species abundance and composition. J Med Entomol. 2008; 45: 517-521. PMID: 18533447
48. Medeiros MC, Boothe EC, Roark EB, Hamer GL. Dispersal of male and female Culex quinquefasciatus and Aedes albopictus mosquitoes using stable isotope enrichment. PLoS neglected tropical diseases. 2017; 11: e0005347. https://doi.org/10.1371/journal.pntd.0005347 PMID: 28135281
49. Marini F, Caputo B, Pombi M, Tarsitani G, Della Torre A. Study of Aedes albopictus dispersal in Rome, Italy, using sticky traps in mark-release-recapture experiments. Med
34 Vet Entomol. 2010; 24: 361-368. https://doi.org/10.1111/j.1365-2915.2010.00898.x PMID: 20666995
50. Medeiros-Sousa AR, Fernandes A, Ceretti-Junior W, Wilke ABB, Marrelli MT. Mosquitoes in urban green spaces: using an island biogeographic approach to identify drivers of species richness and composition. Scientific reports. 2017; 7: 17826. https://doi.org/10.1038/s41598-017-18208-x PMID: 29259304
51. Musso D, Gubler DJ. Zika virus. Clin Microbiol Rev. 2016; 29: 487-524. https://doi.org/10.1128/CMR.00072-15 PMID: 27029595
52. Bara JJ, Parker AT, Muturi EJ. Comparative susceptibility of Ochlerotatus japonicus, Ochlerotatus triseriatus, Aedes albopictus, and Aedes aegypti (Diptera: Culicidae) to La Crosse virus. J Med Entomol. 2016; 53: 1415-1421. https://doi.org/10.1093/jme/tjw097 PMID: 27605372
53. Westby KM, Fritzen C, Paulsen D, Poindexter S, Moncayo AC. La Crosse encephalitis virus infection in field-collected Aedes albopictus, Aedes japonicus, and Aedes triseriatus in Tennessee. J Am Mosq Control Assoc. 2015; 31: 233-241. https://doi.org/10.2987/moco-31-03-233-241.1 PMID: 26375904
54. Dye C. Vectorial capacity: must we measure all its components? Parasitology Today. 1986; 2: 203-209. PMID: 15462840
35 CHAPTER 3
Mosquito Diversity and Dog Heartworm Prevalence in Suburban Areas
ABSTRACT Background: Urbanization is occurring rapidly on a global scale and is altering mosquito communities, creating assemblages that are characteristically less diverse. Despite high rates of urbanization and ample examples of vector-borne diseases transmitted by multiple species, the effects of urbanization-driven mosquito diversity losses on disease transmission has not been well explored. We investigated this question using the dog heartworm, a filarial parasite vectored by numerous mosquito species. Methods: We trapped host-seeking mosquitoes in undeveloped areas and neighborhoods of different ages in Wake County, North Carolina, USA, analyzing captured mosquitoes for heartworm DNA. We compared within-mosquito heartworm infection across land-use types by Kruskal-Wallis and likelihood ratio tests. Using zip code level data acquired from dogs in a local shelter, we performed linear regressions of within-host heartworm prevalence by within- mosquito heartworm prevalence as well as by three mosquito diversity measures. We also determined the best predictor of host-level prevalence among models including within-mosquito infection, mosquito diversity and abundance, and socioeconomic status as predictors. Results: Suburban areas had lower within-mosquito heartworm prevalence and lower likelihood of heartworm positive mosquitoes than did undeveloped field sites, although no differences were seen between suburban and undeveloped wooded sites. No relationships were noted between within-mosquito and within-host heartworm prevalence. However, mosquito diversity metrics were positively correlated with host heartworm prevalence. Model selection revealed within-host prevalence was best predicted by a positive relationship with mosquito Shannon-Wiener diversity and a negative relationship with household income. Conclusions: Our results demonstrate that decreases in mosquito diversity due to urbanization alter vector-borne disease risk. With regard to dog heartworm disease, this loss of mosquito diversity is associated with decreased heartworm prevalence within both the vector and the host. Although the response is likely different for diseases transmitted by one or few species, mosquito
36 diversity losses leading to decreased transmission could be generalizable to other pathogens with multiple vectors. This study contributes to better understanding of the effects of urbanization and the role of vector diversity in multi-vectored pathosystems.
BACKGROUND Mosquitoes are a diverse family of flies that can be pests and vectors of medical and veterinary significance, transmitting a wide array of viruses, protozoa, and parasitic nematodes to both humans and animals. Many important pathogens are transmitted by a limited number of mosquito vectors, and both research and control efforts have focused on targeting these primary vectors. However, there are numerous examples of disease systems in which an assembly of vector species are responsible for pathogen transmission (e.g. [1]-[4]). The focus of studies on primary vectors rather than communities of vectors has led to a general gap in knowledge as to how vector diversity contributes to disease transmission in multi-vectored pathosystems, even though the limited empirical and modeling investigations that have occurred suggest vector diversity can increase transmission [5],[6]. Given that anthropogenic land-use change alters the diversity and composition of mosquito assemblages and is occurring rapidly on a global scale, greater understanding of the role of vector diversity in multi-vectored diseases is critical. The effect of urbanization is well known for certain mosquito species of interest, particularly the container-breeding Aedes [7],[8], but has only recently been examined in the context of effects on mosquito species assemblages. Urbanized areas tend to have distinct mosquito communities that are characteristically less diverse than those in natural habitats [9]- [13]. Generally, anthropogenic disturbance is also associated with increased abundance of vectors of human disease, including container Aedes that transmit dengue, Zika, or chikungunya viruses, and Culex mosquitoes, which transmit West Nile virus and human filarial pathogens [10],[11],[13],[14]. The shift in mosquito assemblage to a lower richness community composed of a high proportion of known vectors likely increases disease transmission for most pathogens [9],[15]. However, other studies have suggested that for diseases transmitted by a variety of vectors, such as malaria or hemorrhagic disease in deer, vector species richness is strongly correlated with disease prevalence, possibly due to functional diversity extending the transmission season [5],[16]. An ideal system for further examining the effects of mosquito diversity on vector-borne disease transmission is that of the dog heartworm, Dirofilaria immitis.
37 The nematode D. immitis is an obligate parasite of mosquitoes and canids. Mosquitoes acquire microfilaria, the mosquito-infective parasite stage, upon ingestion of a bloodmeal from an infectious canine host. The parasite goes through three larval stages within the mosquito before migrating to the thorax and head of the mosquito at the L3 larval stage, which escapes from the mosquito’s proboscis and enters the bite wound of a susceptible host during the next blood feeding. Once within the canine host, the heartworm goes through additional developmental stages before reaching the adult stage, which resides and sexually reproduces in the pulmonary arteries and the heart, causing respiratory distress and eventually congestive heart failure [17]. Dog heartworm disease is global in distribution and is likely the most common vector-borne disease in the United States, with prevalence in domestic dogs between 1% and 12.5% on average nationwide [17], but as high as 48.8% in certain highly endemic regions following natural disasters, like the Gulf Coast post-Hurricane Katrina [18]. Dog heartworm is considered endemic in the contiguous United States, with highest prevalence in the southeastern US [19]. Importantly, the parasite is naturally vectored by at least 25 mosquito species in the US [20] and even greater numbers of species worldwide with varying vectorial capacities. Its multi- vectored status suggests dog heartworm disease transmission is sensitive to changes in mosquito assemblage. As dog heartworm disease is vectored by an assemblage of mosquito species, changes in mosquito diversity will likely affect disease prevalence. Urbanization is one context in which vector assemblage and diversity vary, potentially with implications for heartworm disease transmission. While previous studies have investigated dog heartworm prevalence in urban to rural gradients [21], the effects of urbanization-driven mosquito diversity changes on the pathosystem have not been previously considered. Urban and suburban areas throughout much of the United States are dominated by the peridomestic Aedes albopictus, which is a competent vector of D. immitis [20]. It is possible that heartworm disease risk could be higher in urbanized areas [21], where the majority of mosquito bites are likely to be from Ae. albopictus, versus in natural areas, where bites come from a more diverse mosquito assemblage in which a given species may or may not be a competent D. immitis vector. Although vector biodiversity is our primary interest, another potentially important contributor to dog heartworm disease risk is socioeconomic status. There is effective preventative medication that protects dogs from becoming infected with D. immitis, but these
38 medications are by prescription only, requiring not only the cost of the medication but also the cost of routine veterinary care, presenting a financial barrier for some pet owners. As expected, people of higher socioeconomic status reported greater use of preventative medications, resulting in lower levels of vector-borne pathogens, including D. immitis [22]. In addition to preventative medication usage, lower socioeconomic status could also increase disease risk via increased vector exposure, either through behavioral factors (e.g. dogs spending more time outdoors) or through changes in vector assemblages. Some studies have shown greater prevalence of container-breeding mosquitoes such as Ae. albopictus and Culex pipiens in lower income neighborhoods with high levels of abandonment [23],[24]. But in contrast, similar studies have found either no effect of socioeconomic status [10],[25] or even the opposite trend of greater numbers of mosquitoes in higher socioeconomic areas [26]. While the directionality of its relationship with mosquito communities remains unclear, it is evident that socioeconomic status has the ability to alter susceptibility of hosts, vector exposure, and therefore disease risk. In the current study, we sought to determine the relationship between mosquito diversity and D. immitis prevalence in domestic dogs within the suburban setting. We approached this question by sampling mosquitoes across Wake County, North Carolina, USA, analyzing the mosquitoes for the presence of D. immitis DNA, and comparing heartworm prevalence rates within the mosquito to heartworm prevalence rates within domestic dogs in Wake County. Recognizing that mosquito diversity changes as suburban neighborhoods grow older [10], we stratified our mosquito sampling efforts in suburban areas by neighborhood age, creating a chronosequence with which to test our predictions. We hypothesize that older neighborhoods with less diverse mosquito assemblages dominated by Ae. albopictus have greater dog heartworm prevalence. We also considered that socioeconomic status may affect disease prevalence, hypothesizing that higher income areas have less dog heartworm than low income areas.
METHODS Site Selection and Owner Permission As previously described in Spence Beaulieu et al. 2019 [10], we identified suburban neighborhoods of various ages throughout Wake County, NC. We separated the county into geographical quadrants and selected neighborhoods haphazardly from each quadrant to ensure
39 that mosquitoes were sampled evenly throughout Wake County. We categorized the age of the neighborhood using Google Earth historical images [27], taking the year of first evidence of construction as the year the neighborhood was built. We created categories of neighborhood ages to ensure that neighborhoods of various ages were being sampled: developed before 1993, between 1993 and 2002, between 2003 and 2007, between 2008 and 2012, and from 2013 to present. Each geographical quadrant had at least one neighborhood from each category represented in sampling. We selected 30 neighborhoods in 2015 that were constructed within the last 40 years and added an additional 6 older neighborhoods in 2016 that were developed between 50 and 102 years prior. For each neighborhood, we trapped at a single house with homeowner approval, verifying that homeowners did not intend on doing any mosquito barrier sprays or insecticidal treatments during the study period. We also sampled in three natural habitat sites as controls: Schenck Memorial Forest at North Carolina State University (NC State), NC State’s Equine Educational Unit, and NC State’s Lake Wheeler Beef Unit. Each of these sites contained some wooded habitat and some field or pasture habitat, allowing us to conveniently sample mosquitoes in these two distinct habitat types. We also sampled at 6 additional smaller parcels of land composed of undeveloped woodlots and 5 additional smaller parcels of land composed of undeveloped fields. Each of these smaller natural sites had a radius of at least 100m of undeveloped land around the trap, which is an appropriate radius given previous findings on mosquito habitat fidelity [28]. Overall, we trapped at a total of 9 wooded control sites and 8 field control sites throughout Wake County.
Trapping We sampled overnight with CDC light traps (JW Hock Co., Gainesville, FL, USA) baited with 1kg of dry ice (solid CO2). Lights were removed from the CDC light traps to reduce by- catch. In the suburban neighborhoods, trap placement within the yard was based on the homeowner’s preference. At the control wooded and field sites, traps were placed at least 100m away from any habitat edge, again consistent with previous findings of mosquito habitat fidelity [28]. We trapped at each site biweekly from June through mid-October in 2015, and June through the end of October in 2016.
40 Molecular Analysis We identified all mosquitoes to species using published dichotomous keys [29],[30]. For mosquitoes collected in 2015, we sexed the individuals and females were dissected for parity analysis via ovary tracheation [31]. Because of their inability to be infectious for D. immitis, we excluded males and nulliparous females from further analysis. We then pooled parous females by site, date collected, and species for molecular analysis, with up to 19 individual mosquitoes per pool. We did not dissect mosquitoes collected in the 2016 trapping season, but rather immediately pooled all female mosquitoes by site, date collected, and species, again with up to 19 individual mosquitoes per pool. We extracted DNA from each pool using either the DNeasy Blood and Tissue Kit (Qiagen, Venlo, Netherlands) or the ZR Genomic DNA-Tissue MiniPrep (Zymo Research, Irvine, CA, USA). We homogenized the mosquitoes using sterilized pestles in sterilized microcentrifuge tubes and followed modified versions of the standard kit protocols for DNA extraction (full extraction protocols are presented in Additional file 1). After extraction, we quantified the DNA concentration using the Qubit 2.0 Fluorometer (Invitrogen, Life Technologies Corporation, Waltham, MA, USA), accepting any sample with greater than 0.01 ng/µl of DNA for further analysis. We stored the extracted DNA at -20°C until the time of further analysis. We performed real time quantitative polymerase chain reaction (qPCR) to assess for presence or absence of D. immitis DNA within the previously extracted and quantified mosquito pools. Each qPCR reaction consisted of 0.2 µl each of an established D. immitis-specific COI forward and reverse primer [32], 5 µl of SYBR green master mix, 2 µl of template DNA, and 2.6 µl of water for a total reaction volume of 10 µl. Each 96 well plate run included one reaction with DNA extracted from a known negative lab-reared mosquito combined with a single lab- reared D. immitis L3 for the dual purpose of both a positive control and a sensitivity check to ensure D. immitis could be detected down to the presence of a single larva. Additionally, each 96 well plate included both a reaction with no template and a reaction with DNA extracted from a known negative lab-reared mosquito as negative controls. All samples and controls were run in duplicates to ensure accuracy of results. The qPCR procedure consisted of a denaturing step at 95°C for 30 seconds, followed by 40 cycles of denaturing at 95°C for 5 seconds, annealing at 60°C for 15 seconds, and extension at 72°C for 10 seconds. Because the goal was to determine
41 simple presence or absence of D. immitis within the mosquito pools, CFX Maestro software (Bio-Rad Laboratories, Hercules, CA, USA) was used to visually assess for DNA amplification and therefore presence of D. immitis DNA within the sample prior to 30 cycles.
Within-host Heartworm Data Acquisition We acquired data from the Wake County Animal Center on all dogs entering the shelter either as an owner surrender or as a captured stray between January 2010 and October 2015. These data contained the results of the heartworm test performed at the time of intake as well as the zip code of the prior owner or location of capture. Any dogs without heartworm test results or a designated zip code were removed from analysis.
Statistical Analyses For each mosquito species, we calculated total number of individuals and total number of pools analyzed, the number of pools positive for D. immitis, and the percent D. immitis positive pools. Due to low frequency of positive pools, we did not calculate the minimum infection rate (MIR) of each species. We investigated seasonality of D. immitis transmission within Wake County by plotting the overall percent positive pools across the sampling season. Since parity data for mosquitoes were collected only in 2015, we assessed for correlation of parous mosquitoes with heartworm positive pools in 2015. We performed a Kruskal-Wallis test comparing heartworm prevalence across land-use types (suburban, natural woodlot, natural field) both with and without neighborhood age categories included as different levels within the broader suburban category in R 3.5.0 statistical software [33]. Kruskal-Wallis test was chosen due to the non-normal nature of the data, and Dunn test for multiple comparisons with Bonferroni correction was utilized to identify which land-uses differed in their heartworm prevalence. To further investigate differences between land-use types while accounting for varying mosquito pool sizes during our molecular analysis, we used a likelihood ratio test to compare probability of within-mosquito D. immitis infection across land-use types. We maximized the likelihood function:
*+, (-./+,) *+, /+, "($%) = ((1 − $%) ) ∗ (1 − (1 − $%) ) where 1%2 is the binary response of whether pool 3 in habitat 4 was positive for D. immitis, 5%2 is the number of mosquitoes in pool 3 from habitat 4, and $% is the probability that an individual
42 mosquito in habitat 4 is positive for D. immitis. We then tested the null hypothesis that ∑ $% = $ using a likelihood ratio test, utilizing the Holm method to correct for multiple comparisons when identifying which land-uses differed in their probabilities. We again performed this test for land- use type both with and without neighborhood age categories as levels within the broader suburban category. As the acquired heartworm data within dogs was at the zip code level, we calculated heartworm prevalence within dogs by zip code in Wake County. We then calculated proportion heartworm positive mosquito pools by zip code so that the two datasets were at comparable scales. We created choropleths to visualize risk maps for D. immitis infection within dogs and within mosquitoes in the choroplethrZip package in R [34]. We performed a linear regression of within-mosquito heartworm prevalence versus within-host heartworm prevalence, as well as a student’s t-test comparing the within-host heartworm prevalence by the presence or absence of heartworm positive mosquito pools within a given zip code. Both of these approaches sought to assess whether the heartworm status of mosquitoes was a reliable predictor of infection status within the host. We calculated mosquito rarefied richness, Pielou’s evenness, and Shannon-Wiener diversity as previously described [10] and found the average for a given zip code, then performed linear regressions comparing these diversity measures to within-host heartworm prevalence, testing the hypothesis that mosquito diversity impacts heartworm disease transmission. To disentangle the effects of mosquito diversity and mosquito abundance, we also calculated the average abundance within a zip code per site per trap night and performed a linear regression comparing log average mosquito abundance to within-host heartworm prevalence. We acquired median household income by zip code from the 2013-2017 American Community Survey using U.S. Census Bureau’s American FactFinder tool [35] to investigate whether D. immitis prevalence within dogs and socioeconomic status is correlated. Finally, we performed generalized linear model selection to find the model that best explained within-host heartworm prevalence using all combinations of the following independent variables at the zip code level: presence or absence of heartworm positive mosquito pools, proportion heartworm positive mosquito pools, rarefied richness, evenness, Shannon-Wiener diversity, mosquito abundance, and median household income. We performed a logit transformation of our within-host
43 heartworm prevalence data to account for its proportional nature [36], and used Akaike information criterion (AIC) as the estimator of model quality in our model selection.
RESULTS We collected a total of 10,244 mosquitoes over the two years of sampling. The most prevalent species was Aedes albopictus, the Asian tiger mosquito, which constituted 41.9% of the overall abundance, followed by Culex salinarius and Aedes vexans with 13.9% and 8.3%, respectively. After excluding males and nulliparous females as previously described, 8483 individuals in 2488 pools were tested for the presence of D. immitis DNA.
Table 1: Heartworm prevalence within mosquito species. For each mosquito species, total number of individuals captured throughout the study period, number of pools tested for the presence of Dirofilaria immitis DNA, number of positive pools, and percent positive pools are reported. Culex sp.* were individuals that were too damaged to identify beyond genus. Across all species, we found an overall within-mosquito heartworm prevalence of 0.6%. Species Number of Number of Positive Percent Individuals Pools Pools Positive Aedes albopictus 3429 585 2 0.342 Aedes atlanticus 437 63 0 0 Aedes canadensis 33 13 1 7.692 Aedes cinereus 3 3 0 0 Aedes fulvus pallens 1 1 0 0 Aedes infirmatus 13 7 0 0 Aedes triseriatus 182 95 1 1.053 Aedes vexans 732 296 0 0 Anopheles crucians 38 34 1 2.941 Anopheles punctipennis 377 198 0 0 Anopheles quadrimaculatus 283 176 2 1.136 Coquillettidia perturbans 63 41 0 0 Culex sp.* 6 6 0 0 Culex erraticus 578 232 2 0.862 Culex peccator 1 1 0 0 Culex pipiens 340 148 2 1.351 Culex restuans 1 1 0 0 Culex salinarius 1265 349 0 0 Orthopodomyia signifera 1 1 0 0 Psorophora ciliata 4 3 0 0 Psorophora columbiae 339 137 4 2.920 Psorophora ferox 357 98 0 0 Overall: 8483 2488 15 0.603
44
Figure 1: Species-level heartworm prevalence by land-use type. Eight mosquito species had pools that tested positive for Dirofilaria immitis. Percent positive pools by species is presented, with bars color coded to denote the land-use type where the positive pool originated. Since sample sizes varied among species, number of positive pools and total number of pools tested per species are provided above each bar.
Of the total 2488 pools tested, 15 were positive for the presence of D. immitis DNA (Table 1). Eight mosquito species showed evidence of D. immitis infection, with Aedes canadensis having the highest percentage positive pools at 7.7%, followed by Anopheles crucians and Psorophora columbiae both with approximately 2.9% positive pools (Fig. 1). All other mosquito species had less than 1.4% D. immitis positive pools. The earliest D. immitis positive pool of mosquitoes was collected during the first week of June, which coincided with the beginning of our trapping season, and the latest D. immitis positive pool of mosquitoes was collected during the third week of October. No apparent seasonal trends in D. immitis positive status were noted within the mosquito trapping season (see Additional file 2). However, when analyzing the 2015 data, percent heartworm positive mosquito pools was positively correlated with percent parous mosquitoes across the trapping season (Fig. 2, p = 0.027, Spearman’s r =
45
Figure 2: Relationship between mosquito parity and within-mosquito heartworm prevalence. Percent parous mosquitoes and percent Dirofilaria immitis positive mosquito pools were compared for data from the 2015 trapping season. The two variables were positively correlated (p = 0.027, Spearman’s r = 0.494).
0.494). As detailed in the methods section, parity data were not collected in 2016 and correlation with heartworm positive pools was therefore unable to be assessed for the second trapping year. We found significant differences in within-mosquito D. immitis prevalence between land- use types (Kruskal-Wallis c2 = 8.555, df = 2, p = 0.014). Further investigation with Dunn test revealed that field sites had greater D. immitis prevalence than did suburban sites (Z = 2.925, p = 0.010), but prevalence at wooded sites did not differ from that at suburban sites (Z = -0.630, p = 1.0) (Fig. 3). Neighborhood age was not a significant factor, as Kruskal-Wallis test was not significant when incorporating neighborhood age categories. We also found that the probability of an individual mosquito being positive for D. immitis differed between land-uses (LRT statistic = 6.40, p = 0.041). Field sites had greater likelihood of a mosquito being positive for D. immitis than did suburban sites (LRT statistic = 5.81, Holm adjusted p = 0.048), but the probability at wooded sites did not differ from that at suburban sites (LRT statistic = 0.013, Holm adjusted p = 0.911) or field sites (LRT statistic = 4.99, Holm adjusted p = 0.051). Again, neighborhood age was not a significant factor, as the likelihood ratio test was not significant when incorporating neighborhood age categories into the analysis.
46
Figure 3: Comparison of within-mosquito heartworm prevalence by land-use type. Within-mosquito Dirofilaria immitis prevalence varied by land-use type (Kruskal-Wallis c2 = 8.555, df = 2, p = 0.014). Suburban sites had significantly lower D. immitis prevalence than did undeveloped field sites (Z = 2.925, p = 0.010). The prevalence in undeveloped wooded sites did not significantly differ from that in either suburban (Z = -0.630, p = 1.0) or field sites (Z = 1.884, p = 0.179). Mean prevalence and standard error of the mean are presented for each land-use type.
Figure 4: Visualization of heartworm prevalence by zip code in Wake County, North Carolina. a) Heartworm prevalence within dogs ranged from 3.77% to 15.64% within zip codes where mosquitoes were sampled. All zip codes in Wake County had some level of heartworm infection, but zip codes where mosquitoes were not sampled were omitted from visualization (denoted NA) for clarity. b) Heartworm prevalence within mosquitoes ranged from 0% to 3.09% by zip code. Due to low overall numbers of Dirofilaria immitis positive mosquito pools across all trapping sites, many zip codes had no detected within-mosquito heartworm infection. Zip codes where mosquitoes were not sampled are denoted NA.
47 A risk map for D. immitis infection within the canine host shows a trend of higher prevalence in eastern and southern Wake County zip codes (Fig. 4a). We cannot assess spatial trends for D. immitis within mosquitoes due to low overall prevalence (Fig. 4b). When comparing D. immitis prevalence within dogs to its prevalence within mosquitoes by zip code, no significant relationship was detected (F = 0.511, df = 1 and 16, p = 0.485). Similarly, using presence or absence of D. immitis within mosquitoes, we did not see a significant relationship with the parasite’s within-host prevalence by zip code (t = -0.941, p = 0.374).
Figure 5: Within-host heartworm prevalence increases with mosquito diversity measures. A significant positive correlation was noted between within-host heartworm prevalence and a) mosquito evenness (F = 4.881, df = 1 and 16, p = 0.042, R2 = 0.234) as well as b) mosquito Shannon-Wiener diversity (F = 5.464, df = 1 and 16, p = 0.033, R2 = 0.255). c) While the relationship between within-host heartworm prevalence and mosquito rarefied richness was not significant (F = 4.342, df = 1 and 16, p = 0.054, R2 = 0.213), a similar positive trend was found.
48 When investigating correlations of mosquito diversity metrics at the zip code level to within-host D. immitis prevalence, we detected significant positive relationships for evenness (F = 4.881, df = 1 and 16, p = 0.042, R2 = 0.234) and Shannon-Wiener diversity (F = 5.464, df = 1 and 16, p = 0.033, R2 = 0.255) (Fig. 5a-b). While the relationship with rarefied richness was not significant (F = 4.342, df = 1 and 16, p = 0.054, R2 = 0.213), there was a similar positive trend (Fig. 5c). We did not find a relationship between log mosquito abundance and within-host heartworm prevalence (F = 0.396, df = 1 and 16, p = 0.538). Among models including all combinations of our tested variables (presence or absence of heartworm positive mosquito pools, proportion heartworm positive mosquito pools, rarefied richness, evenness, Shannon-Wiener diversity, mosquito abundance, and median household income), selection revealed that the top model set (all models with DAIC < 2) included two models: 1) mosquito Shannon-Wiener diversity and median household income (AIC = 23.95, residual df = 15, DAIC = 0) and 2) mosquito Shannon-Wiener diversity, mosquito rarefied richness, and median household income (AIC = 25.45, residual df = 14, DAIC = 1.5). Since the latter model is a more complex version of the prior nested model that has greater AIC support [36], the model including only mosquito Shannon-Wiener diversity and median household income is the best predictor of D. immitis prevalence within the canine host. This model was significant, with prevalence positively correlated with diversity and negatively correlated with median household income (DogHWPrev = 0.6454 ShanDiv – 1.035*10-5 Income – 2.524; F = 6.725, df = 2 and 15, p = 0.008, R2 = 0.473).
DISCUSSION We predicted that older neighborhoods with less diverse mosquito assemblages would have greater dog heartworm prevalence based upon the increased abundance of suspected vectors, but instead found that suburban areas generally had the lowest within-mosquito heartworm prevalence. Additionally, we found that mosquito diversity was positively correlated with heartworm prevalence within the canine host. As mosquito diversity decreases with neighborhood age, this again is in contrast to our initial prediction of older neighborhoods exhibiting greater dog heartworm prevalence. With regard to socioeconomic status, we found support for our prediction of a negative relationship with heartworm prevalence.
49 We analyzed entire mosquito bodies for the presence of D. immitis DNA; because of this, we were unable to distinguish between infected and infectious mosquitoes. All pools that were positive for D. immitis DNA were those of mosquito species known to be competent heartworm vectors, so we assume that any positive mosquito pool represents potential transmission. In terms of percent D. immitis positive mosquito pools, Ae. canadensis and An. crucians were implicated as two important local vectors. However, these species only had positive pools detected in undeveloped natural areas. The mosquitoes implicated as heartworm vectors within suburban areas in this study were Ps. columbiae, Cx. pipiens, Anopheles quadrimaculatus, Culex erraticus, and Ae. albopictus. Although uneven sampling size between species makes it difficult to assess which species are important local vectors, Ps. columbiae appears to be a significant contributor to heartworm transmission in suburban Wake County, as it was the only species in the current study to have >1% D. immitis positive pools in suburban neighborhoods. There did not appear to be any seasonal trends in D. immitis infection within mosquitoes, although we could have missed important dynamics in the spring due to our trapping season beginning in June. Throughout the trapping season, percent D. immitis positive pools was positively correlated with mosquito parity data. As the percentage of mosquitoes in the area that have previously laid eggs (and therefore previously taken a bloodmeal) increases, D. immitis presence in mosquitoes also increases, reaffirming that older mosquitoes are the most dangerous from a disease transmission standpoint due to their greater probability of prior pathogen exposure. With so many known heartworm vectors spanning across multiple genera, the apparent lack of seasonality coupled with a positive correlation with parity could be due to vector mosquitoes having different phenologies, making heartworm transmission potential a nearly constant risk throughout the warmer months in North Carolina. Contrary to our initial hypothesis, when comparing the land-use types of suburban neighborhoods, undeveloped woodlots, and undeveloped fields, we found that field areas had significantly higher prevalence and likelihood of heartworm positive mosquitoes than did suburban areas. Given the lower probability of D. immitis positive mosquitoes noted in suburban areas, our focus on sampling mosquitoes predominantly in suburbia could have resulted in lower overall within-mosquito prevalence rates than what has been reported in other studies sampling in more rural landscapes (e.g. [37]). We found that two mosquito diversity metrics were positively correlated with heartworm prevalence within dogs at the zip code level. Our previous
50 work has shown that mosquito rarefied richness, evenness, and Shannon-Wiener diversity is decreased in established suburban neighborhoods when compared with undeveloped natural areas [10]. Taken together with the current findings, this suggests that suburban development is decreasing mosquito diversity, and that the resultant decreased mosquito diversity is linked with lower heartworm disease prevalence. This agrees with findings from a recent study that demonstrated a negative correlation between human population size and within-host heartworm prevalence [38]. While we did not detect any differences in heartworm prevalence within mosquitoes based on neighborhood age, it could still be affecting prevalence within the host indirectly by decreasing mosquito diversity, as mosquito diversity metrics decrease as suburban neighborhoods age [10]. The association between decreased mosquito diversity and decreased heartworm prevalence exists despite the fact that the dominant mosquito species in the sampled suburban areas (e.g. Ae. albopictus, Cx. salinarius, Ae. vexans, Cx. pipiens, and An. quadrimaculatus) are known to be competent heartworm vectors [20]. A similar study of D. immitis prevalence within mosquitoes in an urbanized area of Oklahoma, USA found that urban sites had significantly higher heartworm infection rates than rural sites and implicated Ae. albopictus as the area’s primary vector [21]. These results do not agree with the current study’s findings of suburban areas having lower heartworm infection rates than at least the undeveloped field sites, and of Ae. albopictus not being a primary heartworm vector. One possible explanation for this discrepancy is that vector competence within a single mosquito species is susceptible to selection and can vary among geographically distinct populations [20],[39]. Studies on the vector competence of Ae. albopictus populations in North Carolina are rare, but have suggested that it is likely not a suitable vector for D. immitis in North Carolina [40]. If the local population is indeed refractory to D. immitis infection, that could drive the observed decreased heartworm prevalence in suburban areas, as greater than 40% of our trapped mosquitoes were Ae. albopictus. We found that the best model to predict heartworm prevalence within dogs at the zip code level is one that includes both mosquito Shannon-Wiener diversity and household income. While mosquito diversity had a positive relationship with host heartworm prevalence in the model, household income had a negative relationship with host heartworm prevalence, supporting our hypothesis that higher income areas would have less dog heartworm disease than lower income areas. Interestingly, our previous work in Wake County did not find any effect of
51 socioeconomic status on mosquito diversity measures [10], so the effect of socioeconomic status detected in the current study is likely due to its impact on host-level factors. This could be due to increased preventative medication use in higher income areas [22], or to variation in other factors such as the amount of time a dog spends outside and therefore amount of potential mosquito exposure time. While no host-level factors were explicitly investigated as drivers of dog heartworm prevalence in the current study, these factors are potentially important to dog heartworm disease dynamics and should be addressed in future studies. Another gap in host data is accurate information on wild canid populations that could be serving as reservoirs of dog heartworm. It has been suggested that coyotes are the most significant heartworm reservoirs in North American, with prevalence between 6.5 and 71% nationwide [41] and approximately 47% in North Carolina [42]. Wild host densities are not assessed in this study, but could play an important role in the heartworm transmission dynamics for domestic dogs, particularly if wild hosts that typically serve as primary D. immitis reservoirs are excluded from highly urbanized areas. In addition to unmeasured host factors, this study is limited by the spatial scale of the within-dog heartworm prevalence data that we acquired. The zip code of the surrendered or stray dog was noted at shelter intake, allowing analysis of trends at the scale of zip code level or larger. Mosquitoes show habitat fidelity at a much finer scale of around 100m [28], leading to a separation of geographic scale between mosquito-level factors and host-level factors that could be obscuring some trends. Additionally, the history of the surrendered dogs is largely unknown, including for relevant factors such as travel, prior preventative medication usage, or the surrendering owner’s socioeconomic status. It is not possible to definitively tell whether the dog acquired the heartworm infection within the zip code of its primary residence, nor is it possible to tell whether the positive heartworm test represents a new or chronic infection due to the long lifespan of the parasite within the host. Despite these limitations, shelter data presents a large, readily available dataset that generally cuts across various human demographics, including income and education levels [43], and is the best data currently available with which to test our predictions. Future studies could partner with local veterinarians to get finer scale host data, as collection of detailed travel history for newly heartworm positive dogs would allow for more definitive mapping of spatial and temporal host-level incidence trends.
52 Our results demonstrate that anthropogenic land-use change alters vector-borne disease risk. In the context of dog heartworm disease, the losses in mosquito diversity seen with suburban development are associated with decreased D. immitis prevalence in both the vectors and the host. Previous work with malaria has demonstrated a similar positive relationship between mosquito species richness and disease prevalence [5], suggesting that mosquito diversity losses being linked with decreased disease transmission could be applicable to a variety of multi-vectored diseases. However, decreases in mosquito diversity after human-driven land- use change could be detrimental when considering other disease systems of concern that are vectored by few mosquito species, such as dengue, chikungunya, and Zika with Ae. albopictus [44],[45]. Urban and suburban development is predicted to increase by greater than 100% over the next 50 years in the southeastern United States [46], in line with global trends of increasing urbanization. With suburban and urban development rapidly changing global landscapes and the variable nature of the response of vector-borne diseases to these land-use changes, our understanding of the connection between vector diversity and disease transmission will become increasingly pressing and warrants further investigation.
CONCLUSIONS We identified important vectors of dog heartworm in suburban areas of central North Carolina. We demonstrated that mosquito positive rates did not appear to vary throughout the trapping season, affirming that heartworm transmission risk is consistently present throughout the warmer months, possibly due to complementary phenologies of the many competent mosquito vectors in this system. These findings contribute to an understanding of local disease transmission for this prevalent and devastating disease of domestic and wild canids. We found an overall decrease in heartworm disease within the vector in suburban areas when compared with undeveloped field areas. Similarly, mosquito diversity measures, which are lower in suburban areas than in undeveloped areas, were positively correlated with within-host heartworm prevalence, suggesting that heartworm rates are lower in dogs from suburbanized areas than in dogs from less developed rural areas. Within-host heartworm prevalence was well modeled by mosquito diversity and household income, further underscoring the effect of mosquito diversity on the dog heartworm disease system. This also illustrates the importance of a
53 dog owner’s socioeconomic status, possibly due to differences in administration of preventative medications. To our knowledge, this study represents the first explicit investigation of the effects of urbanization-driven mosquito diversity changes on dog heartworm transmission within both the vector and the host. Our results suggest that decreases in mosquito diversity due to urbanization lead to decreases in dog heartworm prevalence. This information can be utilized to identify areas of high mosquito diversity that may be foci for heartworm transmission. More broadly, our findings can be generalized to other pathogens with multiple vectors, contributing to an understanding of the role of arthropod diversity in multi-vectored disease systems.
ACKNOWLEDGEMENTS This chapter was prepared as a manuscript for publication with Jennifer L. Federico (Wake County Animal Center) and Michael H. Reiskind (NC State Department of Entomology and Plant Pathology) as coauthors. We thank Paul Labadie, Tommy Pleasant, Hannah Jenkins, Cole Keenan, and Dakota Palacio for their work on this project.
ADDITIONAL FILES
Additional file 1: Modified DNA extraction protocols. File contains the full modified DNA extraction protocols used with both the DNeasy Blood and Tissue Kit (Qiagen, Venlo, Netherlands) and the ZR Genomic DNA-Tissue MiniPrep (Zymo Research, Irvine, CA, USA).
54
3
2.5
2
1.5
1 Percent Positive Pools 0.5
0
Jul.1 Jul.2 Jul.3 Jul.4 Jun.1 Jun.2 Jun.3 Jun.4 Jun.5 Sep.1 Sep.2 Sep.3 Sep.4 Oct.1 Oct.2 Oct.3 Oct.4 May.5 Aug.1 Aug.2 Aug.3 Aug.4 Aug.5 Trapping Week
Additional file 2: Within-mosquito heartworm prevalence throughout the trapping season. File contains a figure showing percent of mosquito pools positive for Dirofilaria immitis DNA for each week in the study’s trapping season. Trapping occurred over two years, but both years were analyzed together to get a single average point estimate for each calendar week.
55
REFERENCES
1. Sinka ME, Bangs MJ, Manguin S, Coetzee M, Mbogo CM, Hemingway J, et al. The dominant Anopheles vectors of human malaria in Africa, Europe and the Middle East: occurrence data, distribution maps and bionomic précis. Parasites & vectors. 2010;3 1:117.
2. Linthicum KJ, Britch SC, Anyamba A. Rift Valley fever: an emerging mosquito-borne disease. Annu Rev Entomol. 2016;61:395-415.
3. Evans MV, Dallas TA, Han BA, Murdock CC, Drake JM. Data-driven identification of potential Zika virus vectors. Elife. 2017;6:e22053.
4. Bockarie MJ, Pedersen EM, White GB, Michael E. Role of vector control in the global program to eliminate lymphatic filariasis. Annu Rev Entomol. 2009;54:469-487.
5. Fuller DO, Alimi T, Herrera S, Beier JC, Quiñones ML. Spatial association between malaria vector species richness and malaria in Colombia. Acta Trop. 2016;158:197-200.
6. Roche B, Rohani P, Dobson AP, Guégan J. The impact of community organization on vector-borne pathogens. Am Nat. 2012;181 1:1-11.
7. Reiskind MH, Lounibos LP. Spatial and temporal patterns of abundance of Aedes aegypti L.(Stegomyia aegypti) and Aedes albopictus (Skuse)[Stegomyia albopictus (Skuse)] in southern Florida. Med Vet Entomol. 2013;27 4:421-429.
8. Braks MA, Honório NA, Lourenço-De-Oliveira R, Juliano SA, Lounibos LP. Convergent habitat segregation of Aedes aegypti and Aedes albopictus (Diptera: Culicidae) in southeastern Brazil and Florida. J Med Entomol. 2003;40 6:785-794.
9. Steiger DBM, Ritchie SA, Laurance SG. Land use influences mosquito communities and disease risk on remote tropical islands: a case study using a novel sampling technique. Am J Trop Med Hyg. 2016;94 2:314-321.
10. Spence Beaulieu MR, Hopperstad K, Dunn RR, Reiskind MH. Simplification of vector communities during suburban succession. PloS one. 2019;14 5:e0215485.
11. Thongsripong P, Green A, Kittayapong P, Kapan D, Wilcox B, Bennett S. Mosquito vector diversity across habitats in central Thailand endemic for dengue and other arthropod-borne diseases. PLoS neglected tropical diseases. 2013;7 10:e2507.
12. Ferraguti M, Martínez-de La Puente J, Roiz D, Ruiz S, Soriguer R, Figuerola J. Effects of landscape anthropization on mosquito community composition and abundance. Scientific reports. 2016;6:29002.
56 13. Johnson MF, Gómez A, Pinedo-Vasquez M. Land use and mosquito diversity in the Peruvian Amazon. J Med Entomol. 2008;45 6:1023-1030.
14. de Valdez, Megan R Wise. Mosquito species distribution across urban, suburban, and semi-rural residences in San Antonio, Texas. Journal of Vector Ecology. 2017;42 1:184- 189.
15. Junglen S, Kurth A, Kuehl H, Quan P, Ellerbrok H, Pauli G, et al. Examining landscape factors influencing relative distribution of mosquito genera and frequency of virus infection. Ecohealth. 2009;6 2:239-249.
16. Park AW, Cleveland CA, Dallas TA, Corn JL. Vector species richness increases haemorrhagic disease prevalence through functional diversity modulating the duration of seasonal transmission. Parasitology. 2016;143 7:874-879.
17. Lee AC, Montgomery SP, Theis JH, Blagburn BL, Eberhard ML. Public health issues concerning the widespread distribution of canine heartworm disease. Trends Parasitol. 2010;26 4:168-173.
18. Levy JK, Lappin MR, Glaser AL, Birkenheuer AJ, Anderson TC, Edinboro CH. Prevalence of infectious diseases in cats and dogs rescued following Hurricane Katrina. J Am Vet Med Assoc. 2011;238 3:311-317.
19. American Heartworm Society. Incidence Maps. https://www.heartwormsociety.org/veterinary-resources/incidence-maps. Accessed June 12, 2019.
20. Ledesma N, Harrington L. Mosquito vectors of dog heartworm in the United States: vector status and factors influencing transmission efficiency. Topics in companion animal medicine. 2011;26 4:178-185.
21. Paras KL, O'brien VA, Reiskind MH. Comparison of the vector potential of different mosquito species for the transmission of heartworm, Dirofilaria immitis, in rural and urban areas in and surrounding Stillwater, Oklahoma, USA. Med Vet Entomol. 2014;28 S1:60-67.
22. Figueredo LA, da Silva Sales, Kamila Gaudêncio, Deuster K, Pollmeier M, Otranto D, Dantas-Torres F. Exposure to vector-borne pathogens in privately owned dogs living in different socioeconomic settings in Brazil. Vet Parasitol. 2017;243:18-23.
23. Little E, Biehler D, Leisnham PT, Jordan R, Wilson S, LaDeau SL. Socio-ecological mechanisms supporting high densities of Aedes albopictus (Diptera: Culicidae) in Baltimore, MD. J Med Entomol. 2017;54 5:1183-1192.
24. LaDeau S, Leisnham P, Biehler D, Bodner D. Higher mosquito production in low-income neighborhoods of Baltimore and Washington, DC: understanding ecological drivers and
57 mosquito-borne disease risk in temperate cities. International journal of environmental research and public health. 2013;10 4:1505-1526.
25. Dowling Z, Ladeau SL, Armbruster P, Biehler D, Leisnham PT. Socioeconomic status affects mosquito (Diptera: Culicidae) larval habitat type availability and infestation level. J Med Entomol. 2013;50 4:764-772.
26. Becker B, Leisnham P, LaDeau S. A tale of two city blocks: Differences in immature and adult mosquito abundances between socioeconomically different urban blocks in Baltimore (Maryland, USA). International journal of environmental research and public health. 2014;11 3:3256-3270.
27. Google Inc. Google Earth;7.1.5.1557.
28. Reiskind MH, Griffin RH, Janairo MS, Hopperstad KA. Mosquitoes of field and forest: the scale of habitat segregation in a diverse mosquito assemblage. Med Vet Entomol. 2017;31 1:44-54.
29. Darsie RF. Identification and geographical distribution of the mosquitos of North America, north of Mexico. Gainsville: University Press of Florida; 2005.
30. Burkett-Cadena ND. Mosquitoes of the southeastern United States: University of Alabama Press; 2013.
31. Kardos EH, Bellamy RE. Distinguishing nulliparous from parous female Culex tarsalis by examination of the ovarian tracheation. Ann Entomol Soc Am. 1961;54 3:448-451.
32. Rishniw M, Barr SC, Simpson KW, Frongillo MF, Franz M, Alpizar JLD. Discrimination between six species of canine microfilariae by a single polymerase chain reaction. Vet Parasitol. 2006;135 3-4:303-314.
33. R Core Team. R: A language and environment for statistical computing. 2018.
34. Lamstein A. choroplethrZip: Shapefile, Metadata and Visualization Functions for US Zip Code Tabulated Areas (ZCTAs). 2018;1.5.0.
35. United States Census Bureau. American FactFinder. https://factfinder.census.gov/. Accessed March 1, 2019.
36. Harrison XA, Donaldson L, Correa-Cano ME, Evans J, Fisher DN, Goodwin CE, et al. A brief introduction to mixed effects modelling and multi-model inference in ecology. PeerJ. 2018;6:e4794.
37. Fryxell RTT, Lewis TT, Peace H, Hendricks BB, Paulsen D. Identification of avian malaria (Plasmodium sp.) and canine heartworm (Dirofilaria immitis) in the mosquitoes of Tennessee. J Parasitol. 2014;100 4:455-463.
58 38. McGill E, Berke O, Peregrine AS, Weese JS. Epidemiology of canine heartworm (Dirofilaria immitis) infection in domestic dogs in Ontario, Canada: Geographic distribution, risk factors and effects of climate. Geospatial Health. 2019;14 1.
39. Nayar JK, Knight JW. Aedes albopictus (Diptera: Culicidae): an experimental and natural host of Dirofilaria immitis (Filarioidea: Onchocercidae) in Florida, USA. J Med Entomol. 1999;36 4:441-448.
40. Apperson CS, Engber B, Levine JF. Relative suitability of Aedes albopictus and Aedes aegypti in North Carolina to support development of Dirofilaria immitis. J Am Mosq Control Assoc. 1989;5 3:377-382.
41. Weiss E, Slater M, Garrison L, Drain N, Dolan E, Scarlett J, et al. Large dog relinquishment to two municipal facilities in New York City and Washington, DC: Identifying targets for intervention. Animals. 2014;4 3:409-433.
42. Brown HE, Harrington LC, Kaufman PE, McKay T, Bowman DD, Nelson CT, et al. Key factors influencing canine heartworm, Dirofilaria immitis, in the United States. Parasites & Vectors. 2012;5 245; doi:10.1186/1756-3305-5-245.
43. Chitwood MC, Swingen MB, Lashley MA, Flowers JR, Palamar MB, Apperson CS, et al. Parasitology and serology of free-ranging coyotes (Canis latrans) in North Carolina, USA. J Wildl Dis. 2015;51 3:664-669.
44. Bonizzoni M, Gasperi G, Chen X, James AA. The invasive mosquito species Aedes albopictus: current knowledge and future perspectives. Trends Parasitol. 2013;29 9:460- 468; doi:10.1016/j.pt.2013.07.003.
45. Musso D, Gubler DJ. Zika virus. Clin Microbiol Rev. 2016;29 3:487-524.
46. Terando AJ, Costanza J, Belyea C, Dunn RR, McKerrow A, Collazo JA. The southern megalopolis: using the past to predict the future of urban sprawl in the Southeast US. PLoS One. 2014;9 7:e102261.
59 CHAPTER 4
Comparative Vector Efficiency of Two Prevalent Mosquito Species for Dog Heartworm in
North Carolina
ABSTRACT The dog heartworm, Dirofilaria immitis (Leidy) (Spirurida: Onchocercidae), is a devastating parasite of domestic and wild canines vectored by a multitude of mosquito species. Although many species are implicated as vectors, not all contribute equally to disease transmission, with demonstrated variation in vector competence between species. We investigated the vector competence of North Carolina populations of two known heartworm vectors: a native species, Aedes triseriatus (Say) (Diptera: Culicidae), and an invasive species, Aedes albopictus (Skuse). We compared the parasite developmental times within the mosquito, mosquito longevity and fecundity, and the vector efficiency index between the two species. Overall, we found that the native Ae. triseriatus was an efficient vector of D. immitis in North Carolina, while the invasive Ae. albopictus was a competent but relatively poor vector locally. These results are in contrast to prior studies of geographically distant populations of Ae. albopictus, which have implicated the species as a highly competent heartworm vector. The variation seen for different strains of the same species emphasizes the heritable nature of D. immitis vector competence and highlights the need for local infection studies for accurate transmission risk assessment in a particular locale. Given our findings with local populations of Ae. albopictus and the fact that this species dominates urbanized areas of the state, our results suggest that heartworm transmission may be greater in rural areas than in urban and suburban areas of North Carolina.
INTRODUCTION Dog heartworm disease is a devastating disease of domestic and wild canines caused by the vector-borne nematode Dirofilaria immitis (Leidy) (Spirurida: Onchocercidae). The disease is global in nature and is considered endemic throughout the contiguous United States (American
60 Heartworm Society 2018). It is likely the most common vector-borne disease in the US, with an average prevalence of 1% to 12.5% nationwide (Lee et al. 2010) and an estimated minimum of 1 million domestic dogs affected (American Heartworm Society). Within wild hosts, such as coyotes, infection rates can be even higher, with prevalence estimated between 6.5% and 71% nationwide (Brown et al. 2012). If left untreated after infection, the heartworms travel to the host’s pulmonary arteries and the heart, where they cause respiratory distress and eventually congestive heart failure. In addition to the veterinary health impacts of D. immitis, there is also a significant financial burden associated with prevention and control of the parasite. Dog owners on average spend $102 per year on heartworm prevention (American Pet Products Association 2017). With 60.2 million dog-owning households in the US (American Pet Products Association 2017), this amounts to a large economic impact of heartworm prevention nationwide. Understanding transmission dynamics of D. immitis is important for mitigating the effects of this costly and widespread disease. Of epidemiological importance, D. immitis is competently vectored by at least 25 mosquito species in the United States (Ledesma and Harrington 2011) and even greater numbers of mosquito species worldwide, which contributes to its cosmopolitan distribution. Although D. immitis is transmitted by many species across multiple genera, not all species contribute equally to pathogen transmission; tremendous variation exists in vector competence between known heartworm vector species. These differences can profoundly affect local heartworm transmission dynamics, but for many species they are not well characterized. Contributing to variation in vector competence are differences in physical defenses and immune responses to the filarial parasite during its development within the mosquito. The parasite’s life cycle begins when a mosquito bites an infectious host, ingesting heartworm microfilaria with its bloodmeal. If the microfilaria is not damaged by the mosquito’s cibarial armature during ingestion (Ledesma and Harrington 2011), the parasite then migrates into the Malpighian tubules and undergoes development through its L1 and L2 larval stages. During the microfilarial and first two larval stages, parasite development can be arrested via melanization or encapsulation (Ledesma and Harrington 2011). If it evades these immune responses, the parasite finally reaches the infective L3 larval stage, which migrates to the thorax, head, and proboscis. Both the migration of the microfilaria into the Malpighian tubules and the migration of the L3 toward the proboscis are traumatic events that can increase vector mortality (Christensen 1978). The extrinsic incubation
61 period within the mosquito is temperature dependent, but typically between 10 and 14 days (American Heartworm Society 2018). When the infectious mosquito then bites a susceptible host, L3 larvae exit the proboscis and enter the bite wound. The larvae go through an additional two immature stages before development into adult worms primarily residing in the pulmonary arteries, which sexually reproduce to generate circulating microfilaria in the bloodstream. The intrinsic incubation period within the canine host is between 6 and 9 months (American Heartworm Society 2018). Two potentially important D. immitis vectors in North Carolina are Aedes triseriatus (Say) (Diptera: Culicidae), the eastern treehole mosquito, and Aedes albopictus (Skuse), the Asian tiger mosquito. Aedes triseriatus is a native container breeding species, while Ae. albopictus is an invasive container breeder. Both species have been experimentally implicated as competent heartworm vectors (Intermill 1973, Lai et al. 2000, Tiawsirisup and Kaewthamasorn 2007). Additionally, bloodmeal analyses reveal that between 0% and 35.7% of Ae. triseriatus feed on dogs (Magnarelli 1977, Burkot and DeFollart 1982, Apperson et al. 2002, Apperson et al. 2004, Molaei et al. 2008), while 8.7% to 11.5% of Ae. albopictus feed on dogs (Savage et al. 1993, Faraji et al. 2014). Both species are primarily mammalophilic mosquitoes that will take bloodmeals from canine hosts, making them plausible natural vectors of D. immitis. Their vector status is reinforced by field studies detecting D. immitis DNA in wild-caught individuals (e.g. Licitra et al. 2010, Paras et al. 2014). Despite their similarities in opportunistic feeding behaviors and competence for D. immitis, Ae. triseriatus and Ae. albopictus differ in certain components of their life histories. Aedes triseriatus larvae are typically found in treeholes, while Ae. albopictus utilize primarily man-made containers for their larval habitat (Burkett-Cadena 2013). This leads to Ae. albopictus dominating urban, suburban, and edge habitats and Ae. triseriatus being encountered more frequently in natural wooded areas (Barker et al. 2003, Reiskind et al. 2017, Reed et al. 2018), with potential spatial implications for heartworm transmission if variation in D. immitis competence exists between the two species. One measurement of vector competence for D. immitis is the vector efficiency index (VEI), a ratio expressing the number of infectious L3 larvae produced per the number of microfilaria ingested (Kartman 1954). Laboratory estimates of VEI across multiple species have suggested a range between 0.3% (Lowrie 1991) and as high as 58.3% (Tiawsirisup and Kaewthamasorn 2007) for competent vectors. VEI has previously been estimated at 11% for Ae.
62 triseriatus (Intermill 1973) and between 4.2% and 58.3% for Ae. albopictus, depending on the microfilarial density of the infective bloodmeal (Lai et a. 2000, Tiawsirisup and Kaewthamasorn 2007). In addition to VEI, vector competence can also be affected by changes in survivorship or fecundity with D. immitis infection. Due primarily to the traumatic migration of microfilaria from the midgut to the Malpighian tubules and of L3 larvae from the Malpighian tubules to the head and thorax, infection with D. immitis can increase mosquito mortality, particularly if harboring large worm burdens (Christensen 1978, Nayar and Knight 1999). Fecundity can also decrease with infection, either because of decreased blood intake when feeding on infectious hosts or because of the parasite utilizing vector resources that would otherwise be put toward egg production (Courtney et al. 1985). Since VEI, survivorship, and fecundity all impact either competence for D. immitis or population dynamics and vary between mosquito species, investigating these factors for dominant vectors in a given area is critical for gaining insight into local transmission dynamics. In North Carolina specifically, bloodmeal analysis studies have estimated 2% to 4% blood feeding on dogs for Ae. triseriatus, and 11% for Ae. albopictus, suggesting their potential importance to canine vector-borne disease transmission in the state (Irby and Apperson 1988, Richards et al. 2006). However, we are not aware of any competence studies with local populations of Ae. triseriatus, and the single study with local populations of Ae. albopictus occurred shortly after the species’ introduction to North Carolina (Apperson et al. 1989). It is reasonable that vector competence could have changed with further establishment of Ae. albopictus in the state. In the current study, we sought to compare the dog heartworm vector competence of local strains of these two prevalent North Carolina mosquito species. We approached this by assessing the D. immitis developmental times within the mosquito, the effects of D. immitis infection on mosquito longevity and fecundity, and the VEI for both Ae. triseriatus and Ae. albopictus across three distinct experiments. Based on the prior work with Ae. albopictus in North Carolina (Apperson et al. 1989), we hypothesize that Ae. triseriatus is a better vector of D. immitis locally, characterized at minimum by a higher VEI when compared with Ae. albopictus.
63 MATERIALS AND METHODS Mosquito Rearing For both species and for all experiments, we used F2-F3 generation mosquitoes reared from eggs that were initially collected in various locales across North Carolina, then combined into a single North Carolina mix for lab-rearing. We reared all mosquitoes in ambient conditions of 29.5°C, 73% relative humidity, and a 14:10 light:dark cycle. We reared larvae in densities of 100 larvae per 1L water with 0.6g ground Wardley Pond Pellets fish food. Subsequently, we moved pupae to BugDorm-1 insect rearing cages (MegaView Science Co., Ltd., Taiwan) for eclosion and adult maintenance, offering a 10% sugar solution ad libitum. We sugar starved the mosquitoes for approximately 24 hours prior to blood feeding for each experiment. The D. immitis positive and the D. immitis negative blood used in all experiments was provided by a partner company (Cambrex Corporation, Durham, NC).
D. immitis Developmental Times We used 3 to 7 day old mosquitoes for blood feeding. We fed mosquitoes on D. immitis positive dog blood with a density of 3000 microfilaria per ml. Blood was contained in a petri dish covered in a parafilm feeding membrane with HotHands handwarmers providing heat, and was offered until at least 21 females of each species had blood fed. Blood fed females were separated and maintained in a single cage each for Ae. albopictus and Ae. triseriatus with 10% sugar solution offered ad libitum; all males and unfed females were discarded. For each species, we aimed to dissect 3 individuals at 1 hour, 4 days, 6 days, 8 days, 10 days, 12 days, and 15 days post-blood feeding. We sedated individuals by placing briefly in a freezer, then dissected the midgut, Malpighian tubules, thorax, head, and proboscis to search for the number and developmental stage of D. immitis at each location. We checked daily for mortality, and dissected any dead individuals on the day they were discovered. We calculated the Heartworm Development Units (HDUs) acquired at the time of first L3 larvae detection based on the equation:
7889:9;<=>? ABCD = E
This equation calculates the accumulated degree-days above 14°C, which is the minimum temperature required for D. immitis development (Ledesma and Harrington 2015).
64 Longevity and Fecundity We used 5 to 7 day old mosquitoes for blood feeding. Using the method described above, we fed 50 females of each species on D. immitis positive dog blood with a density of 3000 microfilaria per ml, and 48 female Ae. triseriatus and 50 female Ae. albopictus on dog blood negative for D. immitis. All males and unfed females were discarded. Blood fed females were maintained individually in 50ml conical tubes containing oviposition paper and approximately 10ml water. Tubes were sealed with mesh and a small cotton ball soaked in 10% sugar solution was placed on top as a sugar source. The sugar source was replenished daily to ensure that it remained moist and that sugar was consistently available. We monitored for mortality daily. When death was noted, we counted the number of oviposited eggs and dissected to check for any developed but internally retained eggs. Total egg counts were considered the sum of both the oviposited and retained eggs. We performed survival curves using the survival package (Therneau 2019) in R 3.5.0 statistical software (R Core Team 2018). Visualization and log-rank test comparing the survival curves for D. immitis infected and uninfected mosquitoes of each species was performed using the survminer R package (Kassambara et al. 2019). We compared egg counts between D. immitis infected and uninfected mosquitoes of each species by ANOVA in R. Based on known gonotrophic cycles and to ensure that egg maturation had occurred, we eliminated from fecundity analysis Ae. triseriatus that died prior to 5 days post-blood feeding and Ae. albopictus that died prior to 3 days post-blood feeding (Mather and DeFoliart 1983, Delatte et al. 2009). This resulted in removal of eight D. immitis positive and six D. immitis negative Ae. triseriatus, as well as one D. immitis positive Ae. albopictus from this analysis.
Vector Efficiency Index We used 5 to 7 day old mosquitoes for blood feeding, feeding with 3000 microfilaria per ml D. immitis positive dog blood as described above. Blood was offered until at least 30 females of each species had fed. We retained all blood fed females in a single cage for each species with 10% sugar solution offered ad libitum; all males and unfed females were discarded. At 1 hour post-blood feeding, we dissected 15 individuals of each species to check the midgut for the presence and number of microfilaria. At 15 days post-blood feeding, we dissected all remaining living individuals for the presence and number of L3 larvae in the midgut, Malpighian tubules,
65 abdomen hemocoel, thorax, head, and proboscis. For each species, we calculated the infection rate (IR) and VEI (Kartman 1954) as defined by the equations: 59:M>G NO :NDP94=N>D Q4=ℎ "3D <= 15 ?<1D KL = ∗ 100 59:M>G NO D9GF4F45H :NDP94=N>D <= 15 ?<1D
RESULTS Dirofilaria immitis seemed to develop faster initially within Ae. albopictus, reaching both the L1 and L2 stages one time-point sooner than within Ae. triseriatus. However, D. immitis ultimately achieved the infective L3 stage earlier within Ae. triseriatus (Table 1). L3 larvae were noted at 10 days post-blood feeding within Ae. triseriatus. This equates to 139.5 to 155 HDUs at 29.5°C, which is in line with the previously noted threshold of 130 HDUs for development of L3 D. immitis (Ledesma and Harrington 2015). Within Ae. albopictus, L3 larvae were first noted at 12 days post-blood feeding, which amounts to 170.5 to 186 HDUs acquired. Additionally, Ae. triseriatus yielded higher numbers of L3 larvae in the head and proboscis than Ae. albopictus. Survival curve and log-rank test revealed that longevity was not significantly different between D. immitis positive and D. immitis negative Ae. triseriatus (p = 0.69, Figure 1a). In contrast, D. immitis infected Ae. albopictus suffered greater mortality than uninfected individuals (p = 0.0058, Figure 1b). With regard to fecundity, Ae. triseriatus infected with D. immitis exhibited increased total egg production when compared with uninfected individuals (df = 1, F = 4.552, p = 0.0359, Figure 2a), while Ae. albopictus did not show any differences in egg production between individuals with varying D. immitis infection status (df = 1, F = 0.055, p = 0.815, Figure 2b). For both species, all individuals dissected 1 hour post-blood feeding were positive for microfilaria in the midgut, demonstrating effective inoculation with D. immitis. VEI for Ae. triseriatus was calculated to be 25.92%, with an average ingested microfilaria of 14.467 and average number of L3 larvae at 15 days post-blood feeding of 3.75. Survival rate to 15 days post- blood feeding was 33% (8 surviving of 24 initially blood fed) and IR was 100%. VEI for Ae. albopictus was 5.96%, with an average of 10.067 microfilaria ingested and an average of 0.6 L3 larvae at 15 days post-blood feeding. Survival rate to 15 days post-blood feeding for Ae.
66 albopictus was 18% (5 surviving of 28 initially blood fed) and IR was 40%.
Table 1: Dirofilaria immitis developmental times within Aedes triseriatus and Aedes albopictus. For each day post-infection (DPI) and for each mosquito species, number of individuals dissected, number of heartworms at each developmental stage, and total L3 larvae within the head or proboscis are presented. At each developmental stage, the average number of heartworms per mosquito is given followed by the number within each dissected individual in parentheses. Total L3 larvae within the head and proboscis are summed across all dissected individuals.
Aedes triseriatus Aedes albopictus
Average HW per mosquito (number per individual) Average HW per mosquito (number per individual) No. Total L3s No. Total L3s DPI dissected mf L1 L2 L3 in head dissected mf L1 L2 L3 in head
0 3 12.67 (8,14,16) - - - - 3 2 (2,3,1) - - - -
2 0 - - - - - 1 2 1 - - -
3 1 14 5 - - - 1 - 6 - - -
4 3 5 (5,8,2) 11 (17,7,9) - - - 3 - 11 (8,12,13) 0.67 (2,0,0) - -
6 3 1 (0,0,3) 1.67 (2,1,2) 9 (5,10,12) - - 3 - 0.67 (0,0,2) 9.67 (9,14,6) - -
7 0 - - - - - 1 - 6 7 - -
8 3 - - 10.33 (12,12,7) - - 2 - 1 (1,1) 2 (4,0) - -
9 0 - - - - - 1 - 6 - - -
10 3 - - 3.67 (1,6,4) 7 (13,2,6) 8 2 - 1 (1,1) 5 (7,3) - -
12 3 - 0.67 (0,2,0) - 14 (25,11,6) 37 3 0.33 (0,1,0) 1 (0,1,2) 4.33 (8,5,0) 3.33 (6,4,0) 10
13 1 1 2 1 7 5 0 - - - - -
15 3 - - - 4.33 (8,4,1) 9 3 - 1.33 (3,1,0) - 3.67 (0,0,11) 8
67
Figure 1: Comparison of longevity with and without Dirofilaria immitis infection. Survival curves for heartworm negative and heartworm positive individuals are presented. Shaded areas around the curves signify a 95% confidence interval. Number of surviving individuals are given for every 10 day interval. a) For Ae. triseriatus, survival did not differ between treatments (p = 0.69). b) For Ae. albopictus, D. immitis negative individuals had a significantly greater survival probability than did D. immitis positive individuals (p = 0.0058).
Figure 2: Effects of Dirofilaria immitis infection on fecundity. Estimates of total egg production are presented based on heartworm infection status. Lower and upper hinges denote the first and third quartile respectively. Whiskers extend to the furthest value that is within 1.5x the interquartile range and outliers are plotted individually. a) Total egg production is significantly higher for infected Ae. triseriatus (df = 1, F = 4.552, p = 0.0359). b) Total egg production did not differ between infected and uninfected Ae. albopictus (df = 1, F = 0.055, p = 0.815).
68 DISCUSSION Our results suggest that the invasive Ae. albopictus is a poor vector of D. immitis in North Carolina. Aedes albopictus is competent for D. immitis, as North Carolina strain F3 mosquitoes were able to produce L3 larvae in a laboratory setting. However, the VEI for Ae. albopictus was a meager 5.96%, notably lower than other VEI estimates of 13.5% to 58.3% for this species after comparably microfilaremic bloodmeals (Tiawsirisup and Kaewthamasorn 2007). IR was 40% for Ae. albopictus, suggesting that some individuals were refractory to the complete development of D. immitis. It is possible that the individuals without L3 larvae present at 15 days post-blood feeding simply did not ingest any microfilaria during their infective bloodmeal. The likelihood of this is low, as all individuals dissected shortly after blood feeding were positive for microfilaria. However, if this occurred, it would also indicate a refractoriness to D. immitis infection via avoidance of ingesting microfilaria during blood feeding, as Ae. triseriatus that were offered an identical infective bloodmeal had an IR of 100% at 15 days post-blood feeding and therefore had all ingested microfilaria. With regard to longevity, previous studies have shown that there is no change in survival when mosquitoes are fed a low density microfilaremic bloodmeal, like the 3000 microfilaria per ml used in the current study (Christensen 1978, Silaghi et al. 2017). Even with this low microfilaremia, we were able to detect an increase in mortality for Ae. albopictus. Since mosquitoes must outlive the extrinsic incubation period to successfully transmit a pathogen to a susceptible host, this increased mortality could jeopardize the ability of Ae. albopictus to vector D. immitis in a natural setting. In contrast to our findings with Ae. albopictus, our study suggests that the native Ae. triseriatus is an efficient vector of D. immitis in North Carolina. Local strain F3 Ae. triseriatus were not only able to produce L3 larvae in a laboratory setting, but the VEI for this species was 25.92%, which is approximately 4.3 times greater than that measured for Ae. albopictus. This VEI is also greater than a previous study’s estimate of 11% in a Mississippi population of Ae. triseriatus (Intermill 1973). IR of Ae. triseriatus at 15 days post-blood feeding was 100%, with a markedly higher survival rate of 33% when compared with the 18% noted for Ae. albopictus. Survival curves between infected and uninfected Ae. triseriatus were not significantly different, indicating no increase in mortality associated with D. immitis infection for this species in a laboratory setting. Additionally, D. immitis appeared to reach the infective L3 stage sooner
69 within Ae. triseriatus, requiring fewer HDUs for development. This equates to a shorter extrinsic incubation period and potential for faster transmission of the parasite to a susceptible host. Although no change in fecundity was noted for Ae. albopictus, Ae. triseriatus showed a slight increase in fecundity when infected with D. immitis. This is an atypical finding, as infection is usually associated with decreased fecundity both for D. immitis (Christensen 1981) and for other vector-borne pathogens (e.g. Styer et al. 2007, Vézilier et al. 2012, Breaux et al. 2014). There are, however, some unmeasured factors that could have contributed to the resulting infection-associated fecundity increase. Bloodmeal size positively affects egg production, as does mosquito size. We did not measure blood intake and it is therefore possible that Ae. triseriatus took larger bloodmeals from the D. immitis positive blood, although we would expect a smaller bloodmeal taken from the infected blood if there were differences (Courtney et al. 1985). We attempted to measure wing lengths as a proxy for female size, but unfortunately most wings became severely damaged due to housing individuals in small conical tubes, making accurate measurements of wings impossible. Nevertheless, all individuals came from the same cohort and were randomly assigned to a treatment, so we do not anticipate significant differences in mosquito size between treatments within a species. Egg production could have been affected by differences in nutrient levels in the D. immitis positive and D. immitis negative blood, as the two blood samples necessarily came from different dogs and blood from different humans has been shown to result in fecundity differences in Ae. aegypti (Helinski and Harrington 2011). If the two blood sources had different nutrient profiles, we would expect Ae. albopictus to show the same fecundity patterns as Ae. triseriatus, since both species were offered identical infectious and non-infectious blood, but this was not noted. Although we do not have an obvious explanation for the increased fecundity in infected Ae. triseriatus, our results show at minimum that fecundity is not hindered by infection with D. immitis for either species, in line with previous studies showing decreased fecundity only after a threshold of about 15 heartworms per mosquito has been reached (Christensen 1981). Overall, our results agree with prior findings from Apperson et al. (1989) that Ae. albopictus is not a strong vector of D. immitis in North Carolina. This poor vector competence has landscape level implications for disease transmission, as Ae. albopictus is a peridomestic species that is dominant in urban and suburban areas of the state (Reed et al. 2018, Spence Beaulieu et al. 2019). This could point to a greater risk of D. immitis infection for dogs that
70 reside in rural areas of North Carolina, where mosquito communities are more diverse and bites from efficient vectors such as Ae. triseriatus are more frequent, as compared to the risk for dogs residing in suburban areas where encountered mosquitoes are more likely to be the inefficient vector Ae. albopictus. Additionally, wild hosts of D. immitis, including coyotes, tend to avoid highly urban areas and instead utilize undeveloped areas and corridor habitat (Atwood et al. 2004). This leads to the potential for greater densities of D. immitis reservoirs in rural areas and contributes to a rural nidus for heartworm infection. Despite the lower vector competence of Ae. albopictus for D. immitis when compared to Ae. triseriatus, dog owners in suburban and urban areas of North Carolina should remain vigilant in administering routine heartworm preventative medication to safeguard against infection in this heartworm endemic area. It has been documented that heartworm vector efficiency is heritable, with different species strains having different susceptibility and VEI (Nayar and Knight 1999, Silaghi et al. 2017). As previously mentioned, vector competence studies with populations of Ae. albopictus geographically distant from those in North Carolina have found notably higher VEI, implicating the species as a key vector of D. immitis in certain areas. Owing to the intraspecific variation in D. immitis vector competence, generalization of results is difficult for this pathosystem. Urban areas may be of greater risk for disease transmission than rural areas in locations were Ae. albopictus is a highly efficient vector. Additionally, while the North Carolina population of Ae. albopictus is of lesser concern for heartworm transmission risk, it is worth noting that the species still presents a public health threat due to its competent vector status for emerging pathogens such as dengue virus, chikungunya virus, and Zika virus, as well as for pathogens of focal concern in North Carolina, including West Nile virus and La Cross virus (Bonizzoni et al. 2013, Westby et al. 2015, Musso and Gubler 2016, Bara et al. 2016). Vector competence for a given pathogen can vary by species and population, and competence of a given species is not likely to be equivalent for all pathogens. Local infection studies for all potential vectors of a pathogen of interest are critical to an accurate understanding of disease risk.
ACKNOWLEDGEMENTS This chapter was prepared as a manuscript for publication with Michael H. Reiskind (NC State Department of Entomology and Plant Pathology) as a coauthor. We thank Shawn Janairo and Haley Abernathy for their work on this project.
71 REFERENCES
American Heartworm Society. Heartworm Prevention for Dogs. https://www.heartwormsociety.org/pet-owner-resources/heartworm-prevention-for-dogs.
American Heartworm Society. 2018. Current Canine Guidelines for the Prevention, Diagnosis, and Management of Heartworm (Dirofilaria immitis) Infection in Dogs.
American Pet Products Association. 2017. 2017–2018 APPA national pet owners survey. Stamford, Conn: American Pet Products Association Inc.
Apperson, C. S., B. Engber, and J. F. Levine. 1989. Relative suitability of Aedes albopictus and Aedes aegypti in North Carolina to support development of Dirofilaria immitis. J. Am. Mosq. Control Assoc. 5:377-382.
Apperson, C. S., B. A. Harrison, T. R. Unnasch, H. K. Hassan, W. S. Irby, H. M. Savage, S. E. Aspen, D. Watson, L. M. Rueda, and B. R. Engber. 2002. Host-feeding habits of Culex and other mosquitoes (Diptera: Culicidae) in the Borough of Queens in New York City, with characters and techniques for identification of Culex mosquitoes. J. Med. Entomol. 39:777-785.
Apperson, C. S., H. K. Hassan, B. A. Harrison, H. M. Savage, S. E. Aspen, A. Farajollahi, W. Crans, T. J. Daniels, R. C. Falco, and M. Benedict. 2004. Host feeding patterns of established and potential mosquito vectors of West Nile virus in the eastern United States. Vector-Borne and Zoonotic Diseases 4:71-82.
Atwood, T. C., H. P. Weeks, and T. M. Gehring. 2004. Spatial ecology of coyotes along a suburban-to-rural gradient. The Journal of Wildlife Management 68:1000-1009.
Bara, J. J., A. T. Parker, and E. J. Muturi. 2016. Comparative susceptibility of Ochlerotatus japonicus, Ochlerotatus triseriatus, Aedes albopictus, and Aedes aegypti (Diptera: Culicidae) to La Crosse virus. J. Med. Entomol. 53:1415-1421.
Barker, C. M., C. C. Brewster, and S. L. Paulson. 2003. Spatiotemporal oviposition and habitat preferences of Ochlerotatus triseriatus and Aedes albopictus in an emerging focus of La Crosse virus. J. Am. Mosq. Control Assoc. 19:382-391.
Bonizzoni, M., G. Gasperi, X. Chen, and A. A. James. 2013. The invasive mosquito species Aedes albopictus: current knowledge and future perspectives. Trends Parasitol. 29:460- 468.
Breaux, J. A., M. K. Schumacher, and S. A. Juliano. 2014. What does not kill them makes them stronger: larval environment and infectious dose alter mosquito potential to transmit filarial worms. Proceedings of the Royal Society B: Biological Sciences 281:20140459.
72 Brown, H. E., L. C. Harrington, P. E. Kaufman, T. McKay, D. D. Bowman, C. T. Nelson, D. Wang, and R. Lund. 2012. Key factors influencing canine heartworm, Dirofilaria immitis, in the United States. Parasites & Vectors 5:245.
Burkett-Cadena, N. D. 2013. Mosquitoes of the southeastern United States. University of Alabama Press.
Burkot, T. R., and G. R. DeFollart. 1982. Bloodmeal sources of Aedes triseriatus and Aedes vexans in a southern Wisconsin forest endemic for La Crosse encephalitis virus. Am. J. Trop. Med. Hyg. 31:376-381.
Christensen, B. M. 1978. Dirofilaria immitis: Effect on the longevity of Aedes trivittatus. Experimental Parasitology 44:116-123.
Christensen, B. M. 1981. Effect of Dirofilaria immitis on the fecundity of Aedes trivittatus. Mosq News 41:78-81.
Courtney, C. C., B. M. Christensen, and W. G. Goodman. 1985. Effect of Dirofilaria immitis on Blood Meal Size and Fecundity in Aedes aegypti (Diptera: Culicidae). Journal of Medical Entomology 22:398-400.
Delatte, H., G. Gimonneau, A. Triboire, and D. Fontenille. 2009. Influence of temperature on immature development, survival, longevity, fecundity, and gonotrophic cycles of Aedes albopictus, vector of chikungunya and dengue in the Indian Ocean. J. Med. Entomol. 46:33-41.
Faraji, A., A. Egizi, D. M. Fonseca, I. Unlu, T. Crepeau, S. P. Healy, and R. Gaugler. 2014. Comparative host feeding patterns of the Asian tiger mosquito, Aedes albopictus, in urban and suburban Northeastern USA and implications for disease transmission. PLoS neglected tropical diseases 8:e3037.
Helinski, M. E., and L. C. Harrington. 2011. Male mating history and body size influence female fecundity and longevity of the dengue vector Aedes aegypti. J. Med. Entomol. 48:202-211.
Intermill, R. W. 1973. Development of Dirofilaria immitis in Aedes triseriatus Say. Mosquito News 33:81.
Irby, W. S., and C. S. Apperson. 1988. Hosts of mosquitoes in the coastal plain of North Carolina. J. Med. Entomol. 25:85-93.
Kartman, L. 1954. Suggestions concerning an Index of Experimental Filaria Infection in Mosquitoes. Am. J. Trop. Med. Hyg. 3:329-337.
Kassambara, A., M. Kosinski, P. Biecek, and S. Fabian. 2019. survminer: Drawing Survival Curves using 'ggplot2', vol. 0.4.4 http://www.sthda.com/english/rpkgs/survminer/.
73 Lai, C., K. Tung, H. Ooi, and J. Wang. 2000. Competence of Aedes albopictus and Culex quinquefasciatus as vector of Dirofilaria immitis after blood meal with different microfilarial density. Vet. Parasitol. 90:231-237.
Ledesma, N., and L. Harrington. 2011. Mosquito vectors of dog heartworm in the United States: vector status and factors influencing transmission efficiency. Topics in companion animal medicine 26:178-185.
Ledesma, N., and L. Harrington. 2015. Fine-scale temperature fluctuation and modulation of Dirofilaria immitis larval development in Aedes aegypti. Vet. Parasitol. 209:93-100.
Lee, A. C., S. P. Montgomery, J. H. Theis, B. L. Blagburn, and M. L. Eberhard. 2010. Public health issues concerning the widespread distribution of canine heartworm disease. Trends Parasitol. 26:168-173.
Licitra, B., E. W. Chambers, R. Kelly, and T. R. Burkot. 2010. Detection of Dirofilaria immitis (Nematoda: Filarioidea) by polymerase chain reaction in Aedes albopictus, Anopheles punctipennis, and Anopheles crucians (Diptera: Culicidae) from Georgia, USA. J. Med. Entomol. 47:634-638.
Lowrie, R. C. 1991. Poor vector efficiency of Culex quinquefasciatus following infection with Dirofilaria immitis. J. Am. Mosq. Control Assoc. 7:30-36.
Magnarelli, L. A. 1977. Host feeding patterns of Connecticut mosquitoes (Diptera: Culicidae). Am. J. Trop. Med. Hyg. 26:547-552.
Mather, T. N., and G. R. DeFoliart. 1983. Effect of host blood source on the gonotrophic cycle of Aedes triseriatus. Am. J. Trop. Med. Hyg. 32:189-193.
Molaei, G., T. G. Andreadis, P. M. Armstrong, and M. Diuk-Wasser. 2008. Host-feeding patterns of potential mosquito vectors in Connecticut, USA: molecular analysis of bloodmeals from 23 species of Aedes, Anopheles, Culex, Coquillettidia, Psorophora, and Uranotaenia. J. Med. Entomol. 45:1143-1151.
Musso, D., and D. J. Gubler. 2016. Zika virus. Clin. Microbiol. Rev. 29:487-524.
Nayar, J. K., and J. W. Knight. 1999. Aedes albopictus (Diptera: Culicidae): an experimental and natural host of Dirofilaria immitis (Filarioidea: Onchocercidae) in Florida, USA. J. Med. Entomol. 36:441-448.
Paras, K. L., V. A. O'brien, and M. H. Reiskind. 2014. Comparison of the vector potential of different mosquito species for the transmission of heartworm, Dirofilaria immitis, in rural and urban areas in and surrounding Stillwater, Oklahoma, USA. Med. Vet. Entomol. 28:60-67.
74 R Core Team. 2018. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria.
Reed, E. M., B. D. Byrd, S. L. Richards, M. Eckardt, C. Williams, and M. H. Reiskind. 2018. A statewide survey of container Aedes mosquitoes (Diptera: Culicidae) in North Carolina, 2016: a multiagency surveillance response to Zika using ovitraps. J. Med. Entomol. 56: 483-490.
Reiskind, M. H., R. H. Griffin, M. S. Janairo, and K. A. Hopperstad. 2017. Mosquitoes of field and forest: the scale of habitat segregation in a diverse mosquito assemblage. Med. Vet. Entomol. 31:44-54.
Richards, S. L., L. Ponnusamy, T. R. Unnasch, H. K. Hassan, and C. S. Apperson. 2006. Host-feeding patterns of Aedes albopictus (Diptera: Culicidae) in relation to availability of human and domestic animals in suburban landscapes of central North Carolina. J. Med. Entomol. 43:543-551.
Savage, H. M., M. L. Niebylski, G. C. Smith, C. J. Mitchell, and G. B. Craig Jr. 1993. Host- feeding patterns of Aedes albopictus (Diptera: Culicidae) at a temperate North American site. J. Med. Entomol. 30:27-34.
Silaghi, C., R. Beck, G. Capelli, F. Montarsi, and A. Mathis. 2017. Development of Dirofilaria immitis and Dirofilaria repens in Aedes japonicus and Aedes geniculatus. Parasites & vectors 10:94.
Spence Beaulieu, M. R., K. Hopperstad, R. R. Dunn, and M. H. Reiskind. 2019. Simplification of vector communities during suburban succession. PloS one 14:e0215485.
Styer, L. M., M. A. Meola, and L. D. Kramer. 2007. West Nile virus infection decreases fecundity of Culex tarsalis females. J. Med. Entomol. 44:1074-1085.
Therneau, T. 2019. survival: Survival Analysis, vol. 2.44-1.1 https://github.com/therneau/survival.
Tiawsirisup, S., and M. Kaewthamasorn. 2007. The potential for Aedes albopictus (Skuse)(Diptera: Culicidae) to be a competent vector for canine heartworm, Dirofilaria immitis (Leidy). Southeast Asian J. Trop. Med. Public Health 38:208.
Vézilier, J., A. Nicot, S. Gandon, and A. Rivero. 2012. Plasmodium infection decreases fecundity and increases survival of mosquitoes. Proceedings of the Royal Society B: Biological Sciences 279:4033-4041.
Westby, K. M., C. Fritzen, D. Paulsen, S. Poindexter, and A. C. Moncayo. 2015. La Crosse encephalitis virus infection in field-collected Aedes albopictus, Aedes japonicus, and Aedes triseriatus in Tennessee. J. Am. Mosq. Control Assoc. 31:233-241.
75 CHAPTER 5
The Effects of Vector Diversity and Community Composition on Dog Heartworm
Transmission
ABSTRACT Mathematical models have long been used to investigate the dynamics of and important contributors to vector-borne disease transmission. Despite a plethora of vector-borne diseases being transmitted by an assembly of vectors rather than a single primary vector, models have scarcely addressed the effects of vector diversity on disease transmission. I developed a single host and multi-vector compartmental model of dog heartworm disease, caused by the mosquito- borne nematode Dirofilaria immitis, to assess how changes in vector diversity impact transmission potential. I first used laboratory generated parameters to assess differences in "# between a single highly efficient vector and a single competent but poor vector. I used mosquito community composition field data to compare the "# of a diverse natural assemblage to that of a less diverse suburban assemblage. Finally, I simulated random vector parameters and assessed the average "# of systems with one, three, five, and ten vectors, while holding total abundance constant to isolate the effects of diversity changes from changes in abundance. I validated that the highly efficient vector supports disease spread with an "# > 1, while the competent but poor vector has an "# ≪ 1 and is therefore unlikely to support disease transmission in a natural setting. I demonstrated a greater "# for the natural mosquito assemblage than that of the suburban assemblage, validating previous empirical observations of lower heartworm prevalence in suburban areas. Through simulation, I found that changes in vector diversity did not greatly affect the average "#, but variability in "# was reduced and the highest density of the simulated
"# values increased toward the mean as vector diversity increased. These results suggests that vector diversity serves as a stabilizing force in the transmission of multi-vectored pathogens, while species identity and community composition determine the magnitude of transmission.
76 INTRODUCTION Vector-borne diseases pose a significant threat to global public health. Mosquitoes are the arthropod vectors of greatest concern, transmitting pathogens such as the malaria parasite, filarial worms that cause lymphatic filariasis, and dengue virus. Because of the severe morbidity and mortality associated with these mosquito-borne diseases, many mathematical models have been developed to investigate disease dynamics and test the theoretical efficacy of various control interventions. Most notably, malaria has been studied in depth via mathematical modeling, beginning with Ronald Ross’ publication of a malaria transmission model in the early 1900s, which has been iteratively improved to become what is known as the Ross-Macdonald model (Smith et al. 2012). Despite malaria being transmitted by an assembly of vectors rather than a single species, most of the malaria transmission models to date include only a primary vector in their formulation. Similarly, for other vector-borne diseases, the majority of models do not explicitly address species diversity, even if the pathogen is known to be transmitted by multiple vectors. Species diversity is likely to impact disease transmission for pathogens vectored by an assemblage of mosquitoes, with field studies suggesting that greater species richness is associated with greater disease prevalence (Fuller et al. 2016). However, with a relatively small number of modeling studies incorporating vector species diversity, the effects of diversity on vector-borne disease transmission remain largely unquantified and generally unexplored. Previous modeling studies have investigated the effect of host diversity on disease transmission for both directly transmitted and vector-borne pathogens. Using a multi-host SIR modeling approach, host species diversity was found to decrease epidemic disease outbreaks via significant reductions in "# when pathogen transmission is frequency-dependent or vector-borne (Dobson 2004). This correlates with the dilution effect seen with Lyme disease, where increasing host diversity decreases tick feeding on the primary reservoir of the pathogen, reducing tick nymphal infection rates and therefore disease transmission (LoGiudice et al. 2003, Turney et al. 2014). While these models and empirical results demonstrate the effects of host diversity on disease dynamics, they do not address questions of the effects of vector diversity. Incorporating vector diversity into a multi-host and multi-vector SIR type model, Roche and colleagues found that increases in the mosquito Shannon index, a measure of diversity, led to an increase in the maximum disease prevalence (Roche et al. 2012). However, the study had additional assumptions regarding vector susceptibility and total vector abundance that obscure
77 the assessment of the effects of vector diversity alone. The authors assumed that the most abundant species also had the highest susceptibility to the pathogen, which implicates a primary vector and may not be realistic for some multi-vectored pathogens. Most critically, the model also allowed abundance to change with species richness, leading to the highest total mosquito abundance at the highest levels of vector diversity. In many cases, it may be reasonable to assume that increases in mosquito diversity also lead to higher mosquito abundance, but coupling these two factors means that effects cannot be attributed to vector diversity per se. Indeed, the authors conclude that pathogen transmission is able to overcome any potential dilution effect due to the addition of possibly weakly competent vectors because of the increase in overall mosquito abundance (Roche et al. 2012). Other studies that have incorporated multiple vectors into disease transmission models have been investigating specific questions about a particular disease system’s dynamics rather than the effects of diversity broadly on disease transmission. For example, multi-vector models have been constructed for West Nile virus (WNV), a mosquito-borne pathogen vectored by a diversity of primarily Culex species mosquitoes, to test predictions about seasonality and to mechanistically describe observed dynamics. One such model examined the effects of mosquito activity at different times throughout the year by varying the seasonal abundance pattern of a primary and a secondary WNV vector (Lord 2010). Here, the presence of multiple vector species was found to increase the likelihood of pathogen establishment via extension of the transmission season and greater probability of the pathogen successfully overwintering. Another study modeling WNV dynamics using vectorial capacity as its key metric found that an increase in vector species did not lead to an increase in disease prevalence (McMillan et al. 2019), contradicting previous findings on the effects of vector diversity within the same disease system. The first WNV model held all mosquito parameters constant between species except for seasonal abundance (Lord 2010), while the second model included species-specific survival, biting rate, and extrinsic incubation period, but did not include information about vector competence (McMillan et al. 2019). In both cases, factors potentially important to the effect of vector diversity on disease transmission are excluded from the models, and the contradicting results leave questions of the effects of vector diversity unanswered. The dog heartworm, Dirofilaria immitis, is an ideal pathogen for further exploration into the importance of vector diversity on disease transmission. This filarial parasite is vectored by at
78 least 25 mosquito species across multiple genera in the United States (Ledesma and Harrington 2011) and even greater numbers of species globally. Briefly, its life cycle begins when the mosquito ingests microfilaria when feeding on an infectious host. The heartworm migrates from the midgut to the Malpighian tubules, where it develops through three larval stages (L1, L2, L3). At the L3 stage, the heartworm migrates through the abdomen hemocoel to the thorax, head, and eventually proboscis of the mosquito where it can infect a susceptible host at the time of the next bite. Once inside the host, the heartworm then goes through two more larval stages (L4, juvenile) before finally reaching the adult stage, which resides primarily in the pulmonary arteries and sexually reproduces to generate microfilaria that circulate in the host’s bloodstream. The parasite infects both wild and domestic canines, with average nationwide prevalence of 1 to 12.5% in dogs (Lee et al. 2010) and 6.5 to 71% in coyotes (Brown et al. 2012). Despite its high prevalence, ecological aspects of dog heartworm disease are understudied, with Chapter 3 of my dissertation constituting the first empirical study explicitly investigating the link between mosquito diversity and heartworm transmission (Spence Beaulieu 2019). Additionally, aside from a single catalytic model that ignored all vector attributes except for density (Wada et al. 1989), dog heartworm disease has not been mathematically modeled. In the current study, I sought to quantify the importance of vector diversity and community composition to the transmission of dog heartworm disease. First, I evaluated the transmission potential of an efficient vector compared with a competent but poor D. immitis vector in a single vector system using laboratory generated parameters. Utilizing a single host and multi-vector SIR-type model, I then compared the transmission potential of a realistic suburban mosquito assemblage to that of a realistic natural area mosquito assemblage based on field data from Wake County, North Carolina. Finally, I compared the average "# values of systems with one vector, three vectors, five vectors, and ten vectors via simulation of models with randomly assigned vector parameters. By holding total mosquito abundance constant between the models, the effects of diversity changes are isolated from changes in abundance. Based on prior empirical and theoretical studies linking increased vector diversity to heightened disease prevalence (Roche et al. 2012, Fuller et al. 2016), I hypothesize that mosquito assemblage diversity is positively correlated with the average "# for dog heartworm disease.
79 METHODS Model Formulation I constructed a compartmental model for the dog heartworm system with one host and multiple vectors. While dog heartworm disease in reality has multiple hosts as well as multiple vectors, the model was simplified to include only a single host due to both lack of information available for wild host parameters as well my focus on investigating the effects of mosquito diversity on the disease system. The single host being modeled in this study is owned domestic dogs.
Figure 1: Compartmental model of dog heartworm transmission with a single host and multiple vectors. Host classes are pictured in blue and vector classes are pictured in peach. Transitions between classes are denoted by solid arrows and interactions between classes are denoted by dashed arrows. All susceptible individuals of the ' vector species are infected by interaction with the infectious hosts, and all infectious individuals of the ' vector species can infect susceptible hosts through feeding interactions. This is denoted by susceptible vectors and infectious vectors being grouped together by the shaded boxes.
80
Host classes include susceptible (()), on preventative medications (*)), exposed (+)), and infectious (,)). Dogs that are on heartworm prevention are in effect immune, as preventative medications kill the L3 and L4 larval stages of the heartworm, preventing development to the adult stage and therefore infection. I assume that the proportion of dogs on preventative medication stays relatively stable over time, so a set proportion of dogs, -, enter the prevention class after birth and do not leave the class except by death. Based on what is known about the pathology of dog heartworm infection, exposed individuals do not have added mortality due to the parasite, but infectious individuals do have an additional mortality risk due to infection with the parasite, modeled as .). Infectious dogs can also be treated for heartworms at a rate /), reverting back to the susceptible class with the assumption of no acquired immunity due to prior exposure.
Vector classes include susceptible ((0), exposed (+0), and infectious (,0). Unlike with hosts, there is known to be parasite-induced mortality within the mosquito vector even at the exposed stage, so an additional mortality terms, .0, is present for both the exposed and infectious classes. I assume that there is no interaction among vector species and no competition for bloodmeals from the host. The compartmental model is presented for ' vectors in Fig. 1. The system of ordinary differential equations (ODEs) that describes the system for ' vectors is given by: 1( ) = 4 (1 − -)8 + / , − : ( − ; ( 12 ) ) ) ) ) ) ) ) 1* ) = 4 -8 − ; * 12 ) ) ) ) 1+ ) = : ( − < + − ; + 12 ) ) ) ) ) ) 1, ) = < + − / , − (. , + ; , ) 12 ) ) ) ) ) ) ) )
1(0 = = 4 8 − : ( − ; ( 12 0 0= 0= 0= 0 0=
1+0 = = : ( − < + − >. + + ; + ? 12 0= 0= 0= 0= 0= 0= 0 0=
1,0 = = < + − >. , + ; , ? 12 0= 0= 0= 0= 0 0=
81 where 8) = () + *) + +) + ,), 80= = (0= + +0= + ,0=, and @ = 1, 2, … , '. Force of infections are given by: G EF)=,0= :) = D 8) HIJ
EF0,) :0= = 8) A full list of parameters, their definitions, and their values is given in Table 1.
I used the next generation matrix method (Diekmann et al. 2009) to determine the "# of the multi-vector dog heartworm model. The next generation matrix method is a technique for derivation of "# for compartmental models with heterogeneous populations, including multiple host types (e.g. canine host and multiple mosquito vectors). Through construction of a matrix
Table 1: Parameters of the dog heartworm model. Each parameter is defined and the value used in "# calculations is given. Three of the vector parameters are allowed to vary for each mosquito species; individual values used for each model are given in subsequent sections.
Parameter Definition Value Reference
4) Birth rate 0.0003 New et al. 2004 assumed equal to ; Natural death rate 0.0003 ) births - Proportion of dogs on preventative medications 0.75 Brown et al. 2012 Mwamtobe et al. F Probability of successful transmission to vector 0.1 0 2017 < 0.0056 American Heartworm ) Intrinsic incubation period Society 2018 Host parameters /) Treatment rate 0.0000274 assumed .) Parasite-induced death rate 0.0001 assumed calculated from United States Census Bureau, American ∗ 8) Host population size at DFE (() + *)) 25,889 Veterinary Medical Association 2012, Spence Beaulieu 2019 assumed equal to 4 Birth rate 0.1429 0 deaths Mwamtobe et al. ; Natural death rate 0.1429 0 2017 E Biting rate 0.33 assumed varies for each F Probability of successful transmission to host see Tables 2-3 Vector )L of m species parameters varies for each < Extrinsic incubation period see Tables 2-3 0L of m species varies for each . Parasite-induced death rate see Tables 2-3 0L of m species Total vector abundance for all species at DFE 53,078 for 2:1 calculated from D 8∗ ratio; 265,388 Spence Beaulieu 0L (∑ ( ) 0L for 10:1 ratio 2019
82 containing new infections (N) and a matrix containing the net change by means other than disease transmission (O), the next generation matrix (P) can be calculated. The "# of the disease system is then the dominant eigenvalue of the next generation matrix. The N matrix contains the number of new infections in the @th compartment from QR infectious individuals evaluated at the disease-free equilibrium (DFE). Based on the ODEs above, the matrix is given by: 0 0 0 EF (1 − -) ⋯ 0 EF (1 − -) ⎡ )W )L ⎤ ⎢0 0 0 0 ⋯ 0 0 ⎥ 8∗ ⎢ 0W ⎥ 0 EF0 ∗ 0 0 ⋯ 0 0 ⎢ 8) ⎥ N = ⎢0 0 0 0 ⋯ 0 0 ⎥ ⎢ ⎥ ⎢⋮ ⋮ ⋮ ⋮ ⋱ ⋮ ⋮ ⎥ 8∗ ⎢ 0L ⎥ 0 EF0 ∗ 0 0 ⋯ 0 0 ⎢ 8) ⎥ ⎣0 0 0 0 ⋯ 0 0 ⎦ ∗ ∗ ∗ ∗ Here, 8) and 80= denote the population sizes at the DFE, 8) = () + *) and 80= = (0= . The O matrix contains the net change of individuals in the @th compartment by any means other than disease transmission, and for this model is given by:
<) + ;) 0 0 0 ⋯ 0 0 ⎡ −< / + . + ; 0 0 ⋯ 0 0 ⎤ ⎢ ) ) ) ) ⎥ 0 0 < + . + ; 0 ⋯ 0 0 ⎢ 0W 0W 0 ⎥ ⎢ ⎥ O = 0 0 −<0W <0W + ;0 ⋯ 0 0 ⎢ ⋮ ⋮ ⋮ ⋮ ⋱ ⋮ ⋮ ⎥ ⎢ ⎥ 0 0 0 0 … < + . + ; 0 ⎢ 0L 0L 0 ⎥
⎣ 0 0 0 0 ⋯ −<0L .0L + ;0⎦ The resultant P matrix is then calculated by P = NO^J:
E<0 (1 − -)F) E(1 − -)F) ⎡ 0 0 ⋯ L L L ⎤ ⎢ (;0 + .0L)(;0 + .0L + <0L) ;0 + .0L ⎥ ⎢ 0 0 ⋯ 0 0 ⎥ P = ⎢ ⋮ ⋮ ⋱ ⋮ ⋮ ⎥ ⎢ E< F 8∗ EF 8∗ ⎥ ⎢ ) 0 0L 0 0L ⎥ ∗ ∗ ⋯ 0 0 ⎢(;) + <))(;) + .) + /))8) (;) + .) + /))8) ⎥ ⎣ 0 0 ⋯ 0 0 ⎦
As previously noted, the dominant eigenvalue of the P matrix denotes the "# of the system. For a one vector P matrix of dog heartworm, the dominant eigenvalue is calculated as P_`_a P_a_`.
Since I assume no interaction between vector species, the "# of the multi-vector system is therefore: G
"# = D P_ _ P_ _ ` a= a= ` HIJ
83 Plugging in values from the multi-vector P matrix, this gives: G b ∗ E F0(1 − -)<) <0=F)= 80= "# = ∗ D (;) + <))(;) + .) + /))8 (;0 + .0 )(;0 + .0 + <0 ) ) HIJ = = = Here, Eb represents the two bites necessary for a vector to become infected and subsequently transmit the parasite, F0 and F)= are the probabilities of successful transmission to the vector and the host respectively, (1 − -) is the proportion of hosts not on preventative medications, c` (d`ec`) is the residency time in the exposed host class, J is the residency time in the infectious (d`ef`eg`) c host class, a= is the residency time in the exposed vector class, and J is the (d ef ec ) (d ef ) a a= a= a a= residency time in the infectious vector class. This "# is used to evaluate the efficiency of dog heartworm transmission in three distinct scenarios: 1) utilizing laboratory generated data, comparison of a highly competent D. immitis vector vs. a poor vector, 2) comparison of a suburban mosquito assemblage vs. a mosquito assemblage in natural areas based on diversity and community composition field data in the two habitat types, and 3) a theoretical inquiry into the effects of vector diversity. For all of the models, I assume that births are equal to deaths for both the host and the vectors. I calculated host population density, 8), based on the number of households owning dogs and the average number of dogs per dog-owning household in North Carolina (American Veterinary Medical Association 2012) as well as the total number of occupied households in Wake County, NC based on the 2017 American Community Survey (United States Census Bureau). I used previously acquired heartworm prevalence data from Chapter 3 of my ∗ dissertation (Spence Beaulieu 2019) to estimate the number of infected dogs, ,), so that 8) could be deduced. The value <) is calculated based on a 6 month (180 day) intrinsic incubation period.
Both probability of successful transmission from the host to the vector, F0, and natural death rate of the vector, ;), are based on values from a recent compartmental model of lymphatic filariasis (Mwamtobe et al. 2017), a disease of humans caused by a similar filarial parasite. All other host values are based on literature specific to D. immitis, or are assumed reasonable values when parameters could not be found in the literature. For the vectors, I calculated the biting rate, E, based on a 3 day gonotrophic cycle and all bites being on the hosts in the model. I assume that all vector species have an equivalent biting rate, as well as equivalent birth and natural death rates. The remaining vector parameters, including probability of successful transmission from the
84 vector to the host (F)), the extrinsic incubation period (<0), and the parasite-induced death rate
(.0), are variable between vector species and are detailed in the sections below. All models are tested with a conservative 2:1 mosquito to dog ratio (i.e. ∑ 80L = 28)) and a 10:1 mosquito to dog ratio (i.e. ∑ 80L = 108)) based on observations of mosquitoes per trap night in Chapter 2 of my dissertation (Spence Beaulieu 2019). By holding the total number of individual mosquitoes constant between models in each of these scenarios, I am able to investigate the effects of changes in vector diversity alone on heartworm transmission since variation in vector abundance is removed. I calculated the number of exposed and infectious mosquitoes based on my prior findings of an overall D. immitis prevalence of 0.6% in Chapter 3 of my dissertation (Spence ∗ Beaulieu 2019), and used this to deduce ∑ 80L.
Efficient Vector vs. Poor Vector I generated parameter values for the extrinsic incubation period of D. immitis, longevity with D. immitis infection, and vector efficiency index (VEI) of the eastern treehole mosquito, Aedes triseriatus, and the Asian tiger mosquito, Aedes albopictus. Methodological details for these laboratory experiments can be found in Chapter 4 of my dissertation (Spence Beaulieu 2019). Through these lab experiments, Ae. triseriatus was demonstrated to be a highly competent vector of D. immitis, with a short extrinsic incubation period, no added mortality due to infection, and a VEI approximately 4.3 times greater than Ae. albopictus. In contrast, Ae. albopictus was found to be a relatively poor vector, with a longer extrinsic incubation period, additional mortality due to infection, and a low VEI.
Extrinsic incubation period was directly translated into <0 for the respective species.
Parasite-induced mortality, .0, was established via comparison of longevity curves for the species with and without D. immitis infection. If the longevity curves were not significantly different, .0 was interpreted as 0. If the longevity curves were significantly different, .0 was assessed as the largest percent difference between mortality of uninfected and infected individuals throughout the length of the curve. Finally, I estimated the probability of successful transmission of the parasite from vector to host, F), using the VEI. VEI gives an estimate of the efficiency of parasite development within the vector, calculated by the average number of L3 larvae surviving to 15 days post-infection divided by the average number of microfilaria ingested. I multiplied this VEI by 0.1 to get the overall transmission efficiency to the host, as
85 Table 2: Vector parameters for Aedes triseriatus and Aedes albopictus. Values used in the comparison of an efficient heartworm vector vs. a poor heartworm vector were generated via a previous laboratory experiment.
Parameter Value
F) 0.02592 Ae. triseriatus < 0.1 parameters 0 .0 0 F) 0.00596 Ae. albopictus < 0.08333 parameters 0 .0 0.25 approximately 10% of L3 larvae successfully enter the bite wound at the time of the infectious vector’s bloodmeal (American Heartworm Society 2018). I calculated the "# for Ae. triseriatus and Ae. albopictus individually using these laboratory generated parameters (Table 2) to compare the transmission potential of dog heartworm with a single efficient vector and with a single competent but poor vector.
Suburban Mosquito Assemblage vs. Natural Mosquito Assemblage I determined the mosquito assemblages present in suburban neighborhoods of different ages and undeveloped wooded and field areas of Wake County, NC as previously described in Spence Beaulieu et al. 2019. I found that the most established suburban areas had overall decreased mosquito diversity measures as compared to undeveloped natural areas, despite approximately equal abundance per trap night. On subsequent investigation of disease trends in these habitat types in Chapter 3 of my dissertation, I also found that heartworm prevalence within both the canine host and the mosquito vectors was lower in suburban areas than in at least the undeveloped field areas (Spence Beaulieu 2019). Based on these epidemiological results, I parameterized models of heartworm transmission by the suburban mosquito assemblage and by the natural mosquito assemblage to quantify differences in "# between the two habitat types. I calculated the average species richness in the oldest category of suburban neighborhoods. I also calculated the average species richness in the undeveloped field and wooded areas combined to approximate the average richness in a natural area of unspecified type. Once average richnesses, h, were established, I analyzed the average species abundances in each habitat type, accepting the h most abundant species from each habitat for inclusion in the respective models. For species that are known vectors of D. immitis in the United States (Ledesma and Harrington 2011, Mckay et al. 2013), I surveyed the published literature for
86 Table 3: Vector parameters for species in the suburban assemblage and the natural assemblage. Relative abundance in the assemblage is given for each species. If estimates were unavailable, an average value (denoted with asterisk) was estimated via the mean of the upper and lower limits known for the parameter from any species.
Relative Species Parameter Value Range Reference abundance Spence Beaulieu 2019 <0 0.08333 - Aedes W 82% . 0.25 - Spence Beaulieu 2019 albopictus 0W 0.00596 Spence Beaulieu 2019 F)W - 0.0959* Kartman 1954, Loftin et al. 1995 <0i 0.0667 – 0.125 Culex Lai et al. 2000, Spence Beaulieu 2019 9.7% . 0.15* 0 – 0.3 salinarius 0i Lowrie 1991, Tiawsirisup and Kaewthamasorn 2007 F)i 0.0293* 0.0003 – 0.0583 0.0923 Jankowski and Bickley 1976 <0j - 0.15* Lai et al. 2000, Spence Beaulieu 2019 Aedes vexans 3.7% .0j 0 – 0.3 0.0293* Lowrie 1991, Tiawsirisup and Kaewthamasorn 2007 F)j 0.0003 – 0.0583 Lai et al. 2001 <0 0.1192 - Culex k 2.8% . 0.15* 0 – 0.3 Lai et al. 2000, Spence Beaulieu 2019 pipiens 0k 0.00347 Lai et al. 2001 F)k - 0.0959* Kartman 1954, Loftin et al. 1995 <0l 0.0667 – 0.125 Psorophora Lai et al. 2000, Spence Beaulieu 2019 Suburban assemblage parameters assemblage Suburban 0.9% . 0.15* 0 – 0.3 ferox 0l 0.0293* Lowrie 1991, Tiawsirisup and Kaewthamasorn 2007 F)l 0.0003 – 0.0583 < 0.1065 Kutz and Dobson 1974 Anopheles 0m - 0.25 Kutz and Dobson 1974 quadrimacul 0.9% .0m - atus 0.0293* Lowrie 1991, Tiawsirisup and Kaewthamasorn 2007 F)m 0.0003 – 0.0583 Spence Beaulieu 2019 <0 0.08333 - Aedes W 34.9% . 0.25 - Spence Beaulieu 2019 albopictus 0W Spence Beaulieu 2019 F)W 0.00596 - <0 0 - - Aedes i 12.3% . 0 - - atlanticus 0i - F)i 0 - 0.0923 Jankowski and Bickley 1976 <0j - 0.15* Lai et al. 2000, Spence Beaulieu 2019 Aedes vexans 12.3% .0j 0 – 0.3 0.0293* Lowrie 1991, Tiawsirisup and Kaewthamasorn 2007 F)j 0.0003 – 0.0583
0.0959* Kartman 1954, Loftin et al. 1995 <0k 0.0667 – 0.125 Culex Lai et al. 2000, Spence Beaulieu 2019 10.4% . 0.15* 0 – 0.3 salinarius 0k Lowrie 1991, Tiawsirisup and Kaewthamasorn 2007 F)k 0.0293* 0.0003 – 0.0583 0.0959* Kartman 1954, Loftin et al. 1995 <0l 0.0667 – 0.125 Psorophora Lai et al. 2000, Spence Beaulieu 2019 10.4% . 0.15* 0 – 0.3 ferox 0l 0.0293* Lowrie 1991, Tiawsirisup and Kaewthamasorn 2007 F)l 0.0003 – 0.0583 Kartman 1954, Loftin et al. 1995 <0 0.0959* 0.0667 – 0.125 Psorophora m 7.5% . 0.15* 0 – 0.3 Lai et al. 2000, Spence Beaulieu 2019 columbiae 0m F 0.0293* 0.0003 – 0.0583 Lowrie 1991, Tiawsirisup and Kaewthamasorn 2007 Natural assemblage parameters assemblage Natural )m Kartman 1954, Loftin et al. 1995 <0n 0.0959* 0.0667 – 0.125 Culex Lai et al. 2000, Spence Beaulieu 2019 4.7% . 0.15* 0 – 0.3 erraticus 0n 0.0293* Lowrie 1991, Tiawsirisup and Kaewthamasorn 2007 F)n 0.0003 – 0.0583 <0 0.1 - Spence Beaulieu 2019 Aedes o 3.8% . 0 - Spence Beaulieu 2019 triseriatus 0o Spence Beaulieu 2019 F)o 0.02592 - Lai et al. 2001 <0 0.1192 - Culex p 3.8% . 0.15* 0 – 0.3 Lai et al. 2000, Spence Beaulieu 2019 pipiens 0p 0.00347 Lai et al. 2001 F)p -
87 appropriate <0, .0, and F) values for each species in the models. Since the extrinsic incubation period depends on temperature, I estimated the acquired heartworm development units (Ledesma and Harrington 2015) in the published studies and recalculated the extrinsic incubation period based on ambient temperature of 29.5°C to match the conditions of my lab generated values and provide consistency in the parameter between species. Where specific values could not be found for a given parameter, I used a mean value for the parameter based on the highest and lowest published values of the parameter for any mosquito species. This approach approximates an average vector where the literature does not provide specific values to suggest otherwise. For species that are not known to be vectors of D. immitis in the US, all three parameter values were designated as 0. Finally, I calculated the "# for the models of the two habitat types, comparing the transmission potential of dog heartworm in an empirically based diverse natural assemblage and a less diverse suburban assemblage (Table 3).
Vector Diversity To test predictions regarding the effect of vector diversity on dog heartworm transmission, I investigated the average "# of many iterations of a one vector model, a three vector model, a five vector model, and a ten vector model. For each of ' vectors in the four respective models, I generated 1000 random values of each parameter from a uniform distribution bounded at the parameter’s minimum and maximum found in the literature for any mosquito species (as listed in Table 3). I then used these vectors with randomly generated parameters to calculate 1000 "# values for each of the four models, with each vector’s ∗ ∗ ∗ population size at DFE being equivalent (80W = 80i = ⋯ = 80L). As previously mentioned, the ∗ total vector abundance for all species, ∑ 80L , is constant across all models (Table 1) to disentangle the effects of changes in diversity from the effects of changes in abundance. Finally,
I found the average "# for the four models and used a violin plot to assess the structure of the data for each model given the random vectors (Hintze and Nelson 1998).
RESULTS When comparing the efficient vector Ae. triseriatus to the poor vector Ae. albopictus based on laboratory generated parameters, marked differences were noted. The system with Ae. triseriatus as the sole vector had an "# of 1.229 with a 2:1 mosquito to dog ratio and an "# of
88
6.146 with a 10:1 mosquito to dog ratio, reaching the critical threshold of "# > 1 for disease spread in either case. In contrast, the system with Ae. albopictus as the sole vector had a very low
"# of 0.044 with a 2:1 mosquito to dog ratio and of 0.218 with a 10:1 mosquito to dog ratio. Despite being a competent vector for D. immitis, Ae. albopictus alone is unable to sustain dog heartworm transmission in a natural setting based on the "#. Using the field data to simulate a suburban mosquito assemblage resulted in a system with six vector species dominated by Ae. albopictus, which comprised 82% of the mosquito assemblage (Table 3). The resultant "# of the suburban multi-vectored model was 0.097 with a 2:1 mosquito to dog ratio and 0.487 with a 10:1 mosquito to dog ratio. For the natural mosquito assemblage based on field data in undeveloped areas, I tested a model with nine vector species. Although Ae. albopictus was still the most dominant species in the natural assemblage, the assemblage was much more even than the suburban assemblage, with Ae. albopictus contributing
Figure 2: Average qr for systems with varying vector diversity and a 2:1 mosquito to dog ratio. Mean values are denoted by red dots, with red bars giving the standard deviation. The density trace is depicted by the grey violin plots.
89 only 34.9% to the overall total (Table 3). The "# of the multi-vectored model with a natural area mosquito assemblage was 0.247 with a 2:1 mosquito to dog ratio and 1.233 with a 10:1 mosquito to dog ratio. While the "# values for both the suburban and the natural assemblages were well under the critical value of 1 with a 2:1 mosquito to dog ratio, the natural assemblage’s "# was notably higher than the suburban assemblage’s value. Additionally, when using a 10:1 mosquito to dog ratio, the "# of the suburban assemblage did exceed the critical value of 1, signifying the potential for disease expansion in the higher mosquito density scenario. The one vector, three vector, five vector, and ten vector models generated from 1000 simulations of random parameters for each vector resulted in a similar average "# for all four models. "# values were 0.525 ± 0.015 (mean ± SEM) for the one vector model, 0.511 ± 0.008 for the three vector model, 0.504 ± 0.006 for the five vector model, and 0.5 ± 0.004 for the ten vector model with the 2:1 mosquito to dog ratio. Using a 10:1 mosquito to dog ratio, "# values were 2.625 ± 0.075 for the one vector model, 2.553 ± 0.040 for the three vector model, 2.522 ± 0.032 for the five vector model, and 2.502 ± 0.022 for the ten vector model. In both cases, while the mean remains relatively constant across the models, the standard error of the mean decreases as the number of vectors increases. The distribution of the data also changes with the number of vectors in the model. For the single vector model, the distribution of "# values is right skewed, with the highest density of values at approximately 0.25 for the 2:1 mosquito to dog ratio (Figure 2). As the number of vectors increases, the skewness decreases and the highest density of values gets closer to the mean. With the ten vector model and the 2:1 mosquito to dog ratio, the right skewness is very slight and the peak density is just under the mean of 0.5 (Figure 2). The violin plot is identical for the 10:1 mosquito to dog ratio with the exception of the y-axis scale, which has values 5x those present in the 2:1 scenario. Overall, as the number of vectors increase, there is less variability on whether an outbreak will occur and the extent of the outbreak.
DISCUSSION In this study, I constructed a single host and multi-vector model of dog heartworm disease to investigate the effects of mosquito diversity on disease transmission. I verified that highly competent heartworm vectors have greater transmission potential than competent but poor vectors, as only the efficient vector leads to an "# > 1. I demonstrated a higher "# for the more
90 diverse natural mosquito assemblage than for the less diverse suburban mosquito assemblage in Wake County, NC. Through model simulation, I also found that vector diversity does not affect the average "# but does stabilize the system and bring the peak of the distribution closer to the average as diversity increases. To my knowledge, this is the first study to uncouple diversity from abundance, modeling the effects of vector diversity changes alone on disease transmission for a multi-vectored pathogen. Prior research in Chapter 3 of my dissertation found decreased heartworm prevalence within both the mosquito vector and the domestic dog host in suburban areas of Wake County, NC when compared to undeveloped field areas (Spence Beaulieu 2019). The suburban areas were characterized by a less diverse mosquito assemblage (Spence Beaulieu et al. 2019), and within-host heartworm prevalence was positively correlated with mosquito Shannon-Wiener index and evenness, two measures of diversity. These empirical findings were validated by "# calculations in the current study, using realistic information on the richness and community composition in these two distinct habitat types and the best available species-specific parameter values. With a 2:1 mosquito to dog ratio, the "# of 0.097 for the suburban mosquito assemblage was notably lower than that of the natural mosquito assemblage at 0.247, agreeing with the decreased within-vector and with-host heartworm prevalence in suburban areas. If assuming a higher ratio of vectors to hosts, the relationship between the "# values of the suburban and natural assemblages remain the same, but the "# of the natural assemblage becomes greater than 1. Based on prior findings of the average number of mosquitoes per trap night, I expect that a higher density of mosquitoes such as that modeled in the 10:1 mosquito to dog scenario is more reasonable than a 2:1 ratio. However, using the lower and more conservative mosquito density, both of the "# values are substantially less than the critical threshold for disease proliferation,
"# > 1. In addition to the uncertainty of true mosquito densities, there are of course multiple simplifications that were made during construction of the model that could have missed important contributors to the transmission process and resulted in lower than expected "# values for a known endemic disease. Chief among these simplifications is the inclusion of only owned domestic dogs as hosts in the model. Dirofilaria immitis infects a variety of wild canine hosts in addition to domestic dogs, including feral dogs and coyotes. These wild hosts are likely to be the greatest reservoirs of D. immitis, with coyote infection rates of approximately 47% in North
91 Carolina (Chitwood et al. 2015), notably higher than the estimated 12.14% heartworm prevalence in domestic dogs that was utilized in this model. In addition to increasing the likelihood of mosquito bites on an infectious host, the proportion of immune hosts would also decrease with inclusion of wild hosts in the model. Since -, the proportion of hosts on preventative medication, would decrease directly with inclusion of wild canines, (1 − -) would increase in the numerator of the "# equation and potentially allow the estimate for at least the natural mosquito assemblage to cross unity with a 2:1 mosquito to dog ratio. Unfortunately, the paucity of information on numbers of coyotes or other wild canine hosts in Wake County prevents evidence-based incorporation of these alternate hosts into the model. Aside from the exclusion of wild hosts, other major model assumptions that could impact the determined "# values are a constant biting rate for all mosquito species, the lack of seasonality, and a deterministic framework. All of these assumptions could be relaxed in future studies to further investigate the mechanics of heartworm transmission. Additionally, the model assumes an absence of host heterogeneity despite evidence of differences in heartworm burdens between infected hosts (Capelli et al. 1996). Questions of host heterogeneity would be better addressed in a macroparasitic model. Similarly, spatial processes are ignored in the current model, but could be investigated through an agent-based modeling approach.
Despite the underlying model assumptions potentially hindering the accuracy of the "# for realistic scenarios in the field setting, this SIR-type model framework still allows for comparison of the transmission potential between simulations with differing numbers of vector species. In previous studies of vector diversity and disease transmission, greater diversity led to greater overall disease prevalence (Roche et al. 2012). However, diversity and abundance were linked such that the total mosquito abundance was likely driving the observed changes in disease rather than vector diversity in and of itself. Here, in the absence of increased abundance, vector diversity is a stabilizing force but does not necessarily lead to a heightened "#. The density of simulated "# values increases toward the average as vector diversity increases, but the range of
"# values is also diminished. With only a single vector species, it is possible to have no disease transmission or relatively high transmission, based on the competence of the sole present species for D. immitis. In contrast, the probability of having zero disease transmission in a ten vector system is exceedingly low, but the intensity of transmission with a diverse assemblage will depend on the community composition.
92 The identity and relative abundance of host species in a community assemblage has been identified as an important factor in Lyme disease risk assessment (LoGiudice et al. 2008). My results suggest that community composition within the vector assemblage is equally important to disease transmission. As evidenced by the significant difference in "# between species such as Ae. albopictus and Ae. triseriatus, not all competent vectors contribute equally to transmission; species identity and relative abundance within a diverse vector assemblage is important to disease transmission. Beyond species identity alone, an understanding of the vector competence of local mosquito populations is necessary, as vector efficiency for heartworm is heritable and can vary across geographically distinct populations (Nayar and Knight 1999). This can lead to situations where even presumably similar mosquito assemblages, such as that of urban areas of Oklahoma (Paras et al. 2014) and suburban areas of North Carolina (Spence Beaulieu 2019), can yield differing disease prevalence due to variation in the vectorial capacity of the dominant species between the two locations. While vector diversity alone stabilizes disease transmission and, in conjunction with increases in abundance, can lead to more disease, the magnitude of disease transmission risk will depend on both the vector community composition and the population-specific competences of the species in the assemblage for a given pathogen. Local studies on the landscape-level determinants of mosquito assemblages as well as the vectorial capacity of local populations is critical for accurate prediction of vector-borne disease risk.
ACKNOWLEDGEMENTS I thank Cristina Lanzas and Michael Reiskind for their invaluable insight and guidance related to both the construction of and goals for the model.
93 REFERENCES
American Heartworm Society. 2018. Current Canine Guidelines for the Prevention, Diagnosis, and Management of Heartworm (Dirofilaria immitis) Infection in Dogs.
American Veterinary Medical Association. 2012. U.S. Pet Ownership & Demographics Sourcebook. American Veterinary Medical Association, Schaumburg, IL.
Brown, H. E., L. C. Harrington, P. E. Kaufman, T. McKay, D. D. Bowman, C. T. Nelson, D. Wang, and R. Lund. 2012. Key factors influencing canine heartworm, Dirofilaria immitis, in the United States. Parasites & Vectors 5:245.
Capelli, G., G. Poglayen, F. Bertotti, S. Giupponi, and M. Martini. 1996. The host-parasite relationship in canine heartworm infection in a hyperendemic area of Italy. Vet. Res. Commun. 20:320-330.
Chitwood, M. C., M. B. Swingen, M. A. Lashley, J. R. Flowers, M. B. Palamar, C. S. Apperson, C. Olfenbuttel, C. E. Moorman, and C. S. DePerno. 2015. Parasitology and serology of free-ranging coyotes (Canis latrans) in North Carolina, USA. J. Wildl. Dis. 51:664-669.
Diekmann, O., J. Heesterbeek, and M. G. Roberts. 2009. The construction of next-generation matrices for compartmental epidemic models. Journal of the Royal Society Interface 7:873-885.
Dobson, A. 2004. Population dynamics of pathogens with multiple host species. Am. Nat. 164:S78.
Fuller, D. O., T. Alimi, S. Herrera, J. C. Beier, and M. L. Quiñones. 2016. Spatial association between malaria vector species richness and malaria in Colombia. Acta Trop. 158:197- 200.
Hintze, J. L., and R. D. Nelson. 1998. Violin plots: a box plot-density trace synergism. The American Statistician 52:181-184.
Jankowski, T. J., and W. E. Bickley. 1976. The mosquitoes, Aedes canadensis and A. vexans as potential vectors of Dirofilaria immitis in Maryland. Ann. Entomol. Soc. Am. 69:781- 783.
Kartman, L. 1954. Suggestions concerning an Index of Experimental Filaria Infection in Mosquitoes. Am. J. Trop. Med. Hyg. 3:329-337.
Kutz, F. W., and R. C. Dobson. 1974. Effects of temperature on the development of Dirofilaria immitis (Leidy) in Anopheles quadrimaculatus Say and on vector mortality resulting from this development. Ann. Entomol. Soc. Am. 67:325-331.
94 Lai, C., K. Tung, H. Ooi, and J. Wang. 2001. Susceptibility of mosquitoes in central Taiwan to natural infections of Dirofilaria immitis. Med. Vet. Entomol. 15:64-67.
Lai, C., K. Tung, H. Ooi, and J. Wang. 2000. Competence of Aedes albopictus and Culex quinquefasciatus as vector of Dirofilaria immitis after blood meal with different microfilarial density. Vet. Parasitol. 90:231-237.
Ledesma, N., and L. Harrington. 2011. Mosquito vectors of dog heartworm in the United States: vector status and factors influencing transmission efficiency. Topics in companion animal medicine 26:178-185.
Ledesma, N., and L. Harrington. 2015. Fine-scale temperature fluctuation and modulation of Dirofilaria immitis larval development in Aedes aegypti. Vet. Parasitol. 209:93-100.
Lee, A. C., S. P. Montgomery, J. H. Theis, B. L. Blagburn, and M. L. Eberhard. 2010. Public health issues concerning the widespread distribution of canine heartworm disease. Trends Parasitol. 26:168-173.
Loftin, K. M., R. L. Byford, M. J. Loftin, and M. E. Craig. 1995. Potential mosquito vectors of Dirofilaria immitis in Bernalillo County, New Mexico. J. Am. Mosq. Control Assoc. 11:90-93.
LoGiudice, K., S. T. Duerr, M. J. Newhouse, K. A. Schmidt, M. E. Killilea, and R. S. Ostfeld. 2008. Impact of host community composition on Lyme disease risk. Ecology 89:2841-2849.
LoGiudice, K., R. S. Ostfeld, K. A. Schmidt, and F. Keesing. 2003. The ecology of infectious disease: effects of host diversity and community composition on Lyme disease risk. Proceedings of the National Academy of Sciences 100:567-571.
Lord, C. C. 2010. The Effect of Multiple Vectors on Arbovirus Transmission. Israel Journal of Ecology and Evolution 56:371-392.
Lowrie, R. C. 1991. Poor vector efficiency of Culex quinquefasciatus following infection with Dirofilaria immitis. J. Am. Mosq. Control Assoc. 7:30-36.
Mckay, T., T. Bianco, L. Rhodes, and S. Barnett. 2013. Prevalence of Dirofilaria immitis (Nematoda: Filarioidea) in mosquitoes from northeast Arkansas, the United States. J. Med. Entomol. 50:871-878.
McMillan, J. R., R. A. Blakney, D. G. Mead, W. T. Koval, S. M. Coker, L. A. Waller, U. Kitron, and G. M. Vazquez-Prokopec. 2019. Linking the vectorial capacity of multiple vectors to observed patterns of West Nile virus transmission. J. Appl. Ecol. 56:956-965.
95 Mwamtobe, P. M., S. M. Simelane, S. Abelman, and J. M. Tchuenche. 2017. Mathematical analysis of a lymphatic filariasis model with quarantine and treatment. BMC Public Health 17:265.
Nayar, J. K., and J. W. Knight. 1999. Aedes albopictus (Diptera: Culicidae): an experimental and natural host of Dirofilaria immitis (Filarioidea: Onchocercidae) in Florida, USA. J. Med. Entomol. 36:441-448.
New, J., John C, W. J. Kelch, J. M. Hutchison, M. D. Salman, M. King, J. M. Scarlett, and P. H. Kass. 2004. Birth and death rate estimates of cats and dogs in US households and related factors. Journal of Applied Animal Welfare Science 7:229-241.
Paras, K. L., V. A. O'brien, and M. H. Reiskind. 2014. Comparison of the vector potential of different mosquito species for the transmission of heartworm, Dirofilaria immitis, in rural and urban areas in and surrounding Stillwater, Oklahoma, USA. Med. Vet. Entomol. 28:60-67.
Roche, B., P. Rohani, A. P. Dobson, and J. Guégan. 2012. The impact of community organization on vector-borne pathogens. Am. Nat. 181:1-11.
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.
Spence Beaulieu, M. R. 2019. The Ecology of Dog Heartworm Disease. PhD Dissertation, Department of Entomology and Plant Pathology, North Carolina State University, Raleigh, NC.
Spence Beaulieu, M. R., K. Hopperstad, R. R. Dunn, and M. H. Reiskind. 2019. Simplification of vector communities during suburban succession. PloS one 14:e0215485.
Tiawsirisup, S., and M. Kaewthamasorn. 2007. The potential for Aedes albopictus (Skuse)(Diptera: Culicidae) to be a competent vector for canine heartworm, Dirofilaria immitis (Leidy). Southeast Asian J. Trop. Med. Public Health 38:208.
Turney, S., A. Gonzalez, and V. Millien. 2014. The negative relationship between mammal host diversity and Lyme disease incidence strengthens through time. Ecology 95:3244- 3250.
United States Census Bureau. American FactFinder. https://factfinder.census.gov/.
Wada, Y., Y. Tsuda, and O. Suenaga. 1989. Transmission dynamics of Dirofilaria immitis in a southwestern part of Japan. Tropical medicine 31:35-47.
96 CHAPTER 6
The Role of Parasite Manipulation in Vector-Borne Diseases
(This work was published in Evolution, Medicine, & Public Health: Spence Beaulieu MR (2019) The role of parasite manipulation in vector-borne diseases. Evolution, Medicine, & Public Health 2019(1): 106-107. https://doi.org/10.1093/emph/eoz019)
Definition and Background The parasite manipulation hypothesis posits that parasites can purposefully alter host behaviours, increasing probability of transmission to an uninfected host [1]. An example is Toxoplasma gondii, where infected rodents become less predator averse, increasing the likelihood of infection reaching the feline host [2]. With other behavioural alterations, determination of whether effects are due to manipulations or are secondary outcomes of infection can be difficult [1]. Regardless, parasite-induced changes represented in the manipulation hypothesis have implications for disease transmission. The hypothesis applies to vector-borne diseases, where parasite-induced changes in vector behaviour can increase transmission to the non-arthropod host. Here, a commonly affected behaviour is blood-feeding. Arthropods must blood-feed twice to transmit pathogens, first on an infectious host then again on a susceptible host. This necessity for two blood meals to fulfil the parasite’s life cycle makes blood-feeding a major component to vector-borne disease transmission [3].
Examples in Public Health Infection within both the vector and host can affect blood-feeding. Malaria is a well- studied example that highlights these dual impacts. Mosquitoes harbouring the non-infectious parasitic stage probe less, reducing mortality risks associated with feeding. Yet when the parasite is at the infectious stage, probing and therefore probability of transmission to the host is increased [4]. Similarly, differential requirements for the blood-meal volume that triggers host-seeking inhibition exist depending on
97 the parasite stage, either minimizing feeding risk with low required volume when non-infectious or maximizing transmission through high required volume when infectious [4]. Within humans, infection with malaria can lead to differential mosquito attraction. When hosts carry the infective parasitic stage, specific blood components and volatile compounds on the skin are altered. This changes mosquito behaviour by increasing attractiveness as compared to the same host when non-infectious [5-7]. Although host odour is directly modified by the parasite, the ultimate effect is alteration of mosquito behaviour, which subsequently increases parasite transmission to the vector.
Evolutionary Perspectives and Implications Malaria is but one well-explored example of a vector-borne disease with parasite-induced behavioural changes. In many pathosystems, parasites have evolved over centuries with both their hosts and vectors. This gives opportunity for host, vector, and parasite adaptation, effects of which could all present as manipulations and have similar impacts on disease dynamics regardless of the cause. Consideration of possible manipulations is important for a holistic understanding of vector-borne disease systems, and should be integrated into transmission models to enable accurate depiction of disease dynamics. Knowledge of the types of behavioural alterations commonly seen with parasite manipulations could help identify appropriate targets for vector-borne disease mitigation efforts [8].
98 REFERENCES
1. Poulin R. The evolution of parasite manipulation of host behaviour: a theoretical analysis. Parasitology 1994;109:S118.
2. Berdoy M, Webster JP, Macdonald DW. Fatal attraction in rats infected with Toxoplasma gondii. Proc R Soc B 2000;267:1591-4.
3. Brady OJ, Godfray HCJ, Tatem AJ et al. Vectorial capacity and vector control: reconsidering sensitivity to parameters for malaria elimination. Trans R Soc Trop Med Hyg 2016;110:107-117.
4. Koella JC, Rieu L, Paul RE. Stage-specific manipulation of a mosquito’s host-seeking behavior by the malaria parasite Plasmodium gallinaceum. Behav Ecol 2002;13:816-20.
5. Emami SN, Lindberg BG, Hua S et al. A key malaria metabolite modulates vector blood seeking, feeding, and susceptibility to infection. Science 2017;355:1076-80.
6. De Moraes CM, Stanczyk NM, Betz HS et al. Malaria-induced changes in host odors enhance mosquito attraction. PNAS 2014;111:11079-84.
7. Lacroix R, Mukabana WR, Gouagna LC et al. Malaria infection increases attractiveness of humans to mosquitoes. PLoS Biol 2005;3:e298.
8. Cator LJ, Lynch PA, Thomas MB et al. Alterations in mosquito behaviour by malaria parasites: potential impact on force of infection. Malaria J 2014;13:164.
99 CHAPTER 7
Conclusions
In this dissertation, I investigated the ways in which suburban development affects mosquito species assemblages and subsequent transmission of vector-borne diseases in the context of dog heartworm disease. In the first study (Chapter 2), I found that suburban development alters mosquito assemblages. The resultant suburban assemblage is less diverse than that found in either undeveloped field or wooded areas. Additionally, suburbia is dominated by the invasive Asian tiger mosquito, Aedes albopictus, which is a known vector of many concerning human pathogens as well as my pathogen of interest, the dog heartworm. Although this finding led me to predict that suburban areas would have higher levels of dog heartworm disease, my second study (Chapter 3) suggests the opposite. I found that within both the mosquito vector and the canine host, suburban areas had less heartworm than undeveloped field areas, with wooded areas intermediate in pathogen presence. These landscape-level differences in host heartworm disease risk were associated with changes in vector diversity as well as socioeconomic status, as host heartworm prevalence was well modeled by a positive relationship with mosquito Shannon-Wiener diversity and a negative relationship with median household income. Given that suburban areas in Wake County, NC are dominated by Ae. albopictus and have the lowest heartworm rates, I questioned the vector competence of the local population of Ae. albopictus for dog heartworm. Testing this species alongside Ae. triseriatus, a suspected efficient local heartworm vector based on a relatively high rate of field infection, I found that Ae. albopictus was indeed a poor vector for heartworm in the study area. In my third study (Chapter 4), I demonstrated that Aedes albopictus had significantly decreased longevity when infected with heartworm and a vector efficiency index less than a quarter that of Ae. triseriatus. This study highlighted that vector competence can vary wildly between species and even within different populations of the same species, as Ae. albopictus is known to be an efficient heartworm vector in other areas. It also generated local information on the competence of two
100 prominent mosquito species for dog heartworm, which is useful for disease modeling and heartworm risk assessment. In the final study (Chapter 5), I validated the importance of both vector diversity and community composition to disease risk via mathematical modeling of the dog heartworm disease system. Using laboratory generated mosquito parameter values, the model demonstrated an "# > 1 for a system with Ae. triseriatus as the sole vector, indicating that it is a vector than can support sustained disease transmission and growth. In contrast, an "# ≪ 1 was found for a system with Ae. albopictus as the sole vector, demonstrating its poor vector status locally. Based on field data on the diversity and identity of mosquito assemblages in suburban and natural areas, the model also generated a lower "# for the suburban assemblage than for the natural assemblage, agreeing with the epidemiological findings of the landscape-level heartworm trends in Wake County, NC. Finally, simulations of models with one, three, five, and ten vectors of random vector competence revealed that the average "# for each system was approximately equal, but variation decreased and the peak density of values increased toward the mean as the number of vectors increased, demonstrating the stabilizing effects of vector diversity on a multi- vectored pathogen. Overall, the magnitude of vector-borne disease transmission depends not only on assemblage diversity, but also on the composition of the individuals within the assemblage. Through field experiments, epidemiological studies, laboratory experiments, and mathematical modeling, I have produced new findings on the ecological determinants of dog heartworm disease. However, many questions remain unanswered. There is an obvious need for incorporation of wild hosts into heartworm studies, as they are likely the largest reservoirs of the pathogen and important disease dynamics can be missed without their inclusion. Detailed information on the distribution, abundance, and heartworm prevalence of coyotes and other wild canids would contribute greatly to a holistic picture of the disease system and to more accurate disease models. Also useful for building better disease prediction capabilities are additional evaluations of the heartworm vector efficiency of North Carolina mosquito populations. But of the potential future studies based on my research, most interesting to me is the idea of parasite manipulation, as discussed in Chapter 6 for malaria. I have consistently wondered whether heartworms alter vector blood-feeding behavior through differential host attraction. It is plausible that infection with the heartworm parasite alters the blood chemistry or volatile compounds
101 emitted from the host, increasing mosquito attraction, blood-feeding, and consequently the probability of the parasite completing its life cycle, like is seen with malaria. However, it is also plausible that heartworm disease is so ubiquitous due to the multiple competent hosts and vectors that manipulation is not necessary for successful transmission. Heartworm-induced changes in vector behavior, if they exist, would have obvious importance for disease transmission dynamics and would be a fascinating course of future study. In his book The Sacred Beetle and Others (Fabre 1918), naturalist and early entomologist Jean-Henri Fabre wrote: “… in the domain of instinct, who can claim a harvest that leaves no grain for other gleaners? Sometimes the handful of corn left on the field is of more importance than the reaper’s sheaves. If we had to wait until we knew every detail of the question studied, no one would venture to write the little he knows. From time to time, a few truths are revealed, tiny pieces of the vast mosaic of things.” It is with this in mind that I close my dissertation, in hopes that my studies have contributed to what is known while also sparking new questions about the ecology of dog heartworm disease.
102 REFERENCES
Fabre, J. 1918. The sacred beetle and others. Dodd, Mead, and Company, New York, NY.