Exploring evolution during the Australian cane toad (Rhinella marina) invasion using data on genetics, expression, and immune function

A thesis submitted in fulfilment of the requirements for the degree of

Doctor of Philosophy

Daniel Selechnik

Faculty of Science, School of Life and Environmental Sciences

The University of Sydney

January 2019

DECLARATION

This statement is to certify that to my best knowledge, the content of this thesis is my own work, and that all assistance received in preparing this thesis has been acknowledged. This thesis has not been submitted for any alternative degree or other purposes.

Daniel Selechnik

January 2019

ii PREFACE

In accordance with the current guidelines for a Doctor of Philosophy thesis in the Faculty of Science, School of Life and Environmental Sciences at the University of Sydney, all data chapters of this thesis are presented as stand-alone manuscripts that are either published in peer-reviewed journals, currently under consideration for publication, or in preparation for submission. This method of presentation means that there is some unavoidable repetition of background information and methodology in each chapter.

All chapters in this thesis have been written with various co-authors, including Dr. Lee

Ann Rollins, Dr. Mark Richardson, Professor Richard Shine, Dr. Gregory Brown, Dr.

Crystal Kelehear, Mahalia Martin, Dr. Melanie Hess, Dr. Andrew Hess, Dr. Ken Dodds,

Dr. BriAnne Addison, and Dr. Kerry Fanson. These individuals contributed to the designing of experiments, editing of the manuscripts, and assisting with field and lab work.

However, I conducted all research associated with this thesis, including the field work, laboratory experiments, data analysis, and manuscript drafting and writing.

Approval for this research was provided by the University of Sydney Animal Care and

Ethics Committee (Project Number: 2016/1003).

iii AUTHORSHIP ATTRIBUTION STATEMENT

Parts of Chapter 1 of this thesis is published as a literature review:

Selechnik D, Rollins LA, Brown GP, Kelehear C, Shine R. 2016. The things they carried: the pathogenic effects of old and new parasites following the intercontinental invasion of the Australian cane toad (Rhinella marina). International Journal for Parasitology: Parasites and Wildlife, 6(3): 375-385.

I synthesized the literature and wrote the drafts of the manuscript. I am the first and corresponding author.

Chapter 5 of this thesis is published as:

Selechnik D, West AJ, Brown GP, Fanson KV, Addison B, Rollins LA, Shine R. 2017. Effects of invasion history on physiological responses to immune system activation in invasive Australian cane toads. PeerJ, 5:e3856.

I designed the study, performed the experiments, analysed the data, and wrote the drafts of the manuscript. I am the first and corresponding author.

As the candidate, I certify that the above statements are true.

Daniel Selechnik, January 2019

As supervisor for the candidature upon which this thesis is based, I can confirm that the authorship attribution statements above are correct.

Richard Shine, January 2019

iv

ACKNOWLEDGEMENTS

First, I thank my primary supervisor, Professor Rick Shine. During the process of applying to graduate school, I saw firsthand how many young researchers would do anything for the opportunity to be your student, and I am grateful that you gave me the chance to move to

Australia and travel throughout the world working under your mentorship. Delving into evolutionary questions with the notorious invasive cane toad system has been a rollercoaster. I thank you and Terri for your kindness in taking me into your home when I first arrived in Sydney and showing me the ropes of being an Australian. I have always been awed by your amazing body of work on snakes, so working with you has been a dream come true. All your wisdom has been invaluably helpful, particularly with writing; it only took one round of your manuscript draft edits to make drastic improvements to my writing skills, which have been useful to me ever since, both in and outside of science.

Your advice rings through my head every time I put pen to paper, and I have even passed along some of it to friends whom have not had such meticulous and crafty supervisors.

Finally, thank you so much for arranging joint mentorship with Lee Ann; I feel so lucky to have worked with both of you.

To my second primary supervisor, Dr. Lee Ann Rollins: Over the past three and a half years, you have been a professional mentor, a parental figure, and a friend. Thank you for all the time you have devoted to helping me grow as a scientist and as a person. Your mastery of genetics, which I have always seen as one of the most challenging disciplines within biology, has made me feel confident and excited to use it as a tool to explore big ecological and evolutionary questions. Thank you for setting up practice sessions before every forum or conference presentation, for reading repeated drafts of my manuscripts until all the knots were worked out, and for always finding time to meet and discuss my

v project. Bouncing ideas back and forth when planning experiments or interpreting data has been a huge asset to developing my scientific thought process. In addition to the intellectual challenges posed by a PhD, I feel that there are moments that require mental and emotional strength and composure, and your supportive nature as a supervisor has been critical to getting through those times. Your stories have inspired me and given me perspective, and I share them with others who fear that adversity may prevent them from achieving their goals. Your family has always been so welcoming and inclusive – I thank

Joachim, Adam, and Klara for all the laughs, good company, and delicious food. All your current and future students are lucky to have your mentorship, and I hope that our paths will continue to cross many times in the future.

I owe big thanks to my supervisor and whiz, Dr. Mark Richardson.

Before starting my PhD, I had no coding experience whatsoever and was intimidated by the steep learning curve posed by bioinformatics and statistical analyses. But thanks to your efficient teaching and instructions, I have come to really enjoy navigating my way through the command line. I appreciate your patience with me whenever it took longer than I would have liked to wrap my head around data analyses, and that you always found time to help me with coding questions and manuscript preparations. I feel very fortunate to have had your support throughout my PhD.

I thank my final supervisor, Dr. Greg Brown, for being so generous with your seemingly endless scientific knowledge, ingenuity, and wisdom. In addition to genetics, I was keen to study the immune system, an opportunity that would not have been possible without you and the years of work you have done refining methods to study immunity in cane toads. Living and conducting field work in the Northern Territory for a few months was an amazing experience, and also would not have been possible without your brilliant management of the facility up there.

vi Due to the fortune I have had with joint supervision, I have been part of multiple teams that have made me feel welcome at different universities: the University of Sydney,

Deakin University, and UNSW.

At the University of Sydney: thanks so much to Melanie Elphick for being both a mentor and a friend to me for the past three and a half years. I look forward to seeing you every time I visit the lab; your cheeriness and positivity never fail to brighten my day. I appreciate all the work you put in to make the logistics at the University run so smoothly for all of us students. And thank you to my fellow Shine lab students, past or present, who have become close friends: Georgia Kosmala, Lachlan Pettit, and Cam Hudson. I have really enjoyed all the eating, running, and exploring we have done together over the past few years, and getting to know you all was an absolute pleasure.

At Deakin University: I am grateful for the clever researchers who provided me with guidance and insight – Dr. BriAnne Addison, Dr. Kerry Fanson, and Dr. Ondi Crino.

At UNSW: I am very happy to be surrounded by my fellow ‘Rollins Stones’ – Kat

Stuart, Roshmi Sarma, Jia Zhou, and Natalie Hofmeister. For the past year, the quirky dynamic that you all have brought to our team has made the grind to finish my thesis much more light-hearted and fun, and I will think of you all fondly every time I eat laksa. Thanks to Teagan Gale for your friendship, support, and fun workouts, which have all been super helpful during thesis crunch time.

Finally, I could not have made it through my PhD without my support system back home in the United States – my Mom and Dad, my pup Shiloh, and closest friends: Alvin

Varghese, Anna Vaigast, Brittany George, Chengcheng Song, Cody Long, Helen Choi,

Nikki Subnani, Olivia Schultz, Peter Eisenhauer, Rachel Lee, Soo-Jung Kim, Tatyana

Fedorenko, and Tina Mensa-Kwao. Immediately after finishing my undergraduate coursework I moved to the other side of the world. Adjusting to this change while trying to

vii forge a successful PhD path was initially very difficult for me, and I could not have done it without all of you. Thanks to my Mom and Dad for always checking up on me, for welcoming me back and spoiling me each time I visited home, and for visiting me in my new home and exploring it with me. The efforts of my family and friends to stay in touch

(and even visit in some cases) have helped me to feel like they are always with me, which has been a big morale boost when trucking along with this massive project. I am very excited for our next reunion.

viii TABLE OF CONTENTS DECLARATION...... ii PREFACE ...... iii AUTHORSHIP ATTRIBUTION STATEMENT ...... iv ACKNOWLEDGEMENTS ...... v ABSTRACT ...... 11 CHAPTER 1: Introduction ...... 12 Literature Cited...... 18 CHAPTER 2: Impacts of filtering parameter choices on inferences of population structure are dependent on inherent levels of population differentiation ...... 26 Abstract ...... 27 Introduction ...... 28 Materials and Methods ...... 31 Results ...... 35 Discussion ...... 44 Literature Cited...... 50 Supporting Information ...... 62 CHAPTER 3: Invaders weather the weather: does rapid adaptation to a novel environment represent a genetic paradox? ...... 70 Abstract ...... 71 Introduction ...... 73 Materials and Methods ...... 77 Results ...... 87 Discussion ...... 98 Literature Cited...... 106 Supporting Information ...... 116 CHAPTER 4: Immune and environment-driven gene expression during invasion: An eco- immunological application of RNA-Seq...... 127 Abstract ...... 128 Introduction ...... 129 Materials and Methods ...... 132 Results ...... 142 Discussion ...... 151 Literature Cited...... 157 Supporting Information ...... 165 CHAPTER 5: Effects of invasion history on physiological responses to immune system activation in invasive Australian cane toads ...... 173 Abstract ...... 174 Introduction ...... 175 Materials and Methods ...... 179 Results ...... 185 Discussion ...... 190 Literature Cited...... 195 CHAPTER 6: Conclusions ...... 204

9 APPENDICES ...... 207 PUBLICATIONS ...... 263

10 ABSTRACT

Invasive species are notorious for their negative impacts on the environment and economy, but also provide useful systems for studying evolution in wild populations. Here, I utilize next- generation sequencing (NGS) technologies to approach mechanistic questions about the rapid adaptation to novel environments exhibited by invaders. Bioinformatics analyses of NGS data currently lack standardized parameter choices and there is growing concern that these choices may affect the evaluation of population structure. To address this issue, I analyze simulated GBS datasets and empirical RAD datasets of varying structure using multiple sets of filtering parameter choices. I demonstrate that such choices indeed affect the results, particularly in datasets with low structure; I also identify some common pitfalls to avoid when analyzing and publishing NGS datasets. Bearing these in mind, I approach questions about population structure during invasion by re-analyzing a RADSeq dataset on invasive Australian cane toads (Rhinella marina) in conjunction with our RNA-Seq dataset on this species. Since their introduction, cane toads have dispersed across northern Australia, displaying adaptive variation that allows them to surmount climatic challenges. RNA-Seq data provide transcriptome-wide coverage of single nucleotide polymorphisms (SNPs), allowing us to test whether cane toads represent a genetic paradox of invasion, and to identify loci putatively under selection by climate. Immune function may also change during invasion due to differences in pathogen-mediated selection; to test the assumptions of the enemy release hypothesis, I use differential gene expression analysis, which provides a snapshot of immune pathways found in spleen transcriptomes. I separately test specific responses to immune activation using field assays. Together, this body of work informs methodology, demonstrates the importance of selection in shaping invasion, and furthers our understanding of immune regulation in invaders.

11 CHAPTER 1: Introduction

Parts of this chapter are published in IJP:PAW: Selechnik D, Rollins LA, Brown GP, Kelehear C, Shine R. 2016. The things they carried: the pathogenic effects of old and new parasites following the intercontinental invasion of the Australian cane toad (Rhinella marina). International Journal for Parasitology: Parasites and Wildlife, 6(3): 375-385.

1. Invasion

Invasive species have received much attention due to their negative impacts on the environment and economy, but also offer unique opportunities to study rapid evolution in wild populations.

Several questions regarding the challenges imposed on invaders (sudden introduction to new areas) have yet to be fully answered, including the genetic paradox of invasion and the enemy release hypothesis (ERH).

The genetic paradox of invasion (Allendorf, 2003) describes a phenomenon that challenges widespread beliefs about the relationship between genetic diversity and adaptive potential. Despite the fact that populations with low genetic diversity have been shown to suffer declines due to inbreeding depression and the associated reduction of individual fitness

(Blomqvist et al., 2010; Madsen et al., 1999; Westemeier et al., 1998), and that invasive populations are thought to undergo genetic bottlenecks (due to the translocation of a small number of founders from their native range to an introduced range (Allendorf, 2003; Barrett &

Kohn, 1991), invasive species are characterized by their ability to establish and spread in their introduced ranges. Invasion success is commonly linked to rapid evolution, including adaptation to novel environmental conditions over short timescales (Franks & Munshi-South, 2014; Gao et al., 2018; Leydet et al., 2018; Whitney & Gabler, 2008), which demonstrates that adaptive potential is not necessarily dependent on high genetic diversity, though the data that support this assertion come from neutral loci (Rollins et al., 2013). However, it has recently been suggested

12 that some invasive populations may not actually suffer a reduction in genetic diversity during introduction, and others may not face novel adaptive challenges in their introduced ranges

(Estoup et al., 2016). Furthermore, inadequate estimation of genetic diversity (i.e. too few markers) may lead to inferences about genetic bottlenecks even if they have not occurred (Estoup et al., 2016). Thus, invasions may not be as paradoxical as originally thought, and require further study with appropriate datasets (such as genome-wide data).

The ERH (Colautti et al., 2004) predicts that when organisms are translocated to new areas, they may escape from co-evolved competitors, predators, parasites, and pathogens. The low number of host individuals transferred to the introduced range diminishes the probability of native pathogens/parasites being represented (Lewicki et al., 2014), and pathogens/parasites that do accompany the invasive host often face barriers to transmission such as low host density

(Arneberg et al., 1998; Blakeslee et al., 2012; MacLeod et al., 2010) and a lack of vectors or intermediate hosts needed to complete their life cycles in the introduced range (Blakeslee et al.,

2012; Lewicki et al., 2014). For “enemy release” to be realized, several conditions must be met.

First, co-evolved pathogens and parasites (specialized to the host in its native range) must be absent from the introduced range (Keane & Crawley, 2002; Liu & Stiling, 2006; Prenter et al.,

2004). Second, host-switching of pathogens and parasites from native taxa in the introduced range to the invasive host should be uncommon (Keane & Crawley, 2002; Liu & Stiling, 2006;

Prenter et al., 2004). Third, enemies in the introduced range should be less pathogenic to the invasive host than to native taxa (Keane & Crawley, 2002; Liu & Stiling, 2006; Prenter et al.,

2004). In these instances, enemy release may enable the invasive species to thrive in its introduced range (Colautti et al., 2004). Because of this, Lee & Klasing (2004) predict that invaders may down-regulate powerful immune responses such as systemic due to a

13 decreased need (Cornet et al., 2016; Lee & Klasing, 2004; Martin et al., 2010). Such immune responses are risky due to the energetic costs (the reduction of nutrients available for partitioning across tissues due to their use in mounting immune responses; Klasing & Leshchinsky, 1999) and to the potential for collateral damage (tissue injury due to the effects of the immune response; Martin et al., 2010). Nonetheless, loss of immunocompetence (the ability to mount a normal immune response after exposure to an antigen; Janeway et al., 2001) could render invaders susceptible to infection by novel pathogens and parasites in their introduced range

(Cornet et al., 2016; Lee & Klasing, 2004). Thus, invaders are predicted to exhibit lower investment in costly (but not all) immune responses than seen in their native ranges (Cornet et al., 2016; Lee & Klasing, 2004). However, support for the predicted consequences of ERH has been mixed, and further investigation is required.

2. Cane toad history

The native range of the cane toad (Rhinella marina) extends from southern Texas and western

Mexico to central Brazil (Acevedo et al., 2016; Zug & Zug, 1979). Cane toads were brought from Guyana (directly) and French Guiana (via Martinique) to Barbados in the mid 1800s to control pest beetles that were consuming farmed sugarcane (Easteal, 1981; Turvey, 2013), then translocated from Barbados to Puerto Rico (some directly, and some via Jamaica) in 1920

(Turvey, 2009). In 1932, 149 toads were brought from Puerto Rico to Oahu, Hawai’i in order to control cane beetles (Turvey 2009). Over the following two years, more than 100,000 individuals were distributed across the Hawai’ian Islands (Turvey, 2009). In 1935, 101 Hawai’ian cane toads were brought to Queensland, Australia and bred in captivity; their offspring were later released along the Queensland coast (Turvey, 2009).

14 Cane toads rapidly dispersed across the heterogeneous environments of tropical and subtropical Australia (Urban et al., 2008), expanding beyond their introduction sites in

Queensland (QLD) through the Northern Territory (NT) and into Western Australia (WA, which houses the current invasion front). In the process, they have had major ecological impacts on

Australian native fauna (Shine, 2010); the chemically distinctive toxins of R. marina (Hayes et al., 2009) are fatal if ingested by many Australian predators (which lack a history of evolutionary exposure to bufonid anurans, and thus to their toxins; Llewelyn et al., 2009; Llewelyn et al.,

2014; Shine, 2010). As a result, the spread of cane toads has caused massive declines in populations of anuran-eating predators both in tropical and temperate Australia (Brown et al.,

2011; Jolly et al., 2015, 2016; Letnic et al., 2008; Shine, 2010).

3. Cane toads as models for approaching questions about invasion

As an invader that has exhibited rapid evolution despite apparently low genetic diversity (Rollins et al., 2015), the Australian cane toad serves as a useful model to investigate the genetic paradox of invasion. The cane toad’s native range in Central and South America consists of tropical environments with relatively high mean annual rainfall warm temperatures (Hijmans, 2015). The

Hawai’ian Islands, the source of the Australian population, and coastal QLD, where Australian toads were first introduced, have similar environmental features to the native range. However, the current Australian range is quite heterogeneous in aridity; inland QLD, the NT, and WA are all much hotter and drier than coastal QLD (Bureau of Meteorology, 2018). Heritable phenotypic differences in behaviour (Gruber et al., 2017), thermal performance (Kosmala et al., 2018), morphology (Hudson et al., 2016; Hudson et al., 2018), and immune function (Brown et al.,

2015c) have been observed in cane toads at different localities across the Australian range.

15 Nonetheless, surveys have reported low levels of genetic diversity, based on microsatellites

(Leblois et al., 2000) and MHC (Lillie et al., 2014), and only one mitochondrial haplotype (ND3 gene) in 31 individuals sequenced from Hawai’i and Australia (Slade & Moritz, 1998). Still, these surveys may have sampled too few markers to conclusively determine that overall genetic diversity is low.

Due to the abundance of ecological data on cane toad pathogens/parasites, toads may also serve as a useful system to study the ERH. Consistent with the ERH, many species of bacteria and fungi (Speare, 1990), viruses (Hyatt et al., 2000; Zupanovic et al., 1998), protozoan

(Delvinquier & Freeland, 1988b), and metazoan (Speare, 1990) parasites of cane toads seem to have been left behind in the native range. A major parasite (lungworm Rhabdias pseudosphaerocephala) from the native range that infects toad populations in the Australian range core is absent from toads at the invasion front (Phillips et al., 2010). Conversely, the

Australian soil bacterium Brucella (Ochrobactrum) anthropi causes spinal spondylosis in toads primarily at the invasion front (Brown et al., 2007), which may represent a novel infection that forces invaders to remain immunocompetent. Furthermore, Rhimavirus A has only been detected in transcriptomes of toads from areas that have been colonized relatively recently (Russo et al.,

2018). Studies report various effects of invasion history on toad immunity (Brown et al., 2015c;

Brown & Shine, 2014).

4. Novel methods for approaching questions about invasion

Novel technologies, such as next-generation sequencing (NGS), are making it possible to perform genetic analyses on wild populations (Ellegren, 2014) without fully sequenced genomes

(Shafer et al., 2017). This has enhanced our ability to study invasive species (many of which lack

16 a fully sequenced genome) by allowing us to approach evolutionary questions from a genome- wide perspective. However, when navigating the necessary bioinformatics pipelines, users must choose from a range of possible values for several parameters with which to filter their data.

Results such as the estimation of genetic diversity and population structure are affected by these parameter choices (Shafer et al., 2017). Attention is increasingly being directed towards the effects of thresholds such as missing data tolerance (MDT; the highest percentage of individuals that a can be absent from or poorly sequenced in without that locus being removed from the dataset; Huang & Knowles, 2016) and minimum minor allele frequency (min MAF; Linck &

Battey, 2017). Selection of filtering thresholds may affect ecological and evolutionary interpretations about invasion. Although it is known that filtering thresholds affect results, it is unclear how the magnitude of their effects changes depending on the actual levels of structure of the dataset. My thesis focuses primarily on utilizing novel methods to address existing questions about invasion, and exploring the best practices with which to use these methods.

5. Chapter Overview

Chapter 2 examines the interaction between inherent levels of population structure and filtering parameter choices using both simulated and empirical NGS datasets. We demonstrate that low population structure is only detectable with stringent parameter choices, but high population structure is detectable regardless of parameter choices; thus, decisions involving experimental design and bioinformatics pipelines should be sensitive to the inherent properties of each dataset.

Chapter 3 aims to characterize population structure and genetic diversity in native and invasive cane toads using NGS datasets. We find that although invasive populations have

17 become highly differentiated from native populations, genetic diversity may not have been lost during the invasion. Interesting, invasive cane toad populations separate genetically along a climatic barrier (with loci putatively under selection involved in aridity tolerance), rather than invasion front populations forming a separate group.

Chapter 4 explores the predicted consequences of the ERH by characterizing differential gene expression. We demonstrate that immune in invasive cane toad spleen tissue do not behave as predicted by the ERH, suggesting that other evolutionary forces can obscure the effects of enemy release in wild populations, or that ERH is not supported by empirical data.

Chapter 5 investigates whether there is an effect of invasion history on phenotypic responses to immune activation. We find no differences at the population level (and thus provide no support for the predicted consequences of the ERH), but validate LPS injection as a method of triggering immune responses in cane toads (and likely other amphibians).

Literature Cited

Acevedo A.A., Lampo M., Cipriani R. (2016) The cane or marine toad, Rhinella marina (Anura,

Bufonidae): two genetically and morphologically distinct species. Zootaxa 4103, 574-

586.

Allendorf F.W. (2003) Introduction: Population Biology, Evolution, and Control of Invasive

Species. Conservation Biology 17, 24 - 30.

Arneberg P., Skorping A., Grenfell B., Read A.F. (1998) Host densities. Proc. R. Soc. Lond. 265,

1283 – 1289.

Barrett S.C.H., Kohn J.R. (1991) Genetics and conservation of rare plants. In: Genetic and

evolutionary consequences of small population size in plants: implications for

18 conservation eds. Falk DA, Holsinger KE), pp. 3 - 30. Oxford University Press, New

York.

Blakeslee A.M.H., Altman I., Miller A.W., et al. (2012) Parasites and invasions: a biogeographic

examination of parasites and hosts in native and introduced ranges. J. Biogeogr. 39, 609 -

622.

Blomqvist D., Pauliny A., Larsson M., Flodin L.A. (2010) Trapped in the extinction vortex?

Strong genetic effects in a declining vertebrate population. BMC Evolutionary Biology

10, 33.

Brown G.P., Phillips B.K., Shine R. (2011) The ecological impact of invasive cane toads on

tropical snakes: Field data do not support laboratory‐based predictions. Ecology 92, 422 -

431.

Brown G.P., Phillips B.L., Dubey S., Shine R. (2015c) Invader immunology: invasion history

alters immune system function in cane toads (Rhinella marina) in tropical Australia. Ecol

Lett 18, 57-65.

Brown G.P., Shilton C., Phillips B.L., Shine R. (2007) Invasion, stress, and spinal arthritis in

cane toads. Proceedings of the National Academy of Science USA 104, 17698 - 17700.

Brown G.P., Shine R. (2014) Immune Response Varies with Rate of Dispersal in Invasive Cane

Toads (Rhinella marina). PLOS ONE 9, 1 - 11.

Bureau of Meteorology A.G. (2018) Climate Data Online. Commonwealth of Australia.

http://www.bom.gov.au/

Colautti R.I., Ricciardi A., Grigorovich I.A., MacIsaac H.J. (2004) Is invasion success explained

by the enemy release hypothesis? Ecology Letters 7, 721-733.

19 Cornet S., Brouat C., Diagne C., Charbonnel N. (2016) Eco-immunology and bioinvasion:

revisiting the evolution of increased competitive ability hypotheses. Evol Appl 9, 952-

962.

Delvinquier B.L.J., Freeland W.J. (1988b) Protozoan Parasites of the Cane Toad, Bufo marinus,

in Australia. Australian Journal of Zoology 36, 301 - 316.

Easteal S. (1981) The history of introductions of Bufo marinus; a natural experiment in

evolution. Biological Journal of the Linnean Society 16, 93.

Ellegren H. (2014) Genome sequencing and population genomics in non-model organisms.

Trends in Ecology & Evolution 29, 51 - 63.

Estoup A., Ravigné V., Hufbauer R., et al. (2016) Is There a Genetic Paradox of Biological

Invasion? Annual Review of Ecology, Evolution, and Systematics 47, 51-72.

Franks S.J., Munshi-South J. (2014) Go forth, evolve and prosper: the genetic basis of adaptive

evolution in an invasive species. Molecular Ecology 23, 2137-2140.

Gao Y., Li S., Zhan A. (2018) Genome-wide single nucleotide polymorphisms (SNPs) for a

model invasive ascidian Botryllus schlosseri. Genetica 146, 227-234.

Gruber J., Brown G.P., Whiting M.J., Shine R. (2017) Is the behavioural divergence between

range-core and range-edge populations of cane toads (Rhinella marina) due to

evolutionary change or developmental plasticity? Royal Society Open Science 4.

Hayes R.A., Crossland M.R., Hagman M., Capon R.J., Shine R. (2009) Ontogenetic variation in

the chemical defenses of cane toads (Bufo marinus): toxin profiles and effects on

predators. J Chem Ecol 35, 391-399.

Hijmans R.J. (2015) raster: Geographic data analysis and modeling.

20 Huang H., Knowles L.L. (2016) Unforeseen Consequences of Excluding Missing Data from

Next-Generation Sequences: Simulation Study of RAD Sequences. Syst Biol 65, 357-365.

Hudson C.M., Brown G.P., Shine R. (2016) It is lonely at the front: contrasting evolutionary

trajectories in male and female invaders. R Soc Open Sci 3, 160687.

Hudson C.M., Brown G.P., Stuart K., Shine R. (2018) Sexual and geographical divergence in

head widths of invasive cane toads, Rhinella marina (Anura: Bufonidae), is driven by

both rapid evolution and plasticity. Biological Journal of the Linnean Society 124, 188-

199.

Hyatt A.D., Gould A.R., Zupanovic Z., et al. (2000) Comparative studies of piscine and

amphibian iridoviruses. Arch. Virol. 145, 301 - 331.

Janeway C.A., Travers P., Walport M. (2001) Immunobiology: The Immune System in Health

and Disease. In: Immunobiology. Garland Science, New York.

Jolly C.J., Shine R., Greenlees M.J. (2015) The impact of invasive cane toads on native wildlife

in southern Australia. Ecol. Evol. 5, 3879-3894.

Jolly C.J., Shine R., Greenlees M.J. (2016) The impacts of a toxic invasive prey species (the cane

toad, Rhinella marina) on a vulnerable predator (the lace monitor, Varanus varius).

Biological Invasions 18, 1499 - 1509.

Keane R.M., Crawley M.J. (2002) Exotic plant invasions and the enemy release hypothesis.

Trends Ecol. Evol. 17, 164 - 170.

Klasing K.C., Leshchinsky T.V. (1999) Functions, costs, and benefits of the immune system

during development and growth. International Ornitology Congress, Proceedings 69,

2817 - 2832.

21 Kosmala G.K., Brown G.P., Christian K.A., Hudson C.M., Shine R. (2018) The thermal

dependency of locomotor performance evolves rapidly within an invasive species

Ecology and Evolution 8, 4403-4408.

Leblois R., Rousset F., Tikel D., Moritz C., Estoup A. (2000) Absence of evidence for isolation

by distance in an expanding cane toad (Bufo marinus) population: an individual-based

analysis of microsatellite genotypes. Molecular Ecology 9, 1905 - 1909.

Lee K.A., Klasing K.C. (2004) A role for immunology in invasion biology. Trends Ecol Evol 19,

523-529.

Letnic M., Webb J., Shine R. (2008) Invasive cane toads (Bufo marinus) cause mass mortality of

freshwater crocodiles (Crocodylus johnstoni) in tropical Australia. Biological

Conservation 141, 1773 - 1782.

Lewicki K.E., Huyvaert K.P., Piaggio A.J., Diller L.V., Franklin A.B. (2014) Effects of barred

owl (Strix varia) range expansion on Haemoproteus parasite assemblage dynamics and

transmission in barred and northern spotted owls (Strix occidentalis caurina). Biological

Invasions 17, 1713-1727.

Leydet K.P., Grupstra C.G.B., Coma R., Ribes M., Hellberg M.E. (2018) Host-targeted RAD-

Seq reveals genetic changes in the coral Oculina patagonica associated with range

expansion along the Spanish Mediterranean coast. Mol Ecol 27, 2529-2543.

Lillie M., Shine R., Belov K. (2014) Characterisation of major histocompatibility complex class I

in the Australian cane toad, Rhinella marina. PLOS ONE 9, e102824.

Linck E.B., Battey C.J. (2017) Minor allele frequency thresholds strongly affect population

structure inference with genomic datasets. bioRxiv.

22 Liu H., Stiling P. (2006) Testing the enemy release hypothesis: a review and meta-analysis.

Biological Invasions 8, 1535-1545.

Llewelyn J., Schwarzkopf L., Alford R., Shine R. (2009) Something different for dinner?

Responses of a native Australian predator (the keelback snake) to an invasive prey

species (the cane toad). Biological Invasions 12, 1045 - 1051.

Llewelyn J., Schwarzkopf L., Phillips B.L., Shine R. (2014) After the crash: How do predators

adjust following the invasion of a novel toxic prey type? Aust. Ecol. 39, 190 - 197.

MacLeod C.J., Paterson A.M., Tompkins D.M., Duncan R.P. (2010) Parasites lost - do invaders

miss the boat or drown on arrival? Ecology Letters 13, 516-527.

Madsen T., Shine R., Olsson M., Wittzell H. (1999) Restoration of an inbred adder population.

Nature 402, 34 - 35.

Martin L.B., Hopkins W.A., Mydlarz L.D., Rohr J.R. (2010) The effects of anthropogenic global

changes on immune functions and disease resistance. Ann N Y Acad Sci 1195, 129-148.

Phillips B.L., Kelehear C., Pizzatto L. (2010) Parasites and pathogens lag behind their host

during periods of host range advance. Ecology 91, 872 - 881.

Prenter J., Macneil C., Dick J.T., Dunn A.M. (2004) Roles of parasites in animal invasions.

Trends Ecol Evol 19, 385-390.

Rollins L.A., Moles A.T., Lam S. (2013) High genetic diversity is not essential for successful

introduction. Ecology and Evolution 3, 4501 - 4517.

Rollins L.A., Richardson M.F., Shine R. (2015) A genetic perspective on rapid evolution in cane

toads (Rhinella marina). Mol Ecol 24, 2264-2276.

23 Russo A.G., Eden J.S., Enosi Tuipulotu D., et al. (2018) Viral discovery in the invasive

Australian cane toad (Rhinella marina) using metatranscriptomic and genomic

approaches. Journal of Virology.

Shafer A.B.A., Peart C.R., Tusso S., et al. (2017) Bioinformatic processing of RAD-seq data

dramatically impacts downstream population genetic inference. Methods in Ecology and

Evolution 8, 907-917.

Shine R. (2010) The ecological impact of invasive cane toads (Bufo marinus) in Australia. Q.

Rev. Biol. 85, 253 - 291.

Slade R.W., Moritz C. (1998) Phylogeography of Bufo marinus from its natural and introduced

ranges. Proceedings of the Royal Society of London. Series B: Biological Sciences 265,

769.

Speare R. (1990) A Review of the Diseases of the Cane Toad, Bufo marinus, With Comments on

Biological-Control. Australian Wildlife Research 17, 387 - 410.

Turvey N. (2009) A Toad's Tale. Hot Topics from the Tropics 1, 1 - 10.

Turvey N. (2013) Cane toads: A tale of sugar, politics and flawed science Sydney University

Press, University of Sydney, Australia.

Urban M.C., Phillips B.L., Skelly D.K., Shine R. (2008) A toad more traveled: the heterogeneous

invasion dynamics of cane toads in Australia. The American Naturalist 171, E134 - E148.

Westemeier R.L., Brawn J.D., Simpson S.A., et al. (1998) Tracking the long-term decline and

recovery of an isolated population. Science 282, 1695-1698.

Whitney K.D., Gabler C.A. (2008) Rapid evolution in introduced species, 'invasive traits' and

recipient communities: challenges for predicting invasive potential. Diversity and

Distributions 14, 569-580.

24 Zug G.R., Zug P.B. (1979) The marine toad, Bufo marinus: a natural history resume of native

populations Smithsonian Institution Press, Washington, D.C.

Zupanovic Z., Hyatt A.D., Green B., et al. (1998) Giant toads Bufo marinus in Australia and

Venezuela have antibodies against 'ranaviruses'. Dis. Aquat. Organ. 32, 1-8.

25 CHAPTER 2: Impacts of filtering parameter choices on inferences of population structure are dependent on inherent levels of population differentiation

This chapter is in preparation for submission with the following co-authors: Richardson MF, Martin M, Cardilini A, Sherman C, Hess M, Hess A, Dodds K, Shine R, Rollins LA

26 Abstract

As novel technologies enhance our ability to study population genetics, understanding the interactions between our datasets and experimental design decisions is increasingly important.

Choices in filtering parameter thresholds, such as missing data tolerance (MDT) and minimum minor allele frequency (min MAF) are known to affect inferences of population structure in reduced representation sequencing (RRS) studies. However, it is unclear to what extent these effects vary across datasets. Here, we performed a literature review on filtering choices and levels of genetic differentiation across RRS studies on wild populations, and find that approaches to these data are not standardised. We predicted that choices in filtering thresholds would have the greatest impact when analysing datasets with low levels of genetic differentiation between populations. To test this prediction, we produced seven simulated RRS datasets with varying levels of population structure, and then analysed them using four different combinations of MDT and minimum MAF. We also performed the same analysis on two empirical RRS datasets (one with low population structure, the other with high). Our simulated and empirical results suggest that the effects of filtering choices indeed vary based on inherent levels of differentiation. As a result, experimental design and analysis choices should consider the actual levels of population structure in each specific dataset. Based on our literature review and analyses, we offer recommendations for best practices when analysing RRS data.

Keywords: reduced representation sequencing; bioinformatics; filtering; missing data; minor allele frequency

27 Introduction

As novel technologies expand the field of molecular ecology, more research is needed to fully understand how experimental design and analysis decisions affect inferences of population structure. Restriction site-associated DNA sequencing (RADSeq) and genotyping-by-sequencing

(GBS) are examples of next-generation sequencing (NGS) technologies referred to as reduced representation sequencing (RRS), in which restriction enzymes and ligated primers are used to target sequencing to specific sites spread throughout a genome (Davey & Blaxter, 2011).

Combined with specialized downstream bioinformatics pipelines, these tools have facilitated genome-wide surveys in wild populations (Ellegren, 2014), including those of species without fully sequenced genomes (Shafer et al., 2017). RRS allows for more complete quantification of genetic diversity and differentiation, and thus for more accurate inferences in population genetics and phylogenetics (Parchman et al., 2018), than do previous methods. As a result, RRS approaches have become popular in studies on conservation, invasion, and general evolutionary questions (Parchman et al., 2018; Rius et al., 2015).

Within RRS, there are several techniques that vary the type and number of restriction enzymes used, adapter ligation methods, size selection, barcoding, and type of sequence data provided (Andrews et al., 2016). The laboratory practices best-suited for a given project depend on factors such as the genetic characteristics of the organism and the questions being asked

(Andrews et al., 2016). Three such characteristics are genome size, restriction site density, and levels of linkage disequilibrium (LD; Lowry et al., 2017). It has been suggested that RRS approaches are useful for assessing neutral genetic variation and population structure, but are less well-suited than other NGS technologies for studying adaptation due to their limited sampling method (Lowry et al., 2017). This is because RRS does not provide information on all SNPs

28 across a genome or transcriptome, but rather a subset of SNPs found within randomly produced

‘tags’ (regions of the genome between cleavage sites of selected restriction enzymes; Davey &

Blaxter, 2011). Furthermore, because genomes are mostly composed of non-coding sequences,

SNPs identified by RRS are more likely to represent neutral variation (as opposed to variation under selection) than those identified by other NGS approaches such as RNA-Seq (Wang et al.,

2009). However, other studies have suggested that these techniques are still appropriate because laboratory choices (such as the type and number of restriction enzymes) can be tailored around marker density and levels of LD, thereby allowing for greater numbers of SNPs to be sampled if necessary (Catchen et al., 2017).

Experimental design choices are not limited to the laboratory; when performing bioinformatics analyses, users can filter data based on several parameters, each of which has a range of possible values. To improve genotyping and SNP calling accuracy, users typically filter based on sequencing quality scores (of base calls), read depth, linkage disequilibrium patterns, and strand bias (Nielsen et al., 2011). Previous work has identified that varying some filtering parameters can influence estimations of the levels of genetic diversity and population structure, as well as the number of SNPs detected (Shafer et al., 2017). Two such parameters are the thresholds for missing data tolerance (MDT; the highest percentage of individuals in which the genotype for a locus is allowed to be missing, above which the locus will be filtered from the dataset) and minimum minor allele frequency (min MAF; the rate of occurrence of the less common allele in a biallelic SNP).

Choosing an MDT threshold is challenging because retaining loci that are missing from many individuals may preclude inferences from being drawn due to uninformative data (Arnold et al., 2013), but removing these loci may also skew interpretations. Tags with high mutation

29 rates are the most likely to have mutations within restriction cut sites, which may prevent them from being sequenced (Arnold et al., 2013; Huang & Knowles, 2016). Thus, all loci in such tags may be missing from individuals that have cut site mutations, and filtering them out may lead to underestimates of genetic diversity and differentiation (Huang & Knowles, 2016). Furthermore, allowing a higher MDT threshold may provide more power for population assignment due to the inclusion of more loci (Chattopadhyay et al., 2014). Choosing a min MAF threshold also presents issues: rare minor alleles may suggest population expansion, and removing them may thus affect the accuracy of identifying population limits (Linck & Battey, 2017). Researchers remain unsure of the best practices in selecting values for these parameters, other than to perform analyses multiple times, including or excluding samples or loci with high rates of missing data to see if results are consistent (Grünwald et al., 2017). In cases where the results appear to differ based on the filtering steps, selection of the appropriate filtering thresholds may be important for accurate ecological and evolutionary interpretations, but it is difficult to predict which answers are correct.

Recommendations for bioinformatics analysis of RRS data are still developing, but several suggestions have been made. Many potential issues with datasets can arise during library preparation or locus reconstruction; these should be identified through quality control measures and mitigated with the appropriate filtering steps (O'Leary et al., 2018). Basic per locus statistics provide information on potential issues such as spurious allele calls, private alleles, fixed alleles, and missing data (Grünwald et al., 2017). When characterizing population structure, results should be cross-checked by performing both model-based and nonparametric methods; model- based approaches have traditionally been used, but are more error-prone than nonparametric approaches (Linck & Battey, 2017). One such error is the confounding of population structure

30 inference by SNPs found in only one individual (singletons), which should be filtered out (Linck

& Battey, 2017). Detailed descriptions of methods (including scripts and metadata) should be attached to papers, even if they must be included as supplementary material (O'Leary et al.,

2018). Raw sequence data should be made accessible to readers in addition to filtered data

(O'Leary et al., 2018). These steps must be taken to ensure that results are easily reproducible

(Gilbert et al., 2012).

Although best practice recommendations for RRS bioinformatics are emerging, it is unclear how often they are followed. Furthermore, additional recommendations may be necessary; we know that filtering thresholds can affect results, but we do not know if this is true under all conditions, or how the magnitude of these effects may change depending on the intrinsic properties of a dataset. Here, we investigate how much of the literature complies with the current set of best practice recommendations, and test the effects of filtering thresholds

(MDT and minimum MAF) on datasets that vary in their inherent level of genetic differentiation between populations. To address this, we: (1) performed a literature review documenting the methods (particularly filtering choices) and levels of genetic differentiation and diversity reported among 71 RRS studies on wild populations, and (2) analyzed seven simulated GBS datasets and two empirical datasets with varying levels of population structure. We predicted (1) a high degree of variability in experimental design choices made in RRS studies, and (2) that the effects of filtering choices would be most pronounced in the dataset with the lowest population structure, and would decline in severity in datasets with increasing population structure.

Materials and Methods Literature review of RRS studies on wild populations

31 We compiled RRS studies by performing a Topic Search (TS) on Web of Science, which searches for the provided terms in titles, abstracts, and keywords within all available records.

Our search term was “TS=((“reduced representation sequencing” OR “RADseq” OR “ddRAD”

OR “GBS” OR “epi-GBS” OR “genotyping by sequencing” AND “population gen*”)” between the years of 2013 to 2018. We manually filtered the results of this search to find relevant studies, eliminating those that did not assess genetic structure in wild populations. This yielded a total of

71 papers, from which we extracted information on study taxa, sample size, sequencing methods, bioinformatics pipelines, filtering parameter choices, number of resulting reads and SNPs, number of genetic groups, measures of genetic differentiation and diversity, and availability of supplemental information and raw data.

Simulation of GBS datasets with varied population structure

We used the method and Julia code (Bezanson et al., 2017) of Hess et al. (2018) to create seven

GBS datasets, manipulating the population history events of each to produce varying levels of population structure (Figure 1A, Table S1). All datasets were simulated to comprise four subpopulations of increasing relatedness (Figure 1B-C). For our pilot study, each dataset contained approximately 2,000 SNPs and 5.338 mean coverage across one of 100 centimorgans (cM).

32

Figure 1. (A) Population history events of seven simulated datasets with varying degrees of structure. (B) Cladogram showing three splits occurring in the population history of all simulated datasets. (C) Expected population structure results of all simulated datasets at K=2,3,4.

Acquisition of empirical RAD datasets with high and low structure

To explore whether the findings of our simulation analyses extend to data obtained from wild populations, we downloaded raw reads of two publicly available ddRAD datasets from the NCBI

Sequence Read Archive (SRA). The first dataset (Trumbo et al., 2016), which we call the

“empirical low” dataset, demonstrates relatively low population structure in Australian cane toads (Rhinella marina) and was retrieved via accession PRJNA328156. We subset this dataset to isolate two populations: one in eastern Queensland (QLD; N=179), and the other in the western Northern Territory (NT; N=441). The second dataset (Bell et al., 2015), which we call the “empirical high” dataset, demonstrates relatively high population structure in African reed frogs (Hyperolius molleri) and was retrieved via accession PRJNA268025. We also subset this

33 dataset to isolate two populations: one on the island of Príncipe (N=17), and the other on the island of São Tomé (NT; N=54).

Processing of reads, SNP calling, and filtering

We used functions from Stacks v2.0 (Catchen et al., 2013) to process all RRS data. First, we used the process_radtags function to remove low quality reads from the FASTQ files using the program’s default parameters. Next, we used the denovo_map pipeline to perform an assembly for each sample, align matching DNA regions across samples (called ‘stacks’), and call SNPs using a maximum likelihood framework. We then filtered the results using the populations function, using one random SNP per locus and four combinations of parameter choices (min

MAF=0.05, MDT = 0.5; min MAF = 0.05, MDT = 0.2; min MAF = 0.01, MDT = 0.5; min MAF

= 0.01, MDT = 0.2). The results were written to a VCF file and a STRUCTURE file (for use with fastStructure).

Evaluation of genetic differentiation and diversity

To quantify levels of genetic differentiation and diversity, we computed basic statistics in the hierfstat (Goudet, 2005) and diveRsity (Keenan et al., 2013) packages in R (Team, 2016). We first used PGDSpider v2.1.1.3 (Lischer & Excoffier, 2012) to convert our VCF to the FSTAT and GENEPOP formats, which are readable to hierfstat and diveRsity, respectively. We then used these packages to calculate measures of genetic differentiation, including global FST (mean across all loci) and pairwise FST (between the four simulated genetic groups), and genetic diversity, such as expected heterozygosity (He) and rarefied allelic richness (AR). This was performed for each dataset. We also computed 95% confidence intervals (CIs) for all pairwise

34 FST values using the bootstrapping method (number of bootstraps = 100 across loci) performed by the StAMPP package (Pembleton et al., 2013).

Inference of population structure

We used fastStructure (Raj et al., 2014) to infer population structure from both the simulated and empirical datasets using a variational Bayesian framework for calculating posterior distributions, and to identify the number of genetic clusters in our dataset (K) using heuristic scores (Raj et al.,

2014). We ran fastStructure ten times (K= 1 to 10), took the resulting meanQ files and plotted them using the pophelper package (Francis, 2016) in R (Team, 2016).

To see if inferences of population structure were consistent across approaches (model- based versus non-parametric methods), we also used adegenet (Jombart & Ahmed, 2011) to perform discriminant analysis of principal components (DAPC) on each dataset. DAPC is a multivariate approach that identifies the number of genetic clusters using K-means of principal components and a Bayesian framework (Jombart & Ahmed, 2011).

Results Literature review: Parameter choices made in RRS studies on wild populations

Our literature search (Table 1; full table available in Supplemental File 1) revealed a high degree of variation across studies in thresholds of MDT and minimum MAF, as well as those in Phred scores and read depths. We found that only 20% of studies ran multiple combinations of filtering parameter choices to check the consistency of the results, 4% cited recommendations from other papers, and 76% did not provide justification for their choices. Although many different MDT values were selected across the 71 studies, the most common choice was to allow 11-30% missing data (Figure 2A). Selections of min MAF were much more consistent, and 0.05 was the

35 most common choice (Figure 2B). There was a wide range of sample sizes; approximately 44% of studies removed samples for reasons such as not passing quality filters, whereas 56% of studies retained all original samples. Approximately 50% of studies assessed population structure using both model-based (STRUCTURE, ADMIXTURE) and non-parametric (PCA, DAPC) methods. Reporting of measures such as global and pairwise FST, FIS, and expected and observed heterozygosity (He and Ho, respectively), was inconsistent (Figure 3). In only 66% of studies, an accession number was provided linked to databases such as the SRA or Dryad, where raw sequence files or downstream files could be downloaded.

Table 1. Choices made in bioinformatics analyses of RRS datasets across 71 papers from 2013-2018. A full table with additional information on each paper such as taxon, sequencing method, bioinformatics pipeline, other parameter choices, number of genetic clusters, measures of genetic differentiation and diversity, and availability of raw data is available in Supplemental File 1. Reference Final Minimum minor allele Missing data threshold sample frequency size Savary et al., 2018 262 0.1 0 Zlonis & Gross, 2018 80 0.01 0.2 Zhao et al., 2018 0.25 (in at most 4 of 12 219 0.05 populations) Perez-Portela et al., 2018 162 - 0.16 Krohn et al., 2018 79 0.05 0.1 Sandoval-Castillo et al., 2018 349 0.03 0.2 Maas et al., 2018 125 - 0.3 Xuereb et al., 2018 0.3 (from each 717 0.01 location) Collins et al., 2018 61 - 0.3 Jackson et al., 2018 970 0.01 0.2 McCartney-Melstad et al., 2018 93 - 0.8 Eimanifar et al., 2018 474 0.05 0.15 Cai et al., 2018 60 0.1 0.5 Van Wyngaarden et al., 2017 245 0.05 0.25 Carreras et al., 2017 176 0.01 0.3 Vega et al., 2017 153 - 0.1 Nørgaard et al., 2017 174 0.05 0.15 Xu et al., 2017 59 0.1 0.5 Beheregaray et al., 2017 105 - 0.1 Martinez et al., 2017 0.3 (from each 110 0.05 location)

36 Johansson et al., 2017 104 - - Drury et al., 2017 332 - 0.5 Forsström et al., 2017 36 0.05 0.3 Metivier et al., 2017 60 0.05 0 Roland et al., 2017 0.25 (in each 125 - population) Wosula et al., 2017 72 0.05 0.8 Anderson et al., 2017 128 - 0.9 Kotsakiozi et al., 2017 86 0.05 0.3 Munoz et al., 2016 53 - 0.1 Parchman et al., 2016 231 0.03 0.02 Sekino et al., 2016 0.3 (and no site- coverage depth <20 in 60 0.05 each individual) Clucas et al., 2016 0.2 (and cannot be absent from any of the 64 0.01 four colonies) Shultz et al., 2016 0.25 (in half of the 113 0.1 populations) Brauer et al., 2016 0.3 (and cannot be absent from more than 263 0.05 30% of populations) Peters et al., 2016 117 - 0.1 Jahner et al., 2016 140 0.05 0.1 Gloria-Soria et al., 2016 48 0.05 0.2 Monzon et al., 2016 90 0.01 - Nicotra et al., 2016 177 0.05 0.1 Munshi-South et al., 2016 0.5 (and cannot be absent from more than 191 - 1/23 sampling sites) Kjeldsen et al., 2016 171 0.02 0.94 Lozier et al., 2016 41 0.05 0.2 Lah et al., 2016 0.2 (and cannot be absent from more than 44 - 1 out of 7 sub-regions) Pellegrino et al., 2016 0.94 (and the major allele identified in no less than 50% of the 53 0.05 samples) Stockwell et al., 2016 81 0.1 0.1 Sovic et al., 2016 134 - - Cahill & Levinton, 2016 72 0.01 0.25 Fernandez et al., 2016 112 - - Farrell et al., 2016 839 - - Bolton et al., 2016 251 0.05 0.2 Trumbo et al., 2016 1123 0.005 0.33 Nunez et al., 2015 120 0.05 0.2 Capblancq et al., 2015 174 0.05 0.4

37 Saenz-Agudelo et al., 2015 0.2 (and cannot be absent from more than 115 0.05 4 of 11 sites) Rasic et al., 2015 163 - 0.3 Hecht et al., 2015 1,956 0.01 0.2 Lavretsky et al., 2015 105 - - Deagle et al., 2015 124 0.075 0.2 Hamlin & Arnold, 2015 92 0.02 0.2 Pukk et al., 2015 76 - - Martin et al., 2015 - - 0.5 Bell et al., 2015 0.25 (and cannot be absent from any of the 123 0.2 three lineages) Leache et al., 2015 - - - Schield et al., 2015 42 - 0.5 Boehm et al., 2015 23 - 0.35 Mastretta-Yanes et al., 2015 81 - 0.20 Seeb et al., 2014 140 0.15 0.2 Martin & Feinstein, 2014 73 0.25 0.5 Larson et al., 2014 270 0.05 0.2 Lozier, 2014 45 0.05 0.15 White et al., 2013 281 - 0.2

38

Figure 2. Histogram showing frequency of choices in (A) missing data threshold (MDT) values (%) and (B) minimum minor allele frequency (min MAF) values when filtering reduced representation sequence (RRS) data. A literature review was performed on 71 RRS studies, and choices of MDT and min MAF were extracted from each study (Table 1; full table in Supplemental File 1).

39

Figure 3. Percentages of 71 reduced representation sequencing (RRS) studies that report each of five population genetics statistics. A literature review was performed, and reports of global FIS, global FST, pairwise FST, expected heterozygosity (HE), and observed heterozygosity (HO) were extracted from each study (Table 1; full table in Supplemental File 1).

Experimental determination of effects of filtering choices on inference of population structure

All simulated datasets encompassed four populations each, but varied in the levels of differentiation between these populations. To test the impacts of filtering on inferences of population structure, we filtered the datasets through every combination of two different MDT values (more stringent = 0.2, less stringent = 0.5) and two different min MAF values (more stringent = 0.05, less stringent = 0.01). When using the model-based fastStructure, detection of these four populations depended on the interaction between filtering choices and inherent levels of differentiation in the dataset (Figure 4). In datasets #1-4 (relatively low differentiation), the four populations were only detected when using the more stringent choice of MDT, and structure was otherwise unclear at K=2,3,4; choice of minimum MAF did not appear to affect the results.

In dataset #5 (relatively moderate differentiation), population structure was clearer (particularly

40 when using the more stringent choice of minimum MAF), but the four populations were still only detected when using the more stringent choice of MDT. In datasets #6-7 (relatively high differentiation), the four populations were detected regardless of filtering choices. The findings from our simulations were corroborated by those of the empirical datasets. In the empirical low dataset, which has inherent levels of genetic differentiation comparable to simulated datasets #2-

3, the two populations were only detected when using the stricter choice of MDT (Figure 5). In the empirical high dataset, which has inherent levels of genetic differentiation comparable to datasets #5-6, the two populations were detected regardless of filtering choices (Figure 5).

Figure 4. Genetic structure of four populations in each of seven simulated GBS datasets with varying levels of differentiation. Models were run four times with different combinations of filtering parameter choices (minimum minor allele frequency and missing data threshold). Results are shown for the model at K=2,3,4.

41

Figure 5. Genetic structure of two populations in each of two empirical ddRAD datasets with varying levels of differentiation. Models were run four times with different combinations of filtering parameter choices (minimum minor allele frequency and missing data threshold). Results are shown for the model at K=2.

When using the non-model-based DAPC, detection of the four simulated populations only depended on filtering choices in dataset #1, which had the lowest inherent level of population structure. In this dataset, the four populations were only detected when using the more stringent choice of MDT; otherwise, only three populations were identified (Figure S1).

However, in datasets #2-7, DAPC detected all four populations regardless of filtering choices

(Figure S1), with higher accuracy of assigning individuals to populations in datasets with greater levels of structure. Once again, the findings of the empirical datasets mirrored those of the simulated datasets. In both empirical datasets, the two populations were detected regardless of filtering choices (Figure S2).

42 Experimental determination of effects of filtering choices on calculations of genetic differentiation and diversity

To test the impacts of filtering on estimations of genetic differentiation and diversity, we filtered our simulated and empirical datasets through every combination of two different MDT values

(more stringent = 0.2, less stringent = 0.5) and two different min MAF values (more stringent =

0.05, less stringent = 0.01). Filtering choices did not affect calculations of global and pairwise

FST in any of the simulated datasets (Table 2, full results in Tables S2-S3) or empirical datasets

(Table 3, full results in Tables S4-S5). The more stringent choice of MDT (but not minimum

MAF) produced wider pairwise FST CIs, but because these CIs also became narrower as inherent genetic differentiation increased, these effects were less pronounced in the datasets with higher population structure (Tables S2 & S4). Calculations of AR were less consistent across filtering parameter choices, and were generally lower when using the more stringent choice of MDT (but not minimum MAF); this trend was seen in all simulated and empirical datasets (Table S3 & S5).

In the empirical low dataset, choice of minimum MAF affected which population had the higher estimation of AR; however, this was not seen in the empirical high dataset or any of the simulations, and differences in AR between populations were otherwise generally consistent across filtering choice combinations. Using the more stringent choices of MDT and minimum

MAF generally produced slightly higher values of He in all simulated and empirical datasets

(Tables S3 & S5). In both empirical datasets, there was a more pronounced difference in He between populations when using the more stringent choices of MDT and minimum MAF; however, filtering choices did not affect which population had the higher estimation of He. In the simulated datasets, this pattern was absent (#1,2,3,5) or very minor (#4,6,7).

43 Table 2. Global and pairwise FST in seven simulated GBS datasets with varying levels of population structure. These values were calculated four times with different combinations of filtering parameter choices, and the results were averaged across combinations. Pairwise FST values are displayed between the first two populations (of four total). The full set of global and pairwise FST values, as well as the allelic richness and expected heterozygosity of each population, are available in Tables S2-S3. Dataset Global FST Pairwise FST (Pop #1 vs #2) Mean  SD Mean  SD 1 0.01  0.00 0.01  0.00 2 0.04  0.00 0.03  0.00 3 0.09  0.01 0.04  0.00 4 0.16  0.02 0.11  0.01 5 0.34  0.02 0.23  0.01 6 0.66  0.01 0.55  0.01 7 0.81  0.00 0.72  0.00

Table 3. Global and pairwise FST in two empirical ddRAD datasets with varying levels of population structure. These values were calculated four times with different combinations of filtering parameter choices, and the results were averaged across combinations. Pairwise FST values are displayed between the two genetic clusters. The full set of global and pairwise FST values, as well as the allelic richness and expected heterozygosity of each population, are available in Tables S4-S5. Dataset Global FST Pairwise FST Mean  SD Mean  SD Empirical low 0.04  0.00 0.09  0.00 Empirical high 0.21  0.02 0.46  0.03

Discussion The results of our literature review demonstrate high variation across RRS studies in experimental choices made in the laboratory and during downstream bioinformatics analysis.

Only approximately 24% of the studies we reviewed shared the rationale for their filtering choices (e.g. trialling a range of options or referencing other studies). This is understandable because considerations for ‘best practices’ when analysing NGS data are still developing, and it is easy to overlook them when they are often not followed. However, unjustified choices of filtering thresholds may be dangerous because it is known that they can influence downstream results (Huang & Knowles, 2016; Linck & Battey, 2017).

We hypothesized that the extent to which filtering choices affect inference of population structure depends on the inherent levels of genetic differentiation in a dataset, with more highly

44 differentiated populations being less affected due to their more pronounced differences. To test this prediction, we created seven GBS datasets with varying levels of genetic differentiation, and then analyzed these datasets using every combination of two different MDT values (more stringent = 0.2, less stringent = 0.5) and two different min MAF values (more stringent = 0.05, less stringent = 0.01). The results of our simulations support our hypothesis; in the datasets with low population structure, stringency with MDT is required to detect separate populations (and these populations are undetectable regardless of minimum MAF choice unless the MDT threshold is strict). This is less necessary in datasets with higher differentiation, suggesting that stricter filtering of MDT is important for detection of finer scale population differentiation. In dataset #5, with intermediate to high levels of differentiation, we begin to see clearer results while using the less stringent MDT threshold, and the populations are more accurately detected when using the stricter minimum MAF threshold than when using the less strict choice. This result suggests that choice of minimum MAF affects the results in similar ways to MDT, but that the effects of minimum MAF are masked by the much stronger effects of MDT in datasets with very low levels of differentiation. At sufficiently high levels of differentiation to reduce the effects of MDT, however, we can see the effects of minimum MAF choice as well. Both filtering parameters are likely important for filtering out noise that obscures differences between populations in the dataset.

In dataset #7, which has the highest levels of differentiation, we see clear and almost entirely correct results, regardless of filtering parameter choices. Interestingly, however, in the model at K=3 we see the incorrect assignment of individuals in either population #3 or population #4 to population #2 when using the stricter MDT threshold, regardless of minimum

MAF choice. This error suggests that when inherent levels of differentiation in the dataset are

45 sufficiently high, stringency with MDT is counterproductive. This may be because missing data are unlikely to be randomly distributed across genetic groups; rather, the amount of missing data is likely proportional to genetic distance (Huang & Knowles, 2016). Therefore, removing loci that are not sequenced in some or most individuals biases the data by only retaining loci with low mutation rates, which may obscure differences in some genetically distinct groups (Huang &

Knowles, 2016). In dataset #6, which has slightly lower levels of differentiation than dataset #7, the same errors occur, but only when using the stricter choice of both MDT and minimum MAF.

Once again, this suggests that choice of minimum MAF affects the results in similar ways to that of MDT, but that its effects are masked by the much stronger effects of MDT in datasets with very high levels of differentiation. If differentiation is low enough, however, we see the effects of minimum MAF choice as well.

To ascertain whether the findings of our simulations are consistent in ‘real world’ scenarios, we re-analyzed two empirical ddRAD datasets using the same combinations of filtering choices. The empirical low dataset has differentiation levels in the range at which our simulations indicate that stringency of MDT is essential. The empirical high dataset has differentiation levels in the range at which our simulations are mostly accurate, regardless of filtering choices. As predicted by our simulations, we see that the two populations in the low differentiation dataset are only detectable when using the more stringent MDT choice, whereas the two populations in the high differentiation dataset are detectable across all combinations of

MDT and minimum MAF. Most datasets in our literature review contained lowly differentiated populations; few demonstrated genetic differentiation comparable to that of our most highly differentiated simulations. Rather, most datasets in our literature review contained lowly differentiated populations. This may be because high differentiation between conspecifics is

46 biologically rare, or because populations with high levels of differentiation continue to be assessed with lower cost approaches such as microsatellites.

Filtering choices also affected calculations of genetic diversity (AR and He) regardless of inherent levels of differentiation. Generally, filtering using the more stringent choice of MDT produced lower calculated values of AR across all simulated and empirical datasets. This is likely because the regions of a genome with the highest mutation rates (greatest number of alleles per locus, or AR) are the most likely to have mutations within restriction cut sites, which may prevent the restriction enzymes from cutting at such sites (Huang & Knowles, 2016). If

DNA is not cut at a certain restriction site due to a mutation, then a tag may not be sequenced there, causing SNPs that would have been in that tag to be missing from all individuals that have that mutation (Huang & Knowles, 2016); additionally, if some target SNPs are in high linkage disequilibrium (LD) with the cut site variant, then the sample of alleles at that SNP may be biased. Choosing a stringent MDT value (only retaining loci that are missing from few to no individuals) will likely cause some of the loci with the highest AR to be filtered out (Huang &

Knowles, 2016). However, stringency with MDT and minimum MAF had the opposite effect on

He in the empirical datasets (this effect was minor or absent in the simulated datasets). This may be because the expected heterozygosity of a locus is dependent on the number of alleles at that locus, and the evenness in frequency of the alleles. Choosing a stringent minimum MAF value would remove loci with very uneven allele frequencies, thereby raising the mean He across loci.

Regardless, these results demonstrate that measures of diversity depend on filtering choices, as do inferences of population structure. Thus, when assessing the validity of choices in filtering thresholds, performing these calculations across combinations of choice parameters would provide useful information.

47 In this study, we have clarified another aspect of an issue that is increasingly gaining attention: not only do choices of filtering thresholds affect inferences of population structure, but their impact is dependent on the inherent levels of differentiation in the dataset. Based on our literature review and analyses of simulated and empirical data, we have framed a recommended set of best practices when analyzing RRS data on wild populations (Box 1). As this body of literature grows, we are optimistic that even more data-sensitive recommendations will emerge, allowing us to continue studying wild populations in efficient and robust ways.

48 Box 1: Recommendations for best practices • Test a range of filtering choices, and tailor them to the differentiation levels of the dataset. For datasets with extremely low or high genetic differentiation, choice of MDT threshold is critical for accurate detection of population structure. For datasets with moderate levels of genetic differentiation, choice of both MDT and minimum MAF thresholds are important for accurate detection of population structure. Because global FST is not strongly affected by filtering parameter choices, it can serve as an indicator of how differentiated the samples in a dataset are, which may inform stringency levels during filtering. • Report common measures of differentiation such as global and pairwise FST. Although 77% of studies in our literature review reported pairwise FST values, only 31% reported global FST. In addition to providing guidance as to what filtering parameter ranges are appropriate, these measures help researchers gauge how differentiated the populations in a dataset are relative to those in other studies. • Report the sample sizes and measures of differentiation and diversity of genetic groups determined by STRUCTURE/PCA, not just those of the groups pre- determined by sampling locations. If the geographic patterns or number of genetic clusters does not match the geographic patterns or number of populations from which collections were conducted, providing statistics on the actual genetic clusters may be more biologically meaningful and useful to your readers. • Upload raw reads, not just downstream files in the analysis. From the 66% of studies that made their data publically available, only 59% of those uploaded raw reads (the rest uploaded downstream files only, such as VCF files or structure inputs). Making these data available assists other researchers who may want to replicate analyses or to re-analyse data for other purposes; raw reads are likely more useful than are downstream files. • Provide accession numbers in manuscripts, and ensure that the accession numbers are correct. In our literature review, we found that 27% of studies did not provide an accession number, and another 7% provided an accession number, but no files were available at that accession. • Provide clear and accurate metadata. Accurate metadata are critical to re- analyses. Be sure that every sample is accounted for (even samples/files that are removed from the analysis, particularly if they are still uploaded). Provide coordinates to collection sites, as these are key to some population genetic analyses.

Acknowledgements

This work was supported by the Australian Research Council (FL120100074, DE150101393).

49 Literature Cited Anderson C., Cunha L., Sechi P., Kille P., Spurgeon D. (2017) Genetic variation in populations

of the earthworm, Lumbricus rubellus, across contaminated mine sites. BMC Genet 18,

97.

Andrews K.R., Good J.M., Miller M.R., Luikart G., Hohenlohe P.A. (2016) Harnessing the

power of RADseq for ecological and evolutionary genomics. Nat Rev Genet 17, 81-92.

Arnold B., Corbett‐Detig R.B., Hartl D., Bomblies K. (2013) RADseq underestimates diversity

and introduces genealogical biases due to nonrandom haplotype sampling. Molecular

Ecology 22, 3179-3190.

Beheregaray L.B., Pfeiffer L.V., Attard C.R.M., et al. (2017) Genome-wide data delimits

multiple climate-determined species ranges in a widespread Australian fish, the golden

perch (Macquaria ambigua). Mol Phylogenet Evol 111, 65-75.

Bell R., Drewes R., Zamudio K. (2015) Reed frog diversification in the Gulf of Guinea:

Overseas dispersal, the progression rule, and in situ speciation. Evolution 69, 904-915.

Bezanson J., Edelman A., Karpinski S., Shah V.B. (2017) Julia: A Fresh Approach to Numerical

Computing. SIAM Review 59, 65-98.

Boehm J.T., Waldman J., Robinson J.D., Hickerson M.J. (2015) Population genomics reveals

seahorses (Hippocampus erectus) of the western mid-Atlantic coast to be residents rather

than vagrants. PLOS ONE 10, e0116219.

Bolton P.E., West A.J., Cardilini A.P., et al. (2016) Three Molecular Markers Show No

Evidence of Population Genetic Structure in the Gouldian Finch (Erythrura gouldiae).

PLOS ONE 11, e0167723.

Brauer C.J., Hammer M.P., Beheregaray L.B. (2016) Riverscape genomics of a threatened fish

across a hydroclimatically heterogeneous river basin. Mol Ecol 25, 5093-5113.

50 Cahill A.E., Levinton J.S. (2016) Genetic differentiation and reduced genetic diversity at the

northern range edge of two species with different dispersal modes. Mol Ecol 25, 515-526.

Cai S., Xu S., Liu L., Gao T., Zhou Y. (2018) Development of genome-wide SNPs for

population genetics and population assignment of Sebastiscus marmoratus. Conservation

Genetics Resources 10, 575-578.

Capblancq T., Despres L., Rioux D., Mavarez J. (2015) Hybridization promotes speciation in

Coenonympha butterflies. Mol Ecol 24, 6209-6222.

Carreras C., Ordonez V., Zane L., et al. (2017) Population genomics of an endemic

Mediterranean fish: differentiation by fine scale dispersal and adaptation. Sci Rep 7,

43417.

Catchen J., Hohenlohe P.A., Bassham S., Amores A., Cresko W.A. (2013) Stacks: an analysis

tool set for population genomics. Mol Ecol 22, 3124-3140.

Catchen J.M., Hohenlohe P.A., Bernatchez L., et al. (2017) Unbroken: RADseq remains a

powerful tool for understanding the genetics of adaptation in natural populations. Mol

Ecol Resour 17, 362-365.

Chattopadhyay B., Garg K.M., Ramakrishnan U. (2014) Effect of diversity and missing data on

genetic assignment with RAD-Seq markers. BMC Research Notes 7, 841.

Clucas G.V., Younger J.L., Kao D., et al. (2016) Dispersal in the sub-Antarctic: king penguins

show remarkably little population genetic differentiation across their range. BMC Evol

Biol 16, 211.

Collins E.E., Galanska M.P., Halanych K.M., Mahon A.R. (2018) Population Genomics of

Nymphon australe Hodgson, 1902 (Pycnogonida, Nymphonidae) in the Western

Antarctic. The Biological Bulletin 234, 180-191.

51 Davey J.W., Blaxter M.L. (2011) RADSeq: next-generation population genetics. Briefings in

Functional Genomics 9, 416-423.

Deagle B.E., Faux C., Kawaguchi S., Meyer B., Jarman S.N. (2015) Antarctic krill population

genomics: apparent panmixia, but genome complexity and large population size muddy

the water. Mol Ecol 24, 4943-4959.

Drury C., Schopmeyer S., Goergen E., et al. (2017) Genomic patterns in Acropora cervicornis

show extensive population structure and variable genetic diversity. Ecol Evol 7, 6188-

6200.

Eimanifar A., Brooks S.A., Bustamante T., Ellis J.D. (2018) Population genomics and

morphometric assignment of western honey bees (Apis mellifera L.) in the Republic of

South Africa. BMC Genomics 19, 615.

Ellegren H. (2014) Genome sequencing and population genomics in non-model organisms.

Trends in Ecology & Evolution 29, 51 - 63.

Farrell E.D., Carlsson J.E., Carlsson J. (2016) Next Gen Pop Gen: implementing a high-

throughput approach to population genetics in boarfish (Capros aper). R Soc Open Sci 3,

160651.

Fernandez R., Schubert M., Vargas-Velazquez A.M., et al. (2016) A genomewide catalogue of

single nucleotide polymorphisms in white-beaked and Atlantic white-sided dolphins. Mol

Ecol Resour 16, 266-276.

Forsström T., Ahmad F., Vasemägi A. (2017) Invasion genomics: genotyping-by-sequencing

approach reveals regional genetic structure and signatures of temporal selection in an

introduced mud crab. Marine Biology 164.

52 Francis R.M. (2016) pophelper: an R package and web app to analyse and visualize population

structure Molecular Ecology Resources 17, 27-32.

Gilbert K.J., Andrew R.L., Bock D.G., et al. (2012) Recommendations for utilizing and reporting

population genetic analyses: the reproducibility of genetic clustering using the program

structure. Molecular Ecology 22, 2357-2357.

Gloria-Soria A., Dunn W.A., Telleria E.L., et al. (2016) Patterns of Genome-Wide Variation in

Glossina fuscipes fuscipes Tsetse Flies from Uganda. G3 (Bethesda) 6, 1573-1584.

Goudet J. (2005) HIERFSTAT, a package for R to compute and test hierarchical F-statistics.

Molecular Ecology Notes 5, 184 - 186.

Grünwald N.J., Everhart S.E., Knaus B.J., Kamvar Z.N. (2017) Best Practices for Population

Genetic Analyses. Phytopathology 107, 1000-1010.

Hamlin J.A., Arnold M.L. (2015) Neutral and Selective Processes Drive Population

Differentiation for Iris hexagona. J Hered 106, 628-636.

Hecht B.C., Matala A.P., Hess J.E., Narum S.R. (2015) Environmental adaptation in Chinook

salmon (Oncorhynchus tshawytscha) throughout their North American range. Mol Ecol

24, 5573-5595.

Hess A.S., Hess M.K., Dodds K.G., et al. (2018) A method to simulate low-depth genotyping-

by-sequencing data for testing genomic analyses. Proceedings of the 11th World

Congress on Genetics Applied to Livestock Production, 385.

Huang H., Knowles L.L. (2016) Unforeseen Consequences of Excluding Missing Data from

Next-Generation Sequences: Simulation Study of RAD Sequences. Syst Biol 65, 357-365.

53 Jackson J.M., Pimsler M.L., Oyen K.J., et al. (2018) Distance, elevation and environment as

drivers of diversity and divergence in bumble bees across latitude and altitude. Mol Ecol

27, 2926-2942.

Jahner J.P., Gibson D., Weitzman C.L., et al. (2016) Fine-scale genetic structure among greater

sage-grouse leks in central Nevada. BMC Evol Biol 16, 127.

Johansson F., Halvarsson P., Mikolajewski D.J., Hoglund J. (2017) Genetic differentiation in the

boreal dragonfly Leucorrhinia dubia in the Palearctic region. Biological Journal of the

Linnean Society 121, 294-304.

Jombart T., Ahmed I. (2011) adegenet 1.3-1: new tools for the analysis of genome-wide SNP

data. Bioinformatics 27, 3070–3071.

Keenan K., McGinnity P., Cross T.F., Crozier W.W., Prodohl P. (2013) diveRsity:AnR package

for the estimation and exploration of popula tion genetics parameters and their associated

errors. Methods in Ecology and Evolution 4, 782-788.

Kjeldsen S.R., Zenger K.R., Leigh K., et al. (2016) Genome-wide SNP loci reveal novel insights

into koala (Phascolarctos cinereus) population variability across its range. Conservation

Genetics 17, 337-353.

Kotsakiozi P., Richardson J.B., Pichler V., et al. (2017) Population genomics of the Asian tiger

mosquito, Aedes albopictus: insights into the recent worldwide invasion. Ecol Evol 7,

10143-10157.

Krohn A.R., Conroy C.J., Pesapane R., et al. (2018) Conservation genomics of desert dwelling

California voles (Microtus californicus) and implications for management of endangered

Amargosa voles (Microtus californicus scirpensis). Conservation Genetics 19, 383-395.

54 Lah L., Trense D., Benke H., et al. (2016) Spatially Explicit Analysis of Genome-Wide SNPs

Detects Subtle Population Structure in a Mobile Marine Mammal, the Harbor Porpoise.

PLOS ONE 11, e0162792.

Larson W.A., Seeb L.W., Everett M.V., et al. (2014) Genotyping by sequencing resolves shallow

population structure to inform conservation of Chinook salmon (Oncorhynchus

tshawytscha). Evol Appl 7, 355-369.

Lavretsky P., Dacosta J.M., Hernandez-Banos B.E., et al. (2015) Speciation genomics and a role

for the Z chromosome in the early stages of divergence between Mexican ducks and

mallards. Mol Ecol 24, 5364-5378.

Leache A.D., Chavez A.S., Jones L.N., et al. (2015) Phylogenomics of phrynosomatid lizards:

conflicting signals from sequence capture versus restriction site associated DNA

sequencing. Genome Biol Evol 7, 706-719.

Linck E.B., Battey C.J. (2017) Minor allele frequency thresholds strongly affect population

structure inference with genomic datasets. bioRxiv.

Lischer H.E., Excoffier L. (2012) PGDSpider: an automated data conversion tool for connecting

population genetics and genomics programs. Bioinformatics 28, 298-299.

Lowry D.B., Hoban S., Kelley J.L., et al. (2017) Breaking RAD: an evaluation of the utility of

restriction site-associated DNA sequencing for genome scans of adaptation. Mol Ecol

Resour 17, 142-152.

Lozier J.D. (2014) Revisiting comparisons of genetic diversity in stable and declining species:

assessing genome-wide polymorphism in North American bumble bees using RAD

sequencing. Mol Ecol 23, 788-801.

55 Lozier J.D., Jackson J.M., Dillon M.E., Strange J.P. (2016) Population genomics of divergence

among extreme and intermediate color forms in a polymorphic insect. Ecol Evol 6, 1075-

1091.

Maas D.L., Prost S., Bi K., et al. (2018) Rapid divergence of mussel populations despite

incomplete barriers to dispersal. Mol Ecol 27, 1556-1571.

Martin C.H., Cutler J.S., Friel J.P., et al. (2015) Complex histories of repeated gene flow in

Cameroon crater lake cichlids cast doubt on one of the clearest examples of sympatric

speciation. Evolution 69, 1406-1422.

Martin C.H., Feinstein L.C. (2014) Novel trophic niches drive variable progress towards

ecological speciation within an adaptive radiation of pupfishes. Mol Ecol 23, 1846-1862.

Martinez E., Buonaccorsi V., Hyde J.R., Aguilar A. (2017) Population genomics reveals high

gene flow in grass rockfish (Sebastes rastrelliger). Mar Genomics 33, 57-63.

Mastretta-Yanes A., Arrigo N., Alvarez N., et al. (2015) Restriction site-associated DNA

sequencing, genotyping error estimation and de novo assembly optimization for

population genetic inference. Mol Ecol Resour 15, 28-41.

McCartney-Melstad E., Gidis M., Shaffer H.B. (2018) Population genomic data reveal extreme

geographic subdivision and novel conservation actions for the declining foothill yellow-

legged frog. Heredity (Edinb) 121, 112-125.

Metivier S.L., Kim J.H., Addison J.A. (2017) Genotype by sequencing identifies natural

selection as a driver of intraspecific divergence in Atlantic populations of the high

dispersal marine invertebrate, Macoma petalum. Ecol Evol 7, 8058-8072.

56 Monzon J.D., Atkinson E.G., Henn B.M., Benach J.L. (2016) Population and Evolutionary

Genomics of Amblyomma americanum, an Expanding Arthropod Disease Vector.

Genome Biol Evol 8, 1351-1360.

Munoz J., Chaturvedi A., De Meester L., Weider L.J. (2016) Characterization of genome-wide

SNPs for the water flea Daphnia pulicaria generated by genotyping-by-sequencing

(GBS). Sci Rep 6, 28569.

Munshi-South J., Zolnik C.P., Harris S.E. (2016) Population genomics of the Anthropocene:

urbanization is negatively associated with genome-wide variation in white-footed mouse

populations. Evol Appl 9, 546-564.

Nicotra A.B., Chong C., Bragg J.G., et al. (2016) Population and phylogenomic decomposition

via genotyping-by-sequencing in Australian Pelargonium. Mol Ecol 25, 2000-2014.

Nielsen R., Paul J.S., Albrechtsen A., Song Y.S. (2011) Genotype and SNP calling from next-

generation sequencing data. Nature Reviews Genetics 12, 443-451.

Nørgaard L.S., Mikkelsen D.M.G., Elmeros M., et al. (2017) Population genomics of the

raccoon (Nyctereutes procyonoides) in Denmark: insights into invasion history and

population development. Biological Invasions 19, 1637-1652.

Nunez J.C., Seale T.P., Fraser M.A., et al. (2015) Population Genomics of the Euryhaline

Teleost Poecilia latipinna. PLOS ONE 10, e0137077.

O'Leary S.J., Puritz J.B., Willis S.C., Hollenbeck C.M., Portnoy D.S. (2018) These aren't the loci

you'e looking for: Principles of effective SNP filtering for molecular ecologists. Mol

Ecol.

Parchman T.L., Buerkle C.A., Soria-Carrasco V., Benkman C.W. (2016) Genome divergence

and diversification within a geographic mosaic of coevolution. Mol Ecol 25, 5705-5718.

57 Parchman T.L., Jahner J.P., Uckele K.A., Galland L.M., Eckert A.J. (2018) RADseq approaches

and applications for forest tree genetics. Tree Genetics & Genomes 14, 39.

Pellegrino I., Boatti L., Cucco M., et al. (2016) Development of SNP markers for population

structure and phylogeography characterization in little owl (Athene noctua) using a

genotyping- by-sequencing approach. Conservation Genetics Resources 8, 13-16.

Pembleton L.W., Cogan N.O., Forster J.W. (2013) StAMPP: an R package for calculation of

genetic differentiation and structure of mixed-ploidy level populations. Molecular

Ecology Resources 13, 946-952.

Perez-Portela R., Bumford A., Coffman B., et al. (2018) Genetic homogeneity of the invasive

lionfish across the Northwestern Atlantic and the Gulf of Mexico based on Single

Nucleotide Polymorphisms. Sci Rep 8, 5062.

Peters J.L., Lavretsky P., DaCosta J.M., et al. (2016) Population genomic data delineate

conservation units in mottled ducks ( Anas fulvigula ). Biological Conservation 203, 272-

281.

Pukk L., Ahmad F., Hasan S., et al. (2015) Less is more: extreme genome complexity reduction

with ddRAD using Ion Torrent semiconductor technology. Mol Ecol Resour 15, 1145-

1152.

Raj A., Stephens M., Pritchard J.K. (2014) fastSTRUCTURE: Variational Inference of

Population Structure in Large SNP Data Sets. Genetics 197, 573 - 589.

Rasic G., Endersby-Harshman N., Tantowijoyo W., et al. (2015) Aedes aegypti has spatially

structured and seasonally stable populations in Yogyakarta, Indonesia. Parasit Vectors 8,

610.

58 Rius M., Bourne S., Hornsby H.G., Chapman M.A. (2015) Applications of next-generation

sequencing to the study of biological invasions Current Zoology 61, 488-504.

Roland A.B., Santos J.C., Carriker B.C., et al. (2017) Radiation of the polymorphic Little Devil

poison frog (Oophaga sylvatica) in Ecuador. Ecol Evol 7, 9750-9762.

Saenz-Agudelo P., Dibattista J.D., Piatek M.J., et al. (2015) Seascape genetics along

environmental gradients in the Arabian Peninsula: insights from ddRAD sequencing of

anemonefishes. Mol Ecol 24, 6241-6255.

Sandoval-Castillo J., Robinson N.A., Hart A.M., Strain L.W.S., Beheregaray L.B. (2018)

Seascape genomics reveals adaptive divergence in a connected and commercially

important mollusc, the greenlip abalone (Haliotis laevigata), along a longitudinal

environmental gradient. Mol Ecol 27, 1603-1620.

Savary R., Masclaux F.G., Wyss T., et al. (2018) A population genomics approach shows

widespread geographical distribution of cryptic genomic forms of the symbiotic fungus

Rhizophagus irregularis. ISME J 12, 17-30.

Schield D.R., Card D.C., Adams R.H., et al. (2015) Incipient speciation with biased gene flow

between two lineages of the Western Diamondback Rattlesnake (Crotalus atrox). Mol

Phylogenet Evol 83, 213-223.

Seeb L.W., Waples R.K., Limborg M.T., et al. (2014) Parallel signatures of selection in

temporally isolated lineages of pink salmon. Mol Ecol 23, 2473-2485.

Sekino M., Nakamichi R., Iwasaki Y., et al. (2016) A new resource of single nucleotide

polymorphisms in the Japanese eel Anguilla japonica derived from restriction site-

associated DNA. Ichthyological Research 63, 496-504.

59 Shafer A.B.A., Peart C.R., Tusso S., et al. (2017) Bioinformatic processing of RAD-seq data

dramatically impacts downstream population genetic inference. Methods in Ecology and

Evolution 8, 907-917.

Shultz A.J., Baker A.J., Hill G.E., Nolan P.M., Edwards S.V. (2016) SNPs across time and

space: population genomic signatures of founder events and epizootics in the House

Finch (Haemorhous mexicanus). Ecol Evol 6, 7475-7489.

Sovic M.G., Carstens B.C., Gibbs H.L. (2016) Genetic diversity in migratory bats: Results from

RADseq data for three tree bat species at an Ohio windfarm. PeerJ 4, e1647.

Stockwell B.L., Larson W.A., Waples R.K., et al. (2016) The application of genomics to inform

conservation of a functionally important reef fish (Scarus niger) in the Philippines.

Conservation Genetics 17, 239-249.

Team R.C. (2016) R: A language and environment for statistical computing. R Foundation for

Statistical Computing, Vienna, Austria.

Trumbo D.R., Epstein B., Hohenlohe P.A., et al. (2016) Mixed population genomics support for

the central marginal hypothesis across the invasive range of the cane toad (Rhinella

marina) in Australia. Molecular Ecology 25, 4161–4176

Van Wyngaarden M., Snelgrove P.V., DiBacco C., et al. (2017) Identifying patterns of dispersal,

connectivity and selection in the sea scallop, Placopecten magellanicus, using RADseq-

derived SNPs. Evol Appl 10, 102-117.

Vega R., Vázquez-Domínguez E., White T.A., Valenzuela-Galván D., Searle J.B. (2017)

Population genomics applications for conservation: the case of the tropical dry forest

dweller Peromyscus melanophrys. Conservation Genetics 18, 313-326.

60 Wang Z., Gerstein M., Snyder M. (2009) RNA-Seq: a revolutionary tool for transcriptomics.

Nature Reviews Genetics 10.

White T.A., Perkins S.E., Heckel G., Searle J.B. (2013) Adaptive evolution during an ongoing

range expansion: the invasive bank vole (Myodes glareolus) in Ireland. Mol Ecol 22,

2971-2985.

Wosula E.N., Chen W., Fei Z., Legg J.P. (2017) Unravelling the Genetic Diversity among

Cassava Bemisia tabaci Whiteflies Using NextRAD Sequencing. Genome Biol Evol 9,

2958-2973.

Xu S., Song N., Zhao L., et al. (2017) Genomic evidence for local adaptation in the

ovoviviparous marine fish Sebastiscus marmoratus with a background of population

homogeneity. Sci Rep 7, 1562.

Xuereb A., Benestan L., Normandeau E., et al. (2018) Asymmetric oceanographic processes

mediate connectivity and population genetic structure, as revealed by RADseq, in a

highly dispersive marine invertebrate (Parastichopus californicus). Mol Ecol 27, 2347-

2364.

Zhao Y., Peng W., Guo H., et al. (2018) Population Genomics Reveals Genetic Divergence and

Adaptive Differentiation of Chinese Sea Bass (Lateolabrax maculatus). Mar Biotechnol

(NY) 20, 45-59.

Zlonis K.J., Gross B.L. (2018) Genetic structure, diversity, and hybridization in populations of

the rare arctic relict Euphrasia hudsoniana (Orobanchaceae) and its invasive congener

Euphrasia stricta. Conservation Genetics 19, 43-55.

61 Supporting Information

Table S1. Details of population history simulated in each of seven GBS datasets to create varying levels of population structure (as illustrated in main text, Figure 1). Dataset Simulated population history Dataset #1 • 500 founders (Global FST ~ • Population size increase to 1000 over 100 generations 0.0063) • Constant population size for 900 generations before two-way split between population #1 and populations #2-4 (500 individuals each) • Population #1 undergoes constant population size for 2 generations • Constant population size for 5 generations before two-way split between population #2 and populations #3-4 (250 individuals each) • Population #2 undergoes constant population size for 2 generations • Constant population size for 5 generations before two-way split between population #3 and population #4 (125 individuals each) • Population #3 undergoes constant population size for 2 generations • Population #4 undergoes constant population size for 2 generations

Dataset #2 • 500 founders (Global FST ~ • Population size increase to 1000 over 100 generations 0.0386) • Constant population size for 100 generations before two-way split between population #1 and populations #2-4 (500 individuals each) • Population #1 undergoes constant population size for 20 generations • Constant population size for 15 generations before two-way split between population #2 and populations #3-4 (250 individuals each) • Population #2 undergoes constant population size for 20 generations • Constant population size for 10 generations before two-way split between population #3 and population #4 (125 individuals each) • Population #3 undergoes constant population size for 20 generations • Population #4 undergoes constant population size for 20 generations

Dataset #3 • 500 founders (Global FST ~ • Population size increase to 1000 over 100 generations 0.0856) • Two-way split between population #1 and populations #2-4 (500 individuals each) • Population #1 undergoes constant population size for 50 generations • Constant population size for 15 generations before two-way split between population #2 and populations #3-4 (250 individuals each) • Population #2 undergoes constant population size for 50 generations • Constant population size for 10 generations before two-way split between population #3 and population #4 (125 individuals each) • Population #3 undergoes constant population size for 50 generations • Population #4 undergoes constant population size for 50 generations

Dataset #4 • 500 founders (Global FST ~ • Population size increase to 1000 over 100 generations 0.1576) • Population size decrease to 800 over 20 generations • Population size increase to 1000 over 20 generations

62 • Two-way split between population #1 and populations #2-4 (500 individuals each) • Population #1 undergoes constant population size for 100 generations • Constant population size for 15 generations before two-way split between population #2 and populations #3-4 (250 individuals each) • Population #2 undergoes constant population size for 100 generations • Constant population size for 10 generations before two-way split between population #3 and population #4 (125 individuals each) • Population #3 undergoes constant population size for 100 generations • Population #4 undergoes constant population size for 100 generations

Dataset #5 • 500 founders (Global FST ~ • Population size increase to 1000 over 100 generations 0.3352) • Population size decrease to 400 over 30 generations • Population size increase to 1000 over 80 generations • Two-way split between population #1 and populations #2-4 (500 individuals each) • Population #1 undergoes constant population size for 200 generations • Constant population size for 15 generations before two-way split between population #2 and populations #3-4 (250 individuals each) • Population #2 undergoes constant population size for 200 generations • Constant population size for 10 generations before two-way split between population #3 and population #4 (125 individuals each) • Population #3 undergoes constant population size for 200 generations • Population #4 undergoes constant population size for 200 generations

Dataset #6 • 500 founders (Global FST ~ • Population size increase to 1000 over 100 generations 0.6637) • Population size decrease to 200 over 40 generations • Population size increase to 1000 over 120 generations • Two-way split between population #1 and populations #2-4 (500 individuals each) • Population #1 undergoes constant population size for 500 generations • Constant population size for 20 generations before two-way split between population #2 and populations #3-4 (250 individuals each) • Population #2 undergoes constant population size for 500 generations • Constant population size for 15 generations before two-way split between population #3 and population #4 (125 individuals each) • Population #3 undergoes constant population size for 500 generations • Population #4 undergoes constant population size for 500 generations

Dataset #7 • 500 founders (Global FST ~ • Population size increase to 1000 over 100 generations 0.8103) • Population size decrease to 100 over 50 generations • Population size increase to 1000 over 150 generations • Two-way split between population #1 and populations #2-4 (500 individuals each) • Population #1 undergoes constant population size for 900 generations

63 • Constant population size for 20 generations before two-way split between population #2 and populations #3-4 (250 individuals each) • Population #2 undergoes constant population size for 900 generations • Constant population size for 15 generations before two-way split between population #3 and population #4 (125 individuals each) • Population #3 undergoes constant population size for 900 generations • Population #4 undergoes constant population size for 900 generations

Table S2. Pairwise FST values and their confidence intervals (CI) between four populations, obtained using four different combinations of parameter choices (minimum minor allele frequency=min MAF and missing data tolerance=MDT), on seven simulated GBS datasets with varying levels of population structure. Dataset #1 Pairwise FST (lower CI, upper CI) Parameter choices 1 vs 2 1 vs 3 1 vs 4 2 vs 3 2 vs 4 3 vs 4 MAF=0.05, 0.0050 0.0114 0.0118 0.0108 0.0103 0.0081 MDT=0.5 (0.0043, (0.0098, (0.0099, (0.0094, (0.0091, (0.0071, 0.0058) 0.0125) 0.0133) 0.0120) 0.0114) 0.0093) MAF=0.05, 0.0053 0.0108 0.0107 0.0106 0.0097 0.0094 MDT=0.2 (0.0040, (0.0088, (0.0085, (0.0087, (0.0075, (0.0078, 0.0067) 0.0131) 0.0129) 0.0124) 0.0120) 0.0116) MAF=0.01, 0.0050 0.0113 0.0117 0.0106 0.0103 0.0080 MDT=0.5 (0.0044, (0.0101, (0.0100, (0.0096, (0.0093, (0.0070, 0.0058) 0.0125) 0.0131) 0.0119) 0.0117) 0.0090) MAF=0.01, 0.0053 0.0107 0.0106 0.0104 0.0097 0.0094 MDT=0.2 (0.0040, (0.0086, (0.0089, (0.0088, (0.0077, (0.0072, 0.0067) 0.0124) 0.0128) 0.0124) 0.0116) 0.0117) Mean 0.0052 0.0110 0.0112 0.0106 0.0100 0.0087 SD 0.0002 0.0004 0.0006 0.0002 0.0004 0.0008 Dataset #2 Pairwise FST (lower CI, upper CI) Parameter choices 1 vs 2 1 vs 3 1 vs 4 2 vs 3 2 vs 4 3 vs 4 MAF=0.05, 0.0339 0.0571 0.0539 0.0688 0.0673 0.0732 MDT=0.5 (0.0312, (0.0505, (0.0478, (0.0620, (0.0601, (0.0657, 0.0369) 0.0635) 0.0588) 0.0747) 0.0729) 0.0785) MAF=0.05, 0.0341 0.0449 0.0588 0.0607 0.0710 0.0708 MDT=0.2 (0.0295, (0.0377, (0.0498, (0.0517, (0.0605, (0.0588, 0.0385) 0.0522) 0.0712) 0.0688) 0.0812) 0.0822) MAF=0.01, 0.0332 0.0558 0.0531 0.0677 0.0666 0.0728 MDT=0.5 (0.0303, (0.0507, (0.0492, (0.0619, (0.0608, (0.0660, 0.0358) 0.0623) 0.0580) 0.0745) 0.0725) 0.0783) MAF=0.01, 0.0337 0.0439 0.0579 0.0596 0.0702 0.0699 MDT=0.2 (0.0297, (0.0365, (0.0473, (0.0497, (0.0593, (0.0572, 0.0386) 0.0516) 0.0679) 0.0693) 0.0807) 0.0798) Mean 0.0338 0.0505 0.0559 0.0642 0.0688 0.0717 SD 0.0004 0.0070 0.0028 0.0047 0.0021 0.0016 Dataset #3 Pairwise FST (lower CI, upper CI) Parameter choices 1 vs 2 1 vs 3 1 vs 4 2 vs 3 2 vs 4 3 vs 4

64 MAF=0.05, 0.0703 0.1190 0.1361 0.1298 0.1465 0.1900 MDT=0.5 (0.0627, (0.1092, (0.1216, (0.1206, (0.1335, (0.1741, 0.0779) 0.1268) 0.1513) 0.1407) 0.1599) 0.2067) MAF=0.05, 0.0687 0.1088 0.1280 0.1205 0.1335 0.1667 MDT=0.2 (0.0579, (0.0910, (0.1046, (0.1046, (0.1147, (0.1403, 0.0780) 0.1249) 0.1550) 0.1365) 0.1557) 0.1880) MAF=0.01, 0.0689 0.1339 0.1282 0.1449 0.1890 MDT=0.5 (0.0626, 0.1162 (0.1224, (0.1179, (0.1314, (0.1724, 0.0747) (0.1078,01238) 0.1467) 0.1366) 0.1600) 0.2036) MAF=0.01, 0.0670 0.1066 0.1253 0.1189 0.1314 0.1659 MDT=0.2 (0.0576, (0.0909, (0.1001, (0.0952, (0.1102, (0.1378, 0.0764) 0.1186) 0.1470) 0.1385) 0.1498) 0.1917) Mean 0.0687 0.1127 0.1308 0.1244 0.1391 0.1779 SD 0.0013 0.0059 0.0050 0.0054 0.0077 0.0134 Dataset #4 Pairwise FST (lower CI, upper CI) Parameter choices 1 vs 2 1 vs 3 1 vs 4 2 vs 3 2 vs 4 3 vs 4 MAF=0.05, 0.1302 0.2193 0.1737 0.2593 0.2322 0.3255 MDT=0.5 (0.1184, (0.2014, (0.1605, (0.2427, (0.2167, (0.3056, 0.1413) 0.2372) 0.1879) 0.2763) 0.2499) 0.3510) MAF=0.05, 0.1265 0.2112 0.1578 0.2303 0.1981 0.3160 MDT=0.2 (0.1059, (0.1810, (0.1369, (0.2051, (0.1705, (0.2705, 0.1473) 0.2358) 0.1830) 0.2544) 0.2316) 0.3509) MAF=0.01, 0.1277 0.2134 0.1700 0.2540 0.2299 0.3221 MDT=0.5 (0.1178, (0.1985, (0.1575, (0.2380, (0.2082, (0.2992, 0.1383) 0.2291) 0.1831) 0.2713) 0.2478) 0.3406) MAF=0.01, 0.1237 0.2049 0.1533 0.2239 0.1947 0.3106 MDT=0.2 (0.1028, (0.1706, (0.1315, (0.1929, (0.1710, (0.2687, 0.1371) 0.2316) 0.1763) 0.2528) 0.2194) 0.3528) Mean 0.1270 0.2122 0.1637 0.2419 0.2138 0.3186 SD 0.0027 0.0059 0.0097 0.0174 0.0201 0.0066 Dataset #5 Pairwise FST (lower CI, upper CI) Parameter choices 1 vs 2 1 vs 3 1 vs 4 2 vs 3 2 vs 4 3 vs 4 MAF=0.05, 0.2411 0.3290 0.3647 0.4511 0.4645 0.6005 MDT=0.5 (0.2272, (0.3085, (0.3445, (0.4260, (0.4409, (0.5701, 0.2561) 0.3470) 0.3885) 0.4740) 0.4889) 0.6263) MAF=0.05, 0.2306 0.3175 0.3669 0.4403 0.4704 0.5853 MDT=0.2 (0.2006, (0.2819, (0.3301, (0.4052, (0.4258, (0.5427, 0.2530) 0.3504) 0.3950) 0.4800) 0.5023) 0.6299) MAF=0.01, 0.2372 0.3241 0.3597 0.4400 0.4546 0.5890 MDT=0.5 (0.2232, (0.3009, (0.3406, (0.4138, (0.4297, (0.5625, 0.2529) 0.3437) 0.3772) 0.4698) 0.4744) 0.6122) MAF=0.01, 0.2271 0.3121 0.3609 0.4320 0.4594 0.5744 MDT=0.2 (0.1974, (0.2810, (0.3278, (0.4014, (0.4231, (0.5318, 0.2560) 0.3471) 0.3927) 0.4706) 0.4939) 0.6172) Mean 0.2340 0.3207 0.3630 0.4408 0.4622 0.5873 SD 0.0064 0.0074 0.0033 0.0078 0.0068 0.0108 Dataset #6 Pairwise FST (lower CI, upper CI)

65 Parameter choices 1 vs 2 1 vs 3 1 vs 4 2 vs 3 2 vs 4 3 vs 4 MAF=0.05, 0.5443 0.5675 0.5930 0.7747 0.7884 0.8716 MDT=0.5 (0.5230, (0.5405, (0.5656, (0.7526, (0.7672, (0.8578, 0.5638) 0.5914) 0.6116) 0.7934) 0.8048) 0.8843) MAF=0.05, 0.5546 0.5937 0.6144 0.7845 0.8011 0.8855 MDT=0.2 (0.5167, (0.5655, (0.5762, (0.7531, (0.7779, (0.8677, 0.5808) 0.6238) 0.6408) 0.8115) 0.8218) 0.9023) MAF=0.01, 0.5362 0.5617 0.5863 0.7638 0.7758 0.8621 MDT=0.5 (0.5104, (0.5438, (0.5670, (0.7444, (0.7595, (0.8510, 0.5572) 0.5824) 0.6041) 0.7806) 0.7941) 0.8753) MAF=0.01, 0.5455 0.5876 0.6068 0.7749 0.7905 0.8782 MDT=0.2 (0.5035, (0.5596, (0.5760, (0.7380, (0.7587, (0.8536, 0.5762) 0.6178) 0.6338) 0.8005) 0.8141) 0.8956) Mean 0.5451 0.5776 0.6001 0.7745 0.7889 0.8744 SD 0.0075 0.0154 0.0128 0.0084 0.0104 0.0099 Dataset #7 Pairwise FST (lower CI, upper CI) Parameter choices 1 vs 2 1 vs 3 1 vs 4 2 vs 3 2 vs 4 3 vs 4 MAF=0.05, 0.7245 0.7270 0.7002 0.8971 0.8951 0.9655 MDT=0.5 (0.7088, (0.7131, (0.6814, (0.8867, (0.8834, (0.9567, 0.7430) 0.7478) 0.7163) 0.9101) 0.9063) 0.9714) MAF=0.05, 0.7291 0.7210 0.6929 0.9122 0.9065 0.9656 MDT=0.2 (0.6969, (0.6907, (0.6621, (0.8940, (0.8887, (0.9512, 0.7543) 0.7516) 0.7229) 0.9281) 0.9233) 0.9739) MAF=0.01, 0.7203 0.7227 0.6952 0.8933 0.8916 0.9605 MDT=0.5 (0.7006, (0.7028, (0.6771, (0.8771, (0.8747, (0.9520, 0.7377) 0.7393) 0.7136) 0.9048) 0.9021) 0.9675) MAF=0.01, 0.7243 0.7155 0.6860 0.9079 0.9012 0.9595 MDT=0.2 (0.6980, (0.6886, (0.6520, (0.8822, (0.8765, (0.9443, 0.7429) 0.7373) 0.7159) 0.9231) 0.9178) 0.9701) Mean 0.7246 0.7216 0.6936 0.9026 0.8986 0.9628 SD 0.0036 0.0047 0.0059 0.0089 0.0066 0.0032

Table S3. Global FST, allelic richness (AR), and expected heterozygosity (He) of each of four populations, obtained using four different combinations of parameter choices (minimum minor allele frequency=min MAF and missing data tolerance=MDT), on seven simulated GBS datasets with varying levels of population structure. Dataset #1 Parameter choice Global AR AR AR AR He He He He combination FST Pop#1 Pop#2 Pop#3 Pop#4 Pop#1 Pop#2 Pop#3 Pop#4 MAF=0.05,MDT=0.5 0.0068 1.89 1.90 1.88 1.88 0.31 0.30 0.30 0.30 MAF=0.05,MDT=0.2 0.0058 1.60 1.61 1.57 1.63 0.41 0.40 0.41 0.39 MAF=0.01,MDT=0.5 0.0068 1.87 1.89 1.84 1.84 0.27 0.27 0.27 0.27 MAF=0.01,MDT=0.2 0.0057 1.59 1.61 1.56 1.61 0.38 0.37 0.38 0.36 Mean 0.0063 1.74 1.75 1.71 1.74 0.34 0.34 0.34 0.33 SD 0.0006 0.17 0.17 0.17 0.14 0.06 0.06 0.06 0.05 Dataset #2

66 Parameter choice Global AR AR AR AR He He He He combination FST Pop#1 Pop#2 Pop#3 Pop#4 Pop#1 Pop#2 Pop#3 Pop#4 MAF=0.05,MDT=0.5 0.0428 1.89 1.87 1.74 1.82 0.32 0.30 0.32 0.29 MAF=0.05,MDT=0.2 0.0348 1.59 1.53 1.36 1.59 0.42 0.42 0.45 0.37 MAF=0.01,MDT=0.5 0.0422 1.89 1.81 1.66 1.73 0.28 0.27 0.28 0.25 MAF=0.01,MDT=0.2 0.0344 1.60 1.49 1.32 1.53 0.38 0.39 0.41 0.34 Mean 0.0386 1.74 1.68 1.52 1.67 0.35 0.35 0.36 0.31 SD 0.0046 0.17 0.19 0.21 0.14 0.07 0.07 0.08 0.05 Dataset #3 Parameter choice Global AR AR AR AR He He He He combination FST Pop#1 Pop#2 Pop#3 Pop#4 Pop#1 Pop#2 Pop#3 Pop#4 MAF=0.05,MDT=0.5 0.0989 1.88 1.85 1.67 1.69 0.32 0.30 0.30 0.27 MAF=0.05,MDT=0.2 0.0732 1.48 1.47 1.34 1.47 0.46 0.43 0.40 0.35 MAF=0.01,MDT=0.5 0.0975 1.85 1.79 1.58 1.61 0.29 0.26 0.27 0.24 MAF=0.01,MDT=0.2 0.0726 1.47 1.45 1.27 1.43 0.43 0.39 0.37 0.32 Mean 0.0856 1.67 1.64 1.47 1.55 0.37 0.35 0.34 0.30 SD 0.0146 0.22 0.21 0.19 0.12 0.08 0.08 0.06 0.05 Dataset #4 Parameter choice Global AR AR AR AR He He He He combination FST Pop#1 Pop#2 Pop#3 Pop#4 Pop#1 Pop#2 Pop#3 Pop#4 MAF=0.05,MDT=0.5 0.1751 1.88 1.76 1.51 1.53 0.31 0.29 0.24 0.25 MAF=0.05,MDT=0.2 0.1437 1.53 1.57 1.30 1.25 0.44 0.38 0.32 0.37 MAF=0.01,MDT=0.5 0.1715 1.84 1.69 1.43 1.44 0.28 0.26 0.21 0.22 MAF=0.01,MDT=0.2 0.1399 1.52 1.52 1.26 1.20 0.39 0.34 0.29 0.34 Mean 0.1576 1.70 1.64 1.38 1.35 0.35 0.32 0.26 0.29 SD 0.0183 0.19 0.11 0.12 0.16 0.07 0.05 0.05 0.07 Dataset #5 Parameter choice Global AR AR AR AR He He He He combination FST Pop#1 Pop#2 Pop#3 Pop#4 Pop#1 Pop#2 Pop#3 Pop#4 MAF=0.05,MDT=0.5 0.3548 1.85 1.65 1.39 1.42 0.32 0.26 0.17 0.17 MAF=0.05,MDT=0.2 0.3233 1.67 1.49 1.25 1.24 0.39 0.34 0.25 0.29 MAF=0.01,MDT=0.5 0.3466 1.81 1.61 1.36 1.38 0.29 0.24 0.16 0.16 MAF=0.01,MDT=0.2 0.3162 1.64 1.43 1.23 1.23 0.36 0.33 0.23 0.27 Mean 0.3352 1.74 1.54 1.31 1.32 0.34 0.29 0.20 0.22 SD 0.0184 0.10 0.10 0.08 0.09 0.05 0.05 0.05 0.07 Dataset #6 Parameter choice Global AR AR AR AR He He He He combination FST Pop#1 Pop#2 Pop#3 Pop#4 Pop#1 Pop#2 Pop#3 Pop#4 MAF=0.05,MDT=0.5 0.6628 1.68 1.35 1.17 1.15 0.27 0.13 0.08 0.07 MAF=0.05,MDT=0.2 0.6745 1.49 1.27 1.11 1.05 0.33 0.16 0.13 0.13 MAF=0.01,MDT=0.5 0.6519 1.66 1.35 1.16 1.15 0.25 0.12 0.07 0.07 MAF=0.01,MDT=0.2 0.6655 1.49 1.28 1.11 1.05 0.31 0.16 0.12 0.13 Mean 0.6637 1.58 1.31 1.14 1.10 0.29 0.14 0.10 0.10 SD 0.0093 0.10 0.04 0.03 0.06 0.04 0.02 0.03 0.04 Dataset #7 Parameter choice Global AR AR AR AR He He He He combination FST Pop#1 Pop#2 Pop#3 Pop#4 Pop#1 Pop#2 Pop#3 Pop#4 MAF=0.05,MDT=0.5 0.8102 1.51 1.18 1.02 1.02 0.20 0.09 0.03 0.04 MAF=0.05,MDT=0.2 0.8161 1.37 1.09 1.00 0.96 0.27 0.13 0.07 0.10 MAF=0.01,MDT=0.5 0.8049 1.52 1.18 1.03 1.03 0.19 0.09 0.04 0.04

67 MAF=0.01,MDT=0.2 0.8099 1.39 1.09 1.00 0.97 0.26 0.13 0.08 0.10 Mean 0.8103 1.45 1.13 1.01 1.00 0.23 0.11 0.05 0.07 SD 0.0046 0.08 0.05 0.02 0.03 0.04 0.03 0.02 0.03

Table S4. Pairwise FST values and their confidence intervals (CI) between two populations, obtained using four different combinations of parameter choices (minimum minor allele frequency=min MAF and missing data tolerance=MDT), on two empirical ddRAD datasets with varying levels of population structure. Trumbo et al. (2016) Pairwise FST (lower CI, upper CI) Parameter choices QLD vs NT MAF=0.05, MDT=0.5 0.0932 (0.0906, 0.0949) MAF=0.05, MDT=0.2 0.0970 (0.0934, 0.1009) MAF=0.01, MDT=0.5 0.0910 (0.0888, 0.0935) MAF=0.01, MDT=0.2 0.0953 (0.0920, 0.0993) Mean 0.0941 SD 0.0026 Bell et al. (2015) Pairwise FST (lower CI, upper CI) Parameter choices Príncipe vs São Tomé MAF=0.05, MDT=0.5 0.4800 (0.4736, 0.4856) MAF=0.05, MDT=0.2 0.4957 (0.4860, 0.5053) MAF=0.01, MDT=0.5 0.4303 (0.4238, 0.4350) MAF=0.01, MDT=0.2 0.4445 (0.4334, 0.4518) Mean 0.4626 SD 0.0304

Table S5. Global FST, allelic richness (AR), and expected heterozygosity (He) of each of two populations, obtained using four different combinations of parameter choices (minimum minor allele frequency=min MAF and missing data tolerance=MDT), on two empirical ddRAD datasets with varying levels of population structure. Trumbo et al. (2016) Parameter choice combination Global FST AR QLD AR NT He QLD He NT MAF=0.05,MDT=0.5 0.0424 1.72 1.73 0.50 0.20 MAF=0.05,MDT=0.2 0.0361 1.39 1.40 0.50 0.33 MAF=0.01,MDT=0.5 0.0410 1.67 1.66 0.35 0.27 MAF=0.01,MDT=0.2 0.0352 1.37 1.36 0.47 0.28 Mean 0.0387 1.54 1.54 0.46 0.27 SD 0.0031 0.18 0.18 0.07 0.05 Bell et al. (2015) Parameter choice combination Global FST AR Príncipe AR São Tomé He Príncipe He São Tomé MAF=0.05,MDT=0.5 0.2442 0.87 1.24 0.24 0.37 MAF=0.05,MDT=0.2 0.2164 0.69 1.04 0.21 0.50 MAF=0.01,MDT=0.5 0.2114 0.95 1.21 0.22 0.25 MAF=0.01,MDT=0.2 0.1838 0.76 0.99 0.19 0.40 Mean 0.2140 0.82 1.12 0.22 0.38 SD 0.0214 0.12 0.12 0.02 0.10

68

Figure S1. Discriminant analysis of principal components (DAPC) plot of four populations in each of seven simulated GBS datasets with varying levels of differentiation. PCAs were run four times with different combinations of filtering parameter choices (minimum minor allele frequency and missing data threshold).

Figure S2. Discriminant analysis of principal components (DAPC) plot of two populations in each of two empirical ddRAD datasets with varying levels of differentiation. PCAs were run four times with different combinations of filtering parameter choices (minimum minor allele frequency and missing data threshold).

69 CHAPTER 3: Invaders weather the weather: does rapid adaptation to a novel environment represent a genetic paradox?

This chapter is in preparation for submission with the following co-authors: Richardson MF, Shine R, Rollins LA

70 Abstract Invasive species often evolve rapidly in their introduced ranges despite the genetic bottlenecks that are thought to accompany the translocation of small numbers of founders; however, some invasions may not fit this “genetic paradox” due to a maintenance of genetic diversity or a lack of evolution. The invasive cane toad (Rhinella marina) displays high phenotypic variation across its introduced Australian range, which is more environmentally heterogeneous than its native range. Here, we used three genome-wide datasets to characterize population structure and genetic diversity in invasive toads: (1) RNA-Seq data generated from spleens sampled from the toads’ native range in French Guiana, the introduced population in Hawai’i that was the source of

Australian founders, and Australia; (2) RNA-Seq data generated from brains sampled more extensively in Hawai’i and Australia, and (3) previously published RADSeq data from transects across Australia. We found that toads form three genetic clusters: the first encompassing native range toads, the second encompassing toads from the source population in Hawai’i and long- established areas near introduction sites in Australia, and the third encompassing toads from more recently established northern Australian sites. In addition to strong divergence between native and invasive populations, we find evidence for a reduction in genetic diversity after introduction. However, we do not see this reduction in loci putatively under selection, suggesting that genetic diversity may have been maintained at ecologically relevant traits, or that mutation rates were high enough to maintain adaptive potential. Nonetheless, cane toads encounter novel environmental challenges in Australia and appear to respond to selection across environmental breaks; the transition between genetic clusters occurs at a point along the invasion transect where temperature rises and rainfall decreases. We identify loci known to be involved in resistance to heat and dehydration that show evidence of selection in Australian toads. However, we also find evidence of genetic drift and potentially spatial sorting. Despite well-known predictions

71 regarding genetic drift and spatial sorting during invasion, this study highlights that natural selection occurs rapidly and plays a vital role in shaping the structure of invasive populations.

Keywords: Rhinella marina; Bufo marinus; cane toad; evolution; invasive species; RNA-Seq

72 Introduction The genetic paradox of invasion (Allendorf, 2003) describes a phenomenon that challenges widespread evidence of the relationship between genetic diversity and adaptive potential. High genetic diversity within a population is beneficial because it likely underlies phenotypic variation across individuals, allowing the population to respond to selection imposed by environmental change (Frankham, 2005; Reed & Frankham, 2003). Furthermore, a greater number of alleles confers an increased frequency of heterozygosity, which is often associated with population fitness (Reed & Frankham, 2003). Small or isolated populations with low genetic diversity have been shown to suffer declines due to inbreeding depression and the associated reduction of individual fitness (Blomqvist et al., 2010; Madsen et al., 1999; Westemeier et al., 1998).

Conservation efforts to salvage such populations by introducing conspecific individuals from allopatric populations (thereby introducing new alleles into the population; “genetic rescue”) have been successful, suggesting that the maintenance of genetic diversity can be crucial for population viability (Madsen et al., 1999; Westemeier et al., 1998).

Despite the fact that invasive populations are thought to undergo genetic bottlenecks (due to the translocation of a small number of founders from their native range to an introduced range

(Allendorf, 2003; Barrett & Kohn, 1991)), invasive species are also characterized by their ability to establish and spread in their introduced ranges. Invasion success is commonly linked to rapid evolution, including adaptation to novel environmental conditions over short timescales (Franks

& Munshi-South, 2014; Gao et al., 2018; Leydet et al., 2018; Whitney & Gabler, 2008).

Additionally, some invaders exhibit novel phenotypic traits that enhance invasive potential, such as increased growth and dispersal rates (White et al., 2013; Whitney & Gabler, 2008). There are many examples of evolutionary change during invasion without high levels of genetic diversity

(Rollins et al., 2013).

73 Although low genetic diversity may limit the ability of an invasive population to respond to natural selection, rapid evolution can also occur through non-adaptive processes: (1) genetic drift may occur on range edges, resulting in a reduction in genetic diversity across an introduced range (Rollins et al., 2009); (2) due to spatial sorting, the invasion front may be inhabited exclusively by the individuals with the highest dispersal rates (even if they are not the fittest) because they have arrived first and can only breed with each other (Shine et al., 2011), resulting in a geographic separation of phenotypes (Hudson et al., 2016b; Shine et al., 2011); (3) admixture or hybridization may occur between individuals from different introductions or sources (Mader et al., 2016).

It has recently been suggested that the genetic paradox of invasion may be rare; some invasive populations do not suffer a reduction in genetic diversity during introduction, and others do not face novel adaptive challenges in their introduced ranges (Estoup et al., 2016).

Furthermore, some invasive systems demonstrate a ‘spurious’ paradox due to inadequate estimation of genetic diversity (i.e. too few markers), or to maintenance of genetic diversity only at ecologically relevant traits, or to a reduction in genetic diversity resulting from natural selection rather than from genetic bottlenecks (Estoup et al., 2016). These ideas can be tested with genome-wide data (Wellband et al., 2018; Willoughby et al., 2018).

The Australian cane toad (Rhinella marina) is an invader that has exhibited rapid evolution despite apparently low genetic diversity (Rollins et al., 2015). Cane toads were serially translocated from South America to Puerto Rico to Hawai’i before 101 founders were introduced from the Hawai’ian island of Oahu to Queensland (QLD), Australia in 1935 (Turvey, 2009).

Toads have since spread westward through the Northern Territory (NT) into Western Australia

(WA; Figure 1; Covacevich & Archer, 1975). The native range and source area are tropical

74 environments with relatively high mean annual rainfall (3200 mm in French Guiana, 1200-2000 mm in Hawai’i) and high temperatures (26C in French Guiana, 22-24C in Hawai’i; Hijmans,

2015). The Australian range is heterogeneous in several environmental factors, particularly aridity; on average, sites in QLD receive more annual rainfall than do those in the NT and WA

(2000-3000 mm in QLD, 400-1000 mm in NT and WA; Figure S1A), and have lower annual mean temperatures (21-24C in QLD, 24-27C in NT and WA; Figure S1B; Bureau of

Meteorology, 2018). Heritable phenotypic differences in behavior (Gruber et al., 2017), thermal performance (Kosmala et al., 2018a), morphology (Hudson et al., 2016a; Hudson et al., 2018), and immune function (Brown et al., 2015c) have been documented in cane toads at different localities across the Australian range. Nonetheless, surveys have reported low levels of genetic diversity, based on microsatellites (Leblois et al., 2000) and MHC (Lillie et al., 2014), and only one mitochondrial haplotype in the ND3 gene of 31 individuals sequenced from Hawai’i and

Australia (Slade & Moritz, 1998).

75

Figure 1. Map of two invaded territories (Australia and Hawai’i) and the native range (French Guiana) of the cane toad (Rhinella marina). The shaded region represents the toad’s current Australian range, which is continuing to expand. Black diamonds in French Guiana, Hawai’i, and Australia indicate collection sites for our RNA-Seq experiment. White circles in Australia indicate collection sites for the RADSeq experiment from Trumbo et al. (2016).

Here, we assessed genetic diversity and population structure using single nucleotide polymorphisms (SNPs) identified in RNA-Seq data from samples originating from French

Guiana (the “native range”), Hawai’i (the “source”), QLD (the “range core”), NT

(“intermediate” areas), and WA (the “invasion front”) to test whether the paradox is supported in the cane toad invasion, and to investigate the evolutionary dynamics during introduction. We predicted high divergence between native and invasive populations due to a combination of genetic drift, selection, and spatial sorting, but little genetic differentiation within invasion phases due a putative lack of standing genetic diversity. Analyses to identify loci with outlier FST values or associations with environmental variables allow us to test for selection; because of the increase in aridity at intermediate areas and the invasion front, we predicted that loci under

76 selection would be in genes involved in thermal tolerance, which may underlie the success that toads have had dispersing through these areas.

RNA-Seq data provide genome-wide information, but are limited to transcribed regions of the genome (Davey & Blaxter, 2011). RNA-Seq data may therefore contain variants that exhibit strong signals for both demography and selection. Conversely, reduced representation sequencing approaches such as restriction site-associated DNA sequencing (RADSeq) data provide genome-wide information on a subset of all coding and non-coding sequences (only

SNPs near restriction enzyme cleavage sites are detected); because non-coding sequences make up most of the genome, SNPs detected using this technique are more likely to represent neutral variation (Wang et al., 2009). To focus on how selection may shape genomic diversity in invasive toads, we compared the results from our RNA-Seq data to those from a publicly available RADSeq dataset (Trumbo et al., 2016) that we reanalyzed (the purpose of that study was not to analyze population structure per se, so we modified the analysis for our questions).

The RADSeq dataset was limited to samples from QLD (the “range core” towards

“intermediate” areas) and the border of NT/WA (“intermediate” areas towards the “invasion front”).

Materials and Methods Sample collection, RNA extraction, and sequencing

In November of 2017, we collected samples from French Guiana (native range; Figure 1; Table

S1). In June of 2015, we collected samples from the Hawai’ian Islands (source; Figure 1; Table

S1). In April and May of 2013-2015, we collected samples along an invasion transect in

Australia (range core, intermediate areas, and the invasion front; Figure 1; Table S1). Upon collection, we excised whole spleen tissue (sample sizes in Table S1A) and whole brain tissue

77 (sample sizes in Table S1B) from female toads immediately after euthanasia. Each tissue sample was initially preserved in RNAlater buffer (QIAGEN, USA), initially kept at -20ºC during fieldwork (less than one month), and then drained and transferred to a -80°C freezer for long- term storage.

We carried out RNA extractions using the RNeasy Lipid Tissue Mini Kit (QIAGEN,

USA) following the manufacturer’s instructions, with an additional genomic DNA removal step using on-column RNase-free DNase treatment (QIAGEN, USA). We quantified the total RNA extracted using a Qubit RNA HS assay on a Qubit 3.0 fluorometer (Life Technologies, USA).

Extracts were then stored at -80◦C until sequencing. Sequencing was conducted commercially at

Macrogen (Macrogen Inc., ROK). mRNA libraries were constructed using the TruSeq mRNA v2 sample kit (Illumina Inc., USA), which included a 300bp selection step. We sequenced 46 spleen samples and 72 brain samples in multiple batches (not pooled) across 3 lanes of Illumina HiSeq

2500. Capture of mRNA was performed using the oligo dT method, and size selection parameter choices were made according to the HiSeq2500 manufacturer’s protocol. Each individual sample was ligated with a unique barcode. Overall, this generated 709 million paired-end 2 x 125-bp reads from spleen data and 1.74 billion paired-end 2 x 125-bp reads from brain data. Raw sequence reads are available as FASTQ files in the NCBI short read archive (SRA) under the

BioProject Accession PRJNA395127 (spleen data from Australia) and PRJNA479937 (all brain data).

Data pre-processing and alignment

First, we examined per base raw sequence read quality (Phred scores) and GC content, and checked for the presence of adapter sequences for each sample using FastQC v0.11.5 (Andrews,

78 2010b). We then processed raw reads (FASTQ format) from each sample with Trimmomatic v0.35 (Bolger et al., 2014), using the following parameters: ILLUMINACLIP:TruSeq3-

PE.fa:2:30:10:4 SLIDINGWINDOW:5:30 AVGQUAL:30 MINLEN:36. This removed any adaptor sequences, trimmed any set of five contiguous bases with an average Phred score below

30, and removed any read with an average Phred score below 30 or sequence length below 36 bp.

As a reference, we used the annotated R. marina transcriptome (Richardson et al., 2018), which was constructed from brain, spleen, muscle, liver, ovary, testes, and tadpole tissue. We conducted per sample alignments of reads (FASTQ files) to the reference using STAR v2.5.0a

(Dobin et al., 2013) in basic two-pass mode with default parameters, a runRNGseed of 777, and specifying binary alignment map (BAM) alignment outputs. As STAR-generated BAM files lack read groups (identifiers for reads that specify the individual that they come from and the platform that was used to sequence them), we added them to our BAM files using the

AddOrReplaceReadGroups tool in Picard Tools (Institute, 2018). To avoid making incorrect variant calls, we removed duplicate reads (reads that map to the same position of the same gene in the same individual) using the MarkDuplicates tool in Picard Tools. To split reads containing an N (region of the reference that is skipped in a read) into individual segments, we used the SplitNCigarReads tool in the Genome Analysis Toolkit (GATK) v3.8.0 (McKenna et al.,

2010).

Variant calling and filtering

To call SNPs and insertion-deletions (indels), we used the HaplotypeCaller tool in GATK

(McKenna et al., 2010) on our alignment (BAM) files. This tool works by identifying regions of

79 the reference that show evidence of variation (called ‘active regions’). Variants required a minimum phred-scaled confidence of 20 to be called (marked as passing the quality filters) and emitted (reported in the output), using the stand_call_conf and stand_emit_conf options.

To ensure that SNPs were called as accurately as possible, we performed this in ‘GVCF’ mode (which generates one intermediate genomic variant call format, or ‘gVCF’ file per individual sample). By using this mode, we avoided missing SNPs at loci that match the reference in some but not all individuals. We then used the GenotypeGVCFs tool to merge the gVCF files, re-calculate genotype likelihoods at each SNP locus across all individuals, and re- genotype and re-annotate all SNP loci. The results were written to one merged VCF file.

Although we initially genotyped all spleens and brains together, we discovered during downstream analyses that there was an effect of tissue type on population assignment – even from the same individuals, and using only SNPs from transcripts expressed in both tissues, spleen and brain samples were assigned to separate populations). Thus, we genotyped spleens and brains separately, resulting in two merged VCF files, and subsequently kept them separate for all downstream analyses. We retained both datasets because each provides a unique benefit: the spleen dataset includes native range samples, but the brain dataset has more extensive sampling of the invasive populations.

Rather than following a random distribution, some SNPs are clustered. Clustered SNPs may not be independent, and are thus filtered. We used the VariantFiltration tool to identify and filter ‘clusters’ (sets of 3 SNPs that appear within a window of 35 bases) in each of our merged

VCF files. We also used this tool to filter variants with QualByDepth (QD; variant confidence divided by the unfiltered depth of non-reference samples) less than 2.0, depth of coverage (DP) less than 20.0, and allele frequency less than 0.05. We then subset our VCF files to include only

80 the variants that passed the filters we set in the VariantFiltration step using the SelectVariants tool. This resulted in 803,489 SNPs from spleen data and 818,536 SNPs from brain data. We used bcftools (Li et al., 2009) to further filter the VCF files, only retaining biallelic SNPs. We examined the results of filtering for minimum minor allele frequency (min MAF) thresholds of

0.01 or 0.05, and several missing data tolerance (MDT) thresholds (the maximum percentage of individuals in the dataset in which a genotype for a locus can be absent without that locus being filtered out). Our population structure results were consistent across min MAF and MDT thresholds, so we ultimately chose to filter our data at min MAF = 0.05 and MDT = 0% (no missing data tolerated) because some downstream analyses cannot handle missing data. These filtering steps reduced the number of SNPs to 65,195 in spleen data and 35,842 in brain data

(Table S1A-B).

Inference of population structure

We used PLINK (Purcell et al., 2007) to convert our VCF files to the Browser Extensible Data

(BED) format, which is readable to fastStructure. We then used fastStructure (Raj et al., 2014) to infer population structure using a variational Bayesian framework for calculating posterior distributions, and to identify the number of genetic clusters in our dataset (K) using heuristic scores (Raj et al., 2014). We ran the structure.py ten times each (K= 1 to 10). We then took the resulting meanQ files from fastStructure and plotted them using the pophelper package (Francis,

2016) in R (Team, 2016).

In addition to using fastStructure for population assignment, we also performed a

Redundancy Analysis (RDA) using the vegan package (Oksanen et al., 2018) in R, which fits genetic and environmental data to a multivariate linear regression, and then performs a principal

81 component analysis (PCA) on the fitted values. RDA visualizes both population structure and the effects that environmental variables may have in shaping it. To do this, we downloaded climatic data on French Guiana, Hawai’i, and Australia from the Bioclim database (Hijmans et al., 2005) using the raster package (Hijmans, 2015). Because these areas vary in aridity, we downloaded data on rainfall and temperature; these data are averages of annual statistics over the period of

1970 to 2000. Specifically, we downloaded: annual mean temperature, maximum temperature of the warmest month, minimum temperature of the coldest month, annual precipitation, precipitation of the wettest quarter, and precipitation of the driest quarter. We then used the vcfR package (Knaus & Grünwald, 2017) to convert our VCF files to the GENLIGHT format, which is readable to the vegan package.

Identification of candidate loci under selection

Loci that are under natural selection may have abnormally high or low FST values, causing them to be outliers among all other loci in the transcriptome. Bayescan v2.1 (Foll & Gaggiotti, 2008) detects loci with outlier FST values using the multinomial-Dirichlet model. We used PGDSpider

(Lischer & Excoffier, 2012) to convert our VCF files to the BAYESCAN format, and then ran

Bayescan to detect loci with outlier pairwise FST values between our three genetic clusters in spleen data (native range toads versus source and core toads versus intermediate and frontal toads) and our two genetic clusters in brain data (source and core toads versus intermediate and frontal toads).

Loci under natural selection also may have allele frequencies that are associated with environmental variables. We used the climatic data that we downloaded from the Bioclim database (Hijmans et al., 2005) using the raster package (Hijmans, 2015) on rainfall during the

82 driest quarter and maximum temperature in the warmest month. We then used the lfmm v2.0 package (Frichot & Francois, 2015) in R to perform a latent factor mixed model (LFMM) to test the association between the allele frequencies of every locus and these two environmental variables. We applied a Benjamini-Hochberg correction to all p-values from the LFMM.

Because these scans reveal thousands of loci putatively under selection, and because outlier tests have high false positive rates (Whitlock & Lotterhos, 2015), we took a more conservative approach by cross-matching our list of FST outliers with each of our lists of environmentally-associated loci. For individual investigation, we only looked at loci with both an outlier FST value and an association with a putatively significant environmental variable through

LFMM.

Evaluation of genetic differentiation and diversity

To quantify levels of genetic differentiation and diversity, we computed basic statistics in the hierfstat (Goudet, 2005) and diveRsity (Keenan et al., 2013) packages in R (Team, 2016). We first used PGDSpider v2.1.1.3 (Lischer & Excoffier, 2012) to convert our VCF files to the

FSTAT and GENEPOP formats, which are readable to hierfstat and diveRsity, respectively. We then used these packages to calculate measures of genetic differentiation, including global FST and pairwise FST (by genetic cluster, invasion phase, and collection site), and genetic diversity, such as expected heterozygosity (He) and rarefied allelic richness (AR). We also computed

Shannon’s Information Index (SI) with the dartR package (Gruber et al., 2018). After calculating these measures of diversity across all loci (N=65,195 from spleen data, N=35,842 from brain data), we calculated the same measures in loci putatively under selection: those with outlier FST values (N=648 from spleen data, N=203 from brain data) and those associated with

83 environmental variables (N=4179 from spleen data, N=530 from brain data). In spleen data, this allowed us to investigate the hypothesis that genetic diversity is maintained at ecologically relevant traits even if genome-wide diversity is lost. In brain data, this allowed us to examine the effect of natural selection on genetic diversity within Hawai’i and Australia. We used Kruskal-

Wallis tests to assess the significance of the differences in genetic diversity.

Annotation of SNPs

To visualize the types of genomic regions in which our SNPs lie, we used SnpEff (Cingolani et al., 2012). SnpEff analyzes the information in a VCF file, annotates the SNPs, and estimates their effects. We calculated the relative proportions of each type of SNP for both tissue types using the full dataset of SNPs, and our two selection candidate datasets (containing FST outlier loci associated with an environmental variable). We then used the stats v3.5.0 R package to perform z-tests on the differences in proportions of SNP types between the full and selection candidate datasets. We expected differences in the proportions of certain types of SNPs, such as a lower proportion of synonymous variants in the selection candidates because synonymous mutations do not result in a codon change (and thus are unlikely to change function, i.e. phenotype).

Isolation by distance

To examine the effects of geographic distance on genetic distance across the Australian range, we performed a Mantel test using ade4 v1.7-5 (Thioulouse & Dray, 2007). Because we were focused on isolation by distance through range expansion after introduction (to Australia), samples from Hawai’i and French Guiana were excluded; and because we sampled brain tissue in a greater number of sites within Australia (as opposed to spleen tissue; Table S1A-B), we only

84 performed the Mantel test on brain data. For the geographic distance matrix, we used the dist function in R (Team, 2016) to calculate the Euclidean distances in geographic space between collection sites based on their coordinates. For the genetic distance matrix, we used the collection site-based pairwise FST values generated from hierfstat. We performed four Mantel tests, using:

(1) all loci, (2) loci with outlier FST values, (3) loci with outlier FST values and with an environmental association (temperature or rainfall), (4) loci with outlier FST values but without an environmental association. A significant result across all loci may suggest that the prominent driver of genetic structure is genetic drift. However, clinal variation in allele frequencies may also result from spatial sorting (particularly in loci underlying dispersal ability), or from selection associated with environmental clines; distinguishing these evolutionary processes is difficult.

Thus, we performed the additional tests to attempt to separate the effects of genetic drift and spatial sorting. Loci with outlier FST values are candidates for natural selection and spatial sorting; thus, a significant result from only these loci would less likely be caused by genetic drift.

Outlier loci can be further separated into those with an environmental association, which are likely candidates for natural selection driven by environmental variables, and those without an environmental association, which are possible candidates for spatial sorting. If spatial sorting is indeed influencing genetic differentiation, then we may expect to see different patterns between these two groups of loci (e.g. linearity only in the outlier loci without an environmental association).

Acquisition, processing, and analysis of RADSeq dataset

In addition to our own RNA-Seq dataset, we downloaded a publicly available RADSeq dataset

(Trumbo et al., 2016) to evaluate population structure in Australian cane toads. Unlike RNA-Seq

85 data, which consists of coding sequences only (Wang et al., 2009), RADSeq data include SNPs from across the entire genome, but only around sites that are cleaved by a selected set of restriction enzymes (Davey & Blaxter, 2011). This method allows us to contrast the results of assessing loci potentially under ‘tighter’ selection (e.g., codon bias) with those primarily constituting neutral variation (RADSeq). Individuals in the RADSeq dataset were collected from

QLD (range core towards intermediate areas, N=179) and the border of NT/WA (intermediate areas towards the invasion front, N=441; Figure 1). However, samples were lacking from a large part of the range between those regions; thus, identifying geographic points of differentiation between genetic clusters may be unlikely using this dataset. Furthermore, although there were many collection sites within these regions, metadata with coordinates (or even location names) were unavailable. We downloaded raw 100-bp reads from NCBI short read archive (SRA) under the BioProject Accession PRJNA328156. We then converted SRA files to the FASTQ format using the fastq-dump tool in the SRA toolkit.

We used Stacks v2.0 (Catchen et al., 2013) for all RADSeq data processing. First, we used the process_radtags program to remove low quality reads from the FASTQ files. Next, we used the denovo_map pipeline to perform a de novo assembly for each sample, align matching

DNA regions across samples (called ‘stacks’), and call SNPs using a maximum likelihood framework. We allowed a maximum of five base mismatches between stacks within an individual and three base mismatches between stacks between individuals. This resulted in

499,623 SNPs. We then filtered the results using the populations program, using one random

SNP per locus and four combinations of parameter choices (min MAF=0.05, MDT = 0.5; min

MAF = 0.05, MDT = 0.2; min MAF = 0.01, MDT = 0.5; min MAF = 0.01, MDT = 0.2).

Structure results were consistent across min MAF values, but not MDT values: two distinct

86 populations were only detected using the more stringent MDT choice. Because population structure in this dataset is relatively low, and because it is appropriate to use more stringent filtering choices when analyzing datasets with low population structure (Selechnik et al., unpublished; Chapter 2), we ultimately chose to filter this dataset with min MAF = 0.05 and

MDT = 0.2. Furthermore, using these filtering choices produced results that were most consistent with those reported by Trumbo et al. (2016) and with those of our RNA-Seq dataset. Filtering reduced the number of SNPs to 8,296. The results were written to a VCF file and a

STRUCTURE file (which could be read into fastStructure).

We ran fastStructure using the same methodology as we did for our RNA-Seq dataset. In addition to the model-based fastStructure, we also performed a discriminant analysis of principal components (DAPC) on this dataset using adegenet (Jombart & Ahmed, 2011). DAPC is a multivariate approach that identifies the number of genetic clusters using K-means of principal components and a Bayesian framework. We also converted the VCF to the FSTAT and

GENEPOP formats and computed the same basic statistics for the RADSeq dataset as we did for our RNA-Seq datasets using hierfstat and diveRsity. Because collection site coordinate metadata were unavailable for the RADSeq dataset, we did not perform a Mantel test using these data.

Results Population structure across the Hawai’ian-Australian invasion

We found three genetic clusters in our spleen data (Figure 2A): French Guiana (native range) formed its own genetic cluster, Hawai’i (source) clustered with QLD (core), and NT

(intermediate) clustered with WA (invasion front). These patterns were mirrored by our brain data, which did not include native range individuals but had more extensive sampling in Hawai’i and Australia than did the spleen data. In our brain dataset, population differentiation seemed to

87 align with environmental barriers (Figure 2B-C): Hawai’i (source) clustered with coastal QLD

(core), whereas inland QLD/NT (intermediate) clustered with WA (invasion front). Inference of substructure within these two genetic clusters revealed that source and core toads may further differentiate into two separate groups (Figure 2D); however, intermediate and frontal toads remain one genetic cluster (Figure 2E).

88

Figure 2. (A) Population structure of cane toads from French Guiana, Hawai’i, and Australia, as inferred by SNPs from spleen RNA-Seq data. (B) Rainfall during the driest quarter in each location across collection sites for invasive cane toads in French Guiana, Hawaii, and Australia. A similar environmental barrier occurs in temperature (Figure S1). (C) Population structure of invasive cane toads from Hawai’i and Australia, as inferred by SNPs from brain RNA-Seq data. (D) Substructure within toads from Hawai’i (HI) and Queensland (QLD) as inferred by SNPs from brain RNA-Seq data. (E) Substructure within toads from Northern Territory (NT) and Western Australia (WA) as inferred by SNPs from brain RNA-Seq data.

89 RDA was mostly consistent with fastStructure for both spleen (Figure 3A) and brain

(Figure 3B) data, except for the differentiation of source toads and core toads as two separate genetic clusters. It also suggested that: (1) source toads may have diverged from toads of other clusters due to selection from rainfall during the driest quarter. (2) Core toads may have diverged from toads of other clusters selection from due to rainfall during the wettest quarter, mean annual rainfall, and minimum temperature during the coolest month. (3) Intermediate and frontal toads may have diverged from toads of other clusters due to selection from maximum temperature during the hottest month and mean annual temperature.

Figure 3. (A) Relationship between environmental variables and population structure of cane toads from French Guiana, Hawai’i, and Australia, as inferred by Redundancy Analysis (RDA) on SNPs from spleen RNA-Seq data. Environmental data include: annual mean temperature (Mean Temp), maximum temperature of the warmest month (Max Temp), minimum temperature of the coldest month (Min Temp), annual precipitation (Mean Rain), precipitation of the wettest quarter (Wet Rain), and precipitation of the driest quarter (Dry Rain). (B) Relationship between environmental variables and population structure of invasive cane toads from Hawai’i and Australia, as inferred by RDA on SNPs from brain RNA-Seq data. Points in the centre of the axes in each panel represent loci.

90 Tests for selection

To identify loci putatively under selection, we searched for those with outlier FST values or associations with environmental variables. Out of 65,195 SNPs from spleen data (N=46 individuals across native and invasive populations), 648 were identified as FST outliers, with a mean global FST of 0.89. Mean pairwise FST values were 0.93 between the native and source/core populations, 0.97 between the native and intermediate/front populations, and 0.34 between the source/core and intermediate/front populations. From the same dataset (of 65,195 SNPs), 3088

SNPs were associated with maximum temperature during the warmest month and 1813 were associated with rainfall during the driest quarter. Of these loci, 772 were associated with both environmental variables. We focused on SNPs that were detected both as FST outliers and environmental correlates: there were 64 outlier SNPs associated with maximum temperature and

47 outlier SNPs associated with rainfall in the driest quarter; 18 outliers were associated with both environmental variables.

Most of the 64 outlier SNPs associated with maximum temperature during the hottest month (Table S2A) were in transcripts with functions such as metabolism, transcription regulation, and immune function. There were also two SNPs in HSP4, a gene encoding heat shock protein (HSP) 4 (70 kDa), which is involved in thermal tolerance (on exposure to heat stress) and protein folding and unfolding (Consortium, 2017). Most of the 47 outlier SNPs associated with rainfall during the driest quarter (Table S3A) were in transcripts with functions such as cell signaling and immune function.

Out of 35,842 SNPs from brain data (N=72 individuals across invasive populations only),

203 were identified as FST outliers, with a mean global FST of 0.34. Mean pairwise FST was 0.50 between the source/core and intermediate/front populations. From the same dataset (of 35,842

91 SNPs), 345 SNPs were associated with maximum temperature during the warmest month and

194 were associated with rainfall during the driest quarter. Nine of these loci were associated with both environmental variables. Again, we focused on SNPs that were detected both as FST outliers and environmental correlates: there were 38 outlier SNPs associated with maximum temperature and 33 outlier SNPs associated with rainfall in the driest quarter; no outliers were associated with both environmental variables.

Most of the 38 outlier SNPs associated with maximum temperature during the hottest month (Table S2B) were in genes involved in cell signaling, mitochondrial processes

(metabolism), or gene expression (Consortium, 2017). Similarly to the spleen data, there was one

SNP from brain data in HSP4. Another SNP was in ARNT2, a gene encoding aryl hydrocarbon receptor nuclear translocator 2, a transcription factor involved in the hypothalamic pituitary adrenal (HPA) axis and visual and renal function (Consortium, 2017). Of the 33 outlier SNPs from brain data correlated with rainfall during the driest quarter (Table S3B), 19 were in

MAGED2, a gene encoding melanoma-associated antigen D2, which is involved in renal sodium ion absorption (Consortium, 2017). Four were in STXBP1, a gene encoding syntaxin-binding protein 1, which is involved in platelet aggregation and vesicle fusion (Consortium, 2017). The rest were in genes generally involved in gene expression and cell signaling (Consortium, 2017).

Evaluation of genetic differentiation, diversity, and isolation by distance

Our spleen data suggest that divergence from the native population is strong in the source/core population (pairwise FST = 0.29), and even stronger in the intermediate/frontal population

(pairwise FST = 0.32). Differentiation between the source/core and intermediate/frontal

92 populations is much lower (pairwise FST = 0.08). Overall divergence among natives and invaders is moderately high (global FST = 0.18)

These trends are consistent with our brain data, in which genetic differentiation between the source/core and intermediate/frontal populations is also low (pairwise FST = 0.07). At the level of substructure within the two invasive groups, differentiation between source and core toads was even lower (pairwise FST = 0.04 in brain data). Generally, pairwise FST by invasion phase (Table S4A) and collection site (Table S4B) from brain data revealed that toads from areas that are more closely linked by invasion history are less strongly differentiated. Because the brain data did not include native range samples, overall divergence was estimated to be much lower than it was in the spleen data (global FST = 0.04).

Diversity statistics from spleen data suggest that the native range population is only slightly more genetically diverse than the source/core population (native: He = 0.27, AR = 1.74,

SI = 0.40; source/core: He = 0.26, AR = 1.74, SI = 0.40; p = 5E-4 for He). There is a greater drop in diversity in the intermediate and frontal population (intermediate/front: He = 0.23, AR =

1.68, SI = 0.36; p < 2E-16 for all measures). However, we also estimated these measures by collection site (rather than by genetic cluster) to detect changes in genetic diversity at a finer scale; these calculations reveal a more obvious reduction in genetic diversity from the native population to all invasive collection sites (Table 1).

93 Table 1. Expected heterozygosity (He), allelic richness (AR), and Shannon’s Information Index (SI) estimated using SNPs from spleen RNA-Seq data on native and invasive populations of cane toads (Rhinella marina) at collection sites across French Guiana, Hawai’i, and Australia. Environmentally All loci Outlier FST loci associated loci Collection site (He, AR, SI) (He, AR, SI) (He, AR, SI) French Guiana (native) 0.27, 1.66, 0.40 0.03, 1.25, 0.06 0.17, 1.53, 0.25 Kapolei Park (source) 0.23, 1.60, 0.34 0.07, 1.20, 0.10 0.28, 1.77, 0.41 Haiku Gardens (source) 0.23, 1.60, 0.34 0.07, 1.22, 0.11 0.28, 1.75, 0.41 Gordonvale (core) 0.24, 1.63, 0.36 0.07, 1.21, 0.10 0.30, 1.84, 0.45 Daintree (core) 0.24, 1.62, 0.35 0.08, 1.24, 0.12 0.29, 1.79, 0.44 Cape Crawford (intermediate) 0.22, 1.59, 0.32 0.02, 1.06, 0.03 0.22, 1.59, 0.32 Timber Creek (intermediate) 0.21, 1.58, 0.32 0.02, 1.05, 0.03 0.22, 1.60, 0.32 Halls Creek (front) 0.21, 1.57, 0.32 0.02, 1.05, 0.03 0.19, 1.58, 0.30 Durack River (front) 0.22, 1.58, 0.32 0.03, 1.11, 0.05 0.20, 1.60, 0.31

Despite a slight reduction in genetic diversity from the native to invasive populations genome-wide, our spleen data suggest that diversity was not lost during introduction in subsets of loci putatively under selection. Among loci with outlier FST values, there is no reduction in genetic diversity from the native population to the source/core population; rather, it is higher in the source/core population (native: He = 0.03, AR = 1.25, SI = 0.06; source/core: He = 0.09, AR

= 1.37, SI = 0.14; p < 2E-16 for all measures). However, diversity in these loci is reduced in the intermediate/front population (intermediate/front: He = 0.02, AR = 1.14, SI = 0.04). These trends are also shown in environmentally-associated loci (native: He = 0.17, AR = 1.53, SI = 0.25; source/core: He = 0.37, AR = 1.99, SI = 0.55; intermediate/front: He = 0.23, AR = 1.81, SI =

0.35). Calculations of these measures by collection site further support these findings (Table 1).

Similarly to the spleen data, genome-wide diversity statistics from brain data suggest that the source/core population is more diverse than the intermediate/frontal population (source/core:

He = 0.34, AR = 2.00, SI = 0.51; intermediate/front: He = 0.31, AR = 1.97, SI = 0.47; p < 2E-16 for comparisons of all measures). Calculations by invasion phase (Table S5) suggest a reduction

94 in diversity from core to invasion front. This reduction is greater in loci putatively under selection: in those with outlier FST values, the source/core population is much more diverse than the intermediate/front population (source/core: He = 0.38, AR = 2.00, SI = 0.55; intermediate/front: He = 0.15, AR = 1.70, SI = 0.24; p < 2E-16 for comparisons of all measures).

These trends are also shown in environmentally-associated loci (source/core: He = 0.35, AR =

2.00, SI = 0.53; intermediate/front: He = 0.08, AR = 1.57, SI = 0.15; p < 2E-16 for comparisons of all measures). Calculations of these measures by collection site further support these findings

(Table 2).

Table 2. Expected heterozygosity (He), allelic richness (AR), and Shannon’s Information Index (SI) estimated using SNPs from brain RNA-Seq data on native and invasive populations of cane toads (Rhinella marina) at collection sites across Hawai’i and Australia. Environmentally All loci Outlier FST loci associated loci Collection site (He, AR, SI) (He, AR, SI) (He, AR, SI) Paradise Park (source) 0.28, 1.57, 0.41 0.31, 1.76, 0.45 0.30, 1.84, 0.45 Mauna Lani (source) 0.27, 1.56, 0.40 0.24, 1.61, 0.35 0.29, 1.73, 0.42 Kapolei Park (source) 0.29, 1.57, 0.43 0.29, 1.83, 0.43 0.25, 1.73, 0.38 Haiku Gardens (source) 0.30, 1.59, 0.44 0.31, 1.75, 0.45 0.31, 1.83, 0.46 Gordonvale (core) 0.30, 1.61, 0.44 0.36, 1.82, 0.51 0.22, 1.62, 0.33 Cairns (core) 0.32, 1.60, 0.48 0.38, 1.95, 0.55 0.24, 1.84, 0.38 Daintree (core) 0.29, 1.60, 0.43 0.32, 1.90, 0.48 0.18, 1.57, 0.28 Croydon (intermediate) 0.26, 1.60, 0.37 0.18, 1.45, 0.26 0.10, 1.24, 0.14 Burketown (intermediate) 0.28, 1.57, 0.41 0.18, 1.44, 0.26 0.09, 1.31, 0.14 Cape Crawford (intermediate) 0.28, 1.58, 0.42 0.15, 1.42, 0.23 0.08, 1.32, 0.13 Mataranka (intermediate) 0.28, 1.57, 0.41 0.16, 1.46, 0.24 0.07, 1.25, 0.11 Timber Creek (intermediate) 0.27, 1.55, 0.40 0.05, 1.18, 0.09 0.06, 1.22, 0.10 Purnululu (front) 0.29, 1.55, 0.44 0.20, 1.48, 0.12 0.07, 1.37, 0.11 Halls Creek (front) 0.27, 1.56, 0.40 0.09, 1.25, 0.14 0.06, 1.21, 0.10 Durack River (front) 0.27, 1.56, 0.41 0.13, 1.37, 0.19 0.06, 1.24, 0.10

Geographic distance in collection sites and pairwise FST values by collection site (as estimated with brain data) were significantly associated across all four of our Mantel tests using:

(1) all loci (p = 1E-3, r = 0.76; Figure 4A); (2) all loci with outlier FST values (p = 1E-3, r =

95 0.75; Figure 4B); (3) only loci with outlier FST values and an environmental association (p = 1E-

3, r = 0.62; Figure 4C); (4) only loci with outlier FST values but without an environmental association (p = 1E-3, r = 0.77; Figure 4D).

Figure 4. Relationship between geographic distance in collection sites (km) and genetic distance between cane toads from those sites (pairwise FST) across their invasive Australian range (Figure 1) as inferred by brain RNA-Seq data. (A) represents the full dataset of loci (p = 1E-3, r = 0.76). (B) represents all FST outliers (p = 1E-3, r = 0.75). (C) represents only FST outliers with an association with either rainfall or temperature (p = 1E-3, r = 0.62). (D) represents only FST outliers without an association with either rainfall or temperature (p = 1E-3, r = 0.77).

96 SNP Annotations

Annotation of SNPs revealed that most of our loci were either missense, synonymous, or found in the 3’ untranslated region (UTR). A small percentage were found in the 5’ UTR, and the remaining loci involved the loss or gain of stop or start codons (Figures S2-S3). In the spleen data, the temperature-associated outlier FST subset had a significantly higher proportion of synonymous variants (p = 8E-3) than the full set of SNPs. The outlier loci within HSP4 were both synonymous variants. However, there were no significant differences in proportions between the rainfall-associated outlier FST subset and the full set of SNPs.

In the brain data, there were no significant differences in proportions of SNP variant types between the temperature-associated outlier FST subset and the full set. The locus within

HSP4 was a 3’ UTR variant, and the locus within ARNT2 was synonymous. The rainfall- associated outlier FST subset had a significantly higher proportion of 3’ UTR variants (p = 5E-3) and lower proportion of missense variants (p = 0.04) than did the full set. Eighteen of the nineteen loci within MAGED2 were 3’ UTR variants (the last was missense), and two of the four loci within STXBP1 were missense (the other two were synonymous).

Population structure in Australian toads from the RADSeq experiment

Analysis of the RADSeq dataset through fastStructure and DAPC also showed two genetic clusters within Australia (Figure 5A-B): QLD (core towards intermediate areas) in the first group, and the border of NT/WA (intermediate areas towards the invasion front) in the second group. There was little differentiation between these two groups (pairwise FST = 0.09) or overall

(global FST = 0.04). We found no evidence of significant substructure within either of the two genetic clusters. The two groups were equal in allelic richness (AR = 2.00 for both groups), but

97 expected heterozygosity was higher in core toads than in intermediate/frontal toads (core: He =

0.50; intermediate/front: He = 0.33), which was consistent with our RNA-Seq results.

Figure 5. (A) Population structure of cane toads across their invasive range in Australia, as inferred from running fastStructure on SNPs from RADSeq data. Because coordinate metadata for collection sites were unavailable, the samples could not be arranged from east to west along the Australian range. (B) Population structure of cane toads across their invasive range in Australia, as inferred from DAPC on SNPs from RADSeq data.

Discussion We predicted that genetic diversity would be significantly lower in invasive populations than in native populations due to genetic bottlenecks taking place during the serial introductions of cane toads across Hawai’i and Australia. We also predicted strong divergence between native range and invasive populations due to a combination of genetic drift, selection and spatial sorting. In terms of population structure, we predicted low genetic structure within invasive populations in

98 Hawai’i and Australia resulting from putatively low levels of standing genetic variation. Our analysis shows evidence of a reduction in genetic diversity from native to invasive populations.

We found that the toads form three genetic clusters: (1) native range toads form their own cluster, (2) Hawai’ian source toads cluster with eastern Australian range core toads, and (3) all toads from further west along the recent invasion transect cluster together (intermediate through to the invasion front). Differentiation is much higher between the native population and the invasive populations than between the invasive populations. The loci with the highest differentiation between native and invasive populations (outlier pairwise FST values > 0.9 between native and invasive) have relatively low diversity across all populations; this may be because these loci are homozygous in most individuals with no shared alleles between populations (fixed). Associations between our genetic structure patterns and allele frequencies with environmental variables suggest that natural selection is occurring in the western part of the

Australian range. However, the significant result of our Mantel test across all loci also reveals gradual genetic drift across invasion phases; and the significant result and high correlation coefficient of our Mantel test of FST outliers without environmental associations may be due to spatial sorting. It seems likely that all these forces are responsible for the observed population structure within invasive populations.

Our data suggest that cane toads may fit the model of the genetic paradox; we found evidence of strong divergence between native and invasive cane toad populations, and of a reduction in overall genetic diversity. Previous studies on mtDNA (Slade & Moritz, 1998) and microsatellite (Estoup et al., 2001) data have also suggested losses in genetic diversity from the native range to Hawai’i and Australia. Despite the reduction in genetic diversity from native to invasive populations across the full set of loci in the spleen data, we did not find this reduction in

99 loci putatively under selection (loci with outlier FST values or associations with environmental variables) in the source/core population. This may provide support for the prediction that some invasions do not represent a genetic paradox despite overall loss of genetic diversity and the presence of novel adaptive challenges because genetic diversity at ecologically relevant traits is maintained by balancing selection (Estoup et al., 2016). However, diversity of this subset of loci is not just maintained in the source/core population; it is higher than in the native population.

This may represent a genuine paradox; de novo mutation rates may have been high enough to restore or even enhance adaptive potential in the source/core population (Estoup et al., 2016).

Alternatively, this may reflect admixture between toads from multiple introductions from different areas of the native range to Puerto Rico, where Hawai’ian toads were sourced from

(Turvey, 2009). From the source/core population to the intermediate frontal population, however, diversity of loci putatively under selection is highly reduced; we believe this is because of directional selection, as outlined below.

Our two invasive genetic clusters (source/core and intermediate/front) diverge at the transition from coastal to inland areas within Australia, in an area where temperatures become higher and rainfall becomes scarcer, particularly during the dry season (Bureau of Meteorology,

2018). This correlation suggests that these environmental variables may drive population structure and that toads may be adapting to local conditions. Furthermore, RDA implicates selection from temperature and rainfall in driving divergence between invasive populations.

Heterogeneous climates across introduced or expanding ranges have previously been linked to genetic differentiation, possibly due to climatically-imposed selection (Leydet et al., 2018;

Weber & Schmid, 1998).

100 Diversity estimates from our brain data suggest a reduction in diversity from the source/core population to the intermediate/frontal population in the full set of loci, but a much greater reduction in diversity in the subset of loci putatively under selection. The source

(Hawai’i) and core (QLD) areas represent earlier phases of the invasion and experience similar environmental conditions to those of the native range, whereas intermediate (NT) and invasion front (WA) areas are more arid. Thus, the greater depletion in diversity at loci putatively under selection across the Australian invasion may be due to the effects of directional selection from the harsher climate at more recently colonized areas of the Australian range (Andrews, 2010a;

Estoup et al., 2016).

Among our FST outliers associated with temperature are a few within a gene encoding protein HSP 4 (70 kDa). HSPs protect protein folding during increased temperatures and provide cells with enhanced thermal tolerance (Kiang & Tsokos, 1998), and expression of HSP genes has been shown to underlie adaptive responses to environmental stress (Chen et al., 2018). HSP levels in response to a manipulated thermal environment have been shown to vary between native and invasive cane toads, as well as between populations of toads within Australia

(Kosmala et al., 2018b). The SNPs within HSP4 included two synonymous variants (spleen data) and a 3’ UTR variant (brain data); 3’ UTR variants are known to alter levels of protein functioning through differential expression of genes (Skeeles et al., 2013). In the case of heat shock , adaptation may not necessarily mean a change in function, but rather a change in levels of the function; thus, a 3’ UTR variant is logical. Another notable protein was aryl hydrocarbon receptor nuclear translocator 2 (from the ARNT2 gene), a transcription factor that specifically acts on genes involved in the HPA axis and visual and renal function. The HPA axis is stimulated by high temperatures (Malmkvist et al., 2009), and the secretion of corticosterone

101 can be modulated adaptively in response to thermal change (Telemeco & Addis, 2014).

Elevation of corticosterone secretion in toads increases the rate of desiccation, thus making it maladaptive in arid environments (Jessop et al., 2013). Corticosterone levels in Brazilian and

Australian toads in response to manipulation of thermal environment display similar patterns to those of HSPs (Kosmala et al., 2018b); thus, it is unsurprising that SNPs associated with maximum temperature in the hottest month were found in genes involved with both of these molecules. However, the SNP in ARNT2 is synonymous; thus, this SNP might not change function.

More than half (19 of 33) of the FST outliers associated with rainfall (in brain data) were found in MAGED2 (encodes melanoma-associated antigen D2). This protein regulates the expression and localization of two co-transporters that facilitate renal sodium ion absorption, which prevents excess loss of water and solutes via reabsorption through the kidneys (to be returned to the extracellular fluid and circulatory system; Greger, 2000). Similarly to the SNP within HSP4, 18 of 19 SNPs within MAGED2 were 3’ UTR variants, which can affect levels of protein functioning (Skeeles et al., 2013). This is logical because similarly to HSP4, an adaptive change to the function of MAGED2 would likely be to its level rather than to the process itself.

The high number of SNPs found in the 3’ UTR of MAGED2 is likely responsible for the difference in proportions of variant types between the full dataset and the rainfall-associated outlier dataset. Four more SNPs were found in STXBP1 (encodes syntaxin-binding protein 1, which is involved in vesicle fusion and blood clotting). Dehydration is known to increase blood clotting rate (El-Sabban et al., 1996), and excessive blood clotting can be harmful to the heart, brain, and limbs (PubmedHealth, 2014). Two of the four SNPs in STXBP1 were missense variants, possibly producing change in protein function. Evolved changes in the rates of renal

102 sodium absorption and blood clotting may allow toads from intermediate and invasion front areas to survive in drier environments. The remaining SNPs associated with rainfall or temperature were in genes generally involved in gene expression or signal transduction; these genes may facilitate expression of (or signaling to) the genes we have discussed, or may have independent functions in cellular maintenance.

Invasive species have been shown to rapidly adapt phenotypically to heterogeneous climatic conditions, allowing them to expand their ranges (Colautti & Barrett, 2013). Some invaders tolerate higher temperatures (Braby & Somero, 2006) and water loss (Godoy et al.,

2011) better than do related taxa within the native range (Zerebecki & Sorte, 2011). There is evidence of this in cane toads: wild-caught individuals from Australia exhibit better locomotor performance under dehydrating conditions than do conspecifics from Hawai’i and the native range (Kosmala et al., 2017), and within Australia, toads from semi-arid areas (i.e. NT) exhibit better locomotor performance than do conspecifics from wetter areas (i.e. coastal QLD; Kosmala et al., 2017; Tingley et al., 2012). This trait is heritable; captive-bred toads with parental origins from hotter areas (northwestern Australia) outperform those with parental origins from cooler areas (northeastern Australia) at high (but not low) temperatures in a common-garden setting

(Kosmala et al., 2018a). Coupled with the previously shown heritable phenotypic patterns, our genetic structure results suggest that intermediate and frontal toads in Australia may be evolving enhanced thermal tolerance, thereby facilitating their continued westward range expansion

(Szucs et al., 2017). However, experiments pairing genetic and phenotypic data from the same individuals are required to test this hypothesis.

Although our patterns of genetic structure (cluster separation along a distinct climatic gradient) and the identities of loci putatively under selection suggest that invasive toads are

103 responding to localized natural selection, our data also support the presence of genetic drift. The significant result of our Mantel test using all loci suggests that isolation by distance is occurring in the Australian toad population. Slight reductions in diversity (AR and He) along the invasion trajectory found here are expected and have been seen in other invasions (Rollins et al., 2009).

However, the results of our additional Mantel tests also suggest that another evolutionary force may be at work; spatial sorting is the separation of individuals within a range-expanding population along phases of their expansion based on their dispersive capabilities (Shine et al.,

2011). This may lead to clinal variation in dispersal-related allele frequencies along the range.

Genetic differentiation in our loci with outlier FST values (which are candidates for natural selection and spatial sorting) is significantly associated with geographic distance. Furthermore, separating FST outlier loci with an environmental association (likely candidates for natural selection) from those without (possible candidates for spatial sorting) yields different patterns.

Although the Mantel tests on both datasets are significant, the test on the environmentally associated outliers unsurprisingly shows two discrete groups of pairwise FST values – low pairwise FST values at low geographic distances and high pairwise FST values at high geographic distances. Interestingly, the test on outliers without an environmental association shows a clinal increase in pairwise FST values with geographic distance (like those on the full dataset and the full outlier dataset), which may suggest that some of these loci are influenced by spatial sorting.

Our results from the RADSeq dataset (Trumbo et al., 2016) were similar to those from our RNA-Seq dataset; both showed two genetic clusters within Australia. However, unlike the

RNA-Seq results (which show a divide occurring in inland QLD, where temperature increases and rainfall decreases), the RADSeq results cluster all QLD toads in the first group, and NT/WA toads in the second group. In the RADSeq dataset, the spatial location of the transition between

104 the two groups is difficult to pinpoint due to the absence of intermediate sampling sites. The difference in group divisions between the two datasets may be because selection from environmental variables causes differentiation between toads from coastal and inland QLD (as seen in the RNA-Seq data), whereas demographic processes such as gene flow may lessen differentiation between them (as seen in the RADSeq data). Alternatively, toads from coastal and inland QLD may cluster together in the RADSeq data because the sampling sites in that study were south of those in our study, and the environmental conditions of coastal and inland sites are more similar in that area (Hijmans et al., 2005). Estimates of diversity from the RADSeq dataset were slightly higher than those from the RNA-Seq dataset; this may be because neutral variation is expected to be higher than adaptive variation due to the effects of directional and purifying selection (Andrews, 2010a).

In conclusion, it appears that serial introductions have led to a reduction in genetic diversity of invasive cane toads. Nonetheless, cane toads do indeed face novel adaptive challenges in their introduced ranges, and are able to respond genetically to natural selection, particularly in the harsh environmental conditions of northwestern Australia. The ability of toads to adapt to these conditions may reflect the maintenance of diversity at ecologically relevant loci, or it may reflect sufficiently high mutation rates to bolster adaptive potential; thus, it remains unclear whether the toad invasion represents a true genetic paradox. Natural selection, genetic drift, and spatial sorting are the most plausible mechanisms by which genetic differentiation can occur, but require further investigation to tease apart. It may be useful to measure phenotypic traits with the occurrence of certain SNPs (such as the candidates that we have identified) within the same individuals. Additionally, genetic changes may not be the only mechanisms by which

105 adaptation occurs; studies on plasticity and epigenetic changes may also be useful for uncovering the mysteries of rapid evolution.

Acknowledgements

This work was supported by the Australian Research Council (FL120100074, DE150101393) and the Equity Trustees Charitable Foundation (Holsworth Wildlife Research Endowment). We thank Olivier Francois and Natalie Hofmeister for their input during discussions about our analyses. We thank Cam Hudson, Joachim Ehlenz, Simon Ducatez, Jayna DeVore, Serena Lam, and Chris Jolly for their assistance with sample collection. We thank Andrea West for her assistance during RNA extractions. All procedures involving live animals were approved by the

University of Sydney Animal Care and Ethics Committee.

Literature Cited

Allendorf F.W. (2003) Introduction: Population Biology, Evolution, and Control of Invasive

Species. Conservation Biology 17, 24 - 30.

Andrews C.A. (2010a) Natural Selection, Genetic Drift, and Gene Flow Do Not Act in Isolation

in Natural Populations. Nature Education Knowledge 3, 5.

Andrews S. (2010b) FastQC: a quality control tool for high throughput sequence data.

http://www.bioinformatics.babraham.ac.uk/projects/fastqc

Barrett S.C.H., Kohn J.R. (1991) Genetics and conservation of rare plants. In: Genetic and

evolutionary consequences of small population size in plants: implications for

conservation eds. Falk DA, Holsinger KE), pp. 3 - 30. Oxford University Press, New

York.

106 Blomqvist D., Pauliny A., Larsson M., Flodin L.A. (2010) Trapped in the extinction vortex?

Strong genetic effects in a declining vertebrate population. BMC Evolutionary Biology

10, 33.

Bolger A.M., Lohse M., Usadel B. (2014) Trimmomatic: a flexible trimmer for Illumina

sequence data. Bioinformatics 30, 2114-2120.

Braby C.E., Somero G.N. (2006) Following the heart: temperature and salinity effects on heart

rate in native and invasive species of blue mussels (genus Mytilus). Journal of

Experimental Biology 209, 2554 - 2566.

Brown G.P., Phillips B.L., Dubey S., Shine R. (2015c) Invader immunology: invasion history

alters immune system function in cane toads (Rhinella marina) in tropical Australia. Ecol

Lett 18, 57-65.

Bureau of Meteorology A.G. (2018) Climate Data Online. Commonwealth of Australia.

http://www.bom.gov.au/

Catchen J., Hohenlohe P.A., Bassham S., Amores A., Cresko W.A. (2013) Stacks: an analysis

tool set for population genomics. Mol Ecol 22, 3124-3140.

Chen B., Feder M.E., Kang L. (2018) Evolution of heat-shock protein expression underlying

adaptive responses to environmental stress. Mol Ecol 27, 3040-3054.

Cingolani P., Platts A., Wang L.L., et al. (2012) A program for annotating and predicting the

effects of single nucleotide polymorphisms, SnpEff. Fly (Austin) 6, 80-92.

Colautti R.I., Barrett S.C.H. (2013) Rapid Adaptation to Climate Facilitates Range Expansion of

an Invasive Plant. Science 342, 364 - 366.

Consortium T.U. (2017) UniProt: the universal protein knowledgebase. Nucleic Acids Res. 45,

D158-D169.

107 Covacevich J., Archer M. (1975) The distribution of the Cane Toad, Bufo marinus, in Australia

and its effects on indigenous vertebrates. . Memoirs of the Queensland Museum 17, 305 -

310.

Davey J.W., Blaxter M.L. (2011) RADSeq: next-generation population genetics. Briefings in

Functional Genomics 9, 416-423.

Dobin A., Davis C.A., Schlesinger F., et al. (2013) STAR: ultrafast universal RNA-seq aligner.

Bioinformatics 29, 15-21.

El-Sabban F., Fahim M.A., Al Homsi M.F., Singh S. (1996) Dehydration accelerates in vivo

platelet aggregation in pial arterioles of lead-treated mice. J. therm. Biol. 20, 469-476.

Estoup A., Ravigné V., Hufbauer R., et al. (2016) Is There a Genetic Paradox of Biological

Invasion? Annual Review of Ecology, Evolution, and Systematics 47, 51-72.

Estoup A., Wilson I.J., Sullivan C., Cornuet J.M., Moritz C. (2001) Inferring Population History

From Microsatellite and Enzyme Data in Serially Introduced Cane Toads, Bufo marinus.

Genetics 159, 1671-1687.

Foll M., Gaggiotti O. (2008) A Genome-Scan Method to Identify Selected Loci Appropriate for

Both Dominant and Codominant Markers: A Bayesian Perspective. Genetics 180, 977-

993.

Francis R.M. (2016) pophelper: an R package and web app to analyse and visualize population

structure Molecular Ecology Resources 17, 27-32.

Frankham R. (2005) Genetics and extinction. Biological Conservation 126, 131-140.

Franks S.J., Munshi-South J. (2014) Go forth, evolve and prosper: the genetic basis of adaptive

evolution in an invasive species. Molecular Ecology 23, 2137-2140.

108 Frichot E., Francois O. (2015) LEA: An R package for landscape and ecological association

studies. Methods in Ecology and Evolution 6, 925-929.

Gao Y., Li S., Zhan A. (2018) Genome-wide single nucleotide polymorphisms (SNPs) for a

model invasive ascidian Botryllus schlosseri. Genetica 146, 227-234.

Godoy O., Lemos-Filho J.P., Valladares F. (2011) Invasive species can handle higher leaf

temperature under water stress than Mediterranean natives. Environmental and

Experimental Botany 71, 207 - 214.

Goudet J. (2005) HIERFSTAT, a package for R to compute and test hierarchical F-statistics.

Molecular Ecology Notes 5, 184 - 186.

Greger R. (2000) Physiology of renal sodium transport. Am J Med Sci. 319, 51-62.

Gruber B., Unmack P.J., Berry O.F., Georges A. (2018) dartr: An r package to facilitate analysis

of SNP data generated from reduced representation genome sequencing. Molecular

Ecology Resources 18, 691-699.

Gruber J., Brown G.P., Whiting M.J., Shine R. (2017) Is the behavioural divergence between

range-core and range-edge populations of cane toads (Rhinella marina) due to

evolutionary change or developmental plasticity? Royal Society Open Science 4.

Hijmans R.J. (2015) raster: Geographic data analysis and modeling.

Hijmans R.J., Cameron S.E., Parra J.L., Jones P.G., Jarvis A. (2005) Very high resolution

interpolated climate surfaces for global land areas. International Journal of Climatology

25, 1965–1978.

Hudson C.M., Brown G.P., Shine R. (2016a) It is lonely at the front: contrasting evolutionary

trajectories in male and female invaders. R Soc Open Sci 3, 160687.

109 Hudson C.M., Brown G.P., Shine R. (2016b) It is lonely at the front: contrasting evolutionary

trajectories in male and female invaders. . Royal Society Open Science 3, 160687.

Hudson C.M., Brown G.P., Stuart K., Shine R. (2018) Sexual and geographical divergence in

head widths of invasive cane toads, Rhinella marina (Anura: Bufonidae), is driven by

both rapid evolution and plasticity. Biological Journal of the Linnean Society 124, 188-

199.

Institute B. (2018) Picard Tools. http://broadinstitute.github.io/picard.

Jessop T.S., Letnic M., Webb J.K., Dempster T. (2013) Adrenocortical stress responses influence

an invasive vertebrate's fitness in an extreme environment. Proc Biol Sci 280, 20131444.

Jombart T., Ahmed I. (2011) adegenet 1.3-1: new tools for the analysis of genome-wide SNP

data. Bioinformatics 27, 3070–3071.

Keenan K., McGinnity P., Cross T.F., Crozier W.W., Prodohl P. (2013) diveRsity:AnR package

for the estimation and exploration of popula tion genetics parameters and their associated

errors. Methods in Ecology and Evolution 4, 782-788.

Kiang J.G., Tsokos G.C. (1998) Heat Shock Protein 70 kDa: Molecular Biology, Biochemistry,

and Physiology. Pharmacology & Therapeutics 80, 183 - 201.

Knaus B.J., Grünwald N.J. (2017) vcfr: a package to manipulate and visualize variant call format

data in R. Molecular Ecology Resources 17, 44-53.

Kosmala G., Christian K., Brown G.P., Shine R. (2017) Locomotor performance of cane toads

differs between native-range and invasive populations. The Royal Society Open Science

4, 170517.

110 Kosmala G.K., Brown G.P., Christian K.A., Hudson C.M., Shine R. (2018a) The thermal

dependency of locomotor performance evolves rapidly within an invasive species

Ecology and Evolution 8, 4403-4408.

Kosmala G.K., Brown G.P., Shine R. (2018b) Kicking back Down Under: Invasive populations

of cane toads (Rhinella marina) down-regulate their responses to thermal and hydric

stress in a climatically harsh environment. Submitted.

Leblois R., Rousset F., Tikel D., Moritz C., Estoup A. (2000) Absence of evidence for isolation

by distance in an expanding cane toad (Bufo marinus) population: an individual-based

analysis of microsatellite genotypes. Molecular Ecology 9, 1905 - 1909.

Leydet K.P., Grupstra C.G.B., Coma R., Ribes M., Hellberg M.E. (2018) Host-targeted RAD-

Seq reveals genetic changes in the coral Oculina patagonica associated with range

expansion along the Spanish Mediterranean coast. Mol Ecol 27, 2529-2543.

Li H., Handsaker B., Wysoker A., et al. (2009) The Sequence Alignment/Map format and

SAMtools. Bioinformatics 25, 2078-2079.

Lillie M., Shine R., Belov K. (2014) Characterisation of major histocompatibility complex class I

in the Australian cane toad, Rhinella marina. PLOS ONE 9, e102824.

Lischer H.E., Excoffier L. (2012) PGDSpider: an automated data conversion tool for connecting

population genetics and genomics programs. Bioinformatics 28, 298-299.

Mader G., Castro L., Bonatto S.L., de Freitas L. (2016) Multiple introductions and gene flow in

subtropical South American populations of the fireweed, Senecio

madagascariensis(Asteraceae). Genet Mol Biol 39, 135-144.

Madsen T., Shine R., Olsson M., Wittzell H. (1999) Restoration of an inbred adder population.

Nature 402, 34 - 35.

111 Malmkvist J., Damgaard B.M., Pedersen L.J., et al. (2009) Effects of thermal environment on

hypothalamic-pituitary-adrenal axis hormones, oxytocin, and behavioral activity in

periparturient sows Journal of Animal Science 87, 2796–2805.

McKenna A., Hanna M., Banks E., et al. (2010) The Genome Analysis Toolkit: a MapReduce

framework for analyzing next-generation DNA sequencing data. Genome Research 20,

1297-1303.

Oksanen J., Blanchet F.G., Friendly M., et al. (2018) vegan: Community Ecology Package. R

package.

PubmedHealth (2014) Excessive Blood Clotting. NIH.

https://www.ncbi.nlm.nih.gov/pubmedhealth/PMH0062998/

Purcell S., Neale B., Todd-Brown K., et al. (2007) PLINK: a toolset for whole-genome

association and population-based linkage analysis. American Journal of Human Genetics

81.

Raj A., Stephens M., Pritchard J.K. (2014) fastSTRUCTURE: Variational Inference of

Population Structure in Large SNP Data Sets. Genetics 197, 573 - 589.

Reed D.H., Frankham R. (2003) Correlation between Fitness and Genetic Diversity.

Conservation Biology 17, 230 - 237.

Richardson M.F., Sequeira F., Selechnik D., et al. (2018) Improving amphibian genomic

resources: a multi-tissue reference transcriptome of an iconic invader. GigaScience 7, 1-

7.

Rollins L.A., Moles A.T., Lam S. (2013) High genetic diversity is not essential for successful

introduction. Ecology and Evolution 3, 4501 - 4517.

112 Rollins L.A., Richardson M.F., Shine R. (2015) A genetic perspective on rapid evolution in cane

toads (Rhinella marina). Mol Ecol 24, 2264-2276.

Rollins L.A., Woolnough A.P., Wilton A.N., Sinclair R., Sherwin W.B. (2009) Invasive species

can't cover their tracks: using microsatellites to assist management of starling (Sturnus

vulgaris) populations in Western Australia. Mol Ecol 18, 1560-1573.

Shine R., Brown G.P., Phillips B.L. (2011) An evolutionary process that assembles phenotypes

through space rather than through time. Proceedings of the National Academy of Sciences

of the United States of America 108, 5708 - 5711.

Skeeles L.E., Fleming J.L., Mahler K.L., Toland A.E. (2013) The impact of 3'UTR variants on

differential expression of candidate cancer susceptibility genes. PLOS ONE 8, e58609.

Slade R.W., Moritz C. (1998) Phylogeography of Bufo marinus from its natural and introduced

ranges. Proceedings of the Royal Society of London. Series B: Biological Sciences 265,

769.

Szucs M., Vahsen M.L., Melbourne B.A., et al. (2017) Rapid adaptive evolution in novel

environments acts as an architect of population range expansion. Proc Natl Acad Sci U S

A 114, 13501-13506.

Team R.C. (2016) R: A language and environment for statistical computing. R Foundation for

Statistical Computing, Vienna, Austria.

Telemeco R.S., Addis E.A. (2014) Temperature has species-specific effects on corticosterone in

alligator lizards. Gen. Comp. Endocrinol. 206, 184-192.

Thioulouse J., Dray S. (2007) Interactive Multivariate Data Analysis in R with the ade4 and

ade4TkGUI Packages. Journal of Statistical Software 22, 1 - 14.

113 Tingley R., Greenlees M.J., Shine R. (2012) Hydric balance and locomotor performance of an

anuran (Rhinella marina) invading the Australian arid zone. Oikos 121, 1959-1965.

Trumbo D.R., Epstein B., Hohenlohe P.A., et al. (2016) Mixed population genomics support for

the central marginal hypothesis across the invasive range of the cane toad (Rhinella

marina) in Australia. Molecular Ecology 25, 4161–4176

Turvey N. (2009) A Toad's Tale. Hot Topics from the Tropics 1, 1 - 10.

Wang Z., Gerstein M., Snyder M. (2009) RNA-Seq: a revolutionary tool for transcriptomics.

Nature Reviews Genetics 10.

Weber E., Schmid B. (1998) Latitudinal population differentiation in two species of Solidago

(Asteraceae) introduced into Europe. American Journal of Botany 85, 1110-1121.

Wellband K.W., Pettitt-Wade H., Fisk A.T., Heath D.D. (2018) Standing genetic diversity and

selection at functional gene loci are associated with differential invasion success in two

non-native fish species. Mol Ecol 27, 1572-1585.

Westemeier R.L., Brawn J.D., Simpson S.A., et al. (1998) Tracking the long-term decline and

recovery of an isolated population. Science 282, 1695-1698.

White T.A., Perkins S.E., Heckel G., Searle J.B. (2013) Adaptive evolution during an ongoing

range expansion: the invasive bank vole (Myodes glareolus) in Ireland. Mol Ecol 22,

2971-2985.

Whitlock M.C., Lotterhos K.E. (2015) Reliable Detection of Loci Responsible for Local

Adaptation: Inference of a Null Model through Trimming the Distribution of FST. The

American Naturalist 186, 24-36.

114 Whitney K.D., Gabler C.A. (2008) Rapid evolution in introduced species, 'invasive traits' and

recipient communities: challenges for predicting invasive potential. Diversity and

Distributions 14, 569-580.

Willoughby J.R., Harder A.M., Tennessen J.A., Scribner K.T., Christie M.R. (2018) Rapid

genetic adaptation to a novel environment despite a genome-wide reduction in genetic

diversity. Mol Ecol 27, 4041-4051.

Zerebecki R.A., Sorte C.J.B. (2011) Temperature Tolerance and Stress Proteins as Mechanisms

of Invasive Species Success. PLOS ONE 6, e14806.

115 Supporting Information

Figure S1. Maximum temperature during the warmest month in each location across collection sites for invasive cane toads in French Guiana, Hawaii, and Australia.

116 Table S1. (A) Collection sites for an RNA-Seq experiment on spleen tissue from invasive cane toads from French Guiana, Hawaii, and Australia (N=8 native range, N=10 source, N=10 core, N=8 intermediate, N=10 front). Site Phase Coordinates Plage de Montjoly, French Guiana Native Range 4.9173, -52.2664 Kapolei Regional Park, Oahu Source 21.3356, -158.0779 Haiku Gardens, Oahu Source 21.4151, -157.8157 Gordonvale, QLD Core -17.0832, 145.7961 Daintree, QLD Core -16.25, 145.3167 Cape Crawford, NT Intermediate -16.6667, 135.8 Timber Creek, NT Intermediate -15.6432, 130.4666 Halls Creek, WA Front -18.2247, 127.6729 Durack River, WA Front -15.9068, 128.184

(B) Collection sites for an RNA-Seq experiment on brain tissue from invasive cane toads from Hawaii and Australia (N=18 for each phase of the invasion). Site Phase Coordinates Paradise Park, Big Island Source 19.5724, -154.9577 Mauna Lani Point, Big Island Source 19.9385, -155.8754 Kapolei Regional Park, Oahu Source 21.3356, -158.0779 Haiku Gardens, Oahu Source 21.4151, -157.8157 Gordonvale, QLD Core -17.0832, 145.7961 Cairns, QLD Core -16.9186, 145.7781 Daintree, QLD Core -16.25, 145.3167 Croydon, QLD Intermediate -18.206, 142.24 Burketown, QLD Intermediate -17.8522, 139.633 Cape Crawford, NT Intermediate -16.6667, 135.8 Mataranka, NT Intermediate -14.9234, 133.066 Timber Creek, NT Intermediate -15.6432, 130.4666 Purnululu, WA Front -17.4484, 128.5467 Halls Creek, WA Front -18.2247, 127.6729 Durack River, WA Front -15.9068, 128.184

117 Table S2. (A) SNPs with outlier FST values and significant associations with maximum temperature during the warmest month of the year (per collection site) from RNA-Seq data from spleens in cane toads from French Guiana, Hawaii, and Australia. base gene position protein CDK12 3267 Cyclin-dependent kinase 12 TRM1L 966 TRMT1-like protein THOC1 2284 THO complex subunit 1 TRM1L 2088 TRMT1-like protein HSPA4 2006 Heat shock 70 kDa protein 4 TRM1L 951 TRMT1-like protein HSPA4 1043 Heat shock 70 kDa protein 4 SMG1 9837 Serine/threonine-protein kinase SMG1 NFIP1 2254 NEDD4 family-interacting protein 1 NFIP1 1188 NEDD4 family-interacting protein 1 DREB 574 Drebrin CYLD 871 Ubiquitin carboxyl-terminal hydrolase CYLD TRM1L 2100 TRMT1-like protein TYSY 758 Thymidylate synthase PIGT 1937 GPI transamidase component PIG-T TYSY 677 Thymidylate synthase FUBP3 1638 Far upstream element-binding protein 3 NFIP1 2629 NEDD4 family-interacting protein 1 NFIP1 1725 NEDD4 family-interacting protein 1 TOB1 882 Protein Tob1 UBF1B 2159 Nucleolar transcription factor 1-B NMRL1 129 NmrA-like family domain-containing protein 1 YES 2756 Tyrosine-protein kinase Yes TCPZ 100 T-complex protein 1 subunit zeta NOMO3 743 Nodal modulator 3 TYSY 800 Thymidylate synthase VATO 538 V-type proton ATPase 21 kDa proteolipid subunit PYR1 6103 CAD protein MANEA 1162 Glycoprotein endo-alpha-1,2-mannosidase ZBT7B 1709 and BTB domain-containing protein 7B AL9A1 588 4-trimethylaminobutyraldehyde dehydrogenase

118 AL9A1 582 4-trimethylaminobutyraldehyde dehydrogenase SON 12805 Protein SON SBDS 2613 Ribosome maturation protein SBDS VATO 368 V-type proton ATPase 21 kDa proteolipid subunit SON 9429 Protein SON PYR1 6973 CAD protein PTN6 2024 Tyrosine-protein phosphatase non-receptor type 6 WLSB 292 Protein wntless homolog B PTN6 2362 Tyrosine-protein phosphatase non-receptor type 6 PYR1 5474 CAD protein TCRG1 957 Transcription elongation regulator 1 SON 16145 Protein SON AL9A1 578 4-trimethylaminobutyraldehyde dehydrogenase DHX8 1534 ATP-dependent RNA helicase DHX8 STRN 3715 Striatin RBM34 276 RNA-binding protein 34 WLSB 357 Protein wntless homolog B PYR1 1993 CAD protein AT11C 6666 Phospholipid-transporting ATPase IG PAXB1 2726 PAX3- and PAX7-binding protein 1 SAP 1499 Prosaposin SON 12673 Protein SON Serine/threonine-protein phosphatase 2A 65 kDa 2AAA 964 regulatory subunit A alpha isoform PRLD1 330 PRELI domain-containing protein 1, mitochondrial PRLD1 70 PRELI domain-containing protein 1, mitochondrial . 733 . DDX41 2017 Probable ATP-dependent RNA helicase DDX41 ODO1 3515 2-oxoglutarate dehydrogenase, mitochondrial LEMD2 928 LEM domain-containing protein 2 TTC31 1157 Tetratricopeptide repeat protein 31 SYNRG 2895 Synergin gamma LEMD2 2168 LEM domain-containing protein 2 NBR1 3313 Next to BRCA1 gene 1 protein

119

(B) SNPs with outlier FST values and significant associations with maximum temperature during the warmest month of the year (per collection site) from RNA-Seq data from brains in invasive cane toads from Hawaii and Australia. base gene position protein RASA1 3118 Ras GTPase-activating protein 1 SYLC 3425 Leucine--tRNA ligase, cytoplasmic NFIP1 1452 NEDD4 family-interacting protein 1 NFIP1 2703 NEDD4 family-interacting protein 1 NFIP1 3135 NEDD4 family-interacting protein 1 NFIP1 360 NEDD4 family-interacting protein 1 NFIP1 915 NEDD4 family-interacting protein 1 NFIP1 1188 NEDD4 family-interacting protein 1 NFIP1 1725 NEDD4 family-interacting protein 1 NFIP1 1942 NEDD4 family-interacting protein 1 NFIP1 2214 NEDD4 family-interacting protein 1 NFIP1 2254 NEDD4 family-interacting protein 1 NFIP1 2629 NEDD4 family-interacting protein 1 SMG9 1270 Protein SMG9 SMG9 1285 Protein SMG9 HSPA4 2688 Heat shock 70 kDa protein 4 DREB 574 Drebrin RNF14 1152 E3 ubiquitin-protein ligase RNF14 PRELID1 70 PRELI domain-containing protein 1, mitochondrial PRELID1 330 PRELI domain-containing protein 1, mitochondrial RNF14 1757 E3 ubiquitin-protein ligase RNF14 AT2B4 8002 Plasma membrane calcium-transporting ATPase 4 AT2B4 8123 Plasma membrane calcium-transporting ATPase 4 QCR2 1954 Cytochrome b-c1 complex subunit 2, mitochondrial DDX41 2017 Probable ATP-dependent RNA helicase DDX41 - 596 - QCR2 1987 Cytochrome b-c1 complex subunit 2, mitochondrial ARNT2 1008 Aryl hydrocarbon receptor nuclear translocator 2 BDP1 383 Transcription factor TFIIIB component B'' homolog QCR2 1952 Cytochrome b-c1 complex subunit 2, mitochondrial SERF1 378 Small EDRK-rich factor 1

120 QCR2 2010 Cytochrome b-c1 complex subunit 2, mitochondrial QCR2 2038 Cytochrome b-c1 complex subunit 2, mitochondrial RUVB1 2921 RuvB-like 1 MO4L1 2128 Mortality factor 4-like protein 1 AB17C 443 Protein ABHD17C PIGV 2619 GPI mannosyltransferase 2 AN32E 491 Acidic leucine-rich nuclear phosphoprotein 32 family member E

Table S3. (A) SNPs with outlier FST values and significant associations with rainfall during the driest quarter of the year (per collection site) from RNA-Seq data from spleen in cane toads from French Guiana, Hawaii, and Australia. base gene position protein PTPRC 6803 Receptor-type tyrosine-protein phosphatase C TPRN 2983 Taperin TPRN 3800 Taperin PCNT 12271 Pericentrin TCPD 2105 T-complex protein 1 subunit delta ANR10 1164 Ankyrin repeat domain-containing protein 10 B3GN2 1554 N-acetyllactosaminide beta-1,3-N-acetylglucosaminyltransferase 2 SMG1 9837 Serine/threonine-protein kinase SMG1 GANP 5116 Germinal-center associated nuclear protein GANP 7035 Germinal-center associated nuclear protein PCNT 12168 Pericentrin GANP 5852 Germinal-center associated nuclear protein SFRP3 1044 Secreted frizzled-related protein 3 MYPOP 1291 Myb-related transcription factor, partner of profilin PCNT 13249 Pericentrin SMUF2 3385 E3 ubiquitin-protein ligase SMURF2 PCNT 12106 Pericentrin GANP 6952 Germinal-center associated nuclear protein LHFP 2029 Lipoma HMGIC fusion partner Serine/threonine-protein phosphatase 2A 65 kDa regulatory subunit A alpha 2AAA 2017 isoform TOB1 882 Protein Tob1 BECN1 1151 Beclin-1

121 BTBDA 1531 BTB/POZ domain-containing protein 10 . 639 . NMRL1 129 NmrA-like family domain-containing protein 1 SREK1 1694 Splicing regulatory /lysine-rich protein 1 TCPQ 971 T-complex protein 1 subunit theta TCPQ 1088 T-complex protein 1 subunit theta VATO 538 V-type proton ATPase 21 kDa proteolipid subunit PYR1 6103 CAD protein AL9A1 588 4-trimethylaminobutyraldehyde dehydrogenase AL9A1 582 4-trimethylaminobutyraldehyde dehydrogenase SON 12805 Protein SON VATO 368 V-type proton ATPase 21 kDa proteolipid subunit SON 9429 Protein SON PYR1 6973 CAD protein PYR1 5474 CAD protein SON 16145 Protein SON AL9A1 578 4-trimethylaminobutyraldehyde dehydrogenase NUBP2 834 Cytosolic Fe-S cluster assembly factor nubp2 NUBP2 970 Cytosolic Fe-S cluster assembly factor nubp2 PYR1 1993 CAD protein NUBP2 959 Cytosolic Fe-S cluster assembly factor nubp2 SON 12673 Protein SON MRP3 4668 Canalicular multispecific organic anion transporter 2 SYNRG 2895 Synergin gamma LEMD2 2168 LEM domain-containing protein 2

(B) SNPs with outlier FST values and significant associations with rainfall during the driest quarter of the year (per collection site) from RNA-Seq data from brains in invasive cane toads from Hawaii and Australia. base gene position protein MAGED2 1606 Melanoma-associated antigen D2 MAGED2 1608 Melanoma-associated antigen D2 MAGED2 1613 Melanoma-associated antigen D2 MAGED2 1057 Melanoma-associated antigen D2 MAGED2 1065 Melanoma-associated antigen D2 MAGED2 1068 Melanoma-associated antigen D2

122 MAGED2 1071 Melanoma-associated antigen D2 MAGED2 1074 Melanoma-associated antigen D2 MAGED2 1097 Melanoma-associated antigen D2 MAGED2 1492 Melanoma-associated antigen D2 MAGED2 1516 Melanoma-associated antigen D2 MAGED2 1522 Melanoma-associated antigen D2 MAGED2 1577 Melanoma-associated antigen D2 MAGED2 1581 Melanoma-associated antigen D2 MAGED2 1033 Melanoma-associated antigen D2 RBM34 1282 RNA-binding protein 34 EIF3A 834 Eukaryotic translation initiation factor 3 subunit A MAGED2 982 Melanoma-associated antigen D2 BAP31 261 B-cell receptor-associated protein 31 TOB1 882 Protein Tob1 TOB1 1299 Protein Tob1 MAGED2 2045 Melanoma-associated antigen D2 MAGED2 2046 Melanoma-associated antigen D2 TF3C5 1136 General transcription factor 3C polypeptide 5 MAGED2 1998 Melanoma-associated antigen D2 STXBP1 147 Syntaxin-binding protein 1 STXBP1 1672 Syntaxin-binding protein 1 STXBP1 2752 Syntaxin-binding protein 1 STXBP1 3135 Syntaxin-binding protein 1 SMC1A 2486 Structural maintenance of protein 1A SMC1A 2522 Structural maintenance of chromosomes protein 1A SMC1A 2816 Structural maintenance of chromosomes protein 1A GNDS 3242 Ral guanine nucleotide dissociation stimulator

123

Table S4. (A) Pairwise FST by invasion phase based on SNPs in RNA-Seq data from brains in invasive cane toads from Hawaii and Australia. Source = Hawaii; Core = QLD, Australia; Intermediate = NT, Australia; Front = WA, Australia. N=18 for each phase. Source Core Intermediate Front Source Core 0.04 Intermediate 0.08 0.06 Front 0.11 0.08 0.01

(B) Pairwise FST by collection site based on SNPs in RNA-Seq data from brains in invasive cane toads from Hawaii and Australia Kapolei Paradise Mauna Regional Haiku Cape Timber Halls Durack Park Kea Park Gardens Gordonvale Cairns Daintree Croydon Burketown Crawford Mataranka Creek Purnululu Creek River

Paradise Park

Mauna Kea 0.09 Kapolei Regional Park 0.06 0.10

Haiku Gardens 0.06 0.10 0.07

Gordonvale 0.07 0.09 0.08 0.06

Cairns 0.05 0.08 0.07 0.06 0.01

Daintree 0.07 0.11 0.09 0.08 0.04 0.03

Croydon 0.09 0.10 0.10 0.09 0.05 0.05 0.08

Burketown 0.10 0.12 0.12 0.10 0.05 0.06 0.08 0.03

Cape Crawford 0.11 0.12 0.12 0.11 0.06 0.07 0.09 0.03 0.01

Mataranka 0.11 0.13 0.13 0.11 0.06 0.07 0.08 0.03 0.01 -0.01

Timber Creek 0.12 0.16 0.14 0.12 0.08 0.09 0.10 0.06 0.03 0.00 -0.00

Purnululu 0.13 0.14 0.14 0.13 0.08 0.09 0.10 0.05 0.03 0.01 0.01 0.00

Halls Creek 0.13 0.16 0.15 0.14 0.09 0.10 0.10 0.06 0.03 0.01 0.01 0.00 -0.00

Durack River 0.12 0.15 0.14 0.12 0.07 0.08 0.10 0.05 0.02 -0.00 -0.00 0.00 0.00 0.00

124 Table S5. Diversity statistics by invasion phase based on SNPs in RNA-Seq data from brains in invasive cane toads from Hawaii and Australia. He = expected heterozygosity, AR = allelic richness, SI = Shannon’s information index. Source = Hawaii; Core = QLD, Australia; Intermediate = NT, Australia; Front = WA, Australia. He AR SI Source 0.33 1.97 0.45 Core 0.33 1.98 0.48 Intermediate 0.31 1.95 0.44 Front 0.29 1.93 0.44

125

Figure S2. Types of SNPs and in RNA-Seq data from brains in cane toads across sites in French Guiana, Hawai’i, and Australia. (A) represents the full dataset. (B) represents only FST outliers with an association with maximum temperature during the hottest month. (C) represents only FST outliers with an association with rainfall during the driest quarter.

Figure S3. Types of SNPs and in RNA-Seq data from brains in invasive cane toads across sites in Hawai’i and Australia. (A) represents the full dataset. (B) represents only FST outliers with an association with maximum temperature during the hottest month. (C) represents only FST outliers with an association with rainfall during the driest quarter.

126 CHAPTER 4: Immune and environment-driven gene expression during invasion: An eco-immunological application of RNA-Seq

This chapter is in review at Molecular Ecology with the following co-authors: Richardson MF, Shine R, Brown GP, Rollins LA

127 Abstract Host-pathogen associations change rapidly during a biological invasion and are predicted to impose strong selection on immune function. The invader may experience an abrupt reduction in pathogen-mediated selection (‘enemy release’), thereby favoring decreased investment into

‘costly’ immune responses. Across plants and animals, there is mixed support for this prediction.

Pathogens are not the only form of selection imposed on invaders; differences in abiotic environmental conditions between native and introduced ranges are also expected to drive rapid evolution. Here, we assess the expression patterns of immune and environmentally-associated genes in the cane toad (Rhinella marina) across its invasive Australian range. Transcripts encoding mediators of costly immune responses (inflammation, cytotoxicity) showed a curvilinear relationship with invasion history, with highest expression in toads from oldest and newest colonized areas. This pattern is surprising given theoretical expectations of density dynamics in invasive species, and may be because density influences both intraspecific competition and parasite transmission, which may have conflicting effects on the strength of immune responses. Alternatively, this expression pattern may be the result of other evolutionary forces, such as spatial sorting and genetic drift, working simultaneously with natural selection.

Our findings do not support predictions about immune function based on the enemy release hypothesis, and suggest that the effects of enemy release are difficult to isolate in wild populations, which may explain why the literature is inconclusive. Additionally, expression patterns of genes underlying putatively environmentally-associated traits are consistent with previous genetic studies, providing further support that Australian cane toads adapt to their heterogeneous abiotic environment.

Keywords: invasive species; enemy release hypothesis; cane toad; spatial sorting; compositional data analysis

128 Introduction Invasive species pose a massive threat to biodiversity (Bax et al., 2003; Clavero et al., 2009).

The potential for pathogens to limit the impact of invaders, or to exacerbate that impact, makes it critical to understand the effects of host-pathogen dynamics on invader immunity. The enemy release hypothesis (ERH) predicts that the processes of introduction and range expansion decrease rates of infections with co-evolved pathogens and parasites in invasive hosts due to the former’s inability to spread effectively and persist in novel environmental conditions (Colautti et al., 2004). Because of this, Lee & Klasing (2004) predict that invaders may down-regulate powerful immune responses such as systemic inflammation due to a decreased need (Cornet et al., 2016; Lee & Klasing, 2004; Martin et al., 2010). Such immune responses are also risky due to energetic costs (the reduction of nutrients available for partitioning across tissues due to their use in mounting immune responses (Klasing & Leshchinsky, 1999) and to the potential for collateral damage (tissue injury due to the effects of the immune response (Martin et al., 2010).

Reduced energetic investment into these immune responses may enhance invasion success

(McKean & Lazzaro, 2011). Nonetheless, loss of immunocompetence (the ability to mount a normal immune response after exposure to an antigen (Janeway et al., 2001) could render invaders susceptible to infection by novel pathogens and parasites in their introduced range

(Cornet et al., 2016; Lee & Klasing, 2004). Thus, invaders are predicted to exhibit lower investment in costly (but not all) immune responses than are seen in their native ranges (Cornet et al., 2016; Lee & Klasing, 2004).

The predicted consequences of enemy release, as well as other adaptive trait variation in invaders (Colautti & Barrett, 2013; Oduor et al., 2016; White et al., 2013), result from natural selection. However, some traits follow patterns that are not explained by selection (Berthouly-

Salazar et al., 2012; Lowe et al., 2015). This may be because the dispersive tendencies of

129 invaders give rise to additional evolutionary forces: 1) As an invasive population expands, genetic drift may result in a decline in genetic diversity across the range (Rollins et al., 2009) and potentially in phenotypic differences. 2) The concept of spatial sorting predicts that individuals at an expanding invasion front are those with the highest rates of dispersal simply because the fastest arrive at new areas first and can only breed with each other (Shine et al., 2011). Thus, a geographic separation of phenotypes occurs; traits that enhance individual fitness are favored in established populations, and traits that enhance dispersal rate are common in expanding populations (Hudson et al., 2016; Shine et al., 2011). 3) Admixture between individuals from different introductions or sources, as well as hybridization, may also drive change in some invasive populations (Mader et al., 2016).

We examined the effects of range expansion on expression of immune and environmentally-associated genes in the invasive Australian cane toad (Rhinella marina) using

RNA-Seq data from whole spleen tissue from individuals collected from long-established areas in Queensland (QLD, the ‘range core’), geographically ‘intermediate’ areas in the Northern

Territory (NT), and the leading edge of the range expansion in Western Australia (WA, the

‘invasion front’; Figure 1). The expanding range comprises highly varied environments; the range core is similar to the tropical rainforests of the native range (Central and South America), but intermediate areas and the invasion front receive much less annual rainfall (2000-3000 mm in

QLD, 400-1000 mm in NT and WA) and have higher annual mean temperatures (21-24C in

QLD, 24-27C in NT and WA; Bureau of Meteorology, 2018). Toads cluster genetically based on these environmental patterns: toads from the range core are genetically distinct from those from intermediate areas and the invasion front (Selechnik et al., unpublished; Chapter 3).

Furthermore, loci putatively under selection are involved in tolerance of temperature extremes

130 and dehydration (Selechnik et al., unpublished; Chapter 3). Traits such as locomotor performance at high temperatures also follow this pattern (Kosmala et al., 2018), but others do not. For example, behavioral propensity for exploration increases with distance from the introduction site (Gruber et al., 2017). Traits such as leg length (Hudson et al., 2016), spleen size, and fat body mass (Brown et al., 2015b) follow a U-shaped (curvilinear) pattern across the range, in which they are smallest in toads from intermediate areas of the range and larger on the two ends.

Consistent with the ERH, many species of bacteria, protozoan, and metazoan parasites of cane toads seem to have been left behind in the native range (Selechnik et al., 2017a). A major parasite (lungworm Rhabdias pseudosphaerocephala) from the native range that infects toad populations in the Australian range core is absent from toads at the invasion front (Phillips et al.,

2010). Conversely, the Australian soil bacterium Brucella (Ochrobactrum) anthropi causes spinal spondylosis in toads primarily at the invasion front (Brown et al., 2007), which may represent a novel infection that forces invaders to remain immunocompetent. Furthermore,

Rhimavirus A has only been detected in transcriptomes of toads from areas relatively close to the invasion front (Russo et al., 2018). Studies report various effects of invasion history on toad immunity (Brown et al., 2015c; Brown & Shine, 2014; Selechnik et al., 2017b).

Here, we aimed to characterize gene expression patterns in toads across the expanding

Australian range through differential expression (DE) analysis of transcriptome data. In terms of the ERH: release from pathogens depends on a decline in pathogen transmission, and pathogen transmission is typically lower when host densities are lower. The densities of many invasive populations follow a ‘travelling wave,’ in which density is low at recently colonized areas (e.g. the invasion front), high in areas that have been colonized for several years (e.g. intermediate

131 areas), and low at long-colonized areas (e.g. the range core; Hilker et al., 2005; Simberloff &

Gibbons, 2004). Although absolute population densities of cane toads across Australia are unknown, there is evidence that toads follow this trend as well (Brown et al., 2013; Freeland et al., 1986), as does at least one of their major parasites (Rhabdias pseudosphaerocephala). This parasitic lungworm is absent from toads at the invasion front, and is most prevalent in toads from intermediate areas (Brown et al., 2015b; Phillips et al., 2010). Therefore, we predicted that immune genes may follow a curvilinear pattern, in which those encoding mediators of costly immune responses (e.g. inflammation) may be expressed the most in toads from intermediate areas and less in toads on each end. In terms of environmentally-associated genes, we predicted that expression patterns would depend on the functional roles of the genes; for example, genes involved in aridity tolerance may be differentially expressed between toads from the range core and toads throughout the rest of the range.

Materials and Methods Sample collection and RNA extraction

In April and May of 2014 and 2015, we collected adult female cane toads from six locations along an invasion transect (Figure 1): Gordonvale, QLD (N=5, range core, 17.0972S

145.7792E); Daintree, QLD (N=5, range core, 16.25S 145.3167E); Cape Crawford, NT (N=4, intermediate, 16.6667S 135.8E); Timber Creek, NT (N=4, intermediate, 15.6453S 130.4744E);

Halls Creek, WA (N=5, invasion front, 18.2265S 127.759E); and Durack River, WA (N=5, invasion front, 15.9419S 127.2202E). We euthanized toads using 150 mg/kg sodium pentobarbital, decapitated them once they became unresponsive, and excised their spleens immediately following decapitation. We selected spleen tissue for our investigation because

132 mature immune cells travel to secondary lymphoid tissue (spleen, lymph nodes, and MALT) for activation through pathogenic encounter (Janeway et al., 2001). Thus, the cellular compositions of these tissue should reflect the host’s immune functioning. Each spleen was initially preserved in RNAlater (QIAGEN, USA), kept at 4ºC for less than one week, and then drained and transferred to a -80°C freezer for long-term storage.

Figure 1. Geographic distribution of the cane toad in Australia (dark grey region). Since arriving in Queensland in 1935, cane toads have further expanded their range through New South Wales, the Northern Territory, and into Western Australia. Black diamonds indicate our toad collection sites (from east to west): QLD (Gordonvale and Daintree, N=5 each), NT (Cape Crawford and Timber Creek, N=4 each) and WA (Caroline Pool and Durack River, N=5 each). Map adapted from Tingley et al (2017).

Prior to RNA extraction, we flash-froze all spleens individually in liquid nitrogen and ground them with a mortar and pestle to lyse preserved tissue. We carried out RNA extractions using the RNeasy Lipid Tissue Mini Kit (QIAGEN, USA) following the manufacturer’s instructions, with an additional genomic DNA removal step using on-column RNase-free DNase treatment (QIAGEN, USA). We quantified the total RNA extracted using a Qubit RNA HS assay

133 on a Qubit 3.0 fluorometer (Life Technologies, USA). Extracts were then stored at -80°C until sequencing was performed.

Sequencing

Prior to sequencing, we added 4 µL of either mix 1 or mix 2 of External RNA Controls

Consortium (ERCC; Thermo Fisher Science) spike-in solutions diluted 1:100 to 2 µg of RNA to examine the technical performance of sequencing (Table S1). Macrogen (Macrogen Inc., ROK) constructed mRNA libraries using the TruSeq mRNA v2 sample kit (Illumina Inc., USA), which included a 300bp selection step. All samples from the core and the front were sequenced in the same batch (but not pooled), and evenly distributed, across two lanes of Illumina HiSeq 2500

(Illumina Inc., USA); samples from intermediate areas were sequenced in a separate batch on a single lane (also not pooled). Capture of mRNA was performed using the oligo dT method, and size selection parameter choices were made according to the HiSeq2500 manufacturer’s protocol. Each individual sample was ligated with a unique barcode. Overall, this generated 678 million paired-end 2 x 125 bp reads. Raw sequence reads are available as FASTQ files in the

NCBI short read archive (SRA) under the BioProject Accession PRJNA395127.

Data pre-processing, alignment, and expression quantification

First, we examined per base raw sequence read quality (Phred scores) and GC content, and checked for the presence of adapter sequences for each sample using FastQC v0.11.5 (Andrews,

2010). We then processed raw reads (FASTQ files) from each sample with Trimmomatic v0.35

(Bolger et al., 2014), using the following parameters: ILLUMINACLIP:TruSeq3-PE.fa:2:30:10:4

SLIDINGWINDOW:5:20 AVGQUAL:20 MINLEN:36. This removed any adaptor sequences,

134 trimmed any set of 5 contiguous bases with an average Phred score below 20, and removed any read with an average Phred score below 20 or sequence length below 36 bp.

As a reference for alignment, we used the annotated R. marina transcriptome (Richardson et al., 2018), which was constructed from brain, spleen, muscle, liver, ovary, testes, and tadpole tissue. We conducted per sample alignments of our trimmed FASTQ files to this reference using

STAR v2.5.0a (Patro et al., 2017) in basic two-pass mode with default parameters, a runRNGseed of 777, and specifying BAM alignment outputs. We used the BAM outputs to quantify transcript expression using Salmon v0.8.1 (Patro et al., 2017) in alignment mode with libtype=IU, thus producing count files.

Count filtering and log-ratio transformations

Most methods for analyzing RNA-Seq expression data assume that raw read counts represent absolute abundances (Quinn et al., 2017a). However, RNA-Seq count data are not absolute and instead represent relative abundances as a type of compositional count data (Quinn et al., 2017b;

Quinn et al., 2017a). Using methods that assume absolute values is invalid for compositional data (without first including a transformation) because the total number of reads (library size) generated from each sample varies based on factors such as sequencing performance, making comparisons of the actual count values between samples difficult (Fernandes et al., 2014; Quinn et al., 2017b). As such, relationships within RNA-Seq count data make more sense as ratios, either when compared to a reference or to another feature within the dataset. Hence, we analyzed our count data (from the Salmon program) taking the compositional nature into account using the log-ratio transformation (Aitchison & Egozcue, 2005; Erb & Notredame, 2016; Lovell et al.,

2015; Quinn et al., 2017a). Our total amount of expressed transcripts across all toads was 22,930.

135 To filter out transcripts with low expression, we removed transcripts that did not have at least 10 counts in 10 samples. This narrowed our list of expressed transcripts to 18,945. We then used the

R (Team, 2016) package ALDEx2 v1.6.0 (Fernandes et al., 2013) to perform an inter-quantile log-ratio (iqlr) transformation of the transcripts’ counts as the denominator for the geometric mean calculation (rather than centered log-ratio transformation) because it removes the bias of transcripts with very high and low expression that may skew the geometric mean (Quinn et al.,

2017a). To circumvent issues associated with other normalization methods, we used ALDEx2 to model the count values over a multinomial distribution by using 128 Monte Carlo samples to estimate the Dirichlet distribution for each sample (Fernandes et al., 2013). The Dirichlet modelling and iqlr transformation enabled us to perform valid significance tests among samples of different groups for DE analysis. This approach has been shown to be consistent with (but with lower false discovery rates than) those of traditional DE analyses (Quinn et al., 2018).

Technical and diagnostic performance

Because the samples from intermediate areas were sequenced on a different run of the sequencing machine than the core and front samples, we wanted to rule out a batch effect, in which samples from intermediate areas may have had disproportionately higher or lower numbers of reads for each transcript due to technical variation in sequencing performance during different runs. This could result in erroneous DE calls. Thus, we used the previously added

ERCC control mixes (Ambion) to assess whether or not there was a batch effect due to sequencing run. Each set contains four groups of sequences with different ratios between the two mixes, representing ‘known’ differences in abundance (mix 1 versus mix 2 fold changes: 4:1,

1:1, 1:1.5, 1:2). We used the R (Team, 2016) package erccdashboard v1.10.0 (Munro et al.,

136 2014) to analyze the counts of these sequences and generate Receiver Operator Characteristic

(ROC) curves and the Area Under the Curve (AUC) statistic, lower limit of DE detection estimates (LODR), and expression ratio variability and bias measures based on these sequence abundance ratios (Figure S1). In the ROC curves, the AUC for two sets of true-positive ERCC sequences (4:1 and 1:2) was 1.0, indicating perfect diagnostic performance, and the AUC for the third (1:1.5) was 0.89, indicating good diagnostic performance (Figure S1B). The MA plot shows that the measured ratios in our ERCC sequences converge around the rm corrected ratios, indicating low variability (Figure S1C). Finally, the LODR plot indicated that DE p-values were lower for ERCC sequences with wider ratios (i.e. 4:1 has the lowest p-values, then 1:2, then

1:1.5), which is expected because the most pronounced fold-change differences should yield the highest DE significance (Figure S1D). From these results, we inferred that the observed relative abundances between mixes of each set of ERCC sequences were close to the known relative abundances, and thus batch effects do not appear to have occurred.

We also examined the counts of the invariant (1:1) group across all samples. Seven invariant ERCC transcripts remained after count filtering (same as used for the DE testing); we generated a boxplot of their counts, normalized by library size (Figure S2). The consistency of the boxplot distributions of the seven invariant ERCC sequences further indicated that there was no batch effect. Because the erccdashboard package indicated that the sets of true positive ERCC sequences (4:1, 1:2, 1:1.5) existed in observed ratios close to the known ratios, and because the invariant sequences (1:1) exhibited consistency across samples, we proceeded with downstream analyses.

137 Differential gene expression in discrete phases of the invasion

After applying a log-ratio transformation to the count data, we were able to implement statistical tests that would otherwise be invalid for relative data. We grouped populations by phase

(Daintree and Gordonvale in QLD/the core, Cape Crawford and Timber Creek in

NT/intermediate areas, and Durack River and Halls Creek in WA/the front) and used these as groups for DE analysis. We fitted our log-transformed count data to a non-parametric generalized linear model (glm) in ALDEx2. We took a ‘one vs. all’ approach, in which we compared samples from each state to samples from the other two states collectively (e.g. core vs. intermediate + front, intermediate vs. core + front, front vs. core + intermediate) using the

Kruskal Wallis test. This test design allowed us to identify transcripts that were up- and down- regulated in toads from each state relative to those throughout the rest of the range. We only retained transcripts with Benjamini-Hochberg (FDR) corrected p-values < 0.05 (Fernandes et al.,

2013). We detected 1,151 differentially expressed transcripts across all samples. We calculated the effect sizes of the differences between groups for each transcript, with positive values indicating up-regulation, and negative values indicating down-regulation. We further investigated all transcripts with effect sizes greater than 1.5 or less than -1.5.

Spatial gene expression patterns across the range

The DE testing performed in ALDEx2 generates differences between discrete groups; however, our data are sampled across a continuous variable: space. So, to visualize expression patterns across the toad’s Australian range, we performed soft (fuzzy C-means) clustering on our log- transformed count data (with samples grouped by collection site, and sites ordered from east to west) using the R package Mfuzz v2.34.0 (Kumar & Futschik, 2007). The fuzzy C-means

138 algorithm groups transcripts together based on similar expression patterns (using a fuzzifier parameter, m) across conditions to identify prominent, recurring patterns (clusters). Each transcript within a cluster is assigned a membership value, indicating how closely its expression pattern aligns with that of the cluster to which it belongs. To prevent random data from being clustered together, we used the mestimate command in the Mfuzz package to determine the optimal fuzzifier parameter value using a relation proposed for fuzzy c-means clustering

(Schwämmle & Jensen, 2010). We then used the cselection and Dmin commands to determine the optimal number of clusters, c, to generate. The results of both tools suggested using four clusters (c=4); however, these tools need to be used with caution because automatic determination of the optimal value of c is difficult, and it is advised to review the data before choosing (Kumar & Futschik, 2007). For this reason, we manually performed repeated clustering for a range of c (c=3, 4, 5, 6, 7, 8) using the fuzzy c-means algorithm to visualize the differences in clusters of expression patterns across space, and to compare the internal cores (identities and membership values of transcripts within each cluster) across c values. Although internal cores were consistent across all c values, we determined that several uniquely shaped expression patterns were collapsed at c=4, and that these expression patterns only became separate at c=6.

At c > 6, redundant patterns began to emerge. For this reason, we selected c=6 as the final value with which to perform soft clustering. We required a minimum membership value of 0.7 for all transcripts to their respective clusters.

Environmentally-influenced gene expression

Genes affected by natural selection may have expression levels that are associated with environmental variables. We downloaded climatic data from the Bioclim database (Hijmans et

139 al., 2005) using the raster package (Hijmans, 2015) in R. Because different areas in the

Australian range are known to vary in aridity, we downloaded data on rainfall during the driest quarter and maximum temperature in the warmest month; these data are averages of annual statistics over the period of 1970 to 2000. We then used the lfmm v2.0 package (Frichot &

Francois, 2015) in R to perform a latent factor mixed model (LFMM) to test the association between the log-transformed count values of every expressed transcript and these two environmental variables. We applied a Benjamini-Hochberg correction to all p-values from the

LFMM.

Coordination in gene expression

To identify genes with coordinated (co-associated) expression, we calculated proportionality (ρ) between all pairs of transcripts in our dataset using the propr package (Quinn et al., 2017a). A full description of this analysis is available in the Supplementary Information.

Annotation and enrichment

We performed gene ontology (GO) enrichment analysis to identify the most ‘enriched’

(recurring) biological functions in which our transcripts were involved. This allowed us to determine whether certain functional categories from the GO database were overrepresented in our DE, fuzzy clustering (spatial expression), and proportionality (coordination and differential coordination) datasets more than would be expected by chance. We used the Bioconductor tool

(Huber et al., 2015) GOseq v 1.26.0 (Young et al., 2010) because it accounts for bias introduced by variation in transcript lengths. We assessed three sets of GO categories (Biological Process,

Molecular Function, and Cellular Component) for enrichment using the Wallenius

140 approximation (while controlling for transcript length) to test for over-representation, and then

Benjamini-Hochberg corrected the resulting p-values. To visualize the results, we plotted significantly enriched GO categories with REVIGO (Supek et al., 2011), which performs SimRel

(Sæbø et al., 2015) semantic clustering of similar GO functions with annotations sourced from the UniProt database (Consortium, 2017). We then used GO terms to filter through the output lists from our DE, fuzzy clustering, and proportionality datasets to identify transcripts of genes involved in immune function.

Identification of immune genes

To identify additional transcripts that have known functions in the immune system (outside of those with the largest DE effect size or cluster membership), we cross-matched the output transcript lists from all of our analyses with several lists of known immune genes within the database InnateDB (Breuer et al., 2013): Immunology Database and Analysis Portal (ImmPort), the Immunogenetic Related Information Source (IRIS), the MAPK/NFKB Network, and the

Immunome Database. It should be noted that these databases consist entirely of human-mouse- bovine genes, but that the immune systems of mammals and amphibians are largely similar

(Colombo et al., 2015; Robert & Ohta, 2009). We further investigated all transcripts within our datasets that matched a gene within the InnateDB gene lists.

Isolation by distance

To assess the effect of geographic distance on divergence in gene expression (thereby testing for isolation by distance), we performed a Mantel test using the ade4 package (Thioulouse & Dray,

2007). A full description of this analysis is available in the Supplementary Information.

141 Results Identification of differentially expressed genes and their expression patterns

Overall, our DE analysis revealed 1,151 transcripts that are differentially expressed between invasion phases. This consists of 131 transcripts with unique regulation in toads from the range core, 904 transcripts with unique regulation in toads from intermediate areas, and 122 transcripts with unique regulation in toads from the invasion front (list of transcripts in each phase in

Appendix I). Soft clustering analysis revealed six prominent expression patterns (clusters) that our differentially expressed genes follow across the invasion (Figure 2): these clusters correspond to up- and down-regulation of each invasion phase (core, intermediate, front). Only

340 of the 1,151 differentially expressed transcripts had sufficiently high membership values to fit within these six clusters (list of transcripts in each cluster in Appendix I). The first cluster depicts low expression at the core and equally high expression throughout the rest of the range.

The other five clusters all depict curvilinear patterns (in which expression is either highest or lowest in toads from intermediate areas), but vary in the expression levels in toads on the ends of the range. Gene ontology enrichment of each cluster from Figure 2 is shown in Figure 3, and functional characterization of each cluster based on individual transcript investigation is shown in Table 1.

142

Figure 2. Six unique patterns of gene expression in spleen tissue from invasive cane toads (Rhinella marina). We collected samples in populations from areas spanning the invaded range in Australia (QLD = Queensland, NT = Northern Territory, WA = Western Australia). Color indicates membership values of genes to clusters (purple = 0.7-0.8; pink = 0.8-0.9; red = 0.9-1). Tick marks on the x-axis indicate sites across the toad’s Australian range in which spleens were collected (Figure 1).

143

Figure 3. REVIGO plots displaying GO terms depicting biological processes associated with transcripts following six major expression patterns in cane toads (Rhinella marina) across their Australian range (Figure 1). RNA-Seq data from spleens was used to identify differentially expressed transcripts between invasion phases, then soft clustering was performed to visualize the expression patterns that these transcripts follow (Figure 2). Circles represent GO terms; those with the highest statistical significance are labelled. Circle size relates to breadth of GO terms. Colors show log10 p-values.

144 Table 1. Most common functions of transcripts in each of the six major expression patterns in cane toads (Rhinella marina) across their Australian range (Figure 1). RNA-Seq data from spleens was used to identify differentially expressed transcripts between invasion phases, then soft clustering was performed to visualize the expression patterns that these transcripts follow (Figure 2). Cluster Number of Most common biological Most significantly enriched GO term(s) Less common biological functions transcripts function 1 71 blood coagulation/circulation platelet aggregation signal transduction, immune function, (24 transcripts) transcription regulation, viral processes

2 49 none cellular metabolism, metabolism, biosynthesis, biosynthesis cell cycle regulation, protein ubiquitination, translation initiation, transcription regulation

3 66 signal transduction intracellular signal transduction, actin protein transport, immune function, (24 transcripts) cytoskeleton organization smooth muscle contraction, angiogenesis, cell cycle regulation

4 75 translation initiation translation initiation metabolism, transcription regulation, (28 transcripts) protein ubiquitination, cell cycle regulation

5 49 immune function regulation of biological process transcription regulation, (12 transcripts) signal transduction, cell cycle regulation, metabolism

6 30 none biological regulation transcription regulation, signal transduction, cell cycle regulation, immune function

145 Immune genes

The most common immune functions that we found among our differentially expressed transcripts were activation (list of transcripts and their expression patterns/clusters in Table 2a) and suppression (Table 2b) of inflammatory pathways, and cytotoxicity (Table 2c). Although immune genes with other roles were detected, they were too few in number to allow clear inferences to be drawn. Most pro-inflammatory transcripts were down-regulated in toads from intermediate areas relative to toads from the range core and invasion front. This is reflected in the spatial expression data; in the fifth cluster (low expression in intermediate areas, high expression on either end of the range), approximately one fourth of transcripts were identified as relating to immune function (the highest proportion of any cluster), and this is the only cluster in which a GO term directly related to immunity is among the most significant (Figure 3, Table 1).

Furthermore, there are additional immune transcripts in the third and sixth clusters (which also depict lowest expression in toads from intermediate areas). Conversely, in the fourth cluster

(high expression in intermediate areas, low expression on either end of the range), no immune transcripts were identified and all the most significant GO terms are related to translation (Figure

3, Table 1). Nonetheless, there are a few pro- and anti-inflammatory transcripts up-regulated in intermediate toads (Table 2).

146 Table 2. Genes involved in immune function that are differentially expressed across the range of Australian cane toads. Spleens were collected from toads in the range core (QLD: Gordonvale and Daintree, N=5 each), intermediate areas (NT: Cape Crawford and Timber Creek, N=4 each) and invasion front (WA: Caroline Pool and Durack River, N=5 each). Soft clustering was performed to visualize differential expression patterns between different phases of the invasion (Figure 2). (a) Inflammation Gene Protein Expression Pattern MAP3K2 Mitogen-activated protein kinase kinase kinase 2 NT down; Cluster 5 pik3r5 Phosphoinositide 3-kinase regulatory subunit 5 NT down; Cluster 5 CSF2RB Cytokine receptor common subunit beta NT down; Cluster 5 Srf Serum response factor NT down; Cluster 5 PTK2B Protein-tyrosine kinase 2-beta NT down; Cluster 5 PYCARD (isoform 1) -associated speck-like protein containing a CARD (ASC) NT down; Cluster 5 PYCARD (isoform 2) Apoptosis-associated speck-like protein containing a CARD (ASC) NT down; Cluster 6 Nlrp1b NACHT; LRR and PYD domains-containing protein 1b allele 3 (NLRP1b) NT down; Cluster 6 ANKRD17 Ankyrin repeat domain-containing protein 17 NT down; Cluster 6 mapk8 Mitogen-activated protein kinase 8 NT down; Cluster 6 SPAG9 C-Jun-amino-terminal kinase-interacting protein 4 NT down; Cluster 3 Tab1 TGF-beta-activated kinase 1 and MAP3K7-binding protein 1 NT down; Cluster 3 CAMK2G Calcium/calmodulin-dependent protein kinase type II subunit gamma NT down; Cluster 3 TNFAIP2 Tumor necrosis factor alpha-induced protein 2 WA up; NT down; Cluster 3 PYCARD (isoform 3) Apoptosis-associated speck-like protein containing a CARD (ASC) NT down P2RX7 P2X purinoceptor 7 (P2P7) NT down NFATC2 Nuclear factor of activated T-cells; cytoplasmic 2 NT down Rps6ka3 Ribosomal protein S6 kinase alpha-3 NT down PIK3CB Phosphatidylinositol 4;5-bisphosphate 3-kinase catalytic subunit beta NT down isoform NOS2 Nitric oxide synthase; inducible NT down MAP4K5 Mitogen-activated protein kinase kinase kinase kinase 5 NT down HIPK1 Homeodomain-interacting protein kinase 1 (HIP1) NT down STAT1 Signal transducer and activator of transcription 1 (STAT1) NT down Prkcb Protein kinase C beta type NT down RIPK3 Receptor-interacting serine/threonine-protein kinase 3 (RIPK3) NT down Pak2 Serine/threonine-protein kinase PAK 2 NT down TRPC4AP Short transient receptor potential channel 4-associated protein NT down

147 Nsmaf Protein FAN NT down PLEKHG5 Pleckstrin homology domain-containing family G member 5 NT down IL7R Interleukin-7 receptor subunit alpha NT down Tnfrsf21 Tumor necrosis factor receptor superfamily member 21 NT down Rxra Retinoic acid receptor RXR-alpha NT down TRAF2 TNF receptor-associated factor 2 NT down IKBKE Inhibitor of nuclear factor kappa-B kinase subunit epsilon NT down Map3k14 Mitogen-activated protein kinase kinase kinase 14 NT down MTOR Serine/threonine-protein kinase mTOR NT down gtpbp1 GTP-binding protein 1 NT down ADAMTS1 A disintegrin and metalloproteinase with thrombospondin motifs 1 NT down Erc1 ELKS/Rab6-interacting/CAST family member 1 NT down ARHGEF17 Rho guanine nucleotide exchange factor 17 NT down; WA up Dele Death ligand signal enhancer WA up; Cluster 3 Cd84 SLAM family member 5 WA up F2RL2 Proteinase-activated receptor 3 WA up TNFRSF19 Tumor necrosis factor receptor superfamily member 19 WA up IL4R Interleukin-4 receptor subunit alpha WA up mapk1 Mitogen-activated protein kinase 1 WA up Irgc (isoform 1) Interferon-inducible GTPase 5 QLD up Irgc (isoform 2) Interferon-inducible GTPase 5 QLD up Irgc (isoform 3) Interferon-inducible GTPase 5 QLD up Irgc (isoform 4) Interferon-inducible GTPase 5 QLD up Irgc (isoform 5) Interferon-inducible GTPase 5 QLD up mul1a Mitochondrial ubiquitin ligase activator of nfkb 1-A QLD down; Cluster 1 CSF2RA Granulocyte-macrophage colony-stimulating factor receptor subunit alpha QLD down; Cluster 1 TRIM25 E3 ubiquitin/ISG15 ligase TRIM25 QLD down PYCARD (isoform 4) Apoptosis-associated speck-like protein containing a CARD (ASC) NT up Ecsit Evolutionarily conserved signaling intermediate in Toll pathway; NT up mitochondrial NKAP NF-kappa-B-activating protein NT up (b) Anti-Inflammation Gene Protein Expression Pattern

148 Tank TRAF family member-associated NF-kappa-B activator NT down; Cluster 5 ERBIN Erbin NT down; Cluster 5 Itch E3 ubiquitin-protein ligase Itchy NT down; Cluster 5 Sbno1 Protein strawberry notch homolog 1 NT down; Cluster 5 Smad6 Mothers against decapentaplegic homolog 6 NT down; Cluster 3 inpp5d Phosphatidylinositol 3,4,5-trisphosphate 5-phosphatase 1 NT down SBNO2 Protein strawberry notch homolog 2 NT down SYNCRIP Heterogeneous nuclear ribonucleoprotein Q NT down ATF3 Cyclic AMP-dependent transcription factor ATF-3 NT down Dusp4 Dual specificity protein phosphatase 4 NT down Rps6ka4 Ribosomal protein S6 kinase alpha-4 NT down AHR Aryl hydrocarbon receptor NT down PTPRE Receptor-type tyrosine-protein phosphatase epsilon NT down HAX1 HCLS1-associated protein X-1 NT up ppp4c Serine/threonine-protein phosphatase 4 catalytic subunit NT up Nlrc3 (isoform 1) Protein NLRC3 NT up Nlrc3 (isoform 2) Protein NLRC3 NT up ANXA1 Annexin A1 NT up ciapin1 Anamorsin NT up impdh2 Inosine-5'-monophosphate dehydrogenase 2 NT up DHCR24 Delta(24)-sterol reductase NT up CD200R1B Cell surface glycoprotein CD200 receptor 1-B QLD down; Cluster 1 TNFRSF6B Tumor necrosis factor receptor superfamily member 6B QLD down (c) Cytotoxicity Gene Protein Expression Pattern STXBP2 Syntaxin-binding protein 2 NT down; Cluster 5 NCR3LG1 (isoform 1) Natural cytotoxicity triggering receptor 3 ligand 1 NT down; Cluster 5 NCR3LG1 (isoform 2) Natural cytotoxicity triggering receptor 3 ligand 1 NT down NCR3LG1 (isoform 3) Natural cytotoxicity triggering receptor 3 ligand 1 NT down NCR3LG1 (isoform 4) Natural cytotoxicity triggering receptor 3 ligand 1 NT down Slamf7 SLAM family member 7 NT down; WA up

149 Climate-influenced gene expression

Our LFMM revealed eleven transcripts with expression levels associated with maximum temperature during the hottest month, rainfall during the driest quarter, or both (list of transcripts in Appendix II). Three of these transcripts follow the expression pattern of the first cluster (low expression at the range core, high expression throughout the rest of the range): two are involved in cell adhesion and platelet activity, and function of the third is unknown. Two other transcripts are also down-regulated at the core (but not a member of the first cluster); these are involved in transcription regulation and metabolism. Conversely, two transcripts involved in inflammation activation are up-regulated at the core. Another one, involved in cell cycle regulation, is up- regulated in intermediate areas. The three remaining transcripts are not differentially expressed: the first is involved in cell signaling in response to damage, the second is involved in blood circulation and response to nitric oxide (NO), and the third is unknown.

Coordination in gene expression

We tested whether some transcripts were co-associated across invasion phases by examining the expected value of ρ metric (Quinn et al., 2017a), but only found large groups of co-associated transcripts involved in fundamental cellular processes such as translation (Figure S3; list of proportional transcripts in Appendix III).

Isolation by distance

Our Mantel test revealed a significant relationship between geographic distance and gene expression distance (p = 0.005, R2 = 0.02; Figure 4).

150

Figure 4. Correlation between geographic distance and gene expression distance of invasive cane toad (Rhinella marina) populations across their Australian range (Figure 1). Euclidean distances in geographic location and gene expression between populations were calculated using the dist function in R. A mantel test (performed with the ade4 package) confirmed that these were significantly correlated (p = 0.005).

Discussion The conceptual scheme of Lee and Klasing (2004), based on the ERH, predicts that the expression of genes encoding mediators of costly immune responses (e.g. inflammation) would follow a density-driven curvilinear pattern, with highest expression in toads from intermediate areas (where density is presumably highest, which bolsters parasite transmission). However, most immune transcripts (involved pro- or anti-inflammatory signaling) follow the opposite pattern: curvilinear, but with lowest expression in toads from intermediate areas. Although not all immune transcripts follow this pattern, our data largely do not support the enemy release hypothesis. It must be noted that our inferences are based on gene annotations sourced from

151 many different taxa. These genes may not function the same way in cane toads as they do in the organisms from which they were documented. Furthermore, although we only sampled toads that appeared to be healthy, the infection status of each toad was not manipulated nor controlled for.

Thus, spatial heterogeneity in pathogen pressure or environmental conditions may play a role in gene expression; however, parasite prevalence is generally the highest in intermediate-age populations (compared to the youngest and oldest populations; Freeland et al., 1986), and we found that toads from intermediate-age populations largely down-regulate immune transcripts, so individual pathogen loads are unlikely to drive the expression patterns that we observed. Our results may still be due to another environmental variable, or may have a genetic basis through polymorphism in promoter regions, which are not sequenced when using RNA-Seq (Wang et al.,

2009).

Our findings are consistent with previous phenotypic data on toads: spleen sizes and fat body masses also follow a curvilinear pattern in which they are lowest in toads from intermediate parts of the range (Brown et al., 2015b). If toads indeed follow a ‘traveling wave’ density pattern, then the increased densities that occur several years-post colonization would not just heighten pathogen-mediated selection, but also intraspecific competition; reduced access to food resources may thus compromise the toad’s ability to increase investment in immunity, even if they need it (Brown et al., 2015b). This may explain smaller spleen sizes and lower expression of transcripts involved in inflammatory immune responses in toads from intermediate areas.

Availability of energetic resources does not seem to be considered in the predictions of Lee &

Klasing (2004), and although individuals at an invasion front may have a reduced benefit from energetic investment into immunity, they may invest anyway due to reduced intraspecific competition for food. Thus, the strength of immune functioning in invasion front individuals may

152 depend primarily on factors other than enemy release, such as prey abundance and interspecific competition.

Enemy release is not the only potential selective force acting on cane toads as they disperse across Australia; we also searched for evidence of adaptation to aridity. The range core is cooler and receives more rainfall than do intermediate areas and the invasion front. The expression pattern depicted in our first cluster (low expression in the core, high throughout the rest of the range) matches the trend across the Australian range in temperature, and is the opposite of the trend in rainfall. Population genetic structure in toads is driven by these environmental variables (Selechnik et al, unpublished; Chapter 3). Three of the eleven transcripts with expression levels associated with temperature or rainfall are members of this first cluster; the rest are not members of any of the six clusters. Two out of these three are involved in blood clotting, as are most of the transcripts in the first cluster. Proportionality analysis revealed two small groups of clotting-related transcripts with coordinated expression, suggesting coordinated function. Blood clotting is affected by hydration levels (El-Sabban et al., 1996), and excessive blood clotting can impair health (PubmedHealth, 2014). Changes to the rate of blood clotting may serve as an adaptation for intermediate and frontal toads living in drier conditions.

Furthermore, four candidate SNP loci (with outlier FST values and associations with rainfall, indicating they may be under selection) have previously been found in a gene involved in blood clotting (Selechnik et al, unpublished; Chapter 3). However, if lowering blood clotting rates is indeed an adaptation to aridity in toads from intermediate areas and the front, then one might expect that transcripts promoting blood clotting would be down-regulated in these individuals

(yet in our study, they are up-regulated). This may be because toads were collected from the wild and the environment was not controlled for. A common-garden experiment with toads from

153 across the range may reveal whether modification of blooding clotting rates is an evolved adaptation to aridity or simply a physiological reaction to different environments. Nonetheless, detecting expression differences underlying the same traits as previously observed genetic differences across the range likely provides support that cane toads are indeed adapting to their abiotic environment.

Our results suggest that other evolutionary forces are at work as well. The curvilinearity in expression of many of our differentially expressed genes resembles that of other phenotypic traits affected by spatial sorting, such as leg length (Hudson et al., 2016). Physical activity can have transgenerational effects on gene expression (Barres & Zierath, 2016; Murashov et al.,

2016), so it is possible that dispersal affects some expression patterns independently of the environment. Additionally, the significant result of our Mantel test is likely due to either genetic drift or a balance between geographically varying selection and gene flow (Endler, 1977).

Although it is possible that this result is influenced by environmental gradients across the range, it is unlikely that selection driven by environmental factors would act on a genome-wide level

(all expressed transcripts were used in our calculations for gene expression distance). This suggests that genetic drift may be responsible for isolation by distance in Australian cane toads.

Because spatial sorting and genetic drift drive non-adaptive variation, their effects may obscure adaptive variation, particularly when they act on the same traits as selection (i.e. physical activity is also linked to expression of inflammatory genes (Baynard et al., 2012; Gjevestad et al., 2015).

Conclusions

The expression patterns that we have observed in pro- and anti-inflammatory transcripts generally do not support the predictions of Lee & Klasing (2004) based on the ERH. Notably,

154 our study suggests that the predicted consequences of enemy release are difficult to study in the wild because host-parasite dynamics and their impacts are affected by many factors. Increasing host densities bolsters both parasite transmission and host intraspecific competition, but these two factors exert opposite effects on host immune function: while high rates of parasite transmission drive up the need for powerful immune responses, high levels of intraspecific competition limit the energetic resources available to fuel these immune responses. Future studies conducting experimental infections on animals reared under common-garden conditions may clarify the roles played by pathogens versus the environment. Although natural selection engenders adaptive variation in immune function and other traits, this likely co-occurs with non- adaptive variation driven by forces such as spatial sorting and genetic drift in wild populations, particularly in invasive species. These complicated dynamics may explain why support for the predictions of the ERH has been so mixed over the past two decades. Nonetheless, methods such as RNA-Seq remain a powerful tool for uncovering the diverse and sometimes opposing forces that underpin rapid evolution in invasive species.

Acknowledgements

This work was supported by the Australian Research Council (FL120100074, DE150101393) and the Equity Trustees Charitable Foundation (Holsworth Wildlife Research Endowment). We thank BriAnne Addison and Thom Quinn for their input during discussions about our analyses.

We thank John Endler and Lynn B. Martin for their useful comments that improved our manuscript. We thank Cam Hudson, Serena Lam, and Chris Jolly for their assistance with sample collection.

155 Data Accessibility

The dataset supporting the conclusions of this article is available in the NCBI short read archive

(SRA) under the BioProject Accession PRJNA395127 in FASTQ format

(https://www.ncbi.nlm.nih.gov/bioproject/PRJNA395127/).

Authors’ contributions

Dan Selechnik, Mark Richardson, Lee Ann Rollins, Greg Brown, and Richard Shine designed the experiment. Dan Selechnik, Mark Richardson, and Lee Ann Rollins performed data collection. Dan Selechnik, Mark Richardson, and Lee Ann Rollins performed data analysis.

Dan Selechnik wrote the manuscript. Mark Richardson, Lee Ann Rollins, Greg Brown, and

Richard Shine revised the manuscript.

156 Literature Cited Aitchison J., Egozcue J.J. (2005) Compositional Data Analysis: Where Are We and Where

Should We Be Heading? Mathematical Geology 37, 829 - 850.

Andrews S. (2010) FastQC: a quality control tool for high throughput sequence data.

http://www.bioinformatics.babraham.ac.uk/projects/fastqc

Barres R., Zierath J.R. (2016) The role of diet and exercise in the transgenerational epigenetic

landscape of T2DM. Nat Rev Endocrinol 12, 441-451.

Bax N., Williamson A., Aguero M., Gonzalez E., Geeves W. (2003) Marine invasive alien

species: a threat to global biodiversity. Marine Policy 27, 313-323.

Baynard T., Vieira-Potter V.J., Valentine R.J., Woods J.A. (2012) Exercise training effects on

inflammatory gene expression in white adipose tissue of young mice. Mediators Inflamm

2012, 767953.

Berthouly-Salazar C., van Rensburg B.J., Le Roux J.J., van Vuuren B.J., Hui C. (2012) Spatial

sorting drives morphological variation in the invasive bird, Acridotheris tristis. PLOS

ONE 7, e38145.

Bolger A.M., Lohse M., Usadel B. (2014) Trimmomatic: a flexible trimmer for Illumina

sequence data. Bioinformatics 30, 2114-2120.

Breuer K., Foroushani A.K., Laird M.L., et al. (2013) InnateDB: systems biology of innate

immunity and beyond—recent updates and continuing curation. Nucleic Acids Research

41, D1228–D1233.

Brown G.P., Kelehear C., Shilton C.M., Phillips B.L., Shine R. (2015b) Stress and immunity at

the invasion front: a comparison across cane toad (Rhinella marina) populations.

Biological Journal of the Linnean Society 116, 748-760.

157 Brown G.P., Kelehear C., Shine R. (2013) The early toad gets the worm: cane toads at an

invasion front benefit from higher prey availability. J Anim Ecol 82, 854-862.

Brown G.P., Phillips B.L., Dubey S., Shine R. (2015c) Invader immunology: invasion history

alters immune system function in cane toads (Rhinella marina) in tropical Australia. Ecol

Lett 18, 57-65.

Brown G.P., Shilton C., Phillips B.L., Shine R. (2007) Invasion, stress, and spinal arthritis in

cane toads. Proceedings of the National Academy of Science USA 104, 17698 - 17700.

Brown G.P., Shine R. (2014) Immune Response Varies with Rate of Dispersal in Invasive Cane

Toads (Rhinella marina). PLOS ONE 9, 1 - 11.

Bureau of Meteorology A.G. (2018) Climate Data Online. Commonwealth of Australia.

http://www.bom.gov.au/

Clavero M., Brotonsa L., Pons P., Sol D. (2009) Prominent role of invasive species in avian

biodiversity loss. Biological Conservation 142, 2043-2049.

Colautti R.I., Barrett S.C.H. (2013) Rapid Adaptation to Climate Facilitates Range Expansion of

an Invasive Plant. Science 342, 364 - 366.

Colautti R.I., Ricciardi A., Grigorovich I.A., MacIsaac H.J. (2004) Is invasion success explained

by the enemy release hypothesis? Ecology Letters 7, 721-733.

Colombo B.M., Scalvenzi T., Benlamara S., Pollet N. (2015) Microbiota and mucosal immunity

in amphibians. Front Immunol 6, 111.

Consortium T.U. (2017) UniProt: the universal protein knowledgebase. Nucleic Acids Res. 45,

D158-D169.

158 Cornet S., Brouat C., Diagne C., Charbonnel N. (2016) Eco-immunology and bioinvasion:

revisiting the evolution of increased competitive ability hypotheses. Evol Appl 9, 952-

962.

El-Sabban F., Fahim M.A., Al Homsi M.F., Singh S. (1996) Dehydration accelerates in vivo

platelet aggregation in pial arterioles of lead-treated mice. J. therm. Biol. 20, 469-476.

Endler J.A. (1977) Geographic variation, speciation, and clines. In: Monographs in Population

Biology, p. 246. Princeton University Press, University of Minnesota.

Erb I., Notredame C. (2016) How should we measure proportionality on relative gene expression

data? Theory in Biosciences 135, 21 - 36.

Fernandes A.D., Macklaim J.M., Linn T.G., Reid G., Gloor G.B. (2013) ANOVA-Like

Differential Expression (ALDEx) Analysis for Mixed Population RNA-Seq. PLOS ONE

8, e67019.

Fernandes A.D., Reid J.N.S., Macklaim J.M., et al. (2014) Unifying the analysis of high-

throughput sequencing datasets: characterizing RNA-seq, 16S rRNA gene sequencing

and selective growth experiments by compositional data analysis. Microbiome 2, 15.

Freeland W.J., Delvinquier B.L.J., Bonnin B. (1986) Food and Parasitism of the Cane Toad,

Bufo marinus, in Relation to Time Since Colonization. Aust. Wildl. Res. 13, 489 - 499.

Frichot E., Francois O. (2015) LEA: An R package for landscape and ecological association

studies. Methods in Ecology and Evolution 6, 925-929.

Gjevestad G.O., Holven K.B., Ulven S.M. (2015) Effects of Exercise on Gene Expression of

Inflammatory Markers in Human Peripheral Blood Cells: A Systematic Review. Curr

Cardiovasc Risk Rep 9, 34.

159 Gruber J., Brown G., Whiting M.J., Shine R. (2017) Geographic divergence in dispersal-related

behaviour in cane toads from range-front versus range-core populations in Australia.

Behavioral Ecology and Sociobiology 71.

Hijmans R.J. (2015) raster: Geographic data analysis and modeling.

Hijmans R.J., Cameron S.E., Parra J.L., Jones P.G., Jarvis A. (2005) Very high resolution

interpolated climate surfaces for global land areas. International Journal of Climatology

25, 1965–1978.

Hilker F.M., Lewis M.A., Seno H., Langlais M., Malchow H. (2005) Pathogens can Slow Down

or Reverse Invasion Fronts of their Hosts. Biological Invasions 7, 817-832.

Huber W., Carey V.J., Gentleman R., Morgan M. (2015) Orchestrating high-throughput genomic

analysis with Bioconductor. In: Nature Methods, p. 115.

Hudson C.M., Brown G.P., Shine R. (2016) It is lonely at the front: contrasting evolutionary

trajectories in male and female invaders. . Royal Society Open Science 3, 160687.

Ingolia N.T., Ghaemmaghami S., Newman J.R.S., Weissman J.S. (2009) Genome-Wide Analysis

in Vivo of Translation with Nucleotide Resolution Using Ribosome Profiling. Science

324, 218 - 223.

Janeway C.A., Travers P., Walport M. (2001) Immunobiology: The Immune System in Health

and Disease. In: Immunobiology. Garland Science, New York.

Klasing K.C., Leshchinsky T.V. (1999) Functions, costs, and benefits of the immune system

during development and growth. International Ornitology Congress, Proceedings 69,

2817 - 2832.

160 Kosmala G.K., Brown G.P., Christian K.A., Hudson C.M., Shine R. (2018) The thermal

dependency of locomotor performance evolves rapidly within an invasive species

Ecology and Evolution 8, 4403-4408.

Kumar L., Futschik M. (2007) Mfuzz: A software package for soft clustering of microarray data

Bioinformation 2, 5 - 7.

Lee K.A., Klasing K.C. (2004) A role for immunology in invasion biology. Trends Ecol Evol 19,

523-529.

Lovell D., Pawlowsky-Glahn V., Egozcue J.J., Marguerat S., Bahler J. (2015) Proportionality: A

Valid Alternative to Correlation for Relative Data. PLOS Computational Biology 11,

e1004075.

Lowe W.H., Muhlfeld C.C., Allendorf F.W. (2015) Spatial sorting promotes the spread of

maladaptive hybridization. Trends Ecol Evol 30, 456-462.

Mader G., Castro L., Bonatto S.L., de Freitas L. (2016) Multiple introductions and gene flow in

subtropical South American populations of the fireweed, Senecio

madagascariensis(Asteraceae). Genet Mol Biol 39, 135-144.

Martin L.B., Hopkins W.A., Mydlarz L.D., Rohr J.R. (2010) The effects of anthropogenic global

changes on immune functions and disease resistance. Ann N Y Acad Sci 1195, 129-148.

McKean K.A., Lazzaro B. (2011) Mechanisms of Life History Evolution. In: The costs of

immunity and the evolution of immunological defense mechanisms eds. Flatt T, Heyland

A), p. 504. OUP Oxford.

Munro S.A., Lund S.P., Pine P.S., et al. (2014) Assessing technical performance in differential

gene expression experiments with external spike-in RNA control ratio mixtures. Nat

Commun 5, 5125.

161 Murashov A.K., Pak E.S., Koury M., et al. (2016) Paternal long-term exercise programs

offspring for low energy expenditure and increased risk for obesity in mice. FASEB J 30,

775-784.

Oduor A.M.O., Leimu R., van Kleunen M., Mack R. (2016) Invasive plant species are locally

adapted just as frequently and at least as strongly as native plant species. Journal of

Ecology 104, 957-968.

Patro R., Duggal G., Love M.I., Irizarry R.A., Kingsford C. (2017) Salmon provides fast and

bias-aware quantification of transcript expression. Nature Methods.

Phillips B.L., Kelehear C., Pizzatto L. (2010) Parasites and pathogens lag behind their host

during periods of host range advance. Ecology 91, 872 - 881.

PubmedHealth (2014) Excessive Blood Clotting. NIH.

https://www.ncbi.nlm.nih.gov/pubmedhealth/PMH0062998/

Quinn T., Crowley T., Richardson M.F. (2018) Benchmarking differential expression analysis

tools for RNA-Seq: normalization-based vs. log-ratio transformation-based methods.

BMC Bioinformatics 19, 274.

Quinn T., Erb I., Richardson M.F., Crowley T. (2017b) Understanding sequencing data as

compositions: an outlook and review. bioRxiv.

Quinn T., Richardson M.F., Lovell D., Crowley T. (2017a) propr: An R-package for Identifying

Proportionally Abundant Features Using Compositional Data Analysis. Scientific Reports

7, 16252.

Richardson M.F., Sequeira F., Selechnik D., et al. (2018) Improving amphibian genomic

resources: a multi-tissue reference transcriptome of an iconic invader. GigaScience 7, 1-

7.

162 Robert J., Ohta Y. (2009) Comparative and developmental study of the immune system in

Xenopus. Dev Dyn 238, 1249-1270.

Rollins L.A., Woolnough A.P., Wilton A.N., Sinclair R., Sherwin W.B. (2009) Invasive species

can't cover their tracks: using microsatellites to assist management of starling (Sturnus

vulgaris) populations in Western Australia. Mol Ecol 18, 1560-1573.

Russo A.G., Eden J.S., Enosi Tuipulotu D., et al. (2018) Viral discovery in the invasive

Australian cane toad (Rhinella marina) using metatranscriptomic and genomic

approaches. Journal of Virology.

Sæbø S., Almøy T., Helland I.S. (2015) simrel — A versatile tool for linear model data

simulation based on the concept of a relevant subspace and relevant predictors.

Chemometrics and Intelligent Laboratory Systems 146, 128-135.

Schwämmle V., Jensen O.N. (2010) A simple and fast method to determine the parameters for

fuzzy c–means cluster analysis. Bioinformatics 26, 2841–2848.

Selechnik D., Rollins L.A., Brown G.P., Kelehear C., Shine R. (2017a) The things they carried:

The pathogenic effects of old and new parasites following the intercontinental invasion of

the Australian cane toad (Rhinella marina). International Journal for Parasitology:

Parasites and Wildlife 6, 375-385.

Selechnik D., West A.J., Brown G.P., et al. (2017b) Effects of invasion history on physiological

responses to immune system activation in invasive Australian cane toads. PeerJ 5, e3856.

Shine R., Brown G.P., Phillips B.L. (2011) An evolutionary process that assembles phenotypes

through space rather than through time. Proceedings of the National Academy of Sciences

of the United States of America 108, 5708 - 5711.

163 Simberloff D., Gibbons L. (2004) Now you see them, now you don’t! – population crashes of

established introduced species. Biological Invasions 6, 161 - 172.

Supek F., Bosnjak M., Skunca N., Smuc T. (2011) REVIGO Summarizes and Visualizes Long

Lists of Gene Ontology Terms. PLOS ONE 6, e21800.

Team R.C. (2016) R: A language and environment for statistical computing. R Foundation for

Statistical Computing, Vienna, Austria.

Thioulouse J., Dray S. (2007) Interactive Multivariate Data Analysis in R with the ade4 and

ade4TkGUI Packages. Journal of Statistical Software 22, 1 - 14.

Tingley R., Ward-Fear G., Schwarzkopf L., et al. (2017) New Weapons in the Toad Toolkit: A

Review of Methods to Control and Mitigate the Biodiversity Impacts of Invasive Cane

Toads (Rhinella marina). The Quarterly Review of Biology 92, 123 - 149.

Wang Z., Gerstein M., Snyder M. (2009) RNA-Seq: a revolutionary tool for transcriptomics.

Nature Reviews Genetics 10.

White T.A., Perkins S.E., Heckel G., Searle J.B. (2013) Adaptive evolution during an ongoing

range expansion: the invasive bank vole (Myodes glareolus) in Ireland. Mol Ecol 22,

2971-2985.

Young M.D., Wakefield M.J., Smyth G.K., Oshlack A. (2010) Gene ontology analysis for RNA-

seq: accounting for selection bias. Genome Biol 11, R14.

164 Supporting Information Table S1. ERCC spike-in mixes added to RNA from spleens collected from cane toads (Rhinella marina) across the Australian range (main text, Figure 1), RNA integrity numbers (RINs), sexes, accession numbers, RNA concentrations, numbers of reads, and numbers of transcripts retained after quality control filtering (out of 18945 transcripts total). RNA-Seq data from spleens was used to quantify differential expression (DE) analysis between localities (states). ERCC spike-ins are used to assess technical performance of sequencing. RINs indicate the quality of the extracted RNA. Sex SRA Accession RNA conc Reads Transcripts Spleen Location Spike- RIN (ng/L) retained ID In after QC S1 Gordonvale, QLD 1 8.6 F SRX3030411 84.01 23784416 18810 S2 Gordonvale, QLD 1 8.7 F SRX3030410 44.78 24958948 18839 S3 Gordonvale, QLD 2 7.4 F SRX3030409 71.90 24238536 18786 S4 Gordonvale, QLD 2 7.1 F SRX3030408 65.34 25172372 18780 S5 Gordonvale, QLD 1 7.4 F SRX3030397 41.98 23863836 18809 S6 Daintree, QLD 2 7.4 F SRX3030404 80.98 25504510 18825 S7 Daintree, QLD 1 9.5 F SRX3030403 56.05 23452460 18679 S8 Daintree, QLD 1 7.3 F SRX3030405 40.72 24403770 18781 S9 Daintree, QLD 1 9.4 F SRX3030406 71.26 21885152 18811 S10 Daintree, QLD 2 8.9 F SRX3030401 56.55 22693350 18805 S11 Halls Creek, WA 2 8.9 F SRX3030396 70.10 22491142 18851 S12 Halls Creek, WA 2 9.2 F SRX3030395 67.09 24040758 18779 S13 Halls Creek, WA 2 7.0 F SRX3030394 65.81 23746642 18835 S14 Halls Creek, WA 1 9.2 F SRX3030393 77.76 24426716 18852 S15 Halls Creek, WA 1 9.7 F SRX3030392 78.91 23278544 18709 S16 Durack River, WA 2 8.6 F SRX3030385 172.58 24374724 18803 S17 Durack River, WA 1 8.5 F SRX3030389 99.38 23344088 18828 S18 Durack River, WA 2 8.8 F SRX3030407 41.42 24654968 18740 S19 Durack River, WA 1 9.0 F SRX3030384 77.47 23608774 18840 S20 Durack River, WA 2 8.5 F SRX3030402 61.83 25004504 18834 S21 Cape Crawford, NT 1 8.2 F SRX3030387 103.51 25116754 18835 S22 Cape Crawford, NT 2 8.1 F SRX3030388 34.76 19923840 18761 S23 Cape Crawford, NT 1 9.6 F SRX3030390 92.68 26615752 18791 S24 Cape Crawford, NT 2 9.3 F SRX3030391 83.97 27652796 18773 S25 Timber Creek, NT 2 8.7 F SRX3030386 43.59 19004402 18758 S26 Timber Creek, NT 2 9.2 F SRX3030400 102.68 27999430 18820 S27 Timber Creek, NT 1 9.0 F SRX3030399 73.58 29437400 19794 S28 Timber Creek, NT 1 7.1 F SRX3030398 69.74 23455440 18771

165

Figure S1. ERCC technical and diagnostic plots produced by erccdashboard. (A) Signal-abundance plot. Colors represent ERCC sequence sets that are present in abundances of different ratios between mix 1 and mix 2, shape represents the sample type, and error bars represent standard deviation. (B) ROC curves and AUC statistics for each set of true-positive ERCC controls. “Detected” represents the number of controls used in our experiment, and “spiked” represents the total that were present in the ERCC control mixture. (C) MA plot of ERCC ratio measurement variability and bias. Colors represent ERCC sequence sets that are present in abundances of different ratios between mix 1 and mix 2, and error bars represent standard deviation. Grey points represent endogenous transcript ratios. Solid colored lines show the nominal ERCC ratios for each ERCC sequence set, and dashed lines (rm) represent the corrected ratios for each ERCC subset based on the estimate of mRNA fraction differences between samples. (D) Lower limit of differential expression detection estimates (LODR) plot. The black dashed line represents the threshold p- value for the specified FDR (0.05). Colors represent ERCC sequence sets that are present in abundances of different ratios between mix 1 and mix 2. Colored lines are the LODR estimates, indicating if the p- value of each ratio is ever low enough to cross the threshold p-value and be called DE. LODR results and 90% confidence intervals are provided in the table below the plot.

166

Figure S2. Normalized counts of seven invariant ERCC sequences (mix 1 versus mix 2 fold-change 1:1) across all samples. ERCC spike-in mixes were added to RNA from spleens collected from cane toads (Rhinella marina) across their Australian range (main text, Figure 1). Counts of ERCC sequences and endogenous transcripts were generated, then filtered to remove zeroes. Only seven invariant ERCC sequences were retained; count values of these seven ERCC sequences in each sample were divided by the library size of the sample to generate normalized counts.

Supplemental Methods

Available annotation information

In the reference transcriptome (Richardson et al., 2018) that we used for performing alignments, approximately 30% of all transcripts are functionally characterized, and approximately 4% of all transcripts are characterized in Xenopus (the phylogenetically closest model taxa to Rhinella marina). In our list of differentially expressed genes, approximately 85% of all transcripts are

167 functionally characterized, and approximately 17% of all transcripts are characterized in

Xenopus.

Coordination in gene expression

Correlation in expression patterns of different transcripts is of interest because it may suggest that the expression of one transcript is coordinated with the expression of another, which may extend to functional relations between the two; however, correlation is not a statistically valid measure for compositional data such as ours (Quinn et al., 2017a). Proportionality (ρ) is an alternative measure of coordination that is statistically valid for compositional data (Quinn et al.,

2017a); it is a modification of the variance of the log ratios (VLR) that sets a scale on which transcript expression can be compared to look for coordination (Quinn et al., 2017a). Similarly to the correlation coefficient r, ρ exists on a scale from -1 to 1; in this way, they are analogous

(Quinn et al., 2017a). To see if there was coordination among different transcripts in expression, we tested our transcripts for proportionality using propr v2.1.9 (Quinn et al., 2017a). We used the iqlr-transformed counts produced by ALDEx2 as expression values. Proportionality was calculated as:

푣푎푟(퐴푖−퐴푗) 휌(퐴푖, 퐴푗) = 1 − (1) 푣푎푟(퐴푖)+푣푎푟(퐴푗)

In this equation (Erb & Notredame, 2016; Quinn et al., 2017a), Ai and Aj are the log-ratio transformed vectors of two different transcripts (Quinn et al., 2017a). Pairs of transcripts with ρ

> 0.90 or ρ < -0.90 were classified as proportional; we chose the ρ cutoff at 0.90 based on our sample size (28) and number of transcripts (17597), as suggested by Quinn et al (2017). We

168 chose to investigate all groups of proportional transcripts that contained at least one differentially expressed transcript, as well as any group containing more than 50 transcripts (N > 50).

Isolation by distance

To examine the effects of geographic distance on genetic distance, we performed a Mantel test using ade4 v1.7-5 (Thioulouse & Dray, 2007). If gene expression differences are primarily driven by changes in allele proportions due to drift or selection driven by environmental gradients co-occurring with the expanding range, then we might expect a significant linear relationship between geographic distance and genetic distance. However, if gene expression differences are primarily driven by heterogeneous environments at sampling sites, then we would expect a non-significant relationship between geographic distance and genetic distance. We used the dist function in R (Team, 2016) to calculate the Euclidean distances in geographic space between all samples using the coordinates of their collection sites. We then used the dist function to calculate the Euclidean distances in gene expression using the iqlr-transformed counts of all transcripts, thus generating a genetic distance matrix. Genetic distance was calculated as:

푁 2 푑 = √∑1 (푥푖 − 푥푗) (2)

In this equation (Team, 2016), xi is the log-transformed count value of transcript x in sample i, and xj is the log-transformed count value of the same transcript in sample j. The expression difference in each of N transcripts is calculated, and then the square root of the sum of squares of all expression differences is computed.

169 Supplemental Results & Discussion

Coordination in gene expression

We identified many groups of proportionally expressed (co-associated) transcripts; however, few of these groups contained transcripts that were also differentially expressed. We focused only on those that were (list of proportional transcripts in Appendix III). The largest group of proportionally expressed transcripts (N=124; Figure S3A), some of which were part of the fourth cluster (high expression in intermediate areas, equally low expression at the ends of the range) was involved in translation, with functions such as ribosome binding, initiation, elongation, termination, and fidelity. We also identified two small groups of proportionally expressed transcripts mostly involved in platelet activation and adhesion, most of which were part of the first cluster (low expression at the core, equally high expression throughout the rest of the range).

170

Figure S3. Groups of proportionally expressed transcripts in spleen tissue from the invasive Australian cane toad (Rhinella marina). Nodes represent transcripts, and lines indicate proportionality between them. Black fill denotes up-regulation in toads from intermediate areas, gray fill denotes down-regulation in toads from the core, and no fill denotes no differential expression. All transcripts were positively proportional. The propr package in R was used to identify transcripts with expression patterns that were proportional to those of other transcripts from RNA-Seq data obtained from spleens sampled across the Australian range (Figure 1). A REVIGO plot of the largest group of proportionally expressed transcripts is included to show their most common functions.

Although we found few immune transcripts up-regulated in toads from intermediate areas

(and none in our fourth cluster), this cluster does consist of many transcripts with similar functions: approximately half of the transcripts are involved in translation initiation, and proportionality analysis revealed that many of them are coordinated. Because translation serves a wide variety of roles, it is surprising to see a large group of transcripts encoding ribosome components and translation initiation factors up-regulated in intermediate toads. One explanation for this may be that transcripts up-regulated by intermediate toads are, on average, shorter than transcripts up-regulated by toads from the other invasion phases (2555 bases in core toads, 1012 bases in intermediate toads, 1824 bases in frontal toads). Shorter transcripts yield higher

171 ribosome density, and the rate of translation initiation may be higher when transcripts are shorter

(Ingolia et al., 2009). Ribosomes are likely able to begin a new round of translation more frequently if they are translating shorter transcripts. Intermediate toads up-regulate the shortest transcripts of any phase, and down-regulate the longest transcripts; there is no clear adaptive explanation for this trend.

172 CHAPTER 5: Effects of invasion history on physiological responses to immune system activation in invasive Australian cane toads

This chapter is published in PeerJ: Selechnik D, West AJ, Brown GP, Fanson KV, Addison B, Rollins LA, Shine R. 2017. Effects of invasion history on physiological responses to immune system activation in invasive Australian cane toads. PeerJ, 5:e3856.

173 Abstract The cane toad (Rhinella marina) has undergone rapid evolution during its invasion of tropical

Australia. Toads from invasion front populations (in Western Australia) have been reported to exhibit a stronger baseline phagocytic immune response than do conspecifics from range core populations (in Queensland). To explore this difference, we injected wild-caught toads from both areas with the experimental antigen lipopolysaccharide (LPS, to mimic bacterial infection) and measured whole-blood phagocytosis. Because the hypothalamic-pituitary-adrenal axis is stimulated by infection (and may influence immune responses), we measured glucocorticoid response through urinary corticosterone levels. Relative to injection of a control (phosphate- buffered saline, PBS), LPS injection increased both phagocytosis and the proportion of neutrophils in the blood. However, responses were similar in toads from both populations. This null result may reflect the ubiquity of bacterial risks across the toad’s invaded range; utilization of this immune pathway may not have altered during the process of invasion. LPS injection also induced a reduction in urinary corticosterone levels, perhaps as a result of chronic stress.

Keywords: Rhinella marina; cane toad; phagocytosis; eco-immunology; invasive species

174 Introduction Eco-immunological theory predicts that successful invaders will display reduced investment in components of the immune system that produce excessive inflammation, and/or are energetically expensive (Lee & Klasing 2004; White et al. 2012). This prediction is based on the enemy release hypothesis (Colautti et al. 2004): the supposition that invasive hosts lose many co- evolved enemies after translocation (Allendorf 2003; Torchin et al. 2001), potentially reducing pathogen-mediated selection pressures (Lee & Klasing 2004; White et al. 2012). Also, the energetic costs of mounting a strong immune response may reduce the host’s ability to survive, grow, reproduce, and disperse (Hart 1988; Klein & Nelson 1999; Llewelyn et al. 2010); these factors influence invasion success (Chapple et al. 2012; Cote et al. 2010). However, such a down-regulation of immune function may render invaders susceptible to infection by novel pathogens and parasites in their introduced range (Hamrick et al. 1992); for this reason, invaders are also predicted to elevate investment into less energetically costly components of the immune system (Lee & Klasing 2004).

Components of anti-microbial activity within the innate immune system differ in the amount of energy that they require and inflammation that they cause. Systemic mechanisms such as acute phase protein activity, anorexia, lethargy, and fever are highly inflammatory, and thus may be “costly” (Klasing & Leshchinsky 1999). These responses are predicted (Cornet et al.

2016; Lee & Klasing 2004), and have been shown, to be down-regulated in invasive populations of invertebrates (Cornet et al. 2010; Wilson-Rich & Starks 2010), sparrows (Lee et al. 2005), trout (Monzon-Arguello et al. 2014), and deer (Quéméré et al. 2015). Although constitutive innate defences (such as whole-blood phagocytosis of bacteria or yeast) also require substantial energy to activate (McDade et al. 2016), glucose metabolism does not increase during

175 phagocytosis in human neutrophils (Borregaard & Herlin 1982). Thus, it is difficult to predict whether or not these defences are down-regulated in invaders, and further data are required.

The cane toad (Rhinella marina) was brought to Queensland, Australia from Hawai’i in

1935 (Turvey 2009). Reflecting traits such as high fecundity and long-distance dispersal ability

(Urban et al. 2008), toads have expanded their range into New South Wales (Easteal 1981), the

Northern Territory (Urban et al. 2008), and Western Australia (Rollins et al. 2015). Populations have thus been exposed to local pathogens and parasites in Queensland for 81 years, whereas toads near the invasion front in Western Australia may be encountering novel pathogens for the first time. Surveys and common-garden experiments have shown that several phenotypic characteristics (including morphology, physiology, and behaviour) have diverged between populations from the “range core” (QLD) and “invasion front” (WA; Brown et al. 2015c; Gruber et al. 2017; Hudson et al. 2016). Due to differences in selection pressures for established populations (in the range core) versus expanding populations (at the invasion front), comparisons between them provide similar results to those expected between native and invasive populations.

Brown et al. (2015) compared the immunocompetence of captive-raised cane toads whose parents had been collected from QLD and WA. No significant difference was found in

PHA-induced skin swelling, but WA toads exhibited higher bacteria-killing activity and phagocytosis than did QLD toads (Brown et al. 2015c). This result suggests that bacteria-killing activity and phagocytosis may be favored at the invasion front because these responses are less costly. However, Brown et al.’s study measured baseline levels of immune components, rather than the levels elicited by an in vivo immune challenge. One problem with comparing baseline levels of immune responses (such as phagocytosis and bacteria-killing activity) is the amount of variation across individuals in prior exposure to pathogens or antigens. Although individuals

176 within the same population encounter the same types of infection, they may host pathogenic mutants of varying levels of virulence (McCahon et al. 1981). Measuring an immune response before and after experimental infection with agents like lipopolysaccharide (LPS, a bacterial endotoxin produced by Escherichia coli), and treating the change from baseline as the response variable, partially solves this problem by allowing comparisons within individuals through time, as well as among individuals. This is because baseline levels of an immune response (which are reflective of current pathogen load and previous exposure) are accounted for when analysing the response to the antigen. Experimental infection also provides the opportunity to study immune stimulation in a situation where the exact identity and dosage of the experimental antigen are known; this antigen has been isolated and purified, is no longer part of a live organism, and is not a nucleic acid (Gould 2008), so mutation and replication are not possible. Thus, estimates of the strength of the immune response (duration, maximum response, and time to maximum response) are not confounded by the effects of differing pathogenic challenges. We applied the experimental infection method to clarify the relative phagocytic capabilities of toads from the range core versus toads from the invasion front.

Like the immune system, the hypothalamic-pituitary-adrenal (HPA) axis is stimulated by infection (Dunn & Vickers 1994). The HPA axis is a major neuroendocrine system in vertebrates, and results in the release of glucocorticoids, which are closely associated with stress and activity (Janeway et al. 2001). Due to the suppressive effects of glucocorticoids on immune function (Alberts et al. 2002; Cooper 2000; Fanning et al. 1998), glucocorticoids may regulate immune responses, preventing them from being elevated to a level that is harmful to the host

(Ruzek et al. 1999; Stewart et al. 1988). In cane toads, associations between corticosterone and immune responses have previously been documented, with corticosterone having a negative

177 effect on complement lysis activity, but a positive effect on leukocyte oxidative burst (Graham et al. 2012). Furthermore, toads with longer legs (a characteristic of WA toads) exhibited a reduced corticosterone response to stress (capture and confinement) than did their shorter-legged conspecifics (Graham et al. 2012). Because of this potential regulatory interaction and its relevance to immune function modulation in invaders, we also measured the effect of experimental infection on the glucocorticoid response.

Our study compared immune and glucocorticoid responses of wild-caught cane toads from both the invasion front (WA) and range core (QLD) populations after experimental injection with LPS. We expected infection with LPS to cause an increase in phagocytosis through stimulation of the immune system, regardless of population; we did not expect PBS to have an effect. Because infection affects the HPA axis (Dunn et al. 1989), we also expected LPS injection to increase corticosterone levels, regardless of population; we did not expect PBS to have an effect. At the population level, we expected differential effects of LPS on QLD toads and WA toads. If phagocytosis is indeed less costly than other immune responses, then we expected that individuals from the WA population may display higher levels of phagocytosis after injection with LPS than would individuals from the QLD population (as was seen by Brown et al. 2015 in common garden-bred toads). Because Graham et al. (2012) reported that cane toads of different leg lengths differed in glucocorticoid responses, we also expected a population difference in corticosterone levels (with the caveat that LPS injection may elicit a different response than that of capture and confinement, as used in the study by Graham et al.).

178 Materials and Methods Animal Collection and Husbandry

In May of 2016, specimens of R. marina were collected from two locations on opposite ends of the invasion transect. The eastern location (Cairns, QLD; 16.9186S 145.7781E) is the site of the initial release of toads into the wild in 1936 (Turvey 2013). Toads did not arrive at the western location (Oombulgurri, WA; 15.1818S 127.8413E) until 2015; thus, this population represents the invasion front (Figure 1). Ten female toads per location were captured and transported to

Middle Point, Northern Territory (12.5648S, 131.3253E), where they were maintained in a common setting for approximately one month before the experiment began. Only females were collected to eliminate possible sex effects, and for comparison with data on gene expression in female cane toads from a concurrent study. The experiment was conducted during the dry season in the Northern Territory, and thus toads were not breeding at this time. Toads from each location were divided into two groups of five: LPS-injection and PBS-injection (phosphate- buffered saline, control). Specimens were kept separate by their assigned group, and housed in large boxes in groups of two to three individuals. Mesh-covered openings in the boxes provided access to natural light, maintaining specimens on the Australian Central Time Zone light cycle and in outdoor temperatures (nocturnal temperatures ranged from 14 °C to 24.5 °C). Dust-free sawdust was used as a substrate, and plastic containers were provided for shelter. Water was changed daily, and crickets were distributed to each box every third day.

179

Figure 1. Current distribution of the cane toad throughout Australia. Toads were first introduced to Queensland (QLD) in 1935, and have since expanded their range into New South Wales, the Northern Territory (NT), and Western Australia (WA). Black diamonds indicate our toad collection sites: Cairns, QLD and Oombulgurri, WA.

LPS administration

Injections were performed using disposable 25-gauge needles with 1-mL syringes (Livshop,

Rosebery, Australia). Each toad assigned to the LPS-injection group was size-matched (based on mass) with a toad from the same population that had been assigned to the PBS-injection group.

Toads were injected with either 20 mg/kg body mass LPS (Sigma-Aldrich, Castle Hill, Australia) diluted in 100 µL PBS (Sigma-Aldrich, Castle Hill, Australia), or with an equal volume of pure

PBS, to the dorsal lymph sac at 16:00 h. Average toad body mass was 131 g.

Blood sampling

Blood samples were taken for hemocytometry, white blood cell count differentials, and the phagocytosis assay. Cardiac punctures were performed using disposable 25-gauge needles with

1-mL heparinized syringes. Approximately 0.25 mL blood per individual was taken, and immediately transferred to a sterile 1.5-mL microcentrifuge tube. Toads were not anesthetized; all samples were collected within 3 min of disturbance to the toad. This procedure was conducted

180 on each toad twice: three days before and fourteen days after injection, each time at 10:00 h. We allowed three days between blood collection and injection for toads to settle from the disturbance. Fourteen days after injection were allowed for toads to mount an immune response; cellular and humoral immune responses have previously been shown to reach their maximum within this time frame in toads (Diener & Marchalonis 1970).

Hemocytometry

To quantify the concentration of blood cells in each sample, 5 µL whole blood was diluted in

995 µL Natts-Herrick solution (Australian Biostain, Traralgon, Australia) and stored for 24 h at 4

°C. Then, blood cells were resuspended in the solution by inversion before 10 µL of the mixture was loaded into a hemocytometry chamber, and the numbers of erythrocytes (RBCs) and leukocytes were counted.

Counts of White Blood Cell Differentials

Approximately 2 µL whole blood was used to prepare a smear that was then air-dried for an hour, and then stained with Diff-Quik (IHC World, LLC, Woodstock, USA). After 24 h, cover slips were placed on each slide using a thin layer of mounting medium and samples were given another 24 h to dry. Slides were scanned at 100X, and the first 100 leukocytes seen were identified as basophils, eosinophils, neutrophils, lymphocytes, or monocytes. Percentages of each cell type (number of cells of each type divided by 100) were calculated. Because neutrophils are common phagocytes, the relationship between neutrophil percentage and phagocytosis was assessed.

181 Phagocytosis assay

We used a phagocytosis assay in which whole blood samples were stimulated by zymosan in the presence of luminol, generating luminescence (in relative light units, RLU) as a measure of phagocytosis (Martinez & Lynch 2013). Whole blood was first diluted 1:20 in Amphibian

Ringers (AR) solution, and 240 µL of the mixture was added to duplicate wells in a 96-well plate along with 30 µL luminol (Sigma-Aldrich, Castle Hill, Australia) and 10 µL zymosan (Sigma-

Aldrich, Castle Hill, Australia). Another 240 µL of the sample was added to a control well along with 30 µL luminol and 10 µL PBS instead of zymosan. After the addition of zymosan, the 96- well plate was immediately inserted into a luminometer. Light emissions were recorded every 5 min for 200 min. The luminescence value in the control (PBS) well of each sample was subtracted from the luminescence values in the two corresponding zymosan wells to control for variations in light emissions between samples unrelated to the addition of zymosan; duplicates were then averaged together. Because there are multiple facets to the strength of an immune response (duration, maximum, and speed), phagocytosis was assessed via three response variables: mean luminescence across time points, maximum luminescence, and time to reach maximum luminescence. These three response variables were natural log-transformed for data normalization, then run through a principal component analysis (PCA) to determine the best- fitting vector to represent all of the data in a single measure, called principal component 1 (PC1;

Table 1). High PC1 values indicate high average luminescence, high maximum luminescence, and short time to maximum luminescence.

182 Table 1. Loading values for principal component analysis (PCA) formed from three phagocytosis measures. Whole-blood phagocytosis in cane toads was assessed via mean luminescence across time points, maximum luminescence, and time to reach maximum luminescence. These response variables were natural log-transformed, and then run through a PCA. Immune Measure PC1 (70.6%)

Mean luminescence 0.97

Max luminescence 0.93

Time to max luminescence -0.56

Urine sampling

To obtain urine for corticosterone analysis, toads were lifted gently from their boxes and held over a plastic cup for up to 3 minutes. Urine was not collected from toads that did not urinate within this period. Urine was immediately transferred to a 2-mL snap-cap tube and stored at -20

°C. Urine sampling was conducted at seven time points during the experiment: three days (10:00 h and 22:00 h) and two days (10:00 h) prior to injection, as well as six hours (22:00 h), one day

(22:00 h), seven days (22:00 h), and twelve days (10:00 h) after injection. Samples were collected at two different times of day to incorporate periods of activity (22:00 h), when corticosterone levels are relatively high, and periods of inactivity (10:00 h), when corticosterone levels are lower (Jessop et al. 2014).

Creatinine Assay

Creatinine concentration (g/mL) was measured in every urine sample to standardize corticosterone levels by controlling for concentration of urine. Creatinine quantification was based on the Jaffe reaction in which creatinine turns orange in the presence of alkaline picrate.

Briefly, 100 L neat urine was mixed with 50 L 0.75M sodium hydroxide (NaOH) and 50 L

0.04N picric acid in duplicate on a 96-well plate. The plate was then incubated at room

183 temperature for 15 min. Absorbance was measured with a plate reader (Biochrom Anthos 2010,

Biochrom Ltd., UK) using a 405 nm measuring filter and a 620 nm reference filter.

Corticosterone Assay

Urinary corticosterone metabolites were analyzed using an enzyme-immunoassay (EIA) that has been previously validated for cane toads (Narayan et al. 2012). Urinary corticosterone has been shown to lag behind plasma corticosterone by only 1 h (Narayan et al. 2013). The polyclonal corticosterone antibody (CJM06) and corresponding label (corticosterone conjugated with horseradish peroxidase [HRP]) were supplied by Smithsonian National Zoo (Washington D. C.,

USA). Briefly, high binding 96-well plates (Costar) were coated with 150 µl of coating buffer containing goat anti-rabbit IgG (GARG; 2 µg/ml). After 24 h, the coating solution was discarded and 200 µL of Trizma buffer solution rich in bovine serum albumin was added to each well and incubated for at least 4 h. Plates were washed 5 times and immediately loaded with 50 L of standard, control, or neat urine sample, 50 L of corticosterone-HRP (working dilution =

1:80,000), and 50 L of corticosterone antibody (working dilution = 1:100,000). After incubating for 2 h at room temperature, plates were washed and 150 μL of substrate solution (1.6 mM hydrogen peroxide, 0.4 mM azino-bis(3-ethylbenzthiazoline-6-sulfonic acid) in 0.05 M citrate buffer, pH 4.0) was added to each well. The plate was incubated at room temperature for

45 min, and absorbance was quantified using a 450 nm measuring filter and a 620 nm reference filter. All samples were analyzed in duplicate and hormone concentration is expressed as ng/μg creatinine. Urine corticosterone metabolite concentrations were natural log-transformed to meet model assumptions.

184 Statistical Analyses

All statistical analyses were performed using JMP Pro 11.0 (SAS Institute, Cary, NC). We first subtracted each individual’s pre-injection PC1 score from its post-injection PC1 score, and then analyzed the differences using a linear mixed model containing red blood cell concentration, population (QLD vs. WA), and treatment (LPS injection vs. PBS injection) as fixed effects. The interactive effects of population and treatment were also included in the model. We used this model to assess variation in PC1 scores, as well as variation in neutrophil percentages. Changes in PC1 were also tested for correlation with changes in neutrophil percentage due to the putative role of neutrophils in phagocytosis. A p-value less than 0.05 was required to call significance of an effect or relationship.

Hormone data were analyzed using a linear mixed model containing population, treatment, and days post-injection (-3.25, -2.75, -2.25, 0.25, 1.25, 7.25, and 12.25 DPI) as fixed effects. The interactive effects of population, treatment, and DPI were also included in the model. Because several repeated measures were taken for each toad, we also included individual

ID as a random effect in the model. A p-value less than 0.05 was required to call significance of an effect or relationship.

Results Phagocytosis

We found an effect of LPS challenge treatment on phagocytosis PC1 (treatment effect p = 0.02).

LPS-injected toads exhibited a greater increase in phagocytosis than did their PBS-injected counterparts after injection (Figure 2). RBC count also had a significant effect; toads whose blood samples were more concentrated with RBCs exhibited higher levels of phagocytosis.

185 However, population had no significant effect, indicating similarity in the two populations’ phagocytic response to LPS challenge (Table 2).

Figure 2: Phagocytosis curves of cane toads (Rhinella marina) from (A) Queensland (QLD) and (B) Western Australia (WA) both before and after injection with either lipopolysaccharide (LPS) or phosphate-buffered saline (PBS). The average of all samples within a treatment group at each time point (N=5) was calculated to produce the points on the graph. (C) Difference in mean luminescence values between pre-injection and post-injection readings of each population and treatment group.

186 Table 2. Effect of source population and treatment (LPS or PBS) on phagocytic activity in blood of the cane toad, Rhinella marina. Each individual’s pre-injection PC1 score was subtracted from its post- injection PC1 score, and differences were analyzed using a linear mixed model containing red blood cell concentration (RBC), population (Queensland vs. Western Australia), and treatment (LPS injection vs. PBS injection) as fixed effects. Significant effects are in bold. Source Estimate Standard 95% Confidence DF F Ratio p Error Interval RBC 0.01 0.004 [0.004, 0.02] 1,15 8.70 0.01

Population 0.63 0.33 [-0.07, 1.34] 1,15 3.61 0.08

Treatment -0.87 0.34 [-1.58, -0.15] 1,15 6.64 0.02

Population × Treatment -0.06 0.33 [-0.77, 0.65] 1,15 0.03 0.85

White Blood Cell Differentials

Similar to PC1, a strong treatment effect was seen on change in neutrophil percentage (p =

0.0052). Across both populations, LPS-injected toads exhibited a greater increase in the percentages of neutrophils in their blood than did their PBS-injected counterparts after injection

(Figure 3a). Across treatments, the change in neutrophil percentage before versus after injection was positively correlated with the change in PC1 before versus after injection (p = 0.037, R2 =

0.23; Figure 3b).

187

Figure 3: (A) Changes in the percentages of neutrophils (as induced by injection of either lipopolysaccharide [LPS] or phosphate-buffered saline [PBS]) across two treatment groups of cane toads (Rhinella marina) from two populations (invasion front site in Western Australia [WA] and range core site in Queensland [QLD]). The percentage of neutrophils in the toad’s blood pre-injection was subtracted from the percentage of neutrophils in the same toad’s blood post-injection. The average of the difference between pre-injection and post-injection of all samples within a treatment group at each time point (N=5) was calculated to produce the points on the graph. Error bars indicate standard error. (B) Positive correlation between the changes in neutrophil percentage and PC1 before versus after injection.

Urinary Corticosterone Metabolites

A strong treatment × days post-injection (DPI) effect was observed on urinary corticosterone levels (Table 3). Corticosterone increased over time after injection in PBS-injected toads, but decreased over time after injection in LPS-injected toads (Figure 4). However, we found no significant difference between populations.

188 Table 3: Effects of each explanatory variable on corticosterone levels of cane toads after injection with either lipopolysaccharide (LPS) or control (phosphate-buffered saline, PBS). Natural log-transformed corticosterone values were analyzed in a linear mixed model with population (Queensland vs. Western Australia), treatment (LPS injection vs. PBS injection), and days post-injection (DPI) as fixed effects; their interactive effects were also assessed. Significant effects are in bold. Source Estimate Standard DF F Ratio p Error Population -0.14 0.17 1,14 0.67 0.4253

Treatment -0.04 0.17 1,14 0.05 0.8311

DPI -0.02 0.02 1,41 1.83 0.1839

Population × Treatment 0.02 0.17 1,14 0.02 0.8909

Population × DPI -0.02 0.02 1,41 1.12 0.2964

Treatment × DPI -0.05 0.02 1,41 10.53 0.0023

Population × Treatment × DPI 0.02 0.02 1,41 0.89 0.3516

Figure 4: Changes in urinary corticosterone levels across two treatment groups of cane toads (Rhinella marina) from two populations (invasion front site in Western Australia [WA] and range core site in Queensland [QLD]). Toads were injected with either lipopolysaccharide (LPS) or phosphate-buffered saline (PBS). Error bars indicate standard error, and lines are fitted by linear regression to data from each treatment group.

189

Discussion Injection with LPS evoked up-regulation of immune responses in cane toads from both expanding (invasion front) and established (range core) populations. Two weeks after injection with LPS, toads had increased circulating levels of neutrophils, as well as elevated phagocytic ability. Control toads injected with PBS showed no significant changes in these immune measures. A positive correlation between phagocytosis levels and neutrophils was expected, as neutrophils are the most abundant phagocytic cells in the blood (Summers et al. 2010; Wilgus et al. 2013; Wright 2001). Indeed, toads that increased neutrophil production the most also increased phagocytic ability the most, indicating an increased stimulation of neutrophil-mediated phagocytosis by LPS. Thus, LPS is an effective stimulant of the phagocytic response in cane toads via their neutrophils.

Blood samples that were more concentrated with RBCs exhibited higher levels of phagocytosis. However, RBCs had no effect on neutrophil abundance (p = 0.55). Thus, the effect could have potentially arisen through luminescent properties of the RBCs themselves

(autofluorescence; Emmelkamp et al. 2003).

Contrary to our predictions, toads from the invasion front did not exhibit stronger phagocytic responses to LPS exposure than did toads from the range core; there was no difference between populations. Our predictions were based upon the possibility that neutrophil- mediated phagocytosis poses a lower energetic cost than other immune responses, and that this may be favored in toads at the invasion front undergoing enemy release. Brown et al. (2015) reported higher baseline levels of neutrophil-mediated phagocytosis in common garden-raised toads from WA, consistent with the hypothesis that neutrophil-mediated phagocytosis provides an inexpensive way for WA toads to retain some immunocompetence without expending much

190 energy. Thus, we expected phagocytosis to be favoured in the wild-collected adult toads from

WA as well, even with the change in methodology (injection with LPS vs measurement of baseline levels). However, LPS is a common bacterial antigen, and the lack of difference may reflect an equal importance of combating severe bacterial infections across all populations. There may be more than a billion species of bacteria worldwide, and they are found across all environments (Dykhuizen 1998). Because bacteria are able to tolerate many types of abiotic extremes (Dykhuizen 1998), it is likely that the cane toad’s entire Australian range is home to rich bacterial communities. Although microbial species richness follows an aridity gradient

(Yabas et al. 2015) and the particular species of bacteria present at opposite ends of the range may differ, standing innate immune defences such as neutrophils are unspecialized (Janeway et al. 2001), and thus may not differ across populations based on changes in the bacterial species encountered.

However, it is also possible that toads from invasion front versus range core populations differ in phagocytic activity (as reported by Brown et al. 2015), but that those differences were not apparent in our study. Our sample sizes were limited to N=5 per treatment per population.

When a null hypothesis is not rejected (such as in our study, in which a population-level difference was not found), confidence intervals for the effect size are recommended in place of retrospective power analysis to check for validity (Steidl et al. 1997). Our 95% confidence intervals for population as a predictor include zero within their range (Table 2); this indicates that effect could be zero, and thus rejection of the null hypothesis is justified. The same is seen in the 95% confidence intervals for the interactive effect of population and treatment. Sample sizes were low and it is possible that this may not have been sufficient to uncover a population-level

191 difference; however, there are also several biological explanations for the lack of population difference.

Maintaining toads from both populations in a common captive setting, and feeding them the same food, may have eliminated differences in their gut microbiota (Riddell et al. 2014), which in turn influences immune function (Carpenter et al. 2014). Additionally, toads from WA are constantly dispersing into novel environments (Urban et al. 2008) where they will encounter unfamiliar pathogens and parasites. Our study utilized wild-caught toads; prior to our collection, those from WA may have expended more of the energy allocated to neutrophil production and activity than did those from QLD (Brown et al. 2015c), potentially resulting in a diminution of the intrinsically stronger phagocytic response. This idea is supported by a previous study conducted on wild-caught cane toads from the Northern Territory (NT), where the toads were radio-tracked to quantify their movement distances before their immune responses were surveyed (Brown & Shine 2014). Toads that travelled longer distances exhibited decreased standing innate defences, such as neutrophils, compared to less mobile toads (Brown & Shine

2014). Although these toads were not collected from the same areas as those in our experiment,

“more mobile” may be a reasonable proxy for WA, and “less mobile” may represent QLD. If this is the case, a lifetime use of neutrophils and other standing innate effectors may explain why WA toads do not retain the stronger neutrophil responses with which they are born.

Prior energy expenditure is not the only variable that could have obscured a potential difference in phagocytosis levels between populations in our study. Our data suggest that the toads’ immune responses may have also been dampened by abiotic conditions. On average, the

PBS-injection groups from both populations exhibited a decrease in levels of phagocytosis after injection. This decrease may be due to temperature; our experiments were conducted in mid-July

192 to early August, when nocturnal temperatures fall as low as 14 °C. Low temperatures have immunosuppressive effects on amphibians and other ectotherms (Maniero & Carey 1997;

Pxytycz & Jozkowicz 1994; Raffel et al. 2006). Neutrophils and phagocytic activity decrease initially during winter, but return to baseline levels as amphibians acclimate to seasonal temperatures (Raffel et al. 2006). Such seasonal effects in our toads may have masked population differences.

Stress may have also influenced our results. Corticosterone, the primary stress glucocorticoid hormone in amphibians, can induce differential suppression and activation of immune components (Stier et al. 2009). Cane toads exhibit an acute stress response to capture in which their corticosterone levels initially increase (Graham et al. 2012), but decline back to baseline after 7 days of confinement (Narayan et al. 2011); our toads were held in captivity for one month prior to our study. Although our study showed that PBS-injected toads increased in corticosterone levels after injection, we observed the opposite effect in LPS-injected toads. This finding was unexpected; in mammals, glucocorticoids increase during infection, possibly to regulate the immune response and prevent excessive inflammation (Hawes et al. 1992; Ruzek et al. 1999; Stewart et al. 1988; Webster & Sternberg 2004). However, one month in captivity, frequent handling, injection with antigens, and cardiac puncture may have induced a state of chronic stress in the toads (Wingfield & Romero 2001). Adrenal activity is less predictable during chronic stress, and thus the direction of the corticosterone response differs across taxa and stressor types (Dickens & Romero 2013). Because the amount and nature of handling was the same for toads across all treatments, their stress states were likely similar; however, infection apparently made a difference for the LPS-injected toads. Some chronically stressed animals exhibit a decrease in corticosterone levels after infection (Cyr et al. 2007), thereby avoiding the

193 suppressive effects of chronically high corticosterone on immune function. Amphibian immune defences are particularly sensitive to glucocorticoids, and increases in these hormones can raise amphibians’ susceptibility to disease (Rollins-Smith 2001). Additionally, corticosterone is not the only mechanism by which stress can suppress immunity. Stress may also cause leakage of gut microbiota into the bloodstream, triggering immune responses; this results in a lower number of unoccupied effectors (Lambert 2014; Saunders et al. 1994).

Studies on phagocytosis in cane toads have ignored differences in the speed, rather than in the strength, of this immune response. Although phagocytosis levels did not differ significantly between populations, the speed at which the observed up-regulation occurred might have differed. A previous study found that LPS injection triggered a larger metabolic increase after 24 h in QLD toads than in WA toads (Llewellyn et al. 2012). Conceivably, part of this increased metabolic activity seen among QLD toads could have involved increased production of immune components. Our post-injection immune assays were conducted two weeks after LPS injection, by which time toads from both populations had up-regulated immune responses to the same extent. Future studies could explore this question by measuring phagocytosis 24 h after

LPS exposure and monitoring changes in phagocytosis across a shorter time frame.

In this study, we tested phagocytosis in cane toads using the same cell quantification methods and activity assay as those of Brown et al. 2015, but we took repeated measurements before and after injecting LPS in vivo. Our study confirmed that LPS stimulates phagocytosis; however, we did not detect a population-level difference in phagocytosis levels (as had been found in the previous study). Each experiment introduces its own unique confounds; the previous study did not account for inter-individual variation, and ours could not account for differences in environmental effects prior to collection. To definitively compare levels of phagocytosis between

194 individuals from invasion front versus range core populations, a more robust experimental design would employ the experimental antigen methodology simultaneously on wild-caught toads, and on captive-bred toads raised in a common setting from each population.

Acknowledgements

This work was supported by the Australian Research Council (FL120100074, DE150101393) and the Equity Trustees Charitable Foundation (Holsworth Wildlife Research Endowment). All procedures involving live animals were approved by the University of Sydney Animal Care and

Ethics Committee.

Ethics

Research was conducted in accordance with rules set for by the University of Sydney Animal

Ethics Committee. Ethics application was approved under the project number 2016/1003.

Literature Cited Alberts B, Johnson A, and Lewis J. 2002. The Generation of Antibody Diversity. Mol Biol Cell.

4 ed. New York: Garland Science.

Allendorf FW. 2003. Introduction: Population Biology, Evolution, and Control of Invasive

Species. Conservation Biology 17:24 - 30.

Borregaard N, and Herlin T. 1982. Energy Metabolism of Human Neutrophils during

Phagocytosis. J Clin Invest 70:550 - 557.

Brown GP, Phillips BL, Dubey S, and Shine R. 2015c. Invader immunology: invasion history

alters immune system function in cane toads (Rhinella marina) in tropical Australia. Ecol

Lett 18:57-65. 10.1111/ele.12390

195 Brown GP, and Shine R. 2014. Immune Response Varies with Rate of Dispersal in Invasive

Cane Toads (Rhinella marina). PLOS ONE 9:1 - 11. 10.1371/

Carpenter S, Ricci EP, Mercier BC, Moore MJ, and Fitzgerald KA. 2014. Post-transcriptional

regulation of gene expression in innate immunity. Nat Rev Immunol 14:361-376.

10.1038/nri3682

Chapple DG, Simmonds SM, and Wong BB. 2012. Can behavioral and personality traits

influence the success of unintentional species introductions? Trends Ecol Evol 27:57-64.

10.1016/j.tree.2011.09.010

Colautti RI, Ricciardi A, Grigorovich IA, and MacIsaac HJ. 2004. Is invasion success explained

by the enemy release hypothesis? Ecol Lett 7:721-733. 10.1111/j.1461-

0248.2004.00616.x

Cooper GM. 2000. DNA Rearrangements. The Cell: A Molecular Approach. 2 ed. Sinauer

Associates: Sunderland (MA).

Cornet S, Brouat C, Diagne C, and Charbonnel N. 2016. Eco-immunology and bioinvasion:

revisiting the evolution of increased competitive ability hypotheses. Evol Appl 9:952-962.

10.1111/eva.12406

Cornet S, Sorci G, and Moret Y. 2010. Biological invasion and parasitism: Invaders do not suffer

from physiological alterations of the acanthocephalan Pomphorhynchus laevis.

Parasitology 137:137 - 147.

Cote J, Fogarty S, Weinersmith K, Brodin T, and Sih A. 2010. Personality traits and dispersal

tendency in the invasive mosquitofish (Gambusia affinis). Proc Biol Sci 277:1571-1579.

10.1098/rspb.2009.2128

196 Cyr NE, Earle K, Tam C, and Romero LM. 2007. The effect of chronic psychological stress on

corticosterone, plasma metabolites, and immune responsiveness in European starlings.

Gen Comp Endocrinol 154:59-66. 10.1016/j.ygcen.2007.06.016

Dickens MJ, and Romero LM. 2013. A consensus endocrine profile for chronically stressed wild

animals does not exist. Gen Comp Endocrinol 191:177-189. 10.1016/j.ygcen.2013.06.014

Diener E, and Marchalonis J. 1970. Cellular and Humoral Aspects of the Primary Immune

Response of the Toad, Bufo marinus. Immunology 18:279.

Dunn AJ, Powell ML, Meitin C, and Small PA. 1989. Virus Infection as a Stressor: Influenza

Virus Elevates Plasma Concentrations of Corticosterone, and Brain Concentrations of

MHPG and Tryptophan Physiology & Behavior 45:591 - 594.

Dunn AJ, and Vickers SL. 1994. Neurochemical and neuroendocrine responses to Newcastle

disease irus administration in mice Brain Research 645:103 - 112.

Dykhuizen DE. 1998. Santa Rosalia revisited: Why are there so many species of bacteria?

Antonie van Leeuwenhoek Journal of Microbiology 73:25 - 33.

Easteal S. 1981. The history of introductions of Bufo marinus; a natural experiment in evolution.

Biological Journal of the Linnean Society 16:93.

Emmelkamp J, DaCosta R, Andersson H, and van der Berg A. 2003. Intrinsic autofluorescence

of single living cells for label-free cell sorting in a microfluidic system. Transducers

Research Foundation 1:85 - 87.

Fanning L, Bertrand FE, Steinberg C, and Wu GE. 1998. Molecular mechanisms involved in

receptor editing at the Ig heavy chain locus. International Immunology 10:241 - 246.

197 Gould MJ. 2008. Filtration and Purification in the Biopharmaceutical Industry. In: Jornitz MJ,

Jornitz MW, and Meltzer TH, editors. Limulus Amebocyte Lysate Assays and Filter

Applications. 2 ed. Boca Raton, FL: CRC Press. p 425 - 426.

Graham SP, Kelehear C, Brown GP, and Shine R. 2012. Corticosterone-immune interactions

during captive stress in invading Australian cane toads (Rhinella marina). Horm Behav

62:146-153. 10.1016/j.yhbeh.2012.06.001

Gruber J, Brown G, Whiting MJ, and Shine R. 2017. Geographic divergence in dispersal-related

behaviour in cane toads from range-front versus range-core populations in Australia.

Behavioral Ecology and Sociobiology 71. 10.1007/s00265-017-2266-8

Hamrick JL, Godt MJW, and Sherman-Broyles SL. 1992. Factors influencing levels of genetic

diversity in woody plant species. 42:95-124. 10.1007/978-94-011-2815-5_7

Hart BL. 1988. Biological basis of the behavior of sick animals. Neuroscience & Biobehavioral

Reviews 12:123 - 127.

Hawes AS, Rock CS, Keogh CV, Lowry SF, and Calvano SE. 1992. In Vivo Effects of the

Antiglucocorticoid RU 486 on Glucocorticoid and Cytokine Responses to Escherichia

coli Endotoxin. Infect Immun 60:2641 - 2647.

Hudson CM, Brown GP, and Shine R. 2016. It is lonely at the front: contrasting evolutionary

trajectories in male and female invaders. . R Soc Open Sci 3:160687.

Janeway CA, Travers P, and Walport M. 2001. Immunobiology: The Immune System in Health

and Disease. Immunobiology. 5 ed. New York: Garland Science.

Jessop TS, Dempster T, Letnic M, and Webb JK. 2014. Interplay among nocturnal activity,

melatonin, corticosterone and performance in the invasive cane toad (Rhinella marinus).

Gen Comp Endocrinol 206:43-50. 10.1016/j.ygcen.2014.07.013

198 Klasing KC, and Leshchinsky TV. 1999. Functions, costs, and benefits of the immune system

during development and growth. International Ornitology Congress, Proceedings

69:2817 - 2832.

Klein SL, and Nelson RJ. 1999. Activation of the immune–endocrine system with

lipopolysaccharide reduces affiliative behaviors in voles. Behavior Neuroscience

113:1042 - 1048.

Lambert GP. 2014. Stress-induced gastrointestinal barrier dysfunction and its inflammatory

effects. Journal of Animal Science 87:E101-E108.

Lee KA, and Klasing KC. 2004. A role for immunology in invasion biology. Trends Ecol Evol

19:523-529. 10.1016/j.tree.2004.07.012

Lee KA, Martin LB, 2nd, and Wikelski MC. 2005. Responding to inflammatory challenges is

less costly for a successful avian invader, the house sparrow (Passer domesticus), than its

less-invasive congener. Oecologia 145:244-251. 10.1007/s00442-005-0113-5

Llewellyn D, Thompson MB, Brown GP, Phillips BL, and Shine R. 2012. Reduced investment in

immune function in invasion-front populations of the cane toad (Rhinella marina) in

Australia. Biol Invasions 14:999 - 1008.

Llewelyn J, Phillips BL, Alford RA, Schwarzkopf L, and Shine R. 2010. Locomotor

performance in an invasive species: cane toads from the invasion front have greater

endurance, but not speed, compared to conspecifics from a long-colonised area.

Oecologia 162:343-348. 10.1007/s00442-009-1471-1

Maniero GD, and Carey C. 1997. Changes in selected aspects of immune function in the leopard

frog, Rana pipiens, associated with exposure to cold J Comp Physiol B 167:256 - 263.

199 Martinez NM, and Lynch KW. 2013. Control of alternative splicing in immune responses: many

regulators, many predictions, much still to learn. Immunological Reviews 253:216 - 236.

McCahon D, Slade WR, King AMQ, Saunders K, Pullen L, Lake JR, and Priston RAJ. 1981.

Effect of mutation on the virulence in mice of a strain of foot-and-mouth disease virus. J

gen Virol 54:263 - 272.

McDade TW, Georgiev AV, and Kuzawa CW. 2016. Trade-offs between acquired and innate

immune defenses in humans. Evol Med Public Health 2016:1-16. 10.1093/emph/eov033

Monzon-Arguello C, de Leaniz CG, Gajardo G, and Consuegra S. 2014. Eco- immunology of

fish invasions: The role of MHC variation. Immunogenetics 66:393-402.

Narayan EJ, Cockrem J, and Hero JM. 2013. Changes in serum and urinary corticosterone and

testosterone during short-term capture and handling in the cane toad (Rhinella marina).

Gen Comp Endocrinol 191:225-230. 10.1016/j.ygcen.2013.06.018

Narayan EJ, Cockrem JF, and Hero JM. 2011. Urinary corticosterone metabolite responses to

capture and captivity in the cane toad (Rhinella marina). Gen Comp Endocrinol 173:371-

377. 10.1016/j.ygcen.2011.06.015

Narayan EJ, Molinia FC, Cockrem JF, and Hero JM. 2012. Individual variation and repeatability

in urinary corticosterone metabolite responses to capture in the cane toad (Rhinella

marina). Gen Comp Endocrinol 175:284-289.

doi:http://dx.doi.org/10.1016/j.ygcen.2011.11.023

Pxytycz B, and Jozkowicz A. 1994. Differential effects of temperature on macrophages of

ectothermic vertebrates. Journal of Leukocyte Biology 56.

200 Quéméré E, Galan M, Cosson JF, Klein F, Aulagnier S, Gilot-Fromont E, and Charbonnel N.

2015. Immunogenetic heterogeneity in a widespread ungulate: The European roe deer

(Capreolus capreolus). . Mol Ecol 24:3873–3887.

Raffel TR, Rohr JR, Kiesecker JM, and Hudson PJ. 2006. Negative effects of changing

temperature on amphibian immunity under field conditions. Functional Ecology 20:819-

828. 10.1111/j.1365-2435.2006.01159.x

Riddell CE, Lobaton Garces JD, Adams S, Barribeau SM, Twell D, and Mallon EB. 2014.

Differential gene expression and alternative splicing in insect immune specificity. BMC

Genomics 15:1031.

Rollins-Smith LA. 2001. Neuroendocrine-Immune System Interactions in Amphibians.

Immunological Research 23:273 - 280.

Rollins LA, Richardson MF, and Shine R. 2015. A genetic perspective on rapid evolution in cane

toads (Rhinella marina). Mol Ecol 24:2264-2276. 10.1111/mec.13184

Ruzek MC, Pearce BD, Miller AH, and Biron CA. 1999. Endogenous Glucocorticoids Protect

Against Cytokine-Mediated Lethality During Viral Infection. J Immunol 162:3527 -

3533.

Saunders PR, Kosecka U, McKay DM, and Perdue MH. 1994. Acute stressors stimulate ion

secretion and increase epithelial permeability in rat intestine. Am J Physiol

267:G794–G799.

Steidl RJ, Hayes JP, and Schauber E. 1997. Statistical Power Analysis in Wildlife Research. J

Wildl Manage 61.

201 Stewart GL, Mann MA, Ubelaker JE, McCarthy JL, and Wood BG. 1988. A role for elevated

plasma corticosterone in modulation of host response during infection with Trichinella

pseudospiralis. Parasite Immunology 10:139 - 150.

Stier KS, Almasi B, Gasparini J, Piault R, Roulin A, and Jenni L. 2009. Effects of corticosterone

on innate and humoral immune functions and oxidative stress in barn owl nestlings. J Exp

Biol 212:2085-2091. 10.1242/jeb.024406

Summers C, Rankin SM, Condliffe AM, Singh N, Peters AM, and Chilvers ER. 2010.

Neutrophil kinetics in health and disease. Trends Immunol 31:318-324.

10.1016/j.it.2010.05.006

Torchin ME, Lafferty KD, and Kuris AM. 2001. Release from parasites as natural enemies:

increased performance of a globally introduced marine crab. Biological Invasions 3:333 -

345

Turvey N. 2009. A Toad's Tale. Hot Topics from the Tropics 1:1 - 10.

Turvey N. 2013. Cane toads: A tale of sugar, politics and flawed science. University of Sydney,

Australia: Sydney University Press.

Urban MC, Phillips BL, Skelly DK, and Shine R. 2008. A toad more traveled: the heterogeneous

invasion dynamics of cane toads in Australia. The American Naturalist 171:E134 - E148.

Webster JI, and Sternberg EM. 2004. Role of the hypothalamic–pituitary–adrenal axis,

glucocorticoids and glucocorticoid receptors in toxic sequelae of exposure to bacterial

and viral products. Journal of Endocrinology 181:207 - 221.

White TA, Perkins SE, and Dunn A. 2012. The ecoimmunology of invasive species. Functional

Ecology 26:1313-1323. 10.1111/1365-2435.12012

202 Wilgus TA, Roy S, and McDaniel JC. 2013. Neutrophils and Wound Repair: Positive Actions

and Negative Reactions. Adv Wound Care (New Rochelle) 2:379-388.

10.1089/wound.2012.0383

Wilson-Rich N, and Starks PT. 2010. The Polistes war: Weak immune function in the invasive P.

dominulus relative to the native P. fuscatus. Insectes Sociaux 57:47 - 52.

Wingfield JC, and Romero LM. 2001. Adrenocortical Responses to Stress and Their Modulation

in Free-Living Vertebrates. In: McEwen BS, and Goodman HM, eds. Handbook of

Physiology. New York: Oxford University Press 211 - 234.

Wright KM. 2001. Amphibian Medicine and Captive Husbandry. In: Wright KM, and Whitaker

BR, editors. Amphibian Hematology. Florida: Krieger.

Yabas M, Elliott H, and Hoyne GF. 2015. The Role of Alternative Splicing in the Control of

Immune Homeostasis and Cellular Differentiation. Int J Mol Sci 17.

10.3390/ijms17010003

203 CHAPTER 6: Conclusions

In this thesis, I have tested two major evolutionary paradigms: the genetic paradox of invasion and the enemy release hypothesis (ERH). To do this, I utilized immune function, gene expression, and genetic data; the latter two were produced using novel technologies collectively referred to as next-generation sequencing (NGS). I first investigated how variations in the analytical methodology may affect the results, and then framed some suggestions to add to the current body of ‘best practice’ recommendations when analysing these data.

My analyses of simulated NGS datasets revealed that filtering parameter choices do not only affect inferences of population structure, but do so differentially depending on the inherent levels of genetic structure in the dataset. As a result, I recommend (1) analysing each dataset with a variety of filtering parameters and calculating the resulting FST values to estimate levels of genetic differentiation between populations; then (2) choosing an ultimate combination of filtering parameter choices based on the level of genetic differentiation in the specific dataset, as stringency is most important when filtering datasets with relatively low levels of differentiation. I also recommend reporting an array of information that my literature review revealed were reported inconsistently, but would be useful to readers (particularly for replicability). This study helped to inform my methodology and data reporting in the two data chapters in which I analysed NGS data to draw ecological and evolutionary conclusions.

With those recommendations in mind, I used genetic data to test whether invasive

Hawai’ian and Australian cane toads meet the conditions of the genetic paradox of invasion – a reduction of genetic diversity from the native range, and adaptation to novel challenges in the introduced range. I found evidence that toads have suffered a slight loss in overall genetic diversity during invasion, yet respond genetically to novel adaptive challenges in Australia.

204 However, I did not see a decline from the native population to the earlier invasive population in loci putatively under selection. This could be because genetic diversity at ecologically relevant traits was maintained by balancing selection, in which case the cane toad invasion would not represent a true genetic paradox. Alternatively, this could be due to high mutation rates restoring adaptive potential, in which case the cane toad invasion does represent a true paradox. Thus, my data remain equivocal on whether cane toads fit the model of the genetic paradox of invasion.

Similarly, I used gene expression and immune function data to test whether patterns in immunity of invasive Australian cane toads provide support for the conceptual scheme laid out by Lee & Klasing (2004) based on the ERH – a reduction of energetic investment into costly immune responses in the invader due to a release from pathogens and parasites. The results from my gene expression data were opposite to what I predicted – expression of genes encoding mediators of costly immune responses was higher in toads from areas where parasite density was lower. Conversely, the results from my physiological data found no population-level differences in any measures of immune function that I tested. Thus, neither my gene expression study nor my immune function experiment provide support for the predictions of Lee & Klasing (2004).

Although my data across these three chapters provided mixed support for the genetic paradox and little support for the ERH, they all suggested that there are evolutionary forces other than natural selection that are important in shaping this invasion, which may explain why evolutionary questions involving natural selection are difficult to study in wild populations.

Statistically significant results from my isolation by distance (IBD) tests on both genetic and gene expression data suggest that genetic drift and spatial sorting are likely to be occurring as well. Because both forces drive non-adaptive variation, they may be responsible for the seemingly unintuitive gene expression results that I observed. For these reasons, my gene

205 expression and immune function studies highlight the importance of future common-garden experiments coupled with genetic analyses. Common-garden experiments remove many of the experimental confounds present when studying populations in the wild. However, studies like mine remain necessary, as common-garden studies do not always accurately represent what occurs in the wild. Despite the shortcomings of studies conducted in the wild, this body of work is among the first to use NGS to address these ecological and evolutionary questions, and it will inform and guide future studies aiming to elucidate these processes in an invasion context.

206 APPENDICES All appendices are associated with Chapter 4: Immune and environment-driven gene expression during invasion: An eco- immunological application of RNA-Seq

Appendix I Differentially expressed transcripts in cane toads (Rhinella marina) across the Australian range (main text, Figure 1). RNA-Seq data from spleens was used to quantify differential expression (DE) analysis between phases of the invasion.

Table AI1. Core (QLD) up-regulation Gene Protein effect p Irgc Interferon-inducible GTPase 5 2.03 0.00 Irgc Interferon-inducible GTPase 5 1.59 0.00 QRICH1 Glutamine-rich protein 1 1.57 0.00 AIM1L Absent in melanoma 1-like protein 1.35 0.01 stmn1-a Stathmin-1-A 1.34 0.01 . . 1.34 0.01 . . 1.30 0.02 COL4A1 Collagen alpha-1(IV) chain 1.24 0.02 MGEA5 Protein O-GlcNAcase 1.23 0.02 GRK6 G protein-coupled receptor kinase 6 1.17 0.03 Irgc Interferon-inducible GTPase 5 1.16 0.02 ezh2 Histone-lysine N-methyltransferase EZH2 1.16 0.02 Irgc Interferon-inducible GTPase 5 1.15 0.02 AZIN1 Antizyme inhibitor 1 1.14 0.03 . . 1.12 0.03 RBM25 RNA-binding protein 25 1.11 0.03 PTPRJ Receptor-type tyrosine-protein phosphatase eta 1.09 0.02 PRPF4B Serine/threonine-protein kinase PRP4 homolog 1.08 0.02 Xpo1 Exportin-1 1.08 0.03 chmp4b Charged multivesicular body protein 4b 1.08 0.02

207 DNAH11 Dynein heavy chain 11, axonemal 1.08 0.04 GPCPD1 Glycerophosphocholine phosphodiesterase GPCPD1 1.08 0.04 Nav3 Neuron navigator 3 1.06 0.02 ZNF341 Zinc finger protein 341 1.04 0.03 camsap1 Calmodulin-regulated spectrin-associated protein 1 1.04 0.04 Irgc Interferon-inducible GTPase 5 1.04 0.04 . Oocyte zinc finger protein XlCOF22 1.03 0.05 PLEKHA7 Pleckstrin homology domain-containing family A member 7 1.02 0.04 CCDC150 Coiled-coil domain-containing protein 150 1.02 0.05 NOP56 Nucleolar protein 56 1.00 0.02 WDR33 pre-mRNA 3' end processing protein WDR33 1.00 0.03 UPF1 Regulator of nonsense transcripts 1 0.99 0.04 CK095 Uncharacterized protein C11orf95 0.99 0.04 BTAF1 TATA-binding protein-associated factor 172 0.99 0.05 . . 0.98 0.05 NOP58 Nucleolar protein 58 0.96 0.05 impdh1a Inosine-5'-monophosphate dehydrogenase 1a 0.96 0.03 DYR1A Dual specificity tyrosine-phosphorylation-regulated kinase 1A 0.95 0.04 MOV10 Putative helicase MOV-10 0.87 0.05

Table AI2. Core (QLD) down-regulation Gene Protein effect p . LINE-1 reverse transcriptase homolog -1.88 0.00 . . -1.72 0.00 CLUL1 Clusterin-like protein 1 -1.64 0.00 . . -1.60 0.00 ERVW-1 Syncytin-1 -1.60 0.00 CREG1 Protein CREG1 -1.56 0.00 ENDOD1 Endonuclease domain-containing 1 protein -1.55 0.00 CTSK Cathepsin K -1.43 0.01 Hmox2 Heme oxygenase 2 -1.42 0.01

208 . . -1.39 0.00 PXN1 Jeltraxin -1.38 0.01 . . -1.38 0.01 prr5 Proline-rich protein 5 -1.36 0.01 CFH Complement factor H -1.33 0.02 Pafah2 Platelet-activating factor acetylhydrolase 2, cytoplasmic -1.32 0.01 . . -1.31 0.02 NMT2 Phosphomethylethanolamine N-methyltransferase -1.31 0.02 FCGR2 Low affinity immunoglobulin gamma Fc region receptor II -1.30 0.00 Tub Tubby protein -1.28 0.01 itln1 Intelectin-1 -1.27 0.01 CD200R1B Cell surface glycoprotein CD200 receptor 1-B -1.27 0.00 REEP5 Receptor expression-enhancing protein 5 -1.26 0.01 PXN1 Jeltraxin -1.26 0.01 pol Pol polyprotein -1.26 0.02 mul1a Mitochondrial ubiquitin ligase activator of nfkb 1-A -1.25 0.02 CYP3A29 Cytochrome P450 3A29 -1.25 0.01 ST3GAL6 Type 2 lactosamine alpha-2,3-sialyltransferase -1.25 0.01 pol Pol polyprotein -1.25 0.02 MPL Thrombopoietin receptor -1.25 0.02 Gp1bb Platelet glycoprotein Ib beta chain -1.23 0.02 . . -1.21 0.02 U88 Uncharacterized protein U88 -1.20 0.03 Arsa Arylsulfatase A -1.18 0.01 St3gal5 Lactosylceramide alpha-2,3-sialyltransferase -1.18 0.04 PXDC1 PX domain-containing protein 1 -1.17 0.03 ERVPABLB-1 Endogenous retrovirus group PABLB member 1 Env polyprotein -1.16 0.03 LORF2 LINE-1 retrotransposable element ORF2 protein -1.15 0.01 . . -1.15 0.02 . . -1.15 0.01 F10 Coagulation factor X -1.14 0.02

209 CYR61 Protein CYR61 -1.14 0.02 TRIM25 E3 ubiquitin/ISG15 ligase TRIM25 -1.14 0.04 MED12L Mediator of RNA polymerase II transcription subunit 12-like protein -1.13 0.03 Pinlyp phospholipase A2 inhibitor and Ly6/PLAUR domain-containing protein -1.12 0.03 pssA CDP-diacylglycerol--serine O-phosphatidyltransferase -1.11 0.04 ARHGEF19 Rho guanine nucleotide exchange factor 19 -1.11 0.03 TESPA1 Protein TESPA1 -1.10 0.02 Gp9 Platelet glycoprotein IX -1.09 0.03 ITGA2B Integrin alpha-IIb -1.09 0.03 Dhrs9 Dehydrogenase/reductase SDR family member 9 -1.09 0.03 . . -1.08 0.04 PTPRM Receptor-type tyrosine-protein phosphatase mu -1.08 0.02 . . -1.08 0.04 . . -1.08 0.05 . . -1.07 0.03 fn1 Fibronectin -1.07 0.02 Ecm1 Extracellular matrix protein 1 -1.07 0.03 CCDC126 Coiled-coil domain-containing protein 126 -1.06 0.04 Igk-V19-17 Ig kappa chain V19-17 -1.06 0.03 ZNF84 Zinc finger protein 84 -1.06 0.02 VOPP1 Vesicular, overexpressed in cancer, prosurvival protein 1 -1.05 0.05 Mtcl1 Microtubule cross-linking factor 1 -1.05 0.04 c1galt1 Glycoprotein-N-acetylgalactosamine 3-beta-galactosyltransferase 1 -1.04 0.02 Rtn4rl1 Reticulon-4 receptor-like 1 -1.04 0.03 Dcst1 DC-STAMP domain-containing protein 1 -1.03 0.04 liph-a Lipase member H-A -1.03 0.03 . . -1.03 0.04 ME1 NADP-dependent malic enzyme -1.03 0.04 GPR21 Probable G-protein coupled receptor 21 -1.02 0.03 FNDC7 Fibronectin type III domain-containing protein 7 -1.00 0.04 IGKV6-21 Immunoglobulin kappa variable 6-21 -1.00 0.04

210 CSF2RA Granulocyte-macrophage colony-stimulating factor receptor subunit alpha -1.00 0.04 IIGP5 Interferon-inducible GTPase 5 -1.00 0.02 FCN1 Ficolin-1 -1.00 0.04 P2RY12 P2Y purinoceptor 12 -0.99 0.04 Nptn Neuroplastin -0.98 0.04 . . -0.97 0.03 ARHGEF37 Rho guanine nucleotide exchange factor 37 -0.97 0.04 . . -0.97 0.04 Esm1 Endothelial cell-specific molecule 1 -0.96 0.03 Gp5 Platelet glycoprotein V -0.95 0.04 CYP8B1 5-beta-cholestane-3-alpha,7-alpha-diol 12-alpha-hydroxylase -0.95 0.04 Leucine-rich repeat and immunoglobulin-like domain-containing nogo receptor-interacting lingo1 protein 1 -0.95 0.04 PXN1 Jeltraxin -0.93 0.04 . . -0.91 0.04 NUCB2 Nucleobindin-2 -0.90 0.02 NXPE3 NXPE family member 3 -0.90 0.03 . . -0.87 0.04 . . -0.81 0.05 HLA-DRB1 HLA class II histocompatibility antigen, DRB1-4 beta chain -0.80 0.04 . . -0.74 0.04

Table AI3. Intermediate (NT) up-regulation Gene Protein effect p prkra-a Interferon-inducible double-stranded RNA-dependent protein kinase activator A homolog A 2.10 0.00 ISG20L2 Interferon-stimulated 20 kDa exonuclease-like 2 2.03 0.00 . . 1.96 0.00 . . 1.94 0.00 DNAJA2 DnaJ homolog subfamily A member 2 1.87 0.00

211 Dihydrolipoyllysine-residue succinyltransferase component of 2-oxoglutarate dehydrogenase DLST complex, mitochondrial 1.87 0.00 . . 1.83 0.00 . . 1.81 0.00 gsn Gelsolin 1.80 0.00 . . 1.73 0.00 Serbp1 Plasminogen activator inhibitor 1 RNA-binding protein 1.73 0.00 rps8 40S ribosomal protein S8 1.72 0.00 Fip1l1 Pre-mRNA 3'-end-processing factor FIP1 1.71 0.00 SWI/SNF-related matrix-associated actin-dependent regulator of chromatin subfamily E member Smarce1 1 1.70 0.00 TPT1 Translationally-controlled tumor protein homolog 1.67 0.00 tram1l1 . 1.65 0.00 rps2 40S ribosomal protein S2 1.64 0.00 tmem251 Transmembrane protein 251 1.64 0.00 Pnrc1 Proline-rich nuclear receptor coactivator 1 1.64 0.00 MRPL35 39S ribosomal protein L35, mitochondrial 1.64 0.00 KCTD2 BTB/POZ domain-containing protein KCTD2 1.63 0.00 . . 1.63 0.01 . . 1.55 0.01 GLTSCR2 Glioma tumor suppressor candidate region gene 2 protein 1.54 0.00 B4galt3 Beta-1,4-galactosyltransferase 3 1.54 0.00 TMED10 Transmembrane emp24 domain-containing protein 10 1.53 0.00 ppp4c Serine/threonine-protein phosphatase 4 catalytic subunit 1.53 0.00 . . 1.51 0.00 PGM1 Phosphoglucomutase-1 1.51 0.00 Csnk2a2 Casein kinase II subunit alpha' 1.51 0.00 EEF1A Elongation factor 1-alpha 1 1.50 0.00 . . 1.50 0.00 PSMC3 26S protease regulatory subunit 6A 1.49 0.00 Pafah2 Platelet-activating factor acetylhydrolase 2, cytoplasmic 1.49 0.01

212 . . 1.48 0.00 ITM2A Integral membrane protein 2A 1.48 0.00 rpl7a 60S ribosomal protein L7a 1.47 0.00 EEF2 Elongation factor 2 1.47 0.00 METTL9 Methyltransferase-like protein 9 1.47 0.01 . Heterogeneous nuclear ribonucleoprotein A3 homolog 1 1.46 0.00 MCUB Calcium uniporter regulatory subunit MCUb, mitochondrial 1.46 0.01 PSMB5 Proteasome subunit beta type-5 1.45 0.01 pkmyt1 Membrane-associated tyrosine- and threonine-specific cdc2-inhibitory kinase 1.45 0.00 smyd5 SET and MYND domain-containing protein 5 1.44 0.00 . . 1.42 0.01 FBXL3 F-box/LRR-repeat protein 3 1.42 0.01 . . 1.42 0.01 eif3g-b Eukaryotic translation initiation factor 3 subunit G-B 1.42 0.00 . . 1.42 0.01 eef1g-a Elongation factor 1-gamma-A 1.41 0.00 . Oocyte zinc finger protein XlCOF7.1 1.39 0.01 . . 1.38 0.01 eif3h Eukaryotic translation initiation factor 3 subunit H 1.37 0.00 med8-b Mediator of RNA polymerase II transcription subunit 8-B 1.36 0.00 ZNF567 Zinc finger protein 567 1.36 0.00 . . 1.35 0.01 Rack1 Receptor of activated protein C kinase 1 1.35 0.00 ispd Isoprenoid synthase domain-containing protein 1.35 0.01 . . 1.34 0.00 NDUFV2 NADH dehydrogenase [ubiquinone] flavoprotein 2, mitochondrial 1.34 0.01 . . 1.34 0.00 NKAP NF-kappa-B-activating protein 1.34 0.00 . . 1.34 0.01 . Gastrula zinc finger protein XlCGF17.1 1.33 0.01 . . 1.33 0.00

213 PMVK Phosphomevalonate kinase 1.33 0.01 RABEP2 Rab GTPase-binding effector protein 2 1.33 0.02 Gpbp1 Vasculin 1.33 0.01 Xrcc6 X-ray repair cross-complementing protein 6 1.33 0.01 Atp5o ATP synthase subunit O, mitochondrial 1.33 0.01 . . 1.31 0.01 Rps27 40S ribosomal protein S27 1.31 0.00 Rps3 40S ribosomal protein S3 1.30 0.00 Mrps10 28S ribosomal protein S10, mitochondrial 1.30 0.01 Eif2b4 Translation initiation factor eIF-2B subunit delta 1.30 0.02 ART2 Putative uncharacterized protein ART2 1.29 0.00 rps3a 40S ribosomal protein S3a 1.29 0.01 . Tubulin alpha chain 1.29 0.01 Gabarap Gamma-aminobutyric acid receptor-associated protein 1.28 0.01 get4 Golgi to ER traffic protein 4 homolog 1.28 0.01 . . 1.28 0.01 prr5 Proline-rich protein 5 1.28 0.01 rsph9 Radial spoke head protein 9 homolog 1.28 0.01 Rps19 40S ribosomal protein S19 1.27 0.00 eif3a Eukaryotic translation initiation factor 3 subunit A 1.27 0.01 . . 1.27 0.01 . . 1.27 0.00 Eef1d Elongation factor 1-delta 1.27 0.01 rps23 40S ribosomal protein S23 1.27 0.01 nae1 NEDD8-activating enzyme E1 regulatory subunit 1.27 0.01 IFITM10 Interferon-induced transmembrane protein 10 1.27 0.01 . . 1.26 0.01 ART2 Putative uncharacterized protein ART2 1.26 0.00 PIGF Phosphatidylinositol-glycan biosynthesis class F protein 1.26 0.01 PDE12 2',5'-phosphodiesterase 12 1.26 0.03 . . 1.25 0.02

214 . . 1.25 0.02 GORASP2 Golgi reassembly-stacking protein 2 1.25 0.01 . . 1.25 0.02 Cox11 Cytochrome c oxidase assembly protein COX11, mitochondrial 1.25 0.01 . . 1.24 0.02 TCP1 T-complex protein 1 subunit alpha 1.24 0.01 ART2 Putative uncharacterized protein ART2 1.24 0.00 . . 1.23 0.01 FAM47C Putative protein FAM47C 1.23 0.00 Pol LINE-1 retrotransposable element ORF2 protein 1.23 0.00 S100A11 Protein S100-A11 1.22 0.02 rpl4-b 60S ribosomal protein L4-B 1.22 0.01 ZNF300 Zinc finger protein 300 1.22 0.02 pssA CDP-diacylglycerol--serine O-phosphatidyltransferase 1.21 0.01 . . 1.21 0.01 ZNF300 Zinc finger protein 300 1.21 0.01 Glo1 Lactoylglutathione lyase 1.21 0.01 . . 1.21 0.02 . . 1.20 0.02 SLC24A2 Sodium/potassium/calcium exchanger 2 1.20 0.01 yipf3 Protein YIPF3 1.20 0.01 . . 1.19 0.02 Nlrc3 Protein NLRC3 1.18 0.00 MTMR9 Myotubularin-related protein 9 1.18 0.01 Eloc Elongin-C 1.18 0.03 ndor1 NADPH-dependent diflavin oxidoreductase 1 1.18 0.01 . . 1.18 0.02 ERI3 ERI1 exoribonuclease 3 1.18 0.02 pno1 RNA-binding protein PNO1 1.17 0.01 TXNDC5 Thioredoxin domain-containing protein 5 1.17 0.02 RPL3 60S ribosomal protein L3 1.17 0.01

215 . . 1.17 0.03 Agr3 Anterior gradient protein 3 1.17 0.03 Borcs6 BLOC-1-related complex subunit 6 1.17 0.01 NDUFS8 NADH dehydrogenase [ubiquinone] iron-sulfur protein 8, mitochondrial 1.17 0.02 . . 1.17 0.00 NAXD ATP-dependent (S)-NAD(P)H-hydrate dehydratase 1.16 0.01 rnf2-a E3 ubiquitin-protein ligase RING2-A 1.16 0.01 med9 Mediator of RNA polymerase II transcription subunit 9 1.16 0.02 . . 1.16 0.02 ELMO2 Engulfment and cell motility protein 2 1.16 0.02 . Nuclear factor 7, ovary 1.16 0.02 . . 1.16 0.02 ATF4 Cyclic AMP-dependent transcription factor ATF-4 1.16 0.03 Vamp2 Vesicle-associated membrane protein 2 1.15 0.02 HSP90AB1 Heat shock cognate protein HSP 90-beta 1.15 0.01 CCT6 T-complex protein 1 subunit zeta 1.15 0.01 eif3d Eukaryotic translation initiation factor 3 subunit D 1.15 0.01 . . 1.15 0.02 . . 1.14 0.02 . . 1.14 0.03 LCMT1 Leucine carboxyl methyltransferase 1 1.14 0.01 . . 1.14 0.02 . . 1.14 0.03 ybx1 Nuclease-sensitive element-binding protein 1 1.14 0.01 Gtpbp4 Nucleolar GTP-binding protein 1 1.14 0.01 . . 1.14 0.02 ciapin1 Anamorsin 1.14 0.02 . . 1.14 0.02 . . 1.13 0.02 pi4kb Phosphatidylinositol 4-kinase beta 1.13 0.02 . . 1.12 0.03

216 ZNF268 Zinc finger protein 268 1.12 0.03 Pkd1l3 Polycystic kidney disease protein 1-like 3 1.12 0.02 ORF V Enzymatic polyprotein 1.12 0.02 Rpl17 60S ribosomal protein L17 1.12 0.01 fbxl5 F-box/LRR-repeat protein 5 1.12 0.01 rps15 40S ribosomal protein S15 1.12 0.02 . . 1.11 0.01 RPL10 60S ribosomal protein L10 1.11 0.01 Tmed8 Protein TMED8 1.11 0.03 eif3b Eukaryotic translation initiation factor 3 subunit B 1.11 0.01 ACADVL Very long-chain specific acyl-CoA dehydrogenase, mitochondrial 1.11 0.02 . . 1.11 0.02 . . 1.11 0.02 sde2 Protein SDE2 homolog 1.10 0.02 LRRC41 Leucine-rich repeat-containing protein 41 1.10 0.03 QPCTL Glutaminyl-peptide cyclotransferase-like protein 1.10 0.02 GARS Glycine--tRNA ligase 1.10 0.05 Rpl37 60S ribosomal protein L37 1.10 0.02 aad-a Alpha-aspartyl dipeptidase 1.10 0.02 Rpl11 60S ribosomal protein L11 1.09 0.02 RWDD1 RWD domain-containing protein 1 1.09 0.01 eif3l Eukaryotic translation initiation factor 3 subunit L 1.09 0.01 . . 1.09 0.03 rpl18-b 60S ribosomal protein L18-B 1.09 0.02 Psmc5 26S protease regulatory subunit 8 1.09 0.04 rpl18a 60S ribosomal protein L18a 1.09 0.02 . . 1.09 0.03 . . 1.09 0.04 CLPP ATP-dependent Clp protease proteolytic subunit, mitochondrial 1.09 0.03 Rpl23 60S ribosomal protein L23 1.08 0.02 Tuba1b Tubulin alpha-1B chain 1.08 0.03

217 ccdc93 Coiled-coil domain-containing protein 93 1.08 0.02 . . 1.08 0.03 ZNF41 Zinc finger protein 41 1.08 0.04 PSMC4 26S protease regulatory subunit 6B 1.08 0.01 RSPO3 R-spondin-3 1.08 0.02 STARD6 StAR-related lipid transfer protein 6 1.08 0.02 . . 1.08 0.04 . . 1.08 0.03 TOMM40L Mitochondrial import receptor subunit TOM40B 1.07 0.03 . . 1.07 0.03 PCNP PEST proteolytic signal-containing nuclear protein 1.07 0.03 RPL13A 60S ribosomal protein L13a 1.07 0.02 MRPL2 39S ribosomal protein L2, mitochondrial 1.07 0.03 PFDN6 Prefoldin subunit 6 1.07 0.02 . . 1.07 0.03 Dph5 Diphthine methyl ester synthase 1.07 0.02 UBQLN4 Ubiquilin-4 1.07 0.02 . Nucleoside diphosphate kinase A1 1.07 0.01 ARPC1A Actin-related protein 2/3 complex subunit 1A 1.07 0.04 CTDSP2 Carboxy-terminal domain RNA polymerase II polypeptide A small phosphatase 2 1.06 0.03 infB Translation initiation factor IF-2 1.06 0.03 RpL13 60S ribosomal protein L13 1.06 0.02 Abra Actin-binding Rho-activating protein 1.06 0.02 . . 1.06 0.03 ddost Dolichyl-diphosphooligosaccharide--protein glycosyltransferase 48 kDa subunit 1.06 0.01 . . 1.06 0.04 CD200R1B Cell surface glycoprotein CD200 receptor 1-B 1.06 0.04 EIF2B3 Translation initiation factor eIF-2B subunit gamma 1.05 0.02 tmem147 Transmembrane protein 147 1.05 0.03 svbp Small vasohibin-binding protein 1.05 0.03 . . 1.05 0.02

218 PLEK Pleckstrin 1.05 0.04 . . 1.05 0.03 . . 1.05 0.04 aida-b Axin interactor, dorsalization-associated protein B 1.05 0.05 IMMT MICOS complex subunit MIC60 1.05 0.04 TBXAS1 Thromboxane-A synthase 1.05 0.03 . Gastrula zinc finger protein XlCGF57.1 1.04 0.02 EIF3F Eukaryotic translation initiation factor 3 subunit F 1.04 0.04 . . 1.04 0.04 alkbh5 RNA demethylase ALKBH5 1.04 0.01 NDUFB9 NADH dehydrogenase [ubiquinone] 1 beta subcomplex subunit 9 1.04 0.02 PYCARD Apoptosis-associated speck-like protein containing a CARD 1.04 0.04 Ccdc58 Coiled-coil domain-containing protein 58 1.04 0.04 Rpl12 60S ribosomal protein L12 1.04 0.02 Rpl17 60S ribosomal protein L17 1.03 0.03 TAR1 Protein TAR1 1.03 0.05 . . 1.03 0.03 ak2 Adenylate kinase 2, mitochondrial 1.03 0.04 . Oocyte zinc finger protein XlCOF7.1 1.03 0.03 SLC35A1 CMP-sialic acid transporter 1.03 0.02 . Gastrula zinc finger protein XlCGF8.2DB 1.03 0.04 PSMD4 26S proteasome non-ATPase regulatory subunit 4 1.03 0.04 coil Coilin 1.03 0.04 ZNF84 Zinc finger protein 84 1.03 0.04 . Gastrula zinc finger protein XlCGF67.1 1.03 0.03 . . 1.02 0.04 . . 1.02 0.03 Stt3a Dolichyl-diphosphooligosaccharide--protein glycosyltransferase subunit STT3A 1.02 0.02 Tmem115 Transmembrane protein 115 1.02 0.04 MOSPD1 Motile sperm domain-containing protein 1 1.02 0.03 pgap2 Post-GPI attachment to proteins factor 2 1.02 0.03

219 BEND4 BEN domain-containing protein 4 1.02 0.04 hexim Protein HEXIM 1.02 0.03 impdh2 Inosine-5'-monophosphate dehydrogenase 2 1.02 0.03 HAX1 HCLS1-associated protein X-1 1.02 0.02 . . 1.02 0.02 NGB Neuroglobin 1.02 0.03 . . 1.02 0.05 ddx21-b Nucleolar RNA helicase 2-B 1.02 0.02 Paics Multifunctional protein ADE2 1.02 0.05 exosc6 Exosome complex component MTR3 1.02 0.03 GPR21 Probable G-protein coupled receptor 21 1.01 0.03 . . 1.01 0.04 ecsit Evolutionarily conserved signaling intermediate in Toll pathway, mitochondrial 1.01 0.04 . . 1.01 0.03 . . 1.01 0.04 Psmg3 Proteasome assembly chaperone 3 1.00 0.03 . . 1.00 0.05 ZNF8 Zinc finger protein 8 1.00 0.03 rps4 40S ribosomal protein S4 1.00 0.03 Terb2 Telomere repeats-binding bouquet formation protein 2 1.00 0.04 rpl7a 60S ribosomal protein L7a 1.00 0.02 TXNL1 Thioredoxin-like protein 1 1.00 0.03 . . 1.00 0.05 CD2 T-cell surface antigen CD2 1.00 0.04 sh3bp5l SH3 domain-binding protein 5-like 1.00 0.03 . . 1.00 0.03 EIF2S3 Eukaryotic translation initiation factor 2 subunit 3 0.99 0.02 DHCR24 Delta(24)-sterol reductase 0.99 0.03 PLCL1 Inactive phospholipase C-like protein 1 0.99 0.05 LRSAM1 E3 ubiquitin-protein ligase LRSAM1 0.99 0.02 . . 0.99 0.04

220 . . 0.99 0.03 CIR1 Corepressor interacting with RBPJ 1 0.99 0.03 Acads Short-chain specific acyl-CoA dehydrogenase, mitochondrial 0.99 0.04 ZNF484 Zinc finger protein 484 0.99 0.04 . . 0.98 0.03 RPL35 60S ribosomal protein L35 0.98 0.05 . . 0.98 0.02 TTLL9 Probable tubulin polyglutamylase TTLL9 0.98 0.05 ASB6 Ankyrin repeat and SOCS box protein 6 0.98 0.04 SLC27A3 Long-chain fatty acid transport protein 3 0.98 0.02 glmp-a Glycosylated lysosomal membrane protein A 0.98 0.04 . . 0.98 0.03 naxe NAD(P)H-hydrate epimerase 0.98 0.05 MIEN1 Migration and invasion enhancer 1 0.98 0.04 EIF5A2 Eukaryotic translation initiation factor 5A-2 0.97 0.04 RAB24 Ras-related protein Rab-24 0.97 0.03 eif2a Eukaryotic translation initiation factor 2A 0.97 0.03 . . 0.97 0.04 nudt19 Nucleoside diphosphate-linked moiety X motif 19 0.97 0.04 SLC5A3 Sodium/myo-inositol cotransporter 0.97 0.03 PHF1 PHD finger protein 1 0.97 0.04 CCT4 T-complex protein 1 subunit delta 0.97 0.04 nmd3 60S ribosomal export protein NMD3 0.97 0.04 ZNF586 Zinc finger protein 586 0.97 0.04 eif3k Eukaryotic translation initiation factor 3 subunit K 0.97 0.04 Tbl1x F-box-like/WD repeat-containing protein TBL1X 0.97 0.04 PNPO Pyridoxine-5'-phosphate oxidase 0.97 0.04 Atp5i ATP synthase subunit e, mitochondrial 0.97 0.03 . . 0.96 0.03 ostc Oligosaccharyltransferase complex subunit ostc 0.96 0.03 . . 0.96 0.05

221 Ier5 Immediate early response gene 5 protein 0.96 0.04 POLM DNA-directed DNA/RNA polymerase mu 0.96 0.03 . . 0.96 0.04 MOCS2 Molybdopterin synthase catalytic subunit 0.96 0.04 . . 0.96 0.03 . . 0.96 0.05 . . 0.96 0.03 fam32a Protein FAM32A 0.96 0.04 CCDC13 Coiled-coil domain-containing protein 13 0.96 0.03 ORC3 Origin recognition complex subunit 3 0.96 0.03 MVK Mevalonate kinase 0.96 0.04 . . 0.95 0.04 Fxyd1 Phospholemman 0.95 0.04 nubp1-A Cytosolic Fe-S cluster assembly factor nubp1-A 0.95 0.02 PRKACA cAMP-dependent protein kinase catalytic subunit alpha 0.95 0.03 Eef1a2 Elongation factor 1-alpha 2 0.95 0.02 . . 0.95 0.03 hacd4 Very-long-chain (3R)-3-hydroxyacyl-CoA dehydratase 0.95 0.04 . . 0.95 0.04 pxn1 Pentraxin fusion protein 0.95 0.05 Kars Lysine--tRNA ligase 0.95 0.04 Polr2c DNA-directed RNA polymerase II subunit RPB3 0.94 0.04 Was Wiskott-Aldrich syndrome protein homolog 0.94 0.02 Dnajc4 DnaJ homolog subfamily C member 4 0.94 0.04 . . 0.94 0.03 CRLS1 Cardiolipin synthase (CMP-forming) 0.94 0.04 GFM2 Ribosome-releasing factor 2, mitochondrial 0.94 0.05 . . 0.94 0.04 RNF40 E3 ubiquitin-protein ligase BRE1B 0.94 0.04 ZNF300 Zinc finger protein 300 0.94 0.04 RPLP0 60S acidic ribosomal protein P0 0.94 0.05

222 ZCHC3 Zinc finger CCHC domain-containing protein 3 0.94 0.05 . . 0.94 0.04 Ssr4 Translocon-associated protein subunit delta 0.94 0.04 . Cytochrome c oxidase subunit 4 isoform 2, mitochondrial 0.94 0.04 Slc35b2 Adenosine 3'-phospho 5'-phosphosulfate transporter 1 0.93 0.04 RPS14 40S ribosomal protein S14 0.93 0.03 THAP5 THAP domain-containing protein 5 0.93 0.04 MPL Thrombopoietin receptor 0.93 0.04 ELOVL1 Elongation of very long chain fatty acids protein 1 0.93 0.04 Nlrc3 Protein NLRC3 0.93 0.01 . . 0.93 0.03 ZNF300 Zinc finger protein 300 0.92 0.04 rpl8 60S ribosomal protein L8 0.92 0.03 Fhl1 Four and a half LIM domains protein 1 0.92 0.03 VIPAS39 Spermatogenesis-defective protein 39 homolog 0.92 0.04 rps16 40S ribosomal protein S16 0.92 0.04 eif3c Eukaryotic translation initiation factor 3 subunit C 0.92 0.02 Rpl32 60S ribosomal protein L32 0.92 0.03 ANXA1 Annexin A1 0.92 0.03 . . 0.92 0.05 . . 0.91 0.04 Rps9 40S ribosomal protein S9 0.91 0.05 NSD3 Histone-lysine N-methyltransferase NSD3 0.91 0.04 mvp Major vault protein 0.91 0.02 SUGP1 SURP and G-patch domain-containing protein 1 0.91 0.04 . . 0.91 0.05 RAB10 Ras-related protein Rab-10 0.91 0.03 . Gastrula zinc finger protein XlCGF57.1 0.90 0.04 AL3B1 Aldehyde dehydrogenase family 3 member B1 0.90 0.05 RPS5 40S ribosomal protein S5 0.90 0.03 . . 0.90 0.04

223 . . 0.90 0.05 UBXN2A UBX domain-containing protein 2A 0.90 0.04 cct3 T-complex protein 1 subunit gamma 0.90 0.04 Y325 Uncharacterized oxidoreductase TM_0325 0.89 0.05 pld3 Phospholipase D3 0.89 0.05 . . 0.89 0.05 p33monox Putative monooxygenase p33MONOX 0.89 0.04 Rpl27 60S ribosomal protein L27 0.89 0.04 SEC22A Vesicle-trafficking protein SEC22a 0.89 0.04 TESK2 Dual specificity testis-specific protein kinase 2 0.89 0.04 CPA3 Mast cell carboxypeptidase A 0.87 0.05 gkap1 G kinase-anchoring protein 1 0.87 0.04 . . 0.86 0.04 PDCL Phosducin-like protein 0.86 0.05 SLC25A3 Phosphate carrier protein, mitochondrial 0.86 0.04 Imp3 U3 small nucleolar ribonucleoprotein protein IMP3 0.85 0.04 spice1 Spindle and centriole-associated protein 1 0.85 0.04 ABHD16A Protein ABHD16A 0.85 0.03 . . 0.85 0.05 emc3 ER membrane protein complex subunit 3 0.85 0.03 DCTN3 Dynactin subunit 3 0.85 0.05 . . 0.85 0.04 RNF112 RING finger protein 112 0.85 0.04 . . 0.83 0.04

Table AI4. Intermediate (NT) down-regulation Gene Protein effect p . . -7.56 0.00 . . -5.67 0.00 YCX91 Uncharacterized protein ORF91 -4.74 0.00 . . -4.39 0.00

224 YCX91 Uncharacterized protein ORF91 -4.16 0.00 . . -3.99 0.00 . . -3.67 0.00 YCX91 Uncharacterized protein ORF91 -3.38 0.00 . . -3.16 0.00 . . -3.11 0.00 POL Gag-Pol polyprotein -3.06 0.00 . . -2.99 0.00 . . -2.87 0.00 . . -2.55 0.00 TRMD tRNA (guanine-N(1)-)-methyltransferase -2.53 0.00 RPOA DNA-directed RNA polymerase subunit alpha -2.52 0.00 . . -2.36 0.00 RL2 50S ribosomal protein L2 -2.03 0.00 YCX91 Uncharacterized protein ORF91 -1.91 0.00 MAP3K2 Mitogen-activated protein kinase kinase kinase 2 -1.88 0.00 Tbc1d9b TBC1 domain family member 9B -1.88 0.00 SECY Protein translocase subunit SecY -1.87 0.00 SEPT8 Septin-8 -1.84 0.00 Sbno1 Protein strawberry notch homolog 1 -1.81 0.00 PARVA Alpha-parvin -1.67 0.01 MTHFD1L Monofunctional C1-tetrahydrofolate synthase, mitochondrial -1.66 0.00 Larp1 La-related protein 1 -1.64 0.00 ZBTB10 Zinc finger and BTB domain-containing protein 10 -1.63 0.00 SPTBN1 Spectrin beta chain, non-erythrocytic 1 -1.63 0.00 Myo1c Unconventional myosin-Ic -1.58 0.00 CHD6 Chromodomain-helicase-DNA-binding protein 6 -1.58 0.00 CSF2RB Cytokine receptor common subunit beta -1.51 0.00 HTT Huntingtin -1.50 0.00 Ppp6c Serine/threonine-protein phosphatase 6 catalytic subunit -1.50 0.00 B4GALT1 Beta-1,4-galactosyltransferase 1 -1.50 0.00

225 VPS13A Vacuolar protein sorting-associated protein 13A -1.48 0.00 SRCAP Helicase SRCAP -1.48 0.00 SBNO2 Protein strawberry notch homolog 2 -1.47 0.00 Itch E3 ubiquitin-protein ligase Itchy -1.46 0.00 jmjd6-b Bifunctional arginine demethylase and lysyl-hydroxylase JMJD6-B -1.46 0.02 nt5c2 Cytosolic purine 5'-nucleotidase -1.45 0.00 INPP5K Inositol polyphosphate 5-phosphatase K -1.44 0.00 DOCK1 Dedicator of cytokinesis protein 1 -1.43 0.00 MYO18A Unconventional myosin-XVIIIa -1.43 0.00 PEX14 Peroxisomal membrane protein PEX14 -1.42 0.01 . . -1.41 0.01 KRAS GTPase KRas -1.39 0.01 ANO5 Anoctamin-5 -1.38 0.00 Pi4ka Phosphatidylinositol 4-kinase alpha -1.38 0.01 HECTD4 Probable E3 ubiquitin-protein ligase HECTD4 -1.38 0.00 SPAG9 C-Jun-amino-terminal kinase-interacting protein 4 -1.38 0.01 PBXIP1 Pre-B-cell leukemia transcription factor-interacting protein 1 -1.37 0.00 TNRC6C Trinucleotide repeat-containing gene 6C protein -1.37 0.00 ywhag-a 14-3-3 protein gamma-A -1.37 0.00 Rps6ka3 Ribosomal protein S6 kinase alpha-3 -1.37 0.00 TNFAIP2 Tumor necrosis factor alpha-induced protein 2 -1.36 0.01 Meioc Meiosis-specific coiled-coil domain-containing protein MEIOC -1.36 0.01 paxip1 PAX-interacting protein 1 -1.36 0.01 PCNX1 Pecanex-like protein 1 -1.36 0.01 Xpo1 Exportin-1 -1.36 0.01 PPP4R1 Serine/threonine-protein phosphatase 4 regulatory subunit 1 -1.36 0.01 Pak2 Serine/threonine-protein kinase PAK 2 -1.35 0.00 HNRNPLL Heterogeneous nuclear ribonucleoprotein L-like -1.35 0.01 tmem168 Transmembrane protein 168 -1.34 0.01 EP300 Histone acetyltransferase p300 -1.34 0.00 ASXL2 Putative Polycomb group protein ASXL2 -1.34 0.01

226 KIF21A Kinesin-like protein KIF21A -1.34 0.01 KANK1 KN motif and ankyrin repeat domain-containing protein 1 -1.34 0.01 srf Serum response factor -1.34 0.01 FERMT2 Fermitin family homolog 2 -1.33 0.00 pum1 Pumilio homolog 1 -1.33 0.00 Prrc2c Protein PRRC2C -1.32 0.00 TRIM39 E3 ubiquitin-protein ligase TRIM39 -1.32 0.01 slc25a37 Mitoferrin-1 -1.32 0.01 PDS5A Sister chromatid cohesion protein PDS5 homolog A -1.32 0.01 ZNF516 Zinc finger protein 516 -1.32 0.01 PIP4K2B Phosphatidylinositol 5-phosphate 4-kinase type-2 beta -1.31 0.01 . . -1.31 0.00 SDCBP Syntenin-1 -1.31 0.01 DPP8 Dipeptidyl peptidase 8 -1.31 0.00 MPRIP Myosin phosphatase Rho-interacting protein -1.30 0.01 DMD Dystrophin -1.30 0.01 PPP1R12A Protein phosphatase 1 regulatory subunit 12A -1.30 0.01 SETD5 SET domain-containing protein 5 -1.30 0.01 qki-b Protein quaking-B -1.29 0.01 BICD2 Protein bicaudal D homolog 2 -1.29 0.01 timp3 Metalloproteinase inhibitor 3 -1.29 0.00 Foxp1 Forkhead box protein P1 -1.29 0.01 SLC7A3 Cationic transporter 3 -1.29 0.01 ZNF148 Zinc finger protein 148 -1.28 0.01 ADAMTS1 A disintegrin and metalloproteinase with thrombospondin motifs 1 -1.28 0.01 qsox2 Sulfhydryl oxidase 2 -1.28 0.01 NAA25 N-alpha-acetyltransferase 25, NatB auxiliary subunit -1.28 0.01 Nckipsd NCK-interacting protein with SH3 domain -1.28 0.01 FLNB Filamin-B -1.27 0.01 Smad6 Mothers against decapentaplegic homolog 6 -1.27 0.01 PPIP5K2 Inositol hexakisphosphate and diphosphoinositol-pentakisphosphate kinase 2 -1.27 0.01

227 SMARCD2 SWI/SNF-related matrix-associated actin-dependent regulator of chromatin subfamily D member 2 -1.27 0.01 Tcf12 Transcription factor 12 -1.26 0.01 COL4A1 Collagen alpha-1(IV) chain -1.26 0.01 MAP7 Ensconsin -1.26 0.01 Erc1 ELKS/Rab6-interacting/CAST family member 1 -1.26 0.01 ZFHX2 Zinc finger homeobox protein 2 -1.25 0.01 pnrc2-a Proline-rich nuclear receptor coactivator 2 A -1.25 0.01 Ash2l Set1/Ash2 histone methyltransferase complex subunit ASH2 -1.25 0.02 AFAP1 Actin filament-associated protein 1 -1.25 0.01 ZFHX3 Zinc finger homeobox protein 3 -1.25 0.01 MAST4 Microtubule-associated serine/threonine-protein kinase 4 -1.25 0.01 RAB5B Ras-related protein Rab-5B -1.25 0.01 MICU2 Calcium uptake protein 2, mitochondrial -1.25 0.01 Kat6a Histone acetyltransferase KAT6A -1.25 0.01 DAGLB Sn1-specific diacylglycerol lipase beta -1.24 0.01 mrtfb Myocardin-related transcription factor B -1.24 0.01 FAM46A Protein FAM46A -1.24 0.02 slc43a2 Large neutral amino acids transporter small subunit 4 -1.24 0.01 TAF2 Transcription initiation factor TFIID subunit 2 -1.24 0.01 ATP11C Phospholipid-transporting ATPase IG -1.23 0.01 . . -1.23 0.01 HSPA4 Heat shock 70 kDa protein 4 -1.23 0.01 HERC2 E3 ubiquitin-protein ligase HERC2 -1.23 0.01 Slamf7 SLAM family member 7 -1.23 0.01 ATF3 Cyclic AMP-dependent transcription factor ATF-3 -1.22 0.01 Lbr Lamin-B receptor -1.22 0.01 NCR3LG1 Natural cytotoxicity triggering receptor 3 ligand 1 -1.22 0.01 TNRC18 Trinucleotide repeat-containing gene 18 protein -1.22 0.02 cmip C-Maf-inducing protein -1.22 0.02 NFAT5 Nuclear factor of activated T-cells 5 -1.22 0.01 pafah1b1 Lissencephaly-1 homolog -1.21 0.01

228 Rxra Retinoic acid receptor RXR-alpha -1.21 0.01 RUBCN Run domain Beclin-1-interacting and cysteine-rich domain-containing protein -1.21 0.01 ABCC1 Multidrug resistance-associated protein 1 -1.21 0.02 IKBKE Inhibitor of nuclear factor kappa-B kinase subunit epsilon -1.21 0.00 ANKRD17 Ankyrin repeat domain-containing protein 17 -1.21 0.01 Pcnt Pericentrin -1.21 0.01 WWTR1 WW domain-containing transcription regulator protein 1 -1.21 0.02 PLCG1 1-phosphatidylinositol 4,5-bisphosphate phosphodiesterase gamma-1 -1.20 0.01 topbp1-A DNA topoisomerase 2-binding protein 1-A -1.20 0.02 EMSY BRCA2-interacting transcriptional repressor EMSY -1.19 0.01 KIF13B Kinesin-like protein KIF13B -1.18 0.01 RIPK3 Receptor-interacting serine/threonine-protein kinase 3 -1.18 0.01 NFATC2 Nuclear factor of activated T-cells, cytoplasmic 2 -1.18 0.01 WDR11 WD repeat-containing protein 11 -1.18 0.01 MARK2 Serine/threonine-protein kinase MARK2 -1.18 0.02 HERC1 Probable E3 ubiquitin-protein ligase HERC1 -1.18 0.01 ZBTB40 Zinc finger and BTB domain-containing protein 40 -1.17 0.02 BLVRA Biliverdin reductase A -1.17 0.03 mical3a Protein-methionine sulfoxide oxidase mical3a -1.17 0.01 NPR3 Atrial natriuretic peptide receptor 3 -1.17 0.02 dopey2 Protein dopey-2 -1.17 0.01 pum2 Pumilio homolog 2 -1.16 0.01 RALGAPA1 Ral GTPase-activating protein subunit alpha-1 -1.16 0.02 BNIP2 BCL2/adenovirus E1B 19 kDa protein-interacting protein 2 -1.16 0.01 SPEN Msx2-interacting protein -1.16 0.01 prkcb Protein kinase C beta type -1.16 0.01 S1pr1 Sphingosine 1-phosphate receptor 1 -1.16 0.01 CAMK2G Calcium/calmodulin-dependent protein kinase type II subunit gamma -1.16 0.01 Tnfrsf21 Tumor necrosis factor receptor superfamily member 21 -1.16 0.02 FV3-043R Uncharacterized protein 043R -1.16 0.01 GRAMD1A GRAM domain-containing protein 1A -1.16 0.02

229 ABI1 Abl interactor 1 -1.16 0.01 RREB1 Ras-responsive element-binding protein 1 -1.15 0.02 KIAA0232 Uncharacterized protein KIAA0232 -1.15 0.01 Thbd Thrombomodulin -1.15 0.01 CROT Peroxisomal carnitine O-octanoyltransferase -1.15 0.02 BCAT1 Branched-chain-amino-acid aminotransferase, cytosolic -1.15 0.01 coq10b-a Coenzyme Q-binding protein COQ10 homolog A, mitochondrial -1.15 0.03 AKNA AT-hook-containing transcription factor -1.15 0.02 DGKZ Diacylglycerol kinase zeta -1.14 0.02 . . -1.14 0.01 USP34 Ubiquitin carboxyl-terminal hydrolase 34 -1.14 0.03 TRPC4AP Short transient receptor potential channel 4-associated protein -1.14 0.02 rnf44 RING finger protein 44 -1.13 0.03 PYCARD Apoptosis-associated speck-like protein containing a CARD -1.13 0.03 IDE Insulin-degrading enzyme -1.13 0.02 Pigq Phosphatidylinositol N-acetylglucosaminyltransferase subunit Q -1.13 0.03 EEF2K Eukaryotic elongation factor 2 kinase -1.13 0.01 UBR1 E3 ubiquitin-protein ligase UBR1 -1.13 0.02 mlec-b Malectin-B -1.13 0.01 tshz1-b Teashirt homolog 1-B -1.13 0.02 RIC1 RAB6A-GEF complex partner protein 1 -1.12 0.02 Higd1c HIG1 domain family member 1C -1.12 0.01 ZEB1 Zinc finger E-box-binding homeobox 1 -1.12 0.01 NCR3LG1 Natural cytotoxicity triggering receptor 3 ligand 1 -1.12 0.02 ccdc88c Daple-like protein -1.11 0.01 ACACB Acetyl-CoA carboxylase 2 -1.11 0.02 Stard13 StAR-related lipid transfer protein 13 -1.11 0.02 itga5 Integrin alpha-5 -1.11 0.02 TRIM24 Transcription intermediary factor 1-alpha -1.11 0.01 VPS13D Vacuolar protein sorting-associated protein 13D -1.11 0.02 Ahcyl1 S-adenosylhomocysteine hydrolase-like protein 1 -1.11 0.02

230 cwf19l1 CWF19-like protein 1 -1.11 0.01 NR2C2 Nuclear receptor subfamily 2 group C member 2 -1.11 0.03 gtpbp1 GTP-binding protein 1 -1.11 0.02 tubg1 Tubulin gamma-1 chain -1.11 0.02 Znf367 Zinc finger protein 367 -1.11 0.02 Map3k14 Mitogen-activated protein kinase kinase kinase 14 -1.11 0.02 EVI5L EVI5-like protein -1.11 0.01 TIMP2 Metalloproteinase inhibitor 2 -1.11 0.01 Tra2b Transformer-2 protein homolog beta -1.10 0.01 RIC1 Receptor-type tyrosine-protein phosphatase epsilon -1.10 0.02 ST3GAL1 CMP-N-acetylneuraminate-beta-galactosamide-alpha-2,3-sialyltransferase 1 -1.10 0.02 Sik3 Serine/threonine-protein kinase SIK3 -1.10 0.03 UPF1 Regulator of nonsense transcripts 1 -1.10 0.01 SLK STE20-like serine/threonine-protein kinase -1.10 0.02 . . -1.10 0.02 NYAP1 Neuronal tyrosine-phosphorylated phosphoinositide-3-kinase adapter 1 -1.10 0.01 TFEB Transcription factor EB -1.10 0.02 MARE2 Microtubule-associated protein RP/EB family member 2 -1.10 0.05 top1 DNA topoisomerase 1 -1.10 0.01 ITPKB Inositol-trisphosphate 3-kinase B -1.10 0.02 STK17A Serine/threonine-protein kinase 17A -1.09 0.02 HUWE1 E3 ubiquitin-protein ligase HUWE1 -1.09 0.01 Papss1 Bifunctional 3'-phosphoadenosine 5'-phosphosulfate synthase 1 -1.09 0.03 Eml3 Echinoderm microtubule-associated protein-like 3 -1.09 0.02 . . -1.09 0.01 DYSF Dysferlin -1.09 0.02 USP9X Probable ubiquitin carboxyl-terminal hydrolase FAF-X -1.08 0.01 top1 DNA topoisomerase 1 -1.08 0.01 PARP10 Poly [ADP-ribose] polymerase 10 -1.08 0.01 NCOR2 Nuclear receptor corepressor 2 -1.08 0.02 ccdc88c Daple-like protein -1.08 0.03

231 CPNE1 Copine-1 -1.08 0.02 Ube2j1 Ubiquitin-conjugating enzyme E2 J1 -1.08 0.03 Tmco3 Transmembrane and coiled-coil domain-containing protein 3 -1.08 0.03 . . -1.08 0.03 CD40 Tumor necrosis factor receptor superfamily member 5 -1.08 0.03 Zmiz1 Zinc finger MIZ domain-containing protein 1 -1.08 0.01 NBEAL1 Neurobeachin-like protein 1 -1.08 0.02 METTL22 Methyltransferase-like protein 22 -1.07 0.03 TRAF2 TNF receptor-associated factor 2 -1.07 0.02 golim4 Golgi integral membrane protein 4 -1.07 0.02 mybbp1a Myb-binding protein 1A-like protein -1.07 0.02 ralbp1-a RalA-binding protein 1-A -1.07 0.01 DPEP2 Dipeptidase 2 -1.07 0.03 HSPG2 Basement membrane-specific heparan sulfate proteoglycan core protein -1.07 0.02 ATP11B Probable phospholipid-transporting ATPase IF -1.07 0.02 PIAS4 E3 SUMO-protein ligase PIAS4 -1.07 0.01 TRIM39 E3 ubiquitin-protein ligase TRIM39 -1.06 0.03 dym Dymeclin -1.06 0.03 KMT2D Histone-lysine N-methyltransferase 2D -1.06 0.04 Mycbp2 E3 ubiquitin-protein ligase MYCBP2 -1.06 0.04 SEMA6D Semaphorin-6D -1.06 0.03 pacsin2 Protein kinase C and casein kinase substrate in neurons protein 2 -1.06 0.02 Cox10 Protoheme IX farnesyltransferase, mitochondrial -1.06 0.03 NCR3LG1 Natural cytotoxicity triggering receptor 3 ligand 1 -1.06 0.03 TCF4 Transcription factor 4 -1.06 0.02 TNS1 Tensin-1 -1.06 0.03 RSBN1 Round spermatid basic protein 1 -1.05 0.02 TANC1 Protein TANC1 -1.05 0.02 MLXIP MLX-interacting protein -1.05 0.05 Tank TRAF family member-associated NF-kappa-B activator -1.05 0.02 Psd4 PH and SEC7 domain-containing protein 4 -1.05 0.03

232 GNE Bifunctional UDP-N-acetylglucosamine 2-epimerase/N-acetylmannosamine kinase -1.05 0.01 EFNB2 Ephrin-B2 -1.05 0.02 cfap36 Cilia- and flagella-associated protein 36 -1.05 0.03 RBMS1 RNA-binding motif, single-stranded-interacting protein 1 -1.05 0.02 LRP6 Low-density lipoprotein receptor-related protein 6 -1.05 0.02 FOSL2 Fos-related antigen 2 -1.04 0.02 BRAP BRCA1-associated protein -1.04 0.04 ccnd2 G1/S-specific cyclin-D2 -1.04 0.02 MDN1 Midasin -1.04 0.02 Col4a2 Collagen alpha-2(IV) chain -1.04 0.01 Upp1 Uridine phosphorylase 1 -1.04 0.02 G6PD Glucose-6-phosphate 1-dehydrogenase -1.04 0.03 tcf3 Transcription factor E2-alpha -1.04 0.01 SYNCRIP Heterogeneous nuclear ribonucleoprotein Q -1.04 0.01 RIN2 Ras and Rab interactor 2 -1.04 0.04 SP1 Transcription factor Sp1 -1.04 0.02 IFRD1 Interferon-related developmental regulator 1 -1.03 0.02 ascc3 Activating signal cointegrator 1 complex subunit 3 -1.03 0.02 Arid1a AT-rich interactive domain-containing protein 1A -1.03 0.04 PTK2B Protein-tyrosine kinase 2-beta -1.03 0.04 HDAC7 Histone deacetylase 7 -1.03 0.02 UBN2 Ubinuclein-2 -1.03 0.04 Dusp4 Dual specificity protein phosphatase 4 -1.03 0.02 inf2 Inverted formin-2 -1.03 0.03 CDCA7 Cell division cycle-associated protein 7 -1.03 0.02 GPCPD1 Glycerophosphocholine phosphodiesterase GPCPD1 -1.03 0.02 ldlr-b Low-density lipoprotein receptor 2 -1.02 0.02 HEATR3 HEAT repeat-containing protein 3 -1.02 0.03 MESDC2 LDLR chaperone MESD -1.02 0.02 slc23a2 Solute carrier family 23 member 2 -1.02 0.04 odc1-a Ornithine decarboxylase 1 -1.02 0.02

233 Ulk2 Serine/threonine-protein kinase ULK2 -1.02 0.03 PHF8 Histone lysine demethylase PHF8 -1.02 0.03 STAR7 StAR-related lipid transfer protein 7, mitochondrial -1.02 0.04 SP4 Transcription factor Sp4 -1.02 0.03 KIDINS220 Kinase D-interacting substrate of 220 kDa -1.02 0.02 ZMYM2 Zinc finger MYM-type protein 2 -1.02 0.04 RHOBTB2 Rho-related BTB domain-containing protein 2 -1.02 0.02 mboat7 Lysophospholipid acyltransferase 7 -1.01 0.03 Ep400 E1A-binding protein p400 -1.01 0.02 . . -1.01 0.02 TOB2 Protein Tob2 -1.01 0.04 FAM160A2 FTS and Hook-interacting protein -1.01 0.01 NCR3LG1 Natural cytotoxicity triggering receptor 3 ligand 1 -1.01 0.01 Atp2b1 Plasma membrane calcium-transporting ATPase 1 -1.00 0.04 . . -1.00 0.04 FNDC3A Fibronectin type-III domain-containing protein 3a -1.00 0.02 Tab1 TGF-beta-activated kinase 1 and MAP3K7-binding protein 1 -1.00 0.03 HIC1 Hypermethylated in cancer 1 protein -1.00 0.03 EPB41L2 Band 4.1-like protein 2 -1.00 0.04 degs1 Sphingolipid delta(4)-desaturase DES1 -1.00 0.04 pik3r5 Phosphoinositide 3-kinase regulatory subunit 5 -1.00 0.03 KIF16B Kinesin-like protein KIF16B -1.00 0.03 CHSY1 Chondroitin sulfate synthase 1 -1.00 0.03 ddb2 DNA damage-binding protein 2 -1.00 0.03 Smg6 Telomerase-binding protein EST1A -1.00 0.02 SLC46A3 Solute carrier family 46 member 3 -1.00 0.04 PIK3CB Phosphatidylinositol 4,5-bisphosphate 3-kinase catalytic subunit beta isoform -1.00 0.03 STAT1 Signal transducer and activator of transcription 1 -0.99 0.02 Capn15 Calpain-15 -0.99 0.04 TFB2M Dimethyladenosine transferase 2, mitochondrial -0.99 0.03 ZDHHC3 Palmitoyltransferase ZDHHC3 -0.99 0.03

234 Emilin1 EMILIN-1 -0.99 0.03 FCHSD2 F-BAR and double SH3 domains protein 2 -0.99 0.02 ZBTB24 Zinc finger and BTB domain-containing protein 24 -0.99 0.03 KCNQ4 Potassium voltage-gated channel subfamily KQT member 4 -0.99 0.03 loxl2 Lysyl oxidase homolog 2 -0.99 0.03 . . -0.99 0.04 Pycard Apoptosis-associated speck-like protein containing a CARD -0.99 0.03 ADRB4C Beta-4C adrenergic receptor -0.99 0.03 SPRYD3 SPRY domain-containing protein 3 -0.98 0.05 LRCH3 Leucine-rich repeat and calponin homology domain-containing protein 3 -0.98 0.03 LTBP4 Latent-transforming growth factor beta-binding protein 4 -0.98 0.04 xpr1 Xenotropic and polytropic retrovirus receptor 1 homolog -0.98 0.03 MVB12B Multivesicular body subunit 12B -0.98 0.04 pgap1 GPI inositol-deacylase -0.98 0.04 igf2-b Insulin-like growth factor II-B -0.98 0.03 ZFYVE26 Zinc finger FYVE domain-containing protein 26 -0.98 0.03 PDZD8 PDZ domain-containing protein 8 -0.98 0.02 . . -0.98 0.04 ctdspl2 CTD small phosphatase-like protein 2 -0.98 0.04 kcp Kielin/chordin-like protein -0.98 0.04 STK17A Serine/threonine-protein kinase 17A -0.97 0.03 oxr1 Oxidation resistance protein 1 -0.97 0.04 ARHGEF17 Rho guanine nucleotide exchange factor 17 -0.97 0.04 TRI93 E3 ubiquitin-protein ligase TRIM39 -0.97 0.03 TJP1 Tight junction protein ZO-1 -0.97 0.03 CNN2 Calponin-2 -0.97 0.03 TFE3 Transcription factor E3 -0.97 0.05 MAPK6 Mitogen-activated protein kinase 6 -0.97 0.03 ARHGAP4 Rho GTPase-activating protein 4 -0.97 0.03 elp2 Elongator complex protein 2 -0.96 0.05 Pycard Apoptosis-associated speck-like protein containing a CARD -0.96 0.03

235 DIP2B Disco-interacting protein 2 homolog B -0.96 0.03 MED12 Mediator of RNA polymerase II transcription subunit 12 -0.96 0.04 DAAM2 Disheveled-associated activator of morphogenesis 2 -0.96 0.03 SMCHD1 Structural maintenance of chromosomes flexible hinge domain-containing protein 1 -0.96 0.04 RIOX2 Ribosomal oxygenase 2 -0.96 0.04 C2CD2 C2 domain-containing protein 2 -0.96 0.01 pik3r1-a Phosphatidylinositol 3-kinase regulatory subunit alpha -0.96 0.04 inpp5d Phosphatidylinositol 3,4,5-trisphosphate 5-phosphatase 1 -0.96 0.04 Ppp2r5e Serine/threonine-protein phosphatase 2A 56 kDa regulatory subunit epsilon isoform -0.96 0.03 MTMR14 Myotubularin-related protein 14 -0.96 0.03 celf1 CUGBP Elav-like family member 1 -0.96 0.03 SHOC2 Leucine-rich repeat protein SHOC-2 -0.96 0.03 . . -0.96 0.02 THEMIS Protein THEMIS -0.96 0.04 GPR174 Probable G-protein coupled receptor 174 -0.96 0.03 si:dkey- 18l1.1 von Willebrand factor A domain-containing protein 8 -0.96 0.04 NIN Ninein -0.96 0.02 SLC6A15 Sodium-dependent neutral amino acid transporter B(0)AT2 -0.96 0.04 . Oocyte zinc finger protein XlCOF22 -0.96 0.02 Sec23b HEAT repeat-containing protein 5A -0.96 0.04 RAPGEF2 Rap guanine nucleotide exchange factor 2 -0.96 0.03 ICAM4 Intercellular adhesion molecule 4 -0.95 0.03 RASA2 Ras GTPase-activating protein 2 -0.95 0.05 MTMR3 Myotubularin-related protein 3 -0.95 0.05 NOS2 Nitric oxide synthase, inducible -0.95 0.02 SLAIN2 SLAIN motif-containing protein 2 -0.95 0.04 IBP1 Insulin-like growth factor-binding protein 1 -0.95 0.02 IQEC1 IQ motif and SEC7 domain-containing protein 1 -0.95 0.05 ATRX Transcriptional regulator ATRX -0.94 0.03 cdc42se1-a CDC42 small effector protein 1-A -0.94 0.04

236 POLE DNA polymerase epsilon catalytic subunit A -0.94 0.04 twf1 Twinfilin-1 -0.94 0.04 med23 Mediator of RNA polymerase II transcription subunit 23 -0.94 0.03 DGKI Diacylglycerol kinase iota -0.94 0.04 RRN3 RNA polymerase I-specific transcription initiation factor RRN3 -0.94 0.05 FACR1 Fatty acyl-CoA reductase 1 -0.94 0.05 MPZL1 Myelin protein zero-like protein 1 -0.94 0.03 Nsmaf Protein FAN -0.94 0.05 CD37 Leukocyte antigen CD37 -0.93 0.04 HLA-DRB1 HLA class II histocompatibility antigen, DRB1-4 beta chain -0.93 0.02 P2RX7 P2X purinoceptor 7 -0.93 0.05 LPIN2 Phosphatidate phosphatase LPIN2 -0.93 0.03 R3hdm2 R3H domain-containing protein 2 -0.93 0.03 . . -0.93 0.03 kpnb1 Importin subunit beta -0.93 0.05 PLEKHG5 Pleckstrin homology domain-containing family G member 5 -0.93 0.04 Limk2 LIM domain kinase 2 -0.93 0.03 Trim25 E3 ubiquitin/ISG15 ligase TRIM25 -0.93 0.04 . . -0.93 0.02 RARA Retinoic acid receptor alpha -0.93 0.04 ARHGAP31 Rho GTPase-activating protein 31 -0.93 0.04 SIK2 Serine/threonine-protein kinase SIK2 -0.93 0.01 yrbE Uncharacterized oxidoreductase YrbE -0.93 0.04 med24 Mediator of RNA polymerase II transcription subunit 24 -0.93 0.04 PLEKHO2 Pleckstrin homology domain-containing family O member 2 -0.92 0.04 PQLC2 Lysosomal amino acid transporter 1 homolog -0.92 0.05 HIPK1 Homeodomain-interacting protein kinase 1 -0.92 0.04 USP32 Ubiquitin carboxyl-terminal hydrolase 32 -0.92 0.04 IQGAP1 Ras GTPase-activating-like protein IQGAP1 -0.92 0.04 MID1 E3 ubiquitin-protein ligase Midline-1 -0.92 0.03 ZNF268 Zinc finger protein 268 -0.92 0.05

237 ERBIN Erbin -0.91 0.03 agap1 Arf-GAP with GTPase, ANK repeat and PH domain-containing protein 1 -0.91 0.04 ARHGAP12 Rho GTPase-activating protein 12 -0.91 0.03 . . -0.91 0.04 L3MBTL3 Lethal(3)malignant brain tumor-like protein 3 -0.91 0.04 mdk-a Midkine-A -0.91 0.04 DNMBP Dynamin-binding protein -0.91 0.04 SPTLC1 Serine palmitoyltransferase 1 -0.91 0.04 CTSK Cathepsin K -0.91 0.02 heatr6 HEAT repeat-containing protein 6 -0.91 0.03 IL7R Interleukin-7 receptor subunit alpha -0.91 0.04 MAP4K5 Mitogen-activated protein kinase kinase kinase kinase 5 -0.91 0.04 Nlrp1b NACHT, LRR and PYD domains-containing protein 1b allele 3 -0.91 0.04 MED13L Mediator of RNA polymerase II transcription subunit 13-like -0.91 0.03 . . -0.90 0.04 PPFIA3 Liprin-alpha-3 -0.90 0.02 . . -0.90 0.05 PDE7A High affinity cAMP-specific 3',5'-cyclic phosphodiesterase 7A -0.90 0.04 myadm Myeloid-associated differentiation marker homolog -0.90 0.04 ZBED4 Zinc finger BED domain-containing protein 4 -0.90 0.04 Rps6ka4 Ribosomal protein S6 kinase alpha-4 -0.90 0.04 Rras2 Ras-related protein R-Ras2 -0.90 0.04 stk10 Serine/threonine-protein kinase 10 -0.90 0.04 cluh Clustered mitochondria protein homolog -0.89 0.04 FILIP1L Filamin A-interacting protein 1-like -0.89 0.03 . . -0.89 0.02 Sec23b Protein transport protein Sec23B -0.89 0.04 ICAM5 Intercellular adhesion molecule 5 -0.89 0.04 APMAP Adipocyte plasma membrane-associated protein -0.89 0.04 ST2B1 Sulfotransferase family cytosolic 2B member 1 -0.89 0.05 AZIN1 Antizyme inhibitor 1 -0.89 0.05

238 TNS4 Tensin-4 -0.89 0.04 ACIN1 Apoptotic chromatin condensation inducer in the nucleus -0.89 0.05 CRAT Carnitine O-acetyltransferase -0.89 0.02 CA8 Carbonic anhydrase-related protein -0.88 0.03 VPS35 Vacuolar protein sorting-associated protein 35 -0.88 0.05 col4a3bp Collagen type IV alpha-3-binding protein -0.88 0.04 snrpc U1 small nuclear ribonucleoprotein C -0.88 0.04 Cables2 CDK5 and ABL1 enzyme substrate 2 -0.88 0.03 RPRD2 Regulation of nuclear pre-mRNA domain-containing protein 2 -0.88 0.04 UBR4 E3 ubiquitin-protein ligase UBR4 -0.87 0.02 PELI1 E3 ubiquitin-protein ligase pellino homolog 1 -0.87 0.05 ME2 NAD-dependent malic enzyme, mitochondrial -0.87 0.05 . Tanabin -0.87 0.03 Sept7 Septin-7 -0.87 0.03 Chd2 Chromodomain-helicase-DNA-binding protein 2 -0.87 0.04 SLC5A6 Sodium-dependent multivitamin transporter -0.87 0.05 FLO11 Flocculation protein FLO11 -0.87 0.05 NIPA1 Magnesium transporter NIPA1 -0.87 0.04 GCP3 Gamma-tubulin complex component 3 homolog -0.87 0.05 DNMT3A DNA (cytosine-5)-methyltransferase 3A -0.87 0.05 pol Retrovirus-related Pol polyprotein from transposon 17.6 -0.86 0.04 ATF7IP Activating transcription factor 7-interacting protein 1 -0.86 0.04 MYH10 Myosin-10 -0.86 0.03 NCOA6 Nuclear receptor coactivator 6 -0.86 0.04 GPDM Glycerol-3-phosphate dehydrogenase, mitochondrial -0.86 0.05 ITGAV Integrin alpha-V -0.86 0.05 . . -0.86 0.04 GLYM Serine hydroxymethyltransferase, mitochondrial -0.86 0.05 EURL Protein EURL homolog -0.86 0.05 PFKFB3 6-phosphofructo-2-kinase/fructose-2,6-bisphosphatase 3 -0.85 0.04 Nt5c1a Cytosolic 5'-nucleotidase 1A -0.85 0.05

239 AHR Aryl hydrocarbon receptor -0.85 0.04 ANR11 Ankyrin repeat domain-containing protein 11 -0.85 0.05 FAAH1 Fatty-acid amide hydrolase 1 -0.85 0.05 Golm1 Golgi membrane protein 1 -0.85 0.04 Arfip2 Arfaptin-2 -0.85 0.04 apc Adenomatous polyposis coli homolog -0.85 0.03 Arl4a ADP-ribosylation factor-like protein 4A -0.84 0.03 ARHGAP15 Rho GTPase-activating protein 15 -0.84 0.04 PTPN9 Tyrosine-protein phosphatase non-receptor type 9 -0.83 0.05 KANSL1 KAT8 regulatory NSL complex subunit 1 -0.83 0.03 FLT1 Vascular endothelial growth factor receptor 1 -0.83 0.04 CLINT1 Clathrin interactor 1 -0.83 0.04 CAN5 Calpain-5 -0.82 0.05 CCDC50 Coiled-coil domain-containing protein 50 -0.82 0.04 USP24 Ubiquitin carboxyl-terminal hydrolase 24 -0.81 0.05 SMPD1 Sphingomyelin phosphodiesterase -0.81 0.04 PCGF5 Polycomb group RING finger protein 5 -0.81 0.05 col6a1 Collagen alpha-1(VI) chain -0.79 0.05 B3GT1 Beta-1,3-galactosyltransferase 1 -0.79 0.05 Rbm12b1 RNA-binding protein 12B-A -0.79 0.04 PARP11 Poly [ADP-ribose] polymerase 11 -0.79 0.05 stag2 Cohesin subunit SA-2 -0.78 0.05 ARFGEF2 Brefeldin A-inhibited guanine nucleotide-exchange protein 2 -0.77 0.04 HYAL1 Hyaluronidase-1 -0.77 0.04 . . -0.77 0.04 EPG5 Ectopic P granules protein 5 homolog -0.76 0.05 . . -0.75 0.02 Fbxo6 F-box only protein 6 -0.74 0.02 Cbfb Core-binding factor subunit beta -0.73 0.04 USF3 Basic helix-loop-helix domain-containing protein USF3 -0.73 0.03 CCDC162P Coiled-coil domain-containing protein 162 -0.73 0.04

240 HTR5B HEAT repeat-containing protein 5B -0.72 0.05 BFAR Bifunctional apoptosis regulator -0.55 0.04

Table AI5. Front (WA) up-regulation Gene Protein effect p RALGAPA1 Ral GTPase-activating protein subunit alpha-1 1.61 0.01 ICAM4 Intercellular adhesion molecule 4 1.51 0.02 ADAMTSL4 ADAMTS-like protein 4 1.47 0.01 pol Pol polyprotein 1.38 0.01 csnk1d Casein kinase I isoform delta 1.37 0.02 BCAT1 Branched-chain-amino-acid aminotransferase, cytosolic 1.35 0.02 mapk1 Mitogen-activated protein kinase 1 1.35 0.02 mafb Transcription factor MafB 1.33 0.02 Cdc42ep2 Cdc42 effector protein 2 1.31 0.01 qki-b Protein quaking-B 1.28 0.02 FLNB Filamin-B 1.25 0.02 HLA-DRB1 HLA class II histocompatibility antigen, DRB1-13 beta chain 1.25 0.02 ATP13A2 Cation-transporting ATPase 13A2 1.23 0.02 timp3 Metalloproteinase inhibitor 3 1.21 0.02 PRSS23 Serine protease 23 1.20 0.02 . . 1.19 0.03 IDE Insulin-degrading enzyme 1.19 0.03 Tmed2 Transmembrane emp24 domain-containing protein 2 1.18 0.03 Dele Death ligand signal enhancer 1.18 0.03 tppp3 Tubulin polymerization-promoting protein family member 3 1.18 0.03 ITSN1 Intersectin-1 1.17 0.03 ARHGEF17 Rho guanine nucleotide exchange factor 17 1.15 0.03 RBMS1 RNA-binding motif, single-stranded-interacting protein 1 1.15 0.03 Cpt1a Carnitine O-palmitoyltransferase 1, liver isoform 1.15 0.03 gnb1 Guanine nucleotide-binding protein G(I)/G(S)/G(T) subunit beta-1 1.14 0.03

241 MMRN2 Multimerin-2 1.13 0.02 IL4R Interleukin-4 receptor subunit alpha 1.13 0.02 . Class I histocompatibility antigen, F10 alpha chain 1.11 0.03 JAM2 Junctional adhesion molecule B 1.10 0.03 pgap3 Post-GPI attachment to proteins factor 3 1.10 0.04 RAB2A Ras-related protein Rab-2A 1.09 0.03 SDPR Serum deprivation-response protein 1.09 0.04 STK17A Serine/threonine-protein kinase 17A 1.08 0.02 ccnd1 G1/S-specific cyclin-D1 1.06 0.04 Msantd2 Myb/SANT-like DNA-binding domain-containing protein 2 1.06 0.03 NTSR1 Neurotensin receptor type 1 1.06 0.04 BICD2 Protein bicaudal D homolog 2 1.06 0.04 OSBPL2 Oxysterol-binding protein-related protein 2 1.05 0.04 ADA10 Disintegrin and metalloproteinase domain-containing protein 10 1.05 0.03 magt1 Magnesium transporter protein 1 1.05 0.04 Pigq Phosphatidylinositol N-acetylglucosaminyltransferase subunit Q 1.05 0.05 Nagpa N-acetylglucosamine-1-phosphodiester alpha-N-acetylglucosaminidase 1.05 0.04 SPAG9 C-Jun-amino-terminal kinase-interacting protein 4 1.04 0.04 GAA Lysosomal alpha-glucosidase 1.04 0.03 Arsa Arylsulfatase A 1.04 0.04 FKBP8 Peptidyl-prolyl cis-trans isomerase FKBP8 1.03 0.04 GDI1 Rab GDP dissociation inhibitor alpha 1.03 0.04 HLA-DRB1 HLA class II histocompatibility antigen, DRB1-13 beta chain 1.02 0.04 spns2 Protein spinster homolog 2 1.01 0.04 dbnl-a Drebrin-like protein A 1.01 0.04 Cd84 SLAM family member 5 1.01 0.03 Slamf7 SLAM family member 7 1.00 0.04 PRELID3B PRELI domain containing protein 3B 1.00 0.05 pik3r1-a Phosphatidylinositol 3-kinase regulatory subunit alpha 0.99 0.04 EVI5L EVI5-like protein 0.97 0.04 HLA-DRA HLA class II histocompatibility antigen, DR alpha chain 0.94 0.05

242 ATL3 Atlastin-3 0.87 0.05 TNFRSF19 Tumor necrosis factor receptor superfamily member 19 0.84 0.04

Table AI6. Front (WA) down-regulation Gene Protein effect p S100A11 Protein S100-A11 -1.57 0.01 PLCL1 Inactive phospholipase C-like protein 1 -1.34 0.02 . . -1.33 0.02 Kars Lysine--tRNA ligase -1.30 0.03 GPI Glucose-6-phosphate isomerase -1.28 0.02 NOP56 Nucleolar protein 56 -1.27 0.03 Paics Multifunctional protein ADE2 -1.27 0.02 . . -1.20 0.04 rbm8a RNA-binding protein 8A -1.20 0.03 ZNF300 Zinc finger protein 300 -1.18 0.02 eif3a Eukaryotic translation initiation factor 3 subunit A -1.17 0.03 PPP6R1 Serine/threonine-protein phosphatase 6 regulatory subunit 1 -1.16 0.04 SLC24A2 Sodium/potassium/calcium exchanger 2 -1.15 0.02 UBP5 Ubiquitin carboxyl-terminal hydrolase 5 -1.14 0.03 Myg1 UPF0160 protein MYG1, mitochondrial -1.14 0.03 . . -1.14 0.04 ZNF84 Zinc finger protein 84 -1.14 0.03 . -1.12 0.03 ZNF546 Zinc finger protein 546 -1.11 0.03 NDUFB8 NADH dehydrogenase [ubiquinone] 1 beta subcomplex subunit 8, mitochondrial -1.10 0.03 Ppp1ca Serine/threonine-protein phosphatase PP1-alpha catalytic subunit -1.10 0.04 . . -1.08 0.03 . . -1.05 0.02 . . -1.05 0.05 ZBTB49 Zinc finger and BTB domain-containing protein 49 -1.04 0.04 . . -1.04 0.04

243 SNRPA1 U2 small nuclear ribonucleoprotein A' -1.04 0.05 . . -1.03 0.03 . . -1.03 0.04 . . -1.02 0.04 NANS Sialic acid synthase -1.02 0.04 Dapk2 Death-associated protein kinase 2 -1.01 0.04 COX11 Cytochrome c oxidase assembly protein COX11, mitochondrial -1.01 0.05 NUP62 Nuclear pore glycoprotein p62 -1.00 0.04 . . -1.00 0.04 . . -0.99 0.04 Eif2b4 Translation initiation factor eIF-2B subunit delta -0.99 0.04 Cd2bp2 CD2 antigen cytoplasmic tail-binding protein 2 -0.99 0.03 . . -0.98 0.05 COG8 Conserved oligomeric Golgi complex subunit 8 -0.98 0.05 . . -0.98 0.05 Omg Oligodendrocyte-myelin glycoprotein -0.98 0.04 . . -0.97 0.04 . . -0.97 0.05 dcaf13 DDB1- and CUL4-associated factor 13 -0.96 0.05 CC174 Coiled-coil domain-containing protein 174 -0.96 0.04 KCTD2 BTB/POZ domain-containing protein KCTD2 -0.96 0.05 rbm42 RNA-binding protein 42 -0.94 0.04 RPP38 Ribonuclease P protein subunit p38 -0.93 0.05 . . -0.92 0.04 GNL3L Guanine nucleotide-binding protein-like 3-like protein -0.92 0.04 ZG57 Gastrula zinc finger protein XlCGF57.1 -0.92 0.05 . . -0.91 0.05 MRPS2 28S ribosomal protein S2, mitochondrial -0.91 0.04 . . -0.91 0.05 PGM1 Phosphoglucomutase-1 -0.90 0.04 . . -0.89 0.04

244 DPP10 Inactive dipeptidyl peptidase 10 -0.88 0.04 Gatad2b Transcriptional repressor p66-beta -0.88 0.05 NDUFAB1 Acyl carrier protein, mitochondrial -0.87 0.05 . . -0.87 0.05 got2 Aspartate aminotransferase, mitochondrial -0.85 0.04 nxnl1 Nucleoredoxin-like protein 1 -0.84 0.05

Differentially expressed transcripts in cane toads (Rhinella marina) across the Australian range (main text, Figure 5). Soft clustering of RNA-Seq data from spleens was performed to visualize expression patterns across the range (main text, Figure 3). Membership indicates how closely the expression pattern of each transcript fits to that of the cluster it was grouped into.

Table AI7. Cluster 1 Gene Protein Membership PXN1 Jeltraxin 0.97 -- -- 0.96 REEP5 Receptor expression-enhancing protein 5 0.95 FCGR2 Low affinity immunoglobulin gamma Fc region receptor II 0.95 ERVPABLB- 1 Endogenous retrovirus group PABLB member 1 Env polyprotein 0.94 ARHGEF19 Rho guanine nucleotide exchange factor 19 0.94 MPL Thrombopoietin receptor 0.93 PXN1 Jeltraxin 0.93 TESPA1 Protein TESPA1 0.92 lingo1 Leucine-rich repeat and immunoglobulin-like domain-containing nogo receptor-interacting protein 1 0.92 P2RY12 P2Y purinoceptor 12 0.91 PXN1 Jeltraxin 0.91 Gp9 Platelet glycoprotein IX 0.91 Dhrs9 Dehydrogenase/reductase SDR family member 9 0.91 -- -- 0.90 CLUL1 Clusterin-like protein 1 0.89 NMT2 Phosphomethylethanolamine N-methyltransferase 0.88

245 Gp5 Platelet glycoprotein V 0.88 -- -- 0.88 Ecm1 Extracellular matrix protein 1 0.88 mul1a Mitochondrial ubiquitin ligase activator of nfkb 1-A 0.87 -- -- 0.87 St3gal5 Lactosylceramide alpha-2,3-sialyltransferase 0.87 ERVW-1 Syncytin-1 0.86 -- -- 0.86 IGKV6-21 Immunoglobulin kappa variable 6-21 0.86 -- LINE-1 reverse transcriptase homolog 0.86 -- -- 0.85 CYP8B1 5-beta-cholestane-3-alpha,7-alpha-diol 12-alpha-hydroxylase 0.85 CD200R1B Cell surface glycoprotein CD200 receptor 1-B 0.85 -- -- 0.85 itln1 Intelectin-1 0.84 Igk-V19-17 Ig kappa chain V19-17 0.83 CYP3A29 Cytochrome P450 3A29 0.83 Tub Tubby protein 0.82 -- -- 0.82 ST3GAL6 Type 2 lactosamine alpha-2,3-sialyltransferase 0.82 Pol Pol polyprotein 0.82 PXDC1 PX domain-containing protein 1 0.82 ZNF84 Zinc finger protein 84 0.82 GPR21 Probable G-protein coupled receptor 21 0.81 HLA-DRB1 HLA class II histocompatibility antigen, DRB1-4 beta chain 0.81 -- -- 0.81 MED12L Mediator of RNA polymerase II transcription subunit 12-like protein 0.80 FNDC7 Fibronectin type III domain-containing protein 7 0.80 prr5 Proline-rich protein 5 0.80 Gp1bb Platelet glycoprotein Ib beta chain 0.79 -- -- 0.79

246 hacd4 Very-long-chain (3R)-3-hydroxyacyl-CoA dehydratase 0.79 CFH Complement factor H 0.78 ENDOD1 Endonuclease domain-containing 1 protein 0.78 -- -- 0.78 Pafah2 Platelet-activating factor acetylhydrolase 2, cytoplasmic 0.78 ARHGEF37 Rho guanine nucleotide exchange factor 37 0.78 CSF2RA Granulocyte-macrophage colony-stimulating factor receptor subunit alpha 0.77 Rnf130 E3 ubiquitin-protein ligase RNF130 0.76 F10 Coagulation factor X 0.76 PLEK Pleckstrin 0.76 CCDC126 Coiled-coil domain-containing protein 126 0.75 c1galt1 Glycoprotein-N-acetylgalactosamine 3-beta-galactosyltransferase 1 0.75 fn1 Fibronectin 0.74 ITGA2B Integrin alpha-IIb 0.74 -- -- 0.73 STON1 Stonin-1 0.73 -- -- 0.73 Fxyd1 Phospholemman 0.73 Dcst1 DC-STAMP domain-containing protein 1 0.72 Hmox2 Heme oxygenase 2 0.71 -- -- 0.71 rap1b Ras-related protein Rap-1b 0.71 Pinlyp phospholipase A2 inhibitor and Ly6/PLAUR domain-containing protein 0.71 -- -- 0.70 TNFRSF6B Tumor necrosis factor receptor superfamily member 6B 0.70 Esm1 Endothelial cell-specific molecule 1 0.70

Table AI8. Cluster 2 Gene Protein Membership Atp5o ATP synthase subunit O, mitochondrial 0.93

247 pkmyt1 Membrane-associated tyrosine- and threonine-specific cdc2-inhibitory kinase 0.89 -- -- 0.88 eif3a Eukaryotic translation initiation factor 3 subunit A 0.88 CIR1 Corepressor interacting with RBPJ 1 0.87 BEND4 BEN domain-containing protein 4 0.86 -- -- 0.86 ZNF84 Zinc finger protein 84 0.85 Kars Lysine--tRNA ligase 0.85 ccdc93 Coiled-coil domain-containing protein 93 0.83 -- Gastrula zinc finger protein XlCGF57--1 0.83 -- -- 0.82 -- -- 0.82 MVK Mevalonate kinase 0.81 exosc6 Exosome complex component MTR3 0.81 Pkd1l3 Polycystic kidney disease protein 1-like 3 0.81 -- -- 0.80 RNF40 E3 ubiquitin-protein ligase BRE1B 0.80 -- -- 0.80 RNF112 RING finger protein 112 0.80 KCTD2 BTB/POZ domain-containing protein KCTD2 0.79 Csnk2a2 Casein kinase II subunit alpha' 0.79 pgap2 Post-GPI attachment to proteins factor 2 0.78 Cox11 Cytochrome c oxidase assembly protein COX11, mitochondrial 0.78 -- Gastrula zinc finger protein XlCGF67--1 0.77 Eif2b4 Translation initiation factor eIF-2B subunit delta 0.76 -- -- 0.76 PLCL1 Inactive phospholipase C-like protein 1 0.75 -- -- 0.75 ASB6 Ankyrin repeat and SOCS box protein 6 0.75 -- -- 0.75 SLC24A2 Sodium/potassium/calcium exchanger 2 0.74

248 -- -- 0.74 Gabarap Gamma-aminobutyric acid receptor-associated protein 0.74 -- -- 0.74 THAP5 THAP domain-containing protein 5 0.74 -- -- 0.74 -- -- 0.73 -- -- 0.73 LRRC41 Leucine-rich repeat-containing protein 41 0.73 -- -- 0.73 Dapk2 Death-associated protein kinase 2 0.73 VIPAS39 Spermatogenesis-defective protein 39 homolog 0.72 -- -- 0.72 UBP5 Ubiquitin carboxyl-terminal hydrolase 5 0.72 spice1 Spindle and centriole-associated protein 1 0.71 PSMC4 26S protease regulatory subunit 6B 0.71 -- -- 0.70 rpl7a 60S ribosomal protein L7a 0.70 IMMT MICOS complex subunit MIC60 0.70 Ier5 Immediate early response gene 5 protein 0.70 ELOVL1 Elongation of very long chain fatty acids protein 1 0.70

Table AI9. Cluster 3 Gene Protein Membership BICD2 Protein bicaudal D homolog 2 0.95 IDE Insulin-degrading enzyme 0.93 mboat7 Lysophospholipid acyltransferase 7 0.93 APMAP Adipocyte plasma membrane-associated protein 0.91 TNS1 Tensin-1 0.90 SPAG9 C-Jun-amino-terminal kinase-interacting protein 4 0.89 qki-b Protein quaking-B 0.88 CROT Peroxisomal carnitine O-octanoyltransferase 0.88

249 PARVA Alpha-parvin 0.87 RALGAPA1 Ral GTPase-activating protein subunit alpha-1 0.87 Tab1 TGF-beta-activated kinase 1 and MAP3K7-binding protein 1 0.86 Ube2j1 Ubiquitin-conjugating enzyme E2 J1 0.86 RBMS1 RNA-binding motif, single-stranded-interacting protein 1 0.85 Emilin1 EMILIN-1 0.83 IQSEC1 IQ motif and SEC7 domain-containing protein 1 0.83 FLNB Filamin-B 0.83 DPEP2 Dipeptidase 2 0.82 SPTLC1 Serine palmitoyltransferase 1 0.82 csnk1d Casein kinase I isoform delta 0.82 -- -- 0.82 SEPT7 Septin-7 0.82 EVI5L EVI5-like protein 0.82 Tmco3 Transmembrane and coiled-coil domain-containing protein 3 0.81 Larp1 La-related protein 1 0.81 ARHGEF17 Rho guanine nucleotide exchange factor 17 0.81 CNN2 Calponin-2 0.80 ITGAV Integrin alpha-V 0.80 Dele Death ligand signal enhancer 0.79 RIN2 Ras and Rab interactor 2 0.79 Thbd Thrombomodulin 0.78 ZDHHC3 Palmitoyltransferase ZDHHC3 0.78 -- -- 0.77 WWTR1 WW domain-containing transcription regulator protein 1 0.77 SDPR Serum deprivation-response protein 0.76 IQGAP1 Ras GTPase-activating-like protein IQGAP1 0.76 RAB2A Ras-related protein Rab-2A 0.76 FKBP8 Peptidyl-prolyl cis-trans isomerase FKBP8 0.76 SDCBP Syntenin-1 0.76 tppp3 Tubulin polymerization-promoting protein family member 3 0.75

250 ccnd1 G1/S-specific cyclin-D1 0.75 VGLL4 Transcription cofactor vestigial-like protein 4 0.75 CRAT Carnitine O-acetyltransferase 0.75 ATL3 Atlastin-3 0.75 BCAT1 Branched-chain-amino-acid aminotransferase, cytosolic 0.75 kcp Kielin/chordin-like protein 0.75 ITSN1 Intersectin-1 0.74 Golm1 Golgi membrane protein 1 0.74 Rras2 Ras-related protein R-Ras2 0.74 Smad6 Mothers against decapentaplegic homolog 6 0.74 ATP13A2 Cation-transporting ATPase 13A2 0.74 ICAM5 Intercellular adhesion molecule 5 0.74 Camk1 Calcium/calmodulin-dependent protein kinase type 1 0.74 adam10 Disintegrin and metalloproteinase domain-containing protein 10 0.73 Papss1 Bifunctional 3'-phosphoadenosine 5'-phosphosulfate synthase 1 0.73 Clic4 Chloride intracellular channel protein 4 0.72 myadm Myeloid-associated differentiation marker homolog 0.72 C2CD2 C2 domain-containing protein 2 0.72 gnb1 Guanine nucleotide-binding protein G(I)/G(S)/G(T) subunit beta-1 0.72 timp3 Metalloproteinase inhibitor 3 0.72 TNFAIP2 Tumor necrosis factor alpha-induced protein 2 0.72 STK17A Serine/threonine-protein kinase 17A 0.72 TNS4 Tensin-4 0.72 -- -- 0.71 -- -- 0.71 DAAM2 Disheveled-associated activator of morphogenesis 2 0.71 SPPL2A peptidase-like 2A 0.71

Table AI10. Cluster 4 Gene Protein Membership NSD3 Histone-lysine N-methyltransferase NSD3 0.95

251 ZNF300 Zinc finger protein 300 0.95 rps8 40S ribosomal protein S8 0.95 eif3d Eukaryotic translation initiation factor 3 subunit D 0.94 -- -- 0.93 eef1g-a Elongation factor 1-gamma-A 0.92 CTDSP2 Carboxy-terminal domain RNA polymerase II polypeptide A small phosphatase 2 0.92 smyd5 SET and MYND domain-containing protein 5 0.89 MOCS2 Molybdopterin synthase catalytic subunit 0.89 Borcs6 BLOC-1-related complex subunit 6 0.88 Dihydrolipoyllysine-residue succinyltransferase component of 2-oxoglutarate dehydrogenase complex, DLST mitochondrial 0.88 eif3l Eukaryotic translation initiation factor 3 subunit L 0.88 rps2 40S ribosomal protein S2 0.87 -- Gastrula zinc finger protein XlCGF57--1 0.87 pno1 RNA-binding protein PNO1 0.87 LRSAM1 E3 ubiquitin-protein ligase LRSAM1 0.87 rps15 40S ribosomal protein S15 0.86 GFM2 Ribosome-releasing factor 2, mitochondrial 0.86 ISG20L2 Interferon-stimulated 20 kDa exonuclease-like 2 0.86 Rps19 40S ribosomal protein S19 0.85 eif3h Eukaryotic translation initiation factor 3 subunit H 0.85 eif3k Eukaryotic translation initiation factor 3 subunit K 0.85 -- -- 0.84 tmem251 Transmembrane protein 251 0.84 Rps27 40S ribosomal protein S27 0.84 Rpl11 60S ribosomal protein L11 0.84 Rps3 40S ribosomal protein S3 0.84 RPL35 60S ribosomal protein L35 0.83 -- -- 0.82 -- -- 0.82 Rpl12 60S ribosomal protein L12 0.82

252 -- -- 0.81 Rpl23 60S ribosomal protein L23 0.81 ndor1 NADPH-dependent diflavin oxidoreductase 1 0.81 eif3g-b Eukaryotic translation initiation factor 3 subunit G-B 0.81 Psmg3 Proteasome assembly chaperone 3 0.79 -- -- 0.79 ZNF300 Zinc finger protein 300 0.79 eif3b Eukaryotic translation initiation factor 3 subunit B 0.79 svbp Small vasohibin-binding protein 0.79 -- -- 0.78 eif3c Eukaryotic translation initiation factor 3 subunit C 0.78 MTMR9 Myotubularin-related protein 9 0.78 -- -- 0.78 ZNF300 Zinc finger protein 300 0.77 -- -- 0.77 rps23 40S ribosomal protein S23 0.77 Rpl27 60S ribosomal protein L27 0.77 Serbp1 Plasminogen activator inhibitor 1 RNA-binding protein 0.77 RPL3 60S ribosomal protein L3 0.77 -- -- 0.77 Eef1d Elongation factor 1-delta 0.76 -- -- 0.76 -- -- 0.76 -- Oocyte zinc finger protein XlCOF7--1 0.76 -- Cytochrome c oxidase subunit 4 isoform 2, mitochondrial 0.75 rps3a 40S ribosomal protein S3a 0.75 -- -- 0.75 MCUB Calcium uniporter regulatory subunit MCUb, mitochondrial 0.74 EIF5A2 Eukaryotic translation initiation factor 5A-2 0.74 Tmed8 Protein TMED8 0.73 -- -- 0.73

253 rpl18a 60S ribosomal protein L18a 0.73 Ppp1r35 Protein phosphatase 1 regulatory subunit 35 0.72 -- -- 0.72 TPT1 Translationally-controlled tumor protein homolog 0.72 -- -- 0.72 med8-b Mediator of RNA polymerase II transcription subunit 8-B 0.72 get4 Golgi to ER traffic protein 4 homolog 0.72 fam32a Protein FAM32A 0.72 LCMT1 Leucine carboxyl methyltransferase 1 0.71 -- -- 0.71 -- -- 0.70 rnf2-a E3 ubiquitin-protein ligase RING2-A 0.70 NGB Neuroglobin 0.70 SLC35A1 CMP-sialic acid transporter 0.70 rpl18-b 60S ribosomal protein L18-B 0.70 Polr2c DNA-directed RNA polymerase II subunit RPB3 0.70

Table AI11. Cluster 5 Gene Protein Membership -- -- 0.94 NCR3LG1 Natural cytotoxicity triggering receptor 3 ligand 1 0.92 Meioc Meiosis-specific coiled-coil domain-containing protein MEIOC 0.91 ACACB Acetyl-CoA carboxylase 2 0.90 mrtfb Myocardin-related transcription factor B 0.89 Myo1c Unconventional myosin-Ic 0.88 ST3GAL1 CMP-N-acetylneuraminate-beta-galactosamide-alpha-2,3-sialyltransferase 1 0.87 srf Serum response factor 0.87 -- -- 0.86 DMD Dystrophin 0.86 tcf3 Transcription factor E2-alpha 0.85 pum2 Pumilio homolog 2 0.85

254 TFEB Transcription factor EB 0.85 FOSL2 Fos-related antigen 2 0.84 RRN3 RNA polymerase I-specific transcription initiation factor RRN3 0.84 MAP3K2 Mitogen-activated protein kinase kinase kinase 2 0.83 ATP11C Phospholipid-transporting ATPase IG 0.82 DOCK1 Dedicator of cytokinesis protein 1 0.82 pik3r5 Phosphoinositide 3-kinase regulatory subunit 5 0.81 GNE Bifunctional UDP-N-acetylglucosamine 2-epimerase/N-acetylmannosamine kinase 0.81 MICU2 Calcium uptake protein 2, mitochondrial 0.81 stk10 Serine/threonine-protein kinase 10 0.80 Tank TRAF family member-associated NF-kappa-B activator 0.79 SRCAP Helicase SRCAP 0.79 RASA2 Ras GTPase-activating protein 2 0.78 VPS13A Vacuolar protein sorting-associated protein 13A 0.78 ATRX Transcriptional regulator ATRX 0.77 DAGLB Sn1-specific diacylglycerol lipase beta 0.76 AFAP1 Actin filament-associated protein 1 0.76 Pycard Apoptosis-associated speck-like protein containing a CARD 0.76 ZFYVE26 Zinc finger FYVE domain-containing protein 26 0.76 RREB1 Ras-responsive element-binding protein 1 0.75 PTK2B Protein-tyrosine kinase 2-beta 0.75 ABI1 Abl interactor 1 0.75 Sbno1 Protein strawberry notch homolog 1 0.75 PPP1R12A Protein phosphatase 1 regulatory subunit 12A 0.75 HNRNPLL Heterogeneous nuclear ribonucleoprotein L-like 0.74 NFATC2 Nuclear factor of activated T-cells, cytoplasmic 2 0.74 NR2C2 Nuclear receptor subfamily 2 group C member 2 0.74 Ppp6c Serine/threonine-protein phosphatase 6 catalytic subunit 0.74 Lbr Lamin-B receptor 0.74 ERBIN Erbin 0.74 PBXIP1 Pre-B-cell leukemia transcription factor-interacting protein 1 0.73

255 STXBP2 Syntaxin-binding protein 2 0.72 ARHGAP4 Rho GTPase-activating protein 4 0.72 coq10b-a Coenzyme Q-binding protein COQ10 homolog A, mitochondrial 0.71 CSF2RB Cytokine receptor common subunit beta 0.71 Itch E3 ubiquitin-protein ligase Itchy 0.70 PDS5A Sister chromatid cohesion protein PDS5 homolog A 0.70

Table AI12. Cluster 6 Gene Protein Membership ANKRD17 Ankyrin repeat domain-containing protein 17 0.91 MINA Bifunctional lysine-specific demethylase and histidyl-hydroxylase MINA 0.90 odc1-a Ornithine decarboxylase 1 0.90 TRIM39 E3 ubiquitin-protein ligase TRIM39 0.89 paxip1 PAX-interacting protein 1 0.88 Pycard Apoptosis-associated speck-like protein containing a CARD 0.87 TAF2 Transcription initiation factor TFIID subunit 2 0.87 HERC2 E3 ubiquitin-protein ligase HERC2 0.87 RHOBTB2 Rho-related BTB domain-containing protein 2 0.86 SCAF11 Protein SCAF11 0.85 UPF1 Regulator of nonsense transcripts 1 0.84 pol Retrovirus-related Pol polyprotein from transposon 17--6 0.83 VPS13D Vacuolar protein sorting-associated protein 13D 0.81 Arid1a AT-rich interactive domain-containing protein 1A 0.81 KIF21A Kinesin-like protein KIF21A 0.79 MAST4 Microtubule-associated serine/threonine-protein kinase 4 0.79 ZBTB24 Zinc finger and BTB domain-containing protein 24 0.78 HUWE1 E3 ubiquitin-protein ligase HUWE1 0.78 COL4A1 Collagen alpha-1(IV) chain 0.77 ZMYM2 Zinc finger MYM-type protein 2 0.77 -- -- 0.75 -- -- 0.74

256 GPCPD1 Glycerophosphocholine phosphodiesterase GPCPD1 0.73 ccnd2 G1/S-specific cyclin-D2 0.73 DNMT3A DNA (cytosine-5)-methyltransferase 3A 0.73 Nlrp1b NACHT, LRR and PYD domains-containing protein 1b allele 3 0.72 MPZL1 Myelin protein zero-like protein 1 0.72 mapk8 Mitogen-activated protein kinase 8 0.71 Nckipsd NCK-interacting protein with SH3 domain 0.71 HDAC7 Histone deacetylase 7 0.71 FAM46A Protein FAM46A 0.70 Xpo1 Exportin-1 0.70

Appendix II Transcripts in cane toads (Rhinella marina) associated with environmental variables (maximum temperature during the hottest month or rainfall during the driest quarter) across the Australian range (main text, Figure 1). A latent factor mixed model (LFMM) was performed on iqlr-transformed counts produced using RNA-Seq data from spleens in toads across the range (main text, Figure 1). Associated Temperature p- Rain p- Cluster Gene Protein variable(s) value value -- -- temperature & rain 1.52E-11 3.85E-09 1 CLUL1 Clusterin-like protein 1 temperature & rain 1.14E-05 2.01E-07 1 PAFA2 Platelet-activating factor acetylhydrolase 2, cytoplasmic temperature 2.60E-05 none 1 CREG1 Protein CREG1 temperature 5.89E-06 none none ARSA Arylsulfatase A temperature 1.73E-05 none none QRIC1 Glutamine-rich protein 1 temperature & rain 2.21E-08 6.42E-09 none IIGP5 Interferon-inducible GTPase 5 temperature 9.02E-06 none none SEPT8 Septin-8 temperature 2.60E-06 none none DPYL3 Dihydropyrimidinase-related protein 3 temperature 1.76E-05 none none -- -- temperature & rain 1.47E-05 1.10E-05 none GCYB1 Guanylate cyclase soluble subunit beta-1 rain none 1.51E-06 none

257

Appendix III Groups of proportional transcripts in RNA-Seq data obtained from spleens in cane toads (Rhinella marina) sampled across the northern Australian range. The propr package in R was used to identify pairs of transcripts that were coordinated in expression across all sampled states (main text, Figure 1). Expression Fuzzy pattern cluster Group associated with associated Gene Protein (Figure S3) group with group rnaseh2b Ribonuclease H2 subunit B A NT up 4 RPL5 60S ribosomal protein L5 A NT up 4 naca Nascent polypeptide-associated complex subunit alpha A NT up 4 eif3m Eukaryotic translation initiation factor 3 subunit M A NT up 4 -- -- A NT up 4 PLGRKT Plasminogen receptor (KT) A NT up 4 POMP Proteasome maturation protein A NT up 4 rps21 40S ribosomal protein S21 A NT up 4 rps15 40S ribosomal protein S15 A NT up 4 SNRPE Small nuclear ribonucleoprotein E A NT up 4 RPS5 40S ribosomal protein S5 A NT up 4 rps17 40S ribosomal protein S17 A NT up 4 -- -- A NT up 4 RPS21 40S ribosomal protein S21 A NT up 4 ALG2 Alpha-1,3/1,6-mannosyltransferase ALG2 A NT up 4 Atp5g3 ATP synthase F(0) complex subunit C3, mitochondrial A NT up 4 rpl22l1 60S ribosomal protein L22-like 1 A NT up 4 ddx21-b Nucleolar RNA helicase 2-B A NT up 4 eef1b Elongation factor 1-beta A NT up 4 bzw1 Basic leucine zipper and W2 domain-containing protein 1 A NT up 4 Ndufa1 NADH dehydrogenase [ubiquinone] 1 alpha subcomplex subunit 1 A NT up 4

258 -- -- A NT up 4 -- -- A NT up 4 rpl18a 60S ribosomal protein L18a A NT up 4 NDUFA8 NADH dehydrogenase [ubiquinone] 1 alpha subcomplex subunit 8 A NT up 4 NDUFA7 NADH dehydrogenase [ubiquinone] 1 alpha subcomplex subunit 7 A NT up 4 rps4 40S ribosomal protein S4 A NT up 4 RPS25 40S ribosomal protein S25 A NT up 4 Serbp1 Plasminogen activator inhibitor 1 RNA-binding protein A NT up 4 Rps15a 40S ribosomal protein S15a A NT up 4 ATP5J ATP synthase-coupling factor 6, mitochondrial A NT up 4 fmc1 Protein FMC1 homolog A NT up 4 eif3k Eukaryotic translation initiation factor 3 subunit K A NT up 4 Rpl17 60S ribosomal protein L17 A NT up 4 Rpl37 60S ribosomal protein L37 A NT up 4 rps16 40S ribosomal protein S16 A NT up 4 Eef1d Elongation factor 1-delta A NT up 4 FAU Ubiquitin-like protein FUBI A NT up 4 eif3e-a Eukaryotic translation initiation factor 3 subunit E-A A NT up 4 rpsa 40S ribosomal protein SA A NT up 4 RPS14 40S ribosomal protein S14 A NT up 4 ATP5G2 ATP synthase F(0) complex subunit C2, mitochondrial A NT up 4 Rps11 40S ribosomal protein S11 A NT up 4 eif3g-b Eukaryotic translation initiation factor 3 subunit G-B A NT up 4 Tmem258 Transmembrane protein 258 A NT up 4 EIF3F Eukaryotic translation initiation factor 3 subunit F A NT up 4 rpl15 60S ribosomal protein L15 A NT up 4 MTMR9 Myotubularin-related protein 9 A NT up 4 rps20 40S ribosomal protein S20 A NT up 4 rpl22 60S ribosomal protein L22 A NT up 4 rpl8 60S ribosomal protein L8 A NT up 4 RPL36 60S ribosomal protein L36 A NT up 4

259 Eloc Elongin-C A NT up 4 tmem167a Protein kish-A A NT up 4 Rpl17 60S ribosomal protein L17 A NT up 4 RpL13 60S ribosomal protein L13 A NT up 4 rpl10a 60S ribosomal protein L10a A NT up 4 rps6 40S ribosomal protein S6 A NT up 4 RPL10 60S ribosomal protein L10 A NT up 4 Rpl26 60S ribosomal protein L26 A NT up 4 Rps9 40S ribosomal protein S9 A NT up 4 ak2 Adenylate kinase 2, mitochondrial A NT up 4 nap1l1 Nucleosome assembly protein 1-like 1 A NT up 4 -- -- A NT up 4 Rps14 40S ribosomal protein S14 A NT up 4 Rpl38 60S ribosomal protein L38 A NT up 4 Rack1 Receptor of activated protein C kinase 1 A NT up 4 rpl31 60S ribosomal protein L31 A NT up 4 mcts1-a Malignant T-cell-amplified sequence 1-A A NT up 4 Rps8 40S ribosomal protein S8 A NT up 4 PFDN1 Prefoldin subunit 1 A NT up 4 ATP5L ATP synthase subunit g, mitochondrial A NT up 4 eef1g-a Elongation factor 1-gamma-A A NT up 4 RPL13A 60S ribosomal protein L13a A NT up 4 rpl4-b 60S ribosomal protein L4-B A NT up 4 RSL24D1 Probable ribosome biogenesis protein RLP24 A NT up 4 rps23 40S ribosomal protein S23 A NT up 4 rps24 40S ribosomal protein S24 A NT up 4 pno1 RNA-binding protein PNO1 A NT up 4 RPL29 60S ribosomal protein L29 A NT up 4 ATP5C1 ATP synthase subunit gamma, mitochondrial A NT up 4 -- -- A NT up 4 eif3b Eukaryotic translation initiation factor 3 subunit B A NT up 4

260 Rpl23 60S ribosomal protein L23 A NT up 4 rpl19 60S ribosomal protein L19 A NT up 4 Lsm2 U6 snRNA-associated Sm-like protein LSm2 A NT up 4 Prdx5 Peroxiredoxin-5, mitochondrial A NT up 4 Btf3 Transcription factor BTF3 A NT up 4 Rps7 40S ribosomal protein S7 A NT up 4 mtap S-methyl-5'-thioadenosine phosphorylase A NT up 4 CCDC13 Coiled-coil domain-containing protein 13 A NT up 4 Rpl27 60S ribosomal protein L27 A NT up 4 RPL3 60S ribosomal protein L3 A NT up 4 smyd5 SET and MYND domain-containing protein 5 A NT up 4 RPLP0 60S acidic ribosomal protein P0 A NT up 4 RPL34 60S ribosomal protein L34 A NT up 4 Rpl14 60S ribosomal protein L14 A NT up 4 RPL9 60S ribosomal protein L9 A NT up 4 Rpl6 60S ribosomal protein L6 A NT up 4 TPT1 Translationally-controlled tumor protein homolog A NT up 4 SFT2D1 Vesicle transport protein SFT2A A NT up 4 ARPC1A Actin-related protein 2/3 complex subunit 1A A NT up 4 RPL35 60S ribosomal protein L35 A NT up 4 mapk14 Mitogen-activated protein kinase 14 A NT up 4 Rpl11 60S ribosomal protein L11 A NT up 4 EIF5A2 Eukaryotic translation initiation factor 5A-2 A NT up 4 Rps19 40S ribosomal protein S19 A NT up 4 eif2a Eukaryotic translation initiation factor 2A A NT up 4 eif3d Eukaryotic translation initiation factor 3 subunit D A NT up 4 eif3h Eukaryotic translation initiation factor 3 subunit H A NT up 4 Hspe1 10 kDa heat shock protein, mitochondrial A NT up 4 RPL21 60S ribosomal protein L21 A NT up 4 rps27a Ubiquitin-40S ribosomal protein S27a A NT up 4 rps3a 40S ribosomal protein S3a A NT up 4

261 Rps27 40S ribosomal protein S27 A NT up 4 RPL7 60S ribosomal protein L7 A NT up 4 Rpl32 60S ribosomal protein L32 A NT up 4 GFM2 Ribosome-releasing factor 2, mitochondrial A NT up 4 rpl7a 60S ribosomal protein L7a A NT up 4 eif3l Eukaryotic translation initiation factor 3 subunit L A NT up 4 rps2 40S ribosomal protein S2 A NT up 4 Rps3 40S ribosomal protein S3 A NT up 4 snrpf Small nuclear ribonucleoprotein F A NT up 4 Rpl12 60S ribosomal protein L12 A NT up 4 Gp9 Platelet glycoprotein IX B QLD down 1 hacd4 Very-long-chain (3R)-3-hydroxyacyl-CoA dehydratase B QLD down 1 ITGA2B Integrin alpha-IIb B QLD down 1 CYR61 Protein CYR61 B QLD down 1 PLEK Pleckstrin B QLD down 1 -- -- B QLD down 1 PXN1 Jeltraxin C QLD down 1 PXN1 Jeltraxin C QLD down 1 PXN1 Jeltraxin C QLD down 1

262 PUBLICATIONS

This section includes two publications that resulted from this thesis:

Selechnik D, Rollins LA, Brown GP, Kelehear C, Shine R. 2016. The things they carried: the pathogenic effects of old and new parasites following the intercontinental invasion of the Australian cane toad (Rhinella marina). International Journal for Parasitology: Parasites and Wildlife, 6(3): 375-385.

Selechnik D, West AJ, Brown GP, Fanson KV, Addison B, Rollins LA, Shine R. 2017. Effects of invasion history on physiological responses to immune system activation in invasive Australian cane toads. PeerJ, 5:e3856.

263 International Journal for Parasitology: Parasites and Wildlife 6 (2017) 375e385

Contents lists available at ScienceDirect International Journal for Parasitology: Parasites and Wildlife

journal homepage: www.elsevier.com/locate/ijppaw

Invited Review The things they carried: The pathogenic effects of old and new parasites following the intercontinental invasion of the Australian cane toad (Rhinella marina)

* D. Selechnik a, , L.A. Rollins b, G.P. Brown a, C. Kelehear c, R. Shine a a School of Life and Environmental Sciences (SOLES), University of Sydney, Sydney, NSW, 2006, Australia b Centre for Integrative Ecology, School of Life & Environmental Sciences (LES), Deakin University, Pigdons Road, Geelong, VIC, 3217, Australia c Smithsonian Tropical Research Institute, Apartado 0843-03092, Balboa, Ancon, Panama, Panama article info abstract

Article history: Brought to Australia in 1935 to control agricultural pests (from French Guiana, via Martinique, Barbados, Received 16 September 2016 Jamaica, Puerto Rico and Hawai'i), repeated stepwise translocations of small numbers of founders Received in revised form enabled the cane toad (Rhinella marina) to escape many parasites and pathogens from its native range. 17 December 2016 However, the infective organisms that survived the journey continue to affect the dynamics of the toad in Accepted 23 December 2016 its new environment. In Australia, the native-range lungworm Rhabdias pseudosphaerocephala decreases its host's cardiac capacity, as well as growth and survival, but not rate of dispersal. The lungworm is most Keywords: prevalent in long-colonised areas within the toads' Australian range, and absent from the invasion front. Rhabdias pseudosphaerocephala Ecoimmunology Several parasites and pathogens of Australian taxa have host-shifted to cane toads in Australia; for Invasion example, invasion-front toads are susceptible to spinal arthritis caused by the soil bacterium, Ochro- Enemy release hypothesis bactrum anthropi. The pentastome Raillietiella frenata has host-shifted to toads and may thereby expand Immune function its Australian range due to the continued range expansion of the invasive toads. Spill-over and spill-back Bufo of parasites may be detrimental to other host species; however, toads may also reduce parasite loads in Pathogen-mediated selection 1 native taxa by acting as terminal hosts. We review the impact of the toad's parasites and pathogens on the invasive anuran's biology in Australia, as well as collateral effects of toad-borne parasites and pathogens on other host species in Australia. Both novel and co-evolved pathogens and parasites may have played significant roles in shaping the rapid evolution of immune system responses in cane toads within their invaded range. Crown Copyright © 2016 Published by Elsevier Ltd on behalf of Australian Society for Parasitology. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4. 0/).

Contents

1. Introduction ...... 376 2. Impact of translocation to Australia on parasites and pathogens of the cane toad ...... 376 2.1. The enemy release hypothesis ...... 376 2.2. Protozoans ...... 377 2.3. Metazoans ...... 377 2.3.1. Lungworms ...... 378 2.4. Summary ...... 380 3. Impact of range expansion on parasites and pathogens of the cane toad in Australia ...... 380 3.1. Intermediate population disadvantage ...... 380 3.2. Host-parasite lag ...... 380 3.3. Costs of dispersal ...... 380 3.4. Flexibility in the immune system ...... 382

* Corresponding author. E-mail address: [email protected] (D. Selechnik). http://dx.doi.org/10.1016/j.ijppaw.2016.12.001 2213-2244/Crown Copyright © 2016 Published by Elsevier Ltd on behalf of Australian Society for Parasitology. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). 376 D. Selechnik et al. / International Journal for Parasitology: Parasites and Wildlife 6 (2017) 375e385

3.5. Immunogenetic comparisons across the range ...... 383 3.6. Range expansion of lungworms in cane toads ...... 383 4. Conclusion ...... 383 Acknowledgements: ...... 383 References ...... 384

1. Introduction range, thereby increasing infection rates of native species via spill- back mechanisms (Kelly et al., 2009b). On the other hand, invaders The native range of the cane toad (Rhinella marina)(Fig. 1)ex- may reduce rates of parasitism in native hosts by removing infec- tends from southern Texas and western Mexico to central Brazil tive stages of the parasite life cycle from the environment and (Zug and Zug, 1979; Acevedo et al., 2016). Cane toads were brought becoming parasite sinks (Heimpel et al., 2003; Kelly et al., 2009a; from Guyana (directly) and French Guiana (via Martinique) to Lettoof et al., 2013; Nelson et al., 2015a). In the current review, Barbados in the mid 1800s (to control pest beetles that were we summarise available information on the pathogens and para- consuming farmed sugarcane (Easteal, 1981; Turvey, 2013), then sites of cane toads in Australia (compared to those in the toads’ translocated from Barbados to Puerto Rico (some directly, and some native range), and the interaction between invasion dynamics and via Jamaica) in 1920 (Turvey, 2009). In 1932, 149 toads were pathogenic effects as immune function evolves in the toads. brought from Puerto Rico to Oahu, Hawai'i in order to control cane beetles (Turvey, 2009). Over the following two years, more than 100,000 individuals were distributed across the Hawai'ian Islands 2. Impact of translocation to Australia on parasites and (Turvey, 2009). In 1935, 101 Hawai'ian cane toads were brought to pathogens of the cane toad Queensland, Australia and bred in captivity; their offspring were later released along the Queensland coast (Turvey, 2009). 2.1. The enemy release hypothesis Cane toads rapidly spread across tropical and subtropical Australia (Urban et al., 2008), and have had major ecological im- When organisms are translocated to new areas, they may escape pacts on Australian native fauna (Shine, 2010). The chemically from co-evolved competitors, predators, parasites, and pathogens ‘ ’ distinctive toxins of R. marina (Hayes et al., 2009) are fatal if ( enemy release hypothesis, or ERH) (Colautti et al., 2004). The low ingested by many Australian predators (which lack a history of number of host individuals transferred to the introduced range evolutionary exposure to bufonid anurans, and thus to their toxins) diminishes the probability of native pathogens/parasites being (Llewelyn et al., 2009, 2014; Shine, 2010; Llewelyn et al., 2014). As a represented (Lewicki et al., 2014). Pathogens and parasites that do result, the continuing spread of cane toads has caused massive accompany the invasive host often face barriers to transmission declines in populations of anuran-eating predators both in tropical such as low host density (Arneberg et al., 1998; MacLeod et al., and temperate Australia (Letnic et al., 2008; Shine, 2010; Brown 2010; Blakeslee et al., 2012), and a lack of vectors or intermediate et al., 2011; Jolly et al., 2015; Jolly et al., 2016). hosts needed to complete their life cycles in the introduced range “ ” Although scientific studies on the ecological impact of cane (Blakeslee et al., 2012; Lewicki et al., 2014). For enemy release to toads in Australia have focused on lethal toxic ingestion as the be realized, several conditions must be met. First, co-evolved primary mechanism of impact, and have looked primarily at effects pathogens and parasites (specialized to the host in its native on top-order predators, other mechanisms of toad impact may be range) must be absent from the introduced range (Keane and important also. For example, invaders may carry with them para- Crawley, 2002; Prenter et al., 2004; Liu and Stiling, 2006). Second, sites from the native range that can severely affect other host host-switching of pathogens and parasites from native taxa in the species in the invaded range (Raffel et al., 2008). Also, invaders may introduced range to the invasive host should be uncommon (Keane act as additional competent hosts for parasites from the invaded and Crawley, 2002; Prenter et al., 2004; Liu and Stiling, 2006). Third, enemies in the introduced range should be less pathogenic to the invasive host than to native taxa (Keane and Crawley, 2002; Prenter et al., 2004; Liu and Stiling, 2006). If these conditions are met, enemy release may enable the invasive species to thrive in its introduced range (Colautti et al., 2004). The ERH has been sup- ported by many studies on plants (DeWalt et al., 2004; Blumenthal, 2006), but relatively few on animals (Torchin et al., 2002; Torchin, 2003; Roy et al., 2011). The translocation of the cane toad to Australia seems to conform to the ERH. Many pathogens (such as bacteria and fungi) from the toad's native range seem to have been lost along its journey to Australia (Speare, 1990), filtered by the stepwise invasion process. There are two possible exceptions to this, but both are uncertain. The first is a gram-negative bacillus which causes granulomatous lesions in the livers of toads in New South Wales (Speare, 1990). However, this strain of bacteria has not been definitively identified, so its origin remains uncertain (Daszak et al., 1999). The second is the fungus Mucor amphibiorum, which has been found in free- ranging toads throughout Queensland and can cause fatal septi- Fig. 1. Cane toad (Rhinella marina), a large bufonid anuran invasive to Australia. Photo caemia in anurans (Speare et al., 1994). This fungus has not been taken by Dr. Matt Greenlees. detected in native Australian anurans, suggesting that it may have D. Selechnik et al. / International Journal for Parasitology: Parasites and Wildlife 6 (2017) 375e385 377 been introduced with the cane toad (Speare et al., 1994). However, Table 2 the same fungus has been detected in the in Tasmania Species of novel protozoan parasites acquired by the Australian cane toad in Queensland. Prevalence of each parasite within the sample populations of tadpoles, (Speare et al., 1994), a part of Australia to which toads were not juveniles, and adults are reported as percentages. All data from Delvinquier and introduced. This suggests that the parasitic fungus may be endemic Freeland (1988b). to Australia (Speare et al., 1994). Regardless, the toads are suitable Species Bodily location Prevalence in Australian hosts for Mucor amphibiorum, and have the potential to amplify the populations(%) fungal parasite's numbers. Viruses seem to follow a similar pattern, although there are few Tadpoles Juveniles Adults studies documenting them in the toad throughout its worldwide Chilomastix caulleryi Intestine, cloaca 0 0 6 distribution (Speare, 1990). Ranavirus (family Iridoviridae) infects Retortamonas dobelli Intestine, cloaca 0 0 3 < fi Giardia agilis Intestine, cloaca 30 0 1 amphibians and sh (Hyatt et al., 2000). Although six species of Spironucleus elegans Intestine, cloaca 42 0 61 Ranavirus have been isolated in native-range cane toads in Monocercomonas batrachorum Intestine, cloaca 0 0 21 Venezuela (Hyatt et al., 2000), none have been found in Australian Protoopalina australis Intestine, cloaca 58 0 11 toads (Zupanovic et al., 1998). However, antibodies against Rana- Protoopalina hylarum Intestine, cloaca 0 0 <1 virus were detected in blood serum from both Venezuelan and Protoopalina raffae Intestine, cloaca 0 0 4 Nyctotheroides species Cloaca 63 0 27 Australian toads (albeit at low prevalence in Australia), suggesting Trichodina species Tadpole skin 92 0 0 some exposure to Ranavirus in Australia (Whittington et al., 1997; Zupanovic et al., 1998). The viruses encountered by Australian toads may originally have come from South America, or may have shown to exert pathogenic effects on their hosts (Delvinquier and host-shifted from Australian hosts (Zupanovic et al., 1998). Cane Freeland, 1988b), but infections by Zelleriella and Saccamoeba are toads may amplify and disseminate Ranavirus to other susceptible associated with greater susceptibility to other parasites (Freeland Australian anurans (Zupanovic et al., 1998). et al., 1986). Host-shifting of Zelleriella from cane toads to native anurans has been unsuccessful, as the parasite is only able to sur- 2.2. Protozoans vive for a short time within these unfamiliar hosts (Delvinquier and Freeland, 1988a). Within their native South American range, cane toads contain many species of endemic protozoans, reviewed by Delvinquier and Freeland (1988). To test the ERH in Australian cane toads, protozoan 2.3. Metazoans parasite load was surveyed in toads of multiple life stages across Queensland (the original site of introduction: Turvey, 2009) in the In Australia, the cane toad has also escaped from native-range 1980s, approximately 50 years after toads were introduced arthropod parasites. Although several species of ticks that infect (Delvinquier and Freeland, 1988b). Only three species out of cane toads in South America have been lost, local mites and approximately sixty documented South American parasites were mosquitoes utilize cane toads as hosts in Australia (Speare, 1990). found to infect Australian toads (Table 1), indicating that the ma- The same trend is observed in myxozoan parasites, though some jority of native-range pathogens had indeed been lost (Delvinquier South American parasites were thought to have been introduced to and Freeland, 1988b). Blood parasites were completely lost, Australia with the toad (Delvinquier, 1986). When Myxidium para- possibly due to their absence from the small founder population, or sites were detected in the gall bladders of both invasive Australian a lack of vectors in the introduced range (Delvinquier and Freeland, toads and native Australian anurans, the parasite was thought to be 1988b). Myxidium immersum (Delvinquier, 1986), which infects cane toads However, introduction to Australia has also exposed toads to in Brazil (Lutz, 1889). Inspection of anuran museum samples also new protozoan parasites, several of which have been acquired from revealed that no Myxidium were detected in native Australian an- native anurans (Delvinquier and Freeland, 1988b)(Table 2). When urans collected prior to the arrival of toads in Australia (Hartigan an invasive species is physiologically similar to native hosts, it has et al., 2010). However, museum specimens collected after the the potential to amplify parasite numbers (Barton, 1997). This toad introduction revealed that the parasite was found in native process, called spill-back (Kelly et al., 2009a, b), may consequently Australian anurans from areas that the toads had not yet invaded increase incidence of parasitism in native hosts (Lang and Benbow, (Hartigan et al., 2010). Subsequent phylogenetic analyses revealed 2013). Spill-back is a more common phenomenon than its converse, that there were actually two species of Myxidium parasites infecting spill-over, whereby invaders carry with them novel parasites Australian anurans, and both were distinct from the morphologi- (Torchin, 2003) that subsequently infect native hosts (Hartigan cally similar Myxidium immersum found in Brazilian toads (Hartigan et al., 2011). et al., 2011). Moreover, neither of the two Australian Myxidium Other protozoans found to infect Australian toads include Sac- species were found in Hawai'ian toads, further refuting the idea camoeba and flagellates (Diplomonadida and Trichomonadida) that they came to Australia with the cane toads (Hartigan et al., (Freeland et al., 1986). Most of these protozoans have not been 2011). Nonetheless, the toads may have amplified numbers of

Table 1 Species of native-range protozoan parasites retained by the Australian cane toad in Queensland. Prevalence of each parasite within the sample populations of tadpoles, ju- veniles, and adults are reported as percentages.

Species Bodily location Native countries Prevalence in Australian populations (%)

Tadpoles Juveniles Adults

Trichomitus batrachorum Cloaca Costa Rica, Colombia 22 0 77 Zelleriella antilliensis Intestine, Cloaca Jamaica, Bermuda, Brazil, Mexico, Costa Rica, Venezuela, Colombia, Fiji 75 100 38 Hyalodaktylethra renacuajo Cloaca Argentina 26 0 14

All data from Delvinquier and Freeland (1988b). 378 D. Selechnik et al. / International Journal for Parasitology: Parasites and Wildlife 6 (2017) 375e385 these Australian parasites, accounting for their lack of detection (Pizzatto and Shine, 2011c). Unable to eliminate traveling larvae before 1966 (Hartigan et al., 2011). once inside the body, the toads’ histiocytes nonetheless sometimes isolate the pathogen by forming granulomas, which resemble cysts 2.3.1. Lungworms (Pizzatto et al., 2010). Although helminths parasitize toads in South America (Speare, Rhabdias pseudosphaerocephala larvae reduce the survival, 1990; Campiao et al., 2014), only one species (Rhabdias pseudos- growth rate, locomotor activity, and feeding rate of metamorph phaerocephala) has been shown to persist in Australian populations cane toads (Kelehear et al., 2009). This could be through a number (Dubey and Shine, 2008). Initially, the nematode found in the lungs of proposed mechanisms, including the consumption of host of Australian cane toads was identified (based on morphometrics) erythrocytes, physical obstruction of the lung surfaces, and initia- as an endemic Australian species, Rhabdias cf. hylae (Barton, 1997). tion of energetically costly immune responses. Alternatively, these However, subsequent mitochondrial and nuclear genetic analyses impediments may be caused by the imposition of large L3 crawling identified the lungworm as the American taxon through tiny metamorph bodies, but the exact mechanism is still R. pseudosphaerocephala, indicating that this parasite had indeed not fully known (Kelehear et al., 2009). Metamorphs are particu- persisted through several serial translocations (Dubey and Shine, larly vulnerable due to their poor locomotor skills and immuno- 2008). Confusingly, R. pseudosphaerocephala has not been found compromised physiology during this period of development in Hawai'ian cane toads (Barton and Pichelin, 1999; Barton and (Kelehear et al., 2011a, b, c), but they are not the only vulnerable Riley, 2004; Marr et al., 2010), the source of the toads brought to life-stage (Kelehear et al., 2011a, b, c). Infection with Australia (Barton, 1994). The lungworm may have gone extinct in R. pseudosphaerocephala was also associated with reduced growth the Hawai'ian Islands prior to sampling. rates of adult toads, both in the wild and in captivity (Kelehear et al., Adult Rhabdias pseudosphaerocephala attach to the lung 2011a, b, c). Such impacts of parasites on their hosts are often epithelium of the host and consume erythrocytes (Colam, 1971; attributed to changes in host diet or the parasite's directly patho- Barton, 1996). The parasite's life cycle begins in the host's lungs, genic effects (Kelehear et al., 2011a, b, c). However, adult toads where adult worms lay eggs which are carried on a mucous ladder infected with lungworms did not change their feeding rates up the to the throat (Baker, 1979; Anderson, 2000). Eggs are (Kelehear et al., 2011 a, b, c) nor exhibit declining growth rates with then swallowed into the digestive system, and passed into the increasing intensity of infection (Kelehear et al., 2011a, b, c). If en- environment through host faeces (Pizzatto et al., 2010). Newly ergy depletion arising through erythrocyte consumption by hatched larvae escape from faeces into the soil, where they moult R. pseudosphaerocephala was responsible for decreasing growth several times and develop into free-living sexually reproducing rates, more parasites would be expected to produce greater growth adults (Baker, 1979; Anderson, 2000) over the span of 24e48 h reduction (Kelehear et al., 2011a, b, c). Because infection intensity (Kelehear et al., 2012a). Offspring develop within the free-living did not affect growth, the negative impact likely was caused by the mother and eventually consume her (4e10 days after toad defe- costs of mounting an immune response and the mechanical dam- cation: Kelehear et al., 2012a), and enter the environment as age associated with L3 migration (Kelehear et al., 2011a, b, c). infective third-stage larvae (L3) (Baker, 1979; Anderson, 2000). Rhabdias pseudosphaerocephala infection has other impacts on Once the L3 are able to locate a host, they burrow through its skin host physiology also. It is believed that the lungworm impedes the around the eye socket (Kelehear et al., 2011a, b, c). The larvae then process of blood oxygenation in its host (Pizzatto et al., 2012a,b). migrate through host tissues to reach the lungs, where they mature Reduction of cardiac capacity by the lungworm also has implica- into hermaphroditic adults and attach to the epithelia in order to tions for host behavior (Pizzatto et al., 2012a,b; Heise-Pavlov et al., reproduce and repeat the life cycle (Pizzatto et al., 2010). 2013). Although feeding rate is not influenced by infection status or The prevalence (percentage of hosts infected) and intensity intensity, diversity of prey items decreases with increasing para- (number of parasites per infected host) of R. pseudosphaerocephala sitism (Heise-Pavlov et al., 2013). Because foraging can be physically infections in Australian toads vary seasonally (Pizzatto et al., 2013). demanding, individuals with incapacitated cardiovascular systems During the wet season, when the soil is saturated with water, the are more limited in their pursuits. Prey items whose capture pose a mobility of larvae is limited due to their poor swimming abilities greater challenge are likely only attainable to uninfected toads (Pizzatto et al., 2013). Additionally, toads are less likely to aggregate (Heise-Pavlov et al., 2013). Nonetheless, if prey is plentiful, all toads during the wet season, as hydration and shelter sites are wide- may obtain roughly equal quantities of food (Heise-Pavlov et al., spread (Pizzatto et al., 2013). During the dry season, however, toads 2013). are forced to aggregate around receding sources of moisture, increasing their density and facilitating parasite transmission 2.3.1.1. Spill-over. Cane toads overlap with many Australian frogs in (Pizzatto et al., 2013). diet and shelter-site selection, creating opportunities for parasite L3 of R. pseudosphaerocephala commonly enter the toad host by transfer among host species. However, there are no reports of the burrowing through epidermis around the eye (Kelehear et al., lungworm R. pseudosphaerocephala infecting native anurans under 2011a, b, c). Infiltration of helminths through this location may natural conditions (Pizzatto et al., 2012a,b); instead, they are render toads susceptible to eye infections and neurological com- commonly infected by another helminth of the same genus, R. hylae plications, as these effects have been observed by rhabditid nem- (Pizzatto et al., 2010). Laboratory studies have shown that infective atodes in Asian horned frogs (Megophrys montana)(Imai et al., larvae of R. pseudosphaerocephala can penetrate the bodies of at 2009). Although toads attempt to dislodge larvae crawling on least some species of native frogs (Pizzatto et al., 2010), but most their skin (e.g. by kicking, tongue-flicking, and blinking), these may be dead-end hosts (the nematodes are not retained in the measures are largely ineffective (Kelehear et al., 2011a, b, c). lungs) (Pizzatto and Shine, 2011a). Frogs mounted faster immune Furthermore, toads do not actively avoid helminth larvae (Kelehear responses than did toads, indicating that the specialization of this et al., 2011a, b, c). Rather, they approach L3 larvae and attempt to helminth for its preferred host includes evasion strategies for its consume them (Kelehear et al., 2011a, b, c). physiological defences (Pizzatto et al., 2010). Among the seven Toads can also acquire R. pseudosphaerocephala through canni- species of native frogs tested in that study, none exhibited signifi- balism because larger toads often prey upon smaller ones (Pizzatto cant declines in growth, mobility, or survival when exposed to and Shine, 2011c). Within their new host, these lungworms are able R. pseudosphaerocephala (Pizzatto and Shine, 2011a). to survive, continue their life cycle, and reduce host mobility However, a follow-up study found that R. pseudosphaerocephala D. Selechnik et al. / International Journal for Parasitology: Parasites and Wildlife 6 (2017) 375e385 379 is deadly to at least one species of native frog (Litoria splendida) produced illness in toads, while R. hylae did not induce obvious (Pizzatto and Shine, 2011b), diminishing the hope that the parasite pathogenic effects in the frogs (Nelson et al., 2015b). Toads have could be used as a control agent for toads without compromising likely evolved a strong immune response which is also stimulated the safety of native anurans (Pizzatto and Shine, 2011b). Clearly, the by infection with R. hylae (Nelson et al., 2015b). lungworm's impact on frogs differs among species (Pizzatto and All of the parasites (except the native Australian helminths) Shine, 2011b). mentioned above utilize cane toads as a definitive host, meaning that they attain sexual maturity within the toad and thus do not 2.3.1.2. Spill-back. Some Australian helminths from native anurans depend upon any subsequent hosts (Hechinger and Lafferty, 2005). have capitalized on the toad's introduction (Table 3)(Freeland et al., However, some parasites require an intermediate host in which 1986). However, host-switching of parasites from native frogs to they pass through one or more asexual life stages before moving fi introduced toads appears to be rare at the original location of onto their de nitive host (Hechinger and Lafferty, 2005). In these introduction (Townsville area: Freeland et al., 1986; Dubey and systems, transmission between hosts frequently occurs through Shine, 2008), and native lungworms are not known to have host- predation (Hechinger and Lafferty, 2005). Toads are unlikely to be shifted into toads (Pizzatto et al., 2012a,b). Laboratory experi- intermediate hosts in Australia (Kelehear and Jones, 2010) because ments suggest that the immune systems of native frogs can they are eaten by relatively few species of Australian predators “ recognise and destroy larvae of R. pseudosphaerocephala (Nelson (Cabrera-Guzman et al., 2012, 2014, 2015). Toads are thus a dead- ” et al., 2015b). end for several parasites (Kelehear and Jones, 2010; Nelson et al., In general, gastrointestinal parasitism by Australian helminth 2015a). The decline of one native Australian proteocephalid tape- larvae is more prevalent in toads than in native anurans (Kelehear worm has been attributed to increases in toad densities (Freeland, and Jones, 2010), possibly because toads are larger than frogs, and 1994). infection rates tend to be positively correlated with body size (Kelehear and Jones, 2010). Toads may also feed upon a broader 2.3.1.3. Pentastomes. One parasite of cane toads is an arthropod range of prey items, increasing the risk of exposure to new parasites that is neither an Australian native nor brought by the toads. Rather, (Kelehear and Jones, 2010). Among Australian frogs, parasite loads the pentastome Raillietiella frenata was introduced to Australia at were higher in species that were larger or experienced more niche least 40 years ago inside the Asian house gecko (Hemidactylus fre- overlap with toads (Kelehear and Jones, 2010). Although the overall natus)(Kelehear et al., 2013). This pentastome has been detected in incidence of parasitism was higher in toads than in frogs, parasites both native anurans and toads in Australia (Kelehear et al., 2011a, b, in toads were frequently found encapsulated in cysts made up of c). As in the case of R. pseudosphaerocephala, the pathogenesis of toad immune cells, which potentially diminish parasite viability R. frenata includes the consumption of blood cells in the lungs (Kelehear and Jones, 2010). In contrast, coevolution between (Kelehear et al., 2012b). Accumulation of these pentastomes can Australian frogs and their parasites has allowed parasites to cause lung punctures, pneumonia, or blockage of respiratory air- become specialized for the physiological conditions of the available ways (Kelehear et al., 2014). Prevalence of pentastomes was found hosts’ bodies (Kelehear and Jones, 2010). to be higher in male toads than in females, possibly due to sex Also reflecting a lack of long-term coevolution, native helminths differences in microhabitat use or diet (Kelehear et al., 2012b). elicited stronger histological immune reactions in toads than they Toads of intermediate body size exhibited the highest prevalence of did in native frogs (Kelehear and Jones, 2010). The systemic arm of pentastome infection, suggesting that older individuals may the vertebrate immune system, which deals with newly encoun- develop adaptive immune defences against pentastomes (Kelehear tered pathogens, comprises the most inflammatory and stressful et al., 2012b). Although R. frenata infection is correlated with immune responses (Janeway et al., 2001). reduced fat stores in the Mediterranean house gecko (Hemidactylus Cross-infection experiments on toads and several species of turcicus), this trend was not observed in toads, suggesting minimal native frogs, with the toad lungworm (R. pseudosphaerocephala) energetic costs in this novel host (Kelehear et al., 2012b). There is and the frog lungworm (R. hylae), showed that each parasite was also a significant association between increasing pentastome in- more successful at reaching the target tissue in its respective tensity and declining metabolic rate in active house geckos, traditional host; toads exhibited superior resistance to the frog although the same association is not significant in resting house lungworm than did frogs to the toad lungworm (Nelson et al., geckos (Caballero et al., 2015). This relationship has not yet been 2015b). Mirroring earlier findings, R. pseudosphaerocephala tested in toads.

Table 3 Novel helminth parasites acquired by the Australian cane toad from native anurans.

Agent Bodily location Subgroup

Acanthocephalid cysts Not stated Acanthocephala Proteocephalid cysts Not stated Cestoda Maxvachonia flindersi Intestine Nematoda Parathelandros sp. Intestine Nematoda Nematodes (mainly Parathelandros) Intestine Nematoda Trematodes (mainly Mesocoelium and Lecithodendriidae) Intestine Trematoda Spirometra mansoni Small Intestine Cestoda Mesocoelium mesenibrinum Intestine Digenea Dolichosaccus symmetrus Intestine Digenea Dolichosaccus juvenilis Intestine Digenea Zeylanurotrema spearei Intestine Digenea Parathelandros mastigurus Intestine Nematoda Johnpearsonia pearsoni Intestine Nematoda Porrorchis hylae larvae Intestine Acanthocephala Neniatotaenia hylae Intestine Cestoda

All data from Freeland et al. (1986) and Barton (1997). 380 D. Selechnik et al. / International Journal for Parasitology: Parasites and Wildlife 6 (2017) 375e385

Previously, R. frenata was confined to a highly limited range in colonisation of an area. Parasite prevalence was low in the youngest Australia due to the restriction of the Asian house gecko to widely (Westmoreland Station, 2 years old) and oldest (Townsville, 47 separated urban areas (Kelehear et al., 2013). However, infected years old) populations, whereas intermediate-age populations toads could serve as an additional host that carries the pentastome (Burketown and Normanton, 4e19 years old) experienced greater across the toad's entire Australian range, allowing the parasite to parasitism (Freeland et al., 1986). Similarly, rates of parasitism by infect host species with which it did not previously share an R. pseudosphaerocephala were lower in the oldest and youngest overlapping range (Kelehear et al., 2013). populations across eight field sites along the expanding range (Brown et al., 2015a). Interestingly, toads from populations with 2.4. Summary higher parasite prevalence also had smaller spleens and fat bodies, suggesting immune and energetic costs associated with parasite Overall, most pathogens and parasites from the toads' native prevalence (Brown et al., 2015a). range have failed to remain with their hosts during their trans- Intermediate populations of a range-expanding host species location to Australia, presumably because of the multiple sequen- may experience higher energetic and immune costs of parasitism tial founder effects involved in that process (Fig. 2)(Easteal, 1981; (such as reduction in spleen size and fat bodies) because they have Barton, 1997). Not only were the numbers of founders small (e.g., not adapted to adequately suppress parasitic infection. All inter- 101 toads came from Hawai'i to Australia), but it was the progeny of mediate populations are at one time invasion “front” populations, those 101 toads that were released rather than the adult toads in which parasitic infection rates are lower because low host den- (Turvey, 2013). This precaution was taken to prevent the intro- sities make parasite transmission more difficult (Brown et al., duction of pathogens and parasites (Barton, 1997), but in fact the 2015a). However, as new toads arrive to the current front and initial adult toads were kept in a single large enclosure with their move farther westward, host densities (and thus parasite infection offspring (Turvey, 2013), thereby allowing parasite transfer be- rates) likely increase faster than the toads in the intermediate tween generations. populations can adapt. Meanwhile, lower parasitism at the range Intuitively, generalist parasites and those with direct life cycles core may occur through competitive exclusion of other parasites by are more likely to become established because they do not require those which are co-adapted with the host (Freeland et al., 1986). intermediate hosts (Lymbery et al., 2014). This bias may explain why macroparasites (e.g., helminths, arthropods) are generally 3.2. Host-parasite lag more successful than microparasites (e.g., viruses, bacteria, pro- tozoans) during invasion (Barton, 1997, 1999), although there are a Although 100% of Queensland toads surveyed were parasitised substantial number of cases in which parasites with indirect life by the lungworms, this proportion declined westward along the cycles successfully become established in other systems (Lymbery invasion transect, with only 60% in eastern Northern Territory and et al., 2014). Because only one helminth species was retained in 43% in the Darwin area (Dubey and Shine, 2008). At the edge of the the toads (Dubey and Shine, 2008), it is unsurprising that no native- invasion front, lungworms were absent (Phillips et al., 2010). The range viruses, pathogenic bacteria, or fungi have been documented lack of lungworm parasites in invasion-front populations of cane in Australian populations of the invader. toads is likely because host densities are too low for effective transmission (Phillips et al., 2010). 3. Impact of range expansion on parasites and pathogens of Nonetheless, radio-tracking of infected and uninfected toads the cane toad in Australia revealed that R. pseudosphaerocephala does not significantly reduce its host's dispersal rate in the wild; toads with lungworms actually Release from co-evolved predators, competitors, pathogens, and dispersed more rapidly than uninfected conspecifics (Brown et al., parasites may have enhanced the cane toad's ability to spread 2015a). These results are puzzling because R. pseudosphaerocephala through Australia. Rapid evolution of life-history traits and adversely affects the host's cardiovascular system (Pizzatto et al., dispersal ability has enabled toad populations to expand through 2012a,b). Additionally, by virtue of their pathogenic nature, the Queensland (Urban et al., 2008) and the Northern Territory larvae provoke immune responses which can reduce movement by (Covacevich and Archer, 1975), and into New South Wales (Easteal, depleting host energy stores. It seems that neither of these detri- 1981) and Western Australia (Rollins et al., 2015)(Fig. 3). As a result mental effects is strong enough to restrict adult toad mobility. of novel evolutionary forces, and potentially also of genetic drift, Why would infected toads disperse more rapidly? The lung- phenotypic characteristics have diverged between toads in worm might somehow influence the toad to move further (Brown Queensland and those on the invasion front in western regions et al., 2015a), or (more likely) toads that are inherently more mobile (Rollins et al., 2015). Compared to Queensland toads, western toads may be more likely to encounter lungworm larvae, or more sus- are larger in body size and relative size of the parotoid gland ceptible to infection due to immunocompromise associated with (Phillips and Shine, 2005), and have longer legs (Phillips et al., the strenuous dispersal process (Brown et al., 2015a). 2006). Behaviourally, invasion-front toads have more dispersive tendencies (Alford et al., 2009; Lindstrom et al., 2013). The 3.3. Costs of dispersal phenotypic differences between toads in the range core and those on the range edge appear to be due to a combination of natural The advantage of staying ahead of parasites through constant selection (higher fitness of faster dispersers) and spatial sorting (a movement does not come without drawbacks. Approximately 10% non-adaptive process whereby genes for rapid dispersal accumu- of large toads (>110 mm snout-urostyle length) on the invasion late at the range-edge because of interbreeding among the fastest- front are afflicted with spinal spondylosis (Brown et al., 2007). This dispersing animals in each generation (Shine et al., 2011; Perkins condition is caused by Ochrobactrum anthropi, a species of soil et al., 2013). bacteria in Australia that is otherwise only documented as rarely exerting pathogenic effects in immune-compromised humans 3.1. Intermediate population disadvantage (Brown et al., 2007). However, in frontal toads, the bacterium causes bony fusion of the synovial joints between spinal vertebrae, Surveys of different populations of cane toads in Australia pro- leading to arthritis (Brown et al., 2007). The frequency of infection vide evidence on how parasite loads change with time since is positively correlated with toad body size and movement rate .Slcnke l nentoa ora o aaiooy aaie n idie6(07 375 (2017) 6 Wildlife and Parasites Parasitology: for Journal International / al. et Selechnik D. e 385 Fig. 2. Phenomena occurring in pathogen/parasite load during the introduction of exotic host species. All of these concepts are exemplified by the invasive cane toad model. 381 382 D. Selechnik et al. / International Journal for Parasitology: Parasites and Wildlife 6 (2017) 375e385

Fig. 3. Known distribution of the cane toad throughout Australia. Since arriving in Queensland, Australia in 1935, cane toads have further expanded their range through New South Wales, the Northern Territory, and into Western Australia. Map created by Georgia Ward-Fear (Tingley et al., In review).

(Brown et al., 2007). Thus, this unusual affliction appears to reflect et al., 2010). These findings suggest heritable reduced investment the highly stressful lifestyle of dispersive toads and the phenotypic in the immune response to LPS by frontal toads (Llewelyn et al., traits (such as long legs and high levels of activity) responsible for 2010). their high dispersal rate (Brown et al., 2007). To further clarify divergences in immune function, toads were collected from a field site in the Northern Territory and radio- tracked to quantify dispersal rates (Brown and Shine, 2014). With 3.4. Flexibility in the immune system individuals differing in travel distances by up to 10-fold, immune assays were then conducted on the same toads (Brown and Shine, How have these shifts in parasite load shaped the toads them- 2014). Corticosterone levels were not correlated with distance, but selves? It is commonly proposed that individuals within range- more mobile toads exhibited reduced complement-driven bacteria- expanding populations will be under selection to reduce invest- killing and phagocytic capabilities, and enhanced phytohemag- ment in traits that do not directly benefit their ability to disperse glutinin (PHA)-induced skin-swelling relative to their sedentary (such as immune defences, because the body has limited energy counterparts (Brown and Shine, 2014). The activities of comple- stores: Lee and Klasing, 2004; Lee, 2006; Llewelyn et al., 2010). ment proteins and phagocytes are part of innate systemic immu- Nonetheless, down-regulation of immune capacity could imperil a nity (Ochsenbein and Zinkernagel, 2000), whereas PHA is a founding population as it encounters new threats. stimulant of cell-mediated action (Tella et al., 2008; Demas et al., In laboratory trials, toads with longer legs had a reduced stress 2011). These conflicting results require a more nuanced explana- response (Graham et al., 2012). This is further supported by a tion than that of dispersal causing an overall reduction of invest- positive linear relationship between population age and cortico- ment in immunity. Energy is likely reallocated into different sterone levels following a stressful stimulus (Brown et al., 2015b), branches of the immune system based upon individual utility and because toads at the invasion front tend to have longer hind legs cost (Brown and Shine, 2014); but we do not know if movement than do those further behind the invasion front (Phillips et al., patterns have influenced immune function rather than the reverse 2006). Lowered stress responses on the range edge may be adap- (Brown and Shine, 2014). tive, as over-reactions to common stressors would diminish rates of More recently, in an attempt to minimize environmental con- dispersal (Brown et al., 2015b). founds, toads from opposite ends of the species’ current Australian In response to injection with lipopolysaccharide (LPS, an endo- range were bred in a “common garden” setting. Offspring with toxin found in bacteria), captive-bred toads whose parents origi- parental origins from the invasion front displayed higher nated from close to the invasion front exhibited smaller increases in complement-driven bacteria-killing and phagocytic capabilities, metabolic rate than did those with parental origins from long- but no differences in PHA-induced skin-swelling (Brown et al., established populations (Llewelyn et al., 2010). Resting metabolic 2015c). These findings are discordant with the radio-tracking rate was approximately the same across populations (Llewelyn D. Selechnik et al. / International Journal for Parasitology: Parasites and Wildlife 6 (2017) 375e385 383 experiment. Frontal toads are more dispersive, yet mobility was 2012a). Infection intensity, however, did not vary between pop- negatively correlated with the enhanced systemic immune re- ulations (Kelehear et al., 2012a). sponses that they exhibit (Brown et al., 2015c). One plausible The possibility of lungworms and pentastomes expanding their explanation is that the captive-bred toads in the common garden Australian ranges via the toad raises conservation concerns as toads experiment had not undergone the stress imposed by long- move farther westward and southward. Western Australia is home distance movements (Brown et al., 2015c). Thus, the systemic to many endemic species of frogs (Aplin and Smith, 2001), some of component may evolve to be up-regulated in frontal toads because which are threatened (Hero and Roberts, 2004; Roberts and Hero, it will be heavily exhausted during their lifetimes due to their 2004). Because it is difficult to predict how successful these para- arduous lifestyles (Brown et al., 2015c). sites are at infecting novel anuran hosts, the impact that they will have on endemic or endangered frog populations in Western 3.5. Immunogenetic comparisons across the range Australia is unknown. Studies similar to those described in section 2.3.1.1, which examine the ability of lungworms or pentastomes to Despite the phenotypic differences observable between infect different frog species, would be a useful place to begin Queensland and Western Australian toads, which are suggestive of assessing potential damage. However, such studies would only be rapid evolution, genetic diversity is low in all of these invasive predictive of the frog species that are tested. populations (Rollins et al., 2015). The major histocompatibility complex (MHC) is a family of genes encoding glycoproteins which 4. Conclusion function to present foreign antigens that alert other immune ef- fectors (Benacerraf, 1980). MHC class I products primarily target The “enemy release” hypothesis asserts that invasive species viruses that have infiltrated host cells, translocating viral peptides thrive because they escape from most of the co-evolved threats that to the infected cell's surface (Fabre, 1991). This process triggers they have faced in their native range, and face less threat from destruction of that entire cell by cytotoxic T-cells or natural-killer species that are indigenous to the introduced range. Many surveys cells (Fabre, 1991). MHC class II products affect extracellular path- of parasites have been conducted in cane toads around the world, ogens such as bacteria, which can become engulfed by host and only one helminth and three protozoans have been docu- phagocytes (Ting and Baldwin, 1993). With the help of class II MHC mented to persist all the way through to Australia. Although the glycoproteins, macrophages can display the antigens of their protozoans exhibit little pathogenicity, the exotic helminth (Rhab- ingested targets, signalling for helper T-cells to assist them (Ting dias pseudosphaerocephala) significantly reduces toad growth and and Baldwin, 1993). survival. Although it has yet to be shown to infect native anurans in In the course of the toad's range expansion from Queensland to the wild, laboratory studies have indicated that Rhabdias pseudos- Western Australia, MHC class I has lost its remaining allelic di- phaerocephala can infiltrate the bodies of native anurans, causing versity on the invasion front, likely owing to genetic drift rather pathogenesis to widely varying extents in different species. than balancing selection (Lillie et al., 2014). Such relaxed selection Meanwhile, Australian lungworm parasites can also infiltrate toad could be due to the lack of viral challenges (Lillie et al., 2014). MHC bodies, but do not complete their life cycle due to encapsulation of class II diversity, however, has been maintained on the invasion larvae by toad immune defences. These advantages may have aided front (Lillie et al., 2016). In this case, selection may be maintained by the successful establishment and massive range expansion seen in the plethora of extracellular pathogens or parasites (such as the cane toad in Australia. O. anthropi and R. pseudosphaerocephala) at or near the invasion The cane toad invasion of Australia provides many opportunities front (Lillie et al., 2016). Both MHC classes contain very low allelic for study of the dynamics between introduced hosts, native hosts, variation relative to that typically seen in MHC, archetypally introduced parasites, and native parasites. The rapid evolution of demonstrating the effects of bottlenecks (Nei et al., 1975). multiple phenotypic traits in Australian cane toads, apparently in Low levels of genetic variation in Australian toads also manifest response to evolutionary pressures on dispersal rate, also facilitate in their low microsatellite diversity (Leblois et al., 2000), as well as the exploration of how density-dependent disease transmission is their lack of variation in mitochondrial haplotypes (Slade and affected by dispersal. Moritz, 1998). Because of these circumstances, it is unclear Phenotypically, a host's immune defences are moulded by the whether phenotypic differences among toads from different pop- abundance and diversity of pathogens and parasites, as well as their ulations are underpinned by genetic variation, or if heritable infectivity. The invasive cane toad has demonstrated significant epigenetic variation may play a role instead. flexibility in its immune system within 80 years, indicating that rapid evolution has indeed taken place. Such rapid evolution seems 3.6. Range expansion of lungworms in cane toads paradoxical given low levels of genetic diversity in Australian cane toads, resulting from sequential introduction-imposed bottlenecks Toads are not the only organisms to have responded to the se- followed by expansion-driven drift. This situation suggests that lection pressures (and other evolutionary forces such as spatial more is at play than simply genetic variation. One logical next step sorting: Shine et al., 2011) imposed by their dispersal across is to investigate the role of epigenetic changes in driving rapid Australia. Comparisons of R. pseudosphaerocephala on opposite evolution of the cane toad's immune system. Such an approach may ends of the toads’ Australian range revealed life-history differences clarify the mechanisms by which the toad has thrived within its associated with (and putatively driven by) variations in toad pop- new home, and the nature of selective pressures imposed by en- ulation density (Kelehear et al., 2012a). At free-living life stages (i.e., emies from the past and present. between successive hosts), worms closer to the invasion front had larger body sizes, increasing their chances of survival before their Acknowledgements: next toad encounter. That change may offset the density-imposed diminished likelihood of encountering a new host (Kelehear et al., This work was supported by the Australian Research Council 2012a). These lungworm populations also exhibited faster devel- [FL120100074, DE150101393], the Equity Trustees Charitable opment and higher survival to adulthood (Kelehear et al., 2012a). Foundation [Holsworth Wildlife Research Endowment], and the Although range-edge worms laid small numbers of large eggs, Smithsonian Office of Fellowships and Internships [George E. Burch range-core worms laid large numbers of small eggs (Kelehear et al., Postdoctoral Fellowship]. We thank Georgia Ward-Fear for 384 D. Selechnik et al. / International Journal for Parasitology: Parasites and Wildlife 6 (2017) 375e385 providing an updated map of the cane toad distribution. (Protozoa: Opalinata) from the cane toad Bufo marinus in Australia. Aust. J. Ecol. 36, 317e333. Delvinquier, B.L.J., Freeland, W.J., 1988b. Protozoan parasites of the cane toad, Bufo References marinus, in Australia. Aust. J. Zool. 36, 301e316. Demas, G.E., Zysling, D.A., Beechler, B.R., Muehlenbein, M.P., French, S.S., 2011. Acevedo, A.A., Lampo, M., Cipriani, R., 2016. The cane or marine toad, Rhinella Beyond phytohaemagglutinin: assessing vertebrate immune function across marina (Anura, Bufonidae): two genetically and morphologically distinct spe- ecological contexts. J. Anim. Ecol. 80 (4), 710e730. cies. Zootaxa 4103 (6), 574e586. DeWalt, S.J., Denslow, J.S., Ickes, K., 2004. Natural-enemy release facilitates habitat Alford, R.A., Brown, G.P., Schwarzkopf, L., Phillips, B.L., Shine, R., 2009. Comparisons expansion of the invasive tropical shrub Clidemia hirta. Ecology 85 (2), 471e483. through time and space suggest rapid evolution of dispersal behaviour in an Dubey, S., Shine, R., 2008. Origin of the parasites of an invading species, the invasive species. Wildl. Res. 36, 23e28. Australian cane toad (Bufo marinus): are the lungworms Australian or Amer- Anderson, R.C., 2000. Nematode Parasites of Vertebrates: Their Development and ican? Mol. Ecol. 17 (20), 4418e4424. Transmission. CABI Publishing, New York, NY. Easteal, S., 1981. The history of introductions of Bufo marinus; a natural experiment Aplink, K.P., Smith, L.A., 2001. Checklist of the frogs and reptiles of Western in evolution. Biol. J. Linn. Soc. 16 (16), 93. Australia. Rec. W. Aus. Mus. Supp. 63, 51e74. Fabre, J.W., 1991. Regulation of MHC expression. Immunol. Lett. 29, 3e8. Arneberg, P., Skorping, A., Grenfell, B., Read, A.F., 1998. Host densities. Proc. R. Soc. Freeland, W.J., 1994. Parasites, pathogens and the impacts of introduced organisms Lond. 265, 1283e1289. on the balance of nature in Australia. In: Moritz, C., Kikkawa, J. (Eds.), Conser- Baker, M.R., 1979. The free-living and parasitic development of Rhabdias spp. vation Biology in Australia and Oceania. Surrey Beatty and Sons, Sydney, (Nematoda: rhabdiasidae) in amphibians. Can. J. Zool. 57, 161e178. Australia, pp. 171e180. Barton, D., 1996. The cane toad: a new host for helminth parasites in Australia. Aust. Freeland, W.J., Delvinquier, B.L.J., Bonnin, B., 1986. Food and parasitism of the cane J. Ecol. 21, 114e117. toad, Bufo marinus, in relation to time since colonization. Aust. Wildl. Res. 13, Barton, D., 1997. Introduced animals and their parasites: the cane toad, Bufo mar- 489e499. inus, in Australia. Aust. J. Ecol. 22, 316e324. Graham, S.P., Kelehear, C., Brown, G.P., Shine, R., 2012. Corticosterone-immune in- Barton, D., 1999. Ecology of helminth communities in tropical Australian amphib- teractions during captive stress in invading Australian cane toads (Rhinella ians. Int. J. Parasitol. 29, 921e926. marina). Horm. Behav. 62 (2), 146e153. Barton, D.P., 1994. A checklist of helminth parasites of Australian amphibia. Rec. S. Hartigan, A., Phalen, D.N., Slapeta, J., 2010. Museum material reveals a frog parasite Aust. Mus. 27, 13e30. emergence after the invasion of the cane toad in Australia. Parasites Vectors 3, Barton, D.P., Riley, J., 2004. Raillietiella indica (Pentastomida) from the lungs of the 50. giant toad, Bufo marinus (Amphibia) in Hawaii, USA. Comp. Parasitol. 71, Hartigan, A., Fiala, I., Dykova, I., Jirku, M., Okimoto, B., Rose, K., Phalen, D.N., 251e254. Slapeta, J., 2011. A suspected parasite spill-back of two novel Myxidium spp. Barton, D.P., Pichelin, W., 1999. Acanthocephalus bufonis (Acanthocephala) from Bufo (Myxosporea) causing disease in Australian endemic frogs found in the invasive marinus (Bufonidae: Amphibia) in Hawaii. Parasite 6, 269e272. Cane toad. PLoS One 6 (4), e18871. Benacerraf, B., 1980. Role of MHC gene products in immune regulation. Science 212 Hayes, R.A., Crossland, M.R., Hagman, M., Capon, R.J., Shine, R., 2009. Ontogenetic (4500), 1229e1238. variation in the chemical defenses of cane toads (Bufo marinus): toxin profiles Blakeslee, A.M.H., Altman, I., Miller, A.W., Byers, J.E., Hamer, C.E., Ruiz, G.M., 2012. and effects on predators. J. Chem. Ecol. 35 (4), 391e399. Parasites and invasions: a biogeographic examination of parasites and hosts in Hechinger, R.F., Lafferty, K.D., 2005. Host diversity begets parasite diversity: bird native and introduced ranges. J. Biogeogr. 39 (3), 609e622. final hosts and trematodes in snail intermediate hosts. Proc. Biol. Sci. 272 Blumenthal, D.M., 2006. Interactions between resource availability and enemy (1567), 1059e1066. release in plant invasion. Ecol. Lett. 9 (7), 887e895. Heimpel, G.E., Neuhauser, C., Hoogendoorn, M., 2003. Effects of parasitoid fecundity Brown, G.P., Kelehear, C., Pizzatto, L., Shine, R., 2015a. The impact of lungworm and host resistance on indirect interactions among hosts sharing a parasitoid. parasites on rates of dispersal of their anuran host, the invasive cane toad. Biol. Ecol. Lett. 6, 555e566. Inv. 18 (1), 103e114. Heise-Pavlov, S.R., Paleologo, K., Glenny, W., 2013. Effect of Rhabdias pseudos- Brown, G.P., Kelehear, C., Shilton, C.M., Phillips, B.L., Shine, R., 2015b. Stress and phaerocephala on prey consumption of free-ranging cane toads (Rhinella immunity at the invasion front: a comparison across cane toad (Rhinella marina) marina) during Australian tropical wet seasons. J. Pest Sci. 87 (1), 89e97. populations. Biol. J. Linn. Soc. 116 (4), 748e760. Hero, J.M., Roberts, D., 2004. Geocrinia alba. IUCN Red List Threat. Species Brown, G.P., Phillips, B.L., Dubey, S., Shine, R., 2015c. Invader immunology: invasion e.T9031A12952112. history alters immune system function in cane toads (Rhinella marina)in Hyatt, A.D., Gould, A.R., Zupanovic, Z., Cunningham, A.A., Hengstberger, S., tropical Australia. Ecol. Lett. 18 (1), 57e65. Whittington, R.J., Kattenbelt, J., Coupar, B.E.H., 2000. Comparative studies of Brown, G.P., Phillips, B.L., Shine, R., 2011. The ecological impact of invasive cane piscine and amphibian iridoviruses. Arch. Virol. 145 (2), 301e331. toads on tropical snakes: field data do not support laboratory-based pre- Imai, D.M., Nadler, S.A., Brenner, D., Donovan, T.A., Pessier, A.P., 2009. Rhabditid dictions. Ecology 92 (2), 422e431. nematode-associated ophthalmitis and meningoencephalomyelitis in captive Brown, G.P., Shilton, C., Phillips, B.L., Shine, R., 2007. Invasion, stress, and spinal Asian horned frogs (Megophrys Montana). J. Vet. Diagn. Investig. 21, 568e573. arthritis in cane toads. PNAS U. S. A. 104, 17698e17700. Janeway, C.A.J., Travers, P., Walport, M., 2001. Immunobiology: the Immune System Brown, G.P., Shine, R., 2014. Immune response varies with rate of dispersal in in Health and Disease, vol. 5. Garland Science, New York. invasive cane toads (Rhinella marina). PLoS One 9 (6), 1e11. Jolly, C.J., Shine, R., Greenlees, M.J., 2015. The impact of invasive cane toads on native Caballero, I.C., Sakla, A.J., Detwiler, J.T., Le Gall, M., Behmer, S.T., Criscione, C.D., 2015. wildlife in southern Australia.” Ecol. Evol. 5 (18), 3879e3894. Physiological status drives metabolic rate in Mediterranean geckos infected Jolly, C.J., Shine, R., Greenlees, M.J., 2016. The impacts of a toxic invasive prey species with pentastomes. PLoS One 10 (12), e0144477. (the cane toad, Rhinella marina) on a vulnerable predator (the lace monitor, Cabrera-Guzman, E., Crossland, M.R., Pearson, D., Webb, J.K., Shine, R., 2014. Pre- Varanus varius). Biol. Inv. 18 (5), 1499e1509. dation on invasive cane toads (Rhinella marina) by native Australian rodents. Keane, R.M., Crawley, M.J., 2002. Exotic plant invasions and the enemy release J. Pest Sci. 88 (1), 143e153. hypothesis. Trends Ecol. Evol. 17 (4), 164e170. Cabrera-Guzman, E., Crossland, M.R., Shine, R., 2012. Predation on the eggs and Kelehear, C., Brown, G.P., Shine, R., 2011a. Influence of lung parasites on the growth larvae of invasive cane toads (Rhinella marina) by native aquatic invertebrates in rates of free-ranging and captive adult cane toads. Oecologia 165 (3), 585e592. tropical Australia. Biol. Conserv. 153, 1e9. Kelehear, C., Brown, G.P., Shine, R., 2012a. Rapid evolution of parasite life history Cabrera-Guzman, E., Crossland, M.R., Shine, R., 2015. Invasive cane toads as prey for traits on an expanding range-edge. Ecol. Lett. 15 (4), 329e337. native arthropod predators in tropical Australia. Herpetol. Monogr. 29 (1), Kelehear, C., Brown, G.P., Shine, R., 2012b. Size and sex matter: infection dynamics 28e39. of an invading parasite (the pentastome Raillietiella frenatus) in an invading Campiao, K.M., Morais, D.H., Dias, O.T., Aguiar, A., Toledo, G.M., Tavares, L.E., Da host (the cane toad Rhinella marina). Parasitology 139 (12), 1596e1604. Silva, R.J., 2014. Checklist of helminth parasites of Amphibians from south Kelehear, C., Brown, G.P., Shine, R., 2013. Invasive parasites in multiple invasive America. Zootaxa 3843, 1e93. hosts: the arrival of a new host revives a stalled prior parasite invasion. Oikos Colam, B., 1971. Studies on gut ultrastructure and digestive physiology in Rhabdias 122 (9), 1317e1324. bufonis and R. sphaerocephala (Nematoda: rhabditita). Parasitology 62, Kelehear, C., Jones, H.I., 2010. Nematode larvae (order Spirurida) in gastric tissues of 247e258. Australian anurans: a comparison between the introduced cane toad and Colautti, R.I., Ricciardi, A., Grigorovich, I.A., MacIsaac, H.J., 2004. Is invasion success sympatric native frogs. J. Wildl. Dis. 46 (4), 1126e1140. explained by the enemy release hypothesis? Ecol. Lett. 7 (8), 721e733. Kelehear, C., Spratt, D.M., Dubey, S., Brown, G.P., Shine, R., 2011b. Using combined Covacevich, J., Archer, M., 1975. The distribution of the Cane Toad, Bufo marinus,in morphological, allometric and molecular approaches to identify species of the Australia and its effects on indigenous vertebrates. Mem. Qld. Mus. 17 (2), genus Raillietiella (Pentastomida). PLoS One 6 (9), e24936. 305e310. Kelehear, C., Spratt, D.M., O'Meally, D., Shine, R., 2014. Pentastomids of wild snakes Daszak, P., Berger, L., Cunningham, A.A., Hyatt, A.D., Green, D.E., Speare, R., 1999. in the Australian tropics. Int. J. Parasitol. Parasites Wildl. 3 (1), 20e31. Emerging infectious diseases and Amphibian population declines. Emerg. Kelehear, C., Webb, J.K., Hagman, M., Shine, R., 2011c. Interactions between infective Infect. Dis. 5 (6), 735e748. helminth larvae and their anuran hosts. Herpetologia 67 (4), 378e385. Delvinquier, B.L.J., 1986. Myxidium immersum (Protozoa: Myxosporea) of the cane Kelehear, C., Webb, J.K., Shine, R., 2009. Rhabdias pseudosphaerocephala infection in toad, Bufo marinus, in Australian Anura, with synopsis of the genus in Am- Bufo marinus: lung nematodes reduce viability of metamorph cane toads. phibians. Aust. J. Ecol. 34, 843e853. Parasitology 136 (8), 919e927. Delvinquier, B.L.J., Freeland, W.J., 1988a. Observations on Zelleriella antilliensis Kelly, D.W., Paterson, R.A., Townsend, C.R., Poulin, R., Tompkins, D.M., 2009a. Has D. Selechnik et al. / International Journal for Parasitology: Parasites and Wildlife 6 (2017) 375e385 385

the introduction of brown trout altered disease patterns in native New Zealand Phillips, B.L., Kelehear, C., Pizzatto, L., 2010. Parasites and pathogens lag behind their fish? Freshw. Biol. 54, 1805e1818. host during periods of host range advance. Ecology 91, 872e881. Kelly, D.W., Paterson, R.A., Townsend, C.R., Poulin, R., Tompkins, D.M., 2009b. Phillips, B.L., Shine, R., 2005. The morphology, and hence impact, of an invasive Parasite spillback: a neglected concept in invasion ecology? Ecology 90 (8), species (the cane toad, Bufo marinus): changes with time since colonisation. 2047e2056. Anim. Conserv. 8, 407e413. Lang, J.M., Benbow, M.E., 2013. Species interactions and competition. Nat. Edu. 4 (4), Pizzatto, L., Kelehear, C., Dubey, S., Barton, D., Shine, R., 2012a. Host-parasite re- 8. lationships during a biologic invasion: 75 years postinvasion, cane toads and Leblois, R., Rousset, F., Tikel, D., Moritz, C., Estoup, A., 2000. Absence of evidence for sympatric Australian frogs retain separate lungworm faunas. J. Wildl. Dis. 48 isolation by distance in an expanding cane toad (Bufo marinus) population: an (4), 951e961. individual-based analysis of microsatellite genotypes. Mol. Ecol. 9, 1905e1909. Pizzatto, L., Kelehear, C., Shine, R., 2013. Seasonal dynamics of the lungworm, Lee, K.A., 2006. Linking immune defenses and life history at the levels of the in- Rhabdias pseudosphaerocephala, in recently colonised cane toad (Rhinella dividual and the species. Integr. Comp. Biol. 46 (6), 1000e1015. marina) populations in tropical Australia. Int. J. Parasitol. 43 (9), 753e761. Lee, K.A., Klasing, K.C., 2004. A role for immunology in invasion biology. Trends. Pizzatto, L., Shilton, C.M., Shine, R., 2010. Infection dynamics of the lungworm Ecol. Evol. 19 (10), 523e529. Rhabdias pseudosphaerocephala in its natural host, the cane toad (Bufo marinus), Letnic, M., Webb, J., Shine, R., 2008. Invasive cane toads (Bufo marinus) cause mass and in novel hosts (native Australian frogs). J. Wildl. Dis. 46 (4), 1152e1164. mortality of freshwater crocodiles (Crocodylus johnstoni) in tropical Australia. Pizzatto, L., Shine, R., 2011a. Ecological impacts of invading species: do parasites of Biol. Conserv. 141 (7), 1773e1782. the cane toad imperil Australian frogs? Aust. Ecol. 36 (8), 954e963. Lettoof, D.C., Greenlees, M.J., Stockwell, M., Shine, R., 2013. Do invasive cane toads Pizzatto, L., Shine, R., 2011b. The effects of experimentally infecting Australian tree affect the parasite burdens of native Australian frogs? Int. J. Parasitol. Parasites frogs with lungworms (Rhabdias pseudosphaerocephala) from invasive cane Wildl. 2, 155e164. toads. Int. J. Parasitol. 41 (9), 943e949. Lewicki, K.E., Huyvaert, K.P., Piaggio, A.J., Diller, L.V., Franklin, A.B., 2014. Effects of Pizzatto, L., Shine, R., 2011c. You are what you eat: parasite transfer in Cannibalistic barred owl (Strix varia) range expansion on Haemoproteus parasite assemblage cane toads. Herpetologia 67, 118e123. dynamics and transmission in barred and northern spotted owls (Strix occi- Pizzatto, L., Shine, R., Bennett, N., 2012b. Lungworm infection modifies cardiac dentalis caurina). Biol. Inv. 17 (6), 1713e1727. response to exercise in cane toads. J. Zool. 287 (2), 150e155. Lillie, M., Cui, J., Shine, R., Belov, K., 2016. Molecular characterization of MHC class II Prenter, J., Macneil, C., Dick, J.T., Dunn, A.M., 2004. Roles of parasites in animal in- in the Australian invasive cane toad reveals multiple splice variants. vasions. Trends Ecol. Evol. 19 (7), 385e390. Immunogenetics. Raffel, T.R., Martin, L.B., Rohr, J.R., 2008. Parasites as predators: unifying natural Lillie, M., Shine, R., Belov, K., 2014. Characterisation of major histocompatibility enemy ecology. Trends Ecol. Evol. 23 (11), 610e618. complex class I in the Australian cane toad, Rhinella marina. PLoS One 9 (8), Roberts, D., Hero, J.M., 2004. Geocrinia Vitellina. In: The IUCN Red List of Threatened e102824. Species e.T9032A12952365. Lindstrom, T., Brown, G.P., Sisson, S.A., Phillips, B.L., Shine, R., 2013. Rapid shifts in Rollins, L.A., Richardson, M.F., Shine, R., 2015. A genetic perspective on rapid evo- dispersal behavior on an expanding range edge. PNAS U. S. A. 110 (33), lution in cane toads (Rhinella marina). Mol. Ecol. 24 (9), 2264e2276. 13452e13456. Roy, H.E., Lawson Handley, L.J., Schonrogge,€ K., Poland, R.L., Purse, B.V., 2011. Can the Liu, H., Stiling, P., 2006. Testing the enemy release hypothesis: a review and meta- enemy release hypothesis explain the success of invasive alien predators and analysis. Biol. Inv. 8 (7), 1535e1545. parasitoids? BioControl 56 (4), 451e468. Llewelyn, J., Phillips, B.L., Alford, R.A., Schwarzkopf, L., Shine, R., 2010. Locomotor Shine, R., 2010. The ecological impact of invasive cane toads (Bufo marinus)in performance in an invasive species: cane toads from the invasion front have Australia. Q. Rev. Biol. 85 (3), 253e291. greater endurance, but not speed, compared to conspecifics from a long- Shine, R., Brown, G.P., Phillips, B.L., 2011. An evolutionary process that assembles colonised area. Oecologia 162 (2), 343e348. phenotypes through space rather than through time. PNAS U. S. A. 108, Llewelyn, J., Schwarzkopf, L., Alford, R., Shine, R., 2009. Something different for 5708e5711. dinner? Responses of a native Australian predator (the keelback snake) to an Slade, R.W., Moritz, C., 1998. Phylogeography of Bufo marinus from its natural and invasive prey species (the cane toad). Biol. Inv. 12 (5), 1045e1051. introduced ranges. P. Roy. Soc. Lond. B. Biol. 265, 769. Llewelyn, J., Schwarzkopf, L., Phillips, B.L., Shine, R., 2014. After the crash: how do Speare, R., 1990. A review of the diseases of the cane toad, Bufo marinus, with predators adjust following the invasion of a novel toxic prey type? Aust. Ecol. 39 comments on biological-control. Aust. Wildl. Res. 17, 387e410. (2), 190e197. Speare, R., Thomas, A.D., O'Shea, P., Shipton, W.A., 1994. Mucor amphibiorum in the Lutz, A., 1889. Ueber ein Myxosporidium aus der Gallenblase brasilianischer toad, Bufo marinus, in Australia. J. Wildl. Dis. 30 (3), 399e407. Batrachier. Zentralblatt fur Bakteriol. Parasitenkd. 5, 84e88. Tella, J.L., Lemus, J.A., Carrete, M., Blanco, G., 2008. The PHA test reflects acquired T- Lymbery, A.J., Morine, M., Kanani, H.G., Beatty, S.J., Morgan, D.L., 2014. Co-invaders: cell mediated immunocompetence in birds. PLoS One 3 (9), e3295. the effects of alien parasites on native hosts. Int. J. Parasitol. Parasites Wildl. 3 Ting, J.P., Baldwin, A.S., 1993. Regulation of MHC gene expression. Curr. Opin. (2), 171e177. Immunol. 5 (1), 8e16. MacLeod, C.J., Paterson, A.M., Tompkins, D.M., Duncan, R.P., 2010. Parasites lost - do Tingley, R., G. Ward-Fear, L. Schwarzkopf, M.J. Greenlees, B.L. Phillips, G.P. Brown, S. invaders miss the boat or drown on arrival? Ecol. Lett. 13 (4), 516e527. Clulow, J. Webb, R. Capon, A. Sheppard, T. Strive, M. Tizard, and R. Shine. New Marr, S.R., Johnson, S.A., Hara, A.H., McGarrity, M.E., 2010. Preliminary evaluation of weapons in the Toad Toolkit: a review of methods to control and mitigate the the potential of the helminth parasite Rhabdias elegans as a biological control biodiversity impacts of invasive cane toads (Rhinella marina).” (In review). agent for invasive Puerto Rican coquís (Eleutherodactylus coqui) in Hawaii. Biol. Torchin, M.E., Lafferty, K.D., Kuris, A.M., 2002. Parasites and marine invasions. Control 54 (2), 69e74. Parasitology 124 (7), 137e151. Nei, M., Maruyama, T., Chakraborty, R., 1975. The bottleneck effect and genetic Torchin, M. E. e. a, 2003. Introduced species and their missing parasites. Nature 421, variability in populations. Evolution 29, 1e10. 628e630. Nelson, F.B., Brown, G.P., Shilton, C., Shine, R., 2015a. Helpful invaders: can cane Turvey, N., 2009. A Toad's Tale Hot Topics from the Tropics, vol. 1, pp. 1e10. toads reduce the parasite burdens of native frogs? Int. J. Parasitol. Parasites Turvey, N., 2013. Cane Toads: a Tale of Sugar, Politics and Flawed Science. Australia, Wildl. 4 (3), 295e300. Sydney University Press, University of Sydney. Nelson, F.B., Brown, G.P., Shilton, C., Shine, R., 2015b. Host-parasite interactions Urban, M.C., Phillips, B.L., Skelly, D.K., Shine, R., 2008. A toad more traveled: the during a biological invasion: the fate of lungworms (Rhabdias spp.) inside native heterogeneous invasion dynamics of cane toads in Australia. Am. Nat. 171, and novel anuran hosts. Int. J. Parasitol. Parasites Wildl. 4 (2), 206e215. E134eE148. Ochsenbein, A.F., Zinkernagel, R.M., 2000. Natural antibodies and complement link Whittington, R.J., Kearns, C., Speare, R., 1997. Detection of antibodies. J. Virol. Meth. innate and acquired immunity. Immunol. Today 21 (12), 624e630. 68, 105e108. Perkins, T.A., Phillips, B.L., Baskett, M.L., Hastings, A., 2013. Evolution of dispersal Zug, G.R., Zug, P.B., 1979. The Marine Toad, Bufo Marinus: a Natural History Resume and life history interact to drive accelerating spread of an invasive species. Ecol. of Native Populations. Smithsonian Institution Press, Washington, D.C. Lett. 16 (8), 1079e1087. Zupanovic, Z., L. G., Hyatt, A.D., Green, B., Bartran, G., Parkes, H., Whittington, R.J., Phillips, B.L., Brown, G.P., Webb, J.K., Shine, R., 2006. Invasion and the evolution of Speare, R., 1998. Giant toads Bufo marinus in Australia and Venezuela have speed in toads. Nature 439, 803. antibodies against “ranaviruses”. Dis. Aquat. Organ. 32, 1e8. Effects of invasion history on physiological responses to immune system activation in invasive Australian cane toads

Daniel Selechnik1, Andrea J. West2, Gregory P. Brown1, Kerry V. Fanson2, BriAnne Addison2, Lee A. Rollins2 and Richard Shine1 1 School of Life and Environmental Sciences (SOLES), University of Sydney, Sydney, NSW, Australia 2 Centre for Integrative Ecology, School of Life & Environmental Sciences (LES), Deakin University, Geelong, VIC, Australia

ABSTRACT The cane toad (Rhinella marina) has undergone rapid evolution during its invasion of tropical Australia. Toads from invasion front populations (in Western Australia) have been reported to exhibit a stronger baseline phagocytic immune response than do conspecifics from range core populations (in Queensland). To explore this difference, we injected wild-caught toads from both areas with the experimental antigen lipopolysaccharide (LPS, to mimic bacterial infection) and measured whole-blood phagocytosis. Because the hypothalamic-pituitary-adrenal axis is stimulated by infection (and may influence immune responses), we measured glucocorticoid response through urinary corticosterone levels. Relative to injection of a control (phosphate-buffered saline), LPS injection increased both phagocytosis and the proportion of neutrophils in the blood. However, responses were similar in toads from both populations. This null result may reflect the ubiquity of bacterial risks across the toad’s invaded range; utilization of this immune pathway may not have altered during the process of invasion. LPS injection also induced a reduction in urinary corticosterone levels, perhaps as a result of chronic stress.

Submitted 4 August 2017 Subjects Ecology, Evolutionary Studies, Zoology, Immunology Accepted 6 September 2017 Published 6 October 2017 Keywords Rhinella marina, Eco-immunology, Phagocytosis, Cane toad, Invasive species Corresponding author Daniel Selechnik, INTRODUCTION [email protected] Eco-immunological theory predicts that successful invaders will display reduced Academic editor investment in components of the immune system that produce excessive inflammation, John Measey and/or are energetically expensive (Lee & Klasing, 2004; White, Perkins & Dunn, 2012). Additional Information and This prediction is based on the enemy release hypothesis (Colautti et al., 2004): the Declarations can be found on page 14 supposition that invasive hosts lose many co-evolved enemies after translocation (Allendorf, 2003; Torchin, Lafferty & Kuris, 2001), potentially reducing pathogen-mediated DOI 10.7717/peerj.3856 selection pressures (Lee & Klasing, 2004; White, Perkins & Dunn, 2012). Also, the energetic Copyright 2017 Selechnik et al. costs of mounting a strong immune response may reduce the host’s ability to survive, Distributed under grow, reproduce, and disperse (Hart, 1988; Klein & Nelson, 1999; Llewelyn et al., 2010); Creative Commons CC-BY 4.0 these factors influence invasion success (Chapple, Simmonds & Wong, 2012; Cote et al., 2010). However, such a down-regulation of immune function may render invaders

How to cite this article Selechnik et al. (2017), Effects of invasion history on physiological responses to immune system activation in invasive Australian cane toads. PeerJ 5:e3856; DOI 10.7717/peerj.3856 susceptible to infection by novel pathogens and parasites in their introduced range (Hamrick, Godt & Sherman-Broyles, 1992); for this reason, invaders are also predicted to elevate investment into less energetically costly components of the immune system (Lee & Klasing, 2004). Components of anti-microbial activity within the innate immune system differ in the amount of energy that they require and inflammation that they cause. Systemic mechanisms such as acute phase protein activity, anorexia, lethargy, and fever are highly inflammatory, and thus may be ‘costly’ (Klasing & Leshchinsky, 1999). These responses are predicted (Cornet et al., 2016; Lee & Klasing, 2004), and have been shown, to be down-regulated in invasive populations of invertebrates (Cornet, Sorci & Moret, 2010; Wilson-Rich & Starks, 2010), sparrows (Lee, Martin & Wikelski, 2005), trout (Monzon-Arguello et al., 2014), and deer (Que´me´re´ et al., 2015). Although constitutive innate defences (such as whole-blood phagocytosis of bacteria or yeast) also require substantial energy to activate (McDade, Georgiev & Kuzawa, 2016), glucose metabolism does not increase during phagocytosis in human neutrophils (Borregaard & Herlin, 1982). Thus, it is difficult to predict whether or not these defences are down-regulated in invaders, and further data are required. The cane toad (Rhinella marina) was brought to Queensland, Australia from Hawai’i in 1935 (Turvey, 2009). Reflecting traits such as high fecundity and long-distance dispersal ability (Urban et al., 2008), toads have expanded their range into New South Wales (Easteal, 1981), the Northern Territory (Urban et al., 2008), and Western Australia (Rollins, Richardson & Shine, 2015). Populations have thus been exposed to local pathogens and parasites in Queensland for 81 years, whereas toads near the invasion front in Western Australia may be encountering novel pathogens for the first time. Surveys and common-garden experiments have shown that several phenotypic characteristics (including morphology, physiology, and behaviour) have diverged between populations from the ‘range core’ (QLD) and ‘invasion front’ (WA) (Brown et al., 2015; Gruber et al., 2017; Hudson, Brown & Shine, 2016). Due to differences in selection pressures for established populations (in the range core) vs expanding populations (at the invasion front), comparisons between them provide similar results to those expected between native and invasive populations. Brown et al. (2015) compared the immunocompetence of captive-raised cane toads whose parents had been collected from QLD and WA. No significant difference was found in PHA-induced skin swelling, but WA toads exhibited higher bacteria-killing activity and phagocytosis than did QLD toads (Brown et al., 2015). This result suggests that bacteria-killing activity and phagocytosis may be favoured at the invasion front because these responses are less costly. However, Brown et al.’s study measured baseline levels of immune components, rather than the levels elicited by an in vivo immune challenge. One problem with comparing baseline levels of immune responses (such as phagocytosis and bacteria-killing activity) is the amount of variation across individuals in prior exposure to pathogens or antigens. Although individuals within the same population encounter the same types of infection, they may host pathogenic mutants of varying levels of virulence (McCahon et al., 1981). Measuring an immune response before and after

Selechnik et al. (2017), PeerJ, DOI 10.7717/peerj.3856 2/19 experimental infection with agents like lipopolysaccharide (LPS, a bacterial endotoxin produced by Escherichia coli), and treating the change from baseline as the response variable, partially solves this problem by allowing comparisons within individuals through time, as well as among individuals. This is because baseline levels of an immune response (which are reflective of current pathogen load and previous exposure) are accounted for when analysing the response to the antigen. Experimental infection also provides the opportunity to study immune stimulation in a situation where the exact identity and dosage of the experimental antigen are known; this antigen has been isolated and purified, is no longer part of a live organism, and is not a nucleic acid (Gould, 2008), so mutation and replication are not possible. Thus, estimates of the strength of the immune response (duration, maximum response, and time to maximum response) are not confounded by the effects of differing pathogenic challenges. We applied the experimental infection method to clarify the relative phagocytic capabilities of toads from the range core vs toads from the invasion front. Like the immune system, the hypothalamic–pituitary–adrenal (HPA) axis is stimulated by infection (Dunn & Vickers, 1994). The HPA axis is a major neuroendocrine system in vertebrates, and results in the release of glucocorticoids, which are closely associated with stress and activity (Janeway, Travers & Walport, 2001). Due to the suppressive effects of glucocorticoids on immune function (Alberts, Johnson & Lewis, 2002; Cooper, 2000; Fanning et al., 1998), glucocorticoids may regulate immune responses, preventing them from being elevated to a level that is harmful to the host (Ruzek et al., 1999; Stewart et al., 1988). In cane toads, associations between corticosterone and immune responses have previously been documented, with corticosterone having a negative effect on complement lysis activity, but a positive effect on leukocyte oxidative burst (Graham et al., 2012). Furthermore, toads with longer legs (a characteristic of WA toads) exhibited a reduced corticosterone response to stress (capture and confinement) than did their shorter-legged conspecifics (Graham et al., 2012). Because of this potential regulatory interaction and its relevance to immune function modulation in invaders, we also measured the effect of experimental infection on the glucocorticoid response. Our study compared immune and glucocorticoid responses of wild-caught cane toads from both the invasion front (WA) and range core (QLD) populations after experimental injection with LPS. We expected infection with LPS to cause an increase in phagocytosis through stimulation of the immune system, regardless of population; we did not expect phosphate-buffered saline (PBS) to have an effect. Because infection affects the HPA axis (Dunn et al., 1989), we also expected LPS injection to increase corticosterone levels, regardless of population; we did not expect PBS to have an effect. At the population level, we expected differential effects of LPS on QLD toads and WA toads. If phagocytosis is indeed less costly than other immune responses, then we expected that individuals from the WA population may display higher levels of phagocytosis after injection with LPS than would individuals from the QLD population (as was seen by Brown et al. (2015) in common garden-bred toads). Because Graham et al. (2012) reported that cane toads of different leg lengths differed in glucocorticoid responses, we also expected a population difference in corticosterone levels (with the caveat that LPS

Selechnik et al. (2017), PeerJ, DOI 10.7717/peerj.3856 3/19 Figure 1 Map of cane toad distribution in Australia. Current distribution of the cane toad throughout Australia. Toads were first introduced to Queensland (QLD) in 1935, and have since expanded their range into New South Wales, the Northern Territory (NT), and Western Australia (WA). Black diamonds indicate our toad collection sites: Cairns, QLD and Oombulgurri, WA.

injection may elicit a different response than that of capture and confinement, as used in the study by Graham et al. (2012).

MATERIALS AND METHODS Animal collection and husbandry In May of 2016, specimens of R. marina were collected from two locations on opposite ends of the invasion transect. The eastern location (Cairns, QLD; 16.9186S 145.7781E) is the site of the initial release of toads into the wild in 1936 (Turvey, 2013). Toads did not arrive at the western location (Oombulgurri, WA; 15.1818S 127.8413E) until 2015; thus, this population represents the invasion front (Fig. 1). A total of 10 female toads per location were captured and transported to Middle Point, Northern Territory (12.5648S, 131.3253E), where they were maintained in a common setting for approximately one month before the experiment began. Only females were collected to eliminate possible sex effects, and for comparison with data on gene expression in female cane toads from a concurrent study. The experiment was conducted during the dry season in the Northern Territory, and thus toads were not breeding at this time. Toads from each location were divided into two groups of five: LPS-injection and PBS-injection (phosphate-buffered saline, control). Specimens were kept separate by their assigned group, and housed in large boxes in groups of two to three individuals. Mesh-covered openings in the boxes provided access to natural light, maintaining specimens on the Australian Central Time Zone light cycle and in outdoor temperatures (nocturnal temperatures ranged

Selechnik et al. (2017), PeerJ, DOI 10.7717/peerj.3856 4/19 from 14 to 24.5 C). Dust-free sawdust was used as a substrate, and plastic containers were provided for shelter. Water was changed daily, and crickets were distributed to each box every third day.

LPS administration Injections were performed using disposable 25-gauge needles with 1 mL syringes (Livshop, Rosebery, Australia). Each toad assigned to the LPS-injection group was size-matched (based on mass) with a toad from the same population that had been assigned to the PBS-injection group. Toads were injected with either 20 mg/kg body mass LPS (Sigma-Aldrich, Castle Hill, Australia) diluted in 100 mL PBS (Sigma-Aldrich, Castle Hill, Australia), or with an equal volume of pure PBS, to the dorsal lymph sac at 16:00 h. Average toad body mass was 131 g.

Blood sampling Blood samples were taken for haemocytometry, white blood cell count differentials, and the phagocytosis assay. Cardiac punctures were performed using disposable 25-gauge needles with 1 mL heparinised syringes. Approximately 0.25 mL blood per individual was taken, and immediately transferred to a sterile 1.5 mL micro-centrifuge tube. Toads were not anesthetised; all samples were collected within 3 min of disturbance to the toad. This procedure was conducted on each toad twice: three days before and 14 days after injection, each time at 10:00 h. We allowed three days between blood collection and injection for toads to settle from the disturbance. A total of 14 days after injection were allowed for toads to mount an immune response; cellular and humoral immune responses have previously been shown to reach their maximum within this time frame in toads (Diener & Marchalonis, 1970).

Haemocytometry To quantify the concentration of blood cells in each sample, 5 mL whole blood was diluted in 995 mL Natts–Herrick solution (Australian Biostain, Traralgon, Australia) and stored for 24 h at 4 C. Then, blood cells were resuspended in the solution by inversion before 10 mL of the mixture was loaded into a haemocytometry chamber, and the numbers of erythrocytes (RBCs) and leukocytes were counted.

Counts of white blood cell differentials Approximately 2 mL whole blood was used to prepare a smear that was then air-dried for an hour, and then stained with Diff-Quik (IHC World, LLC, Woodstock, MD, USA). After 24 h, cover slips were placed on each slide using a thin layer of mounting medium and samples were given another 24 h to dry. Slides were scanned at 100Â, and the first 100 leukocytes seen were identified as basophils, eosinophils, neutrophils, lymphocytes, or monocytes. Percentages of each cell type (number of cells of each type divided by 100) were calculated. Because neutrophils are common phagocytes, the relationship between neutrophil percentage and phagocytosis was assessed.

Selechnik et al. (2017), PeerJ, DOI 10.7717/peerj.3856 5/19 Table 1 Principal component analysis (PCA) loading values for three phagocytosis measures in cane toads. Immune measure PC1 (70.6%) Mean luminescence 0.97 Max luminescence 0.93 Time to max luminescence -0.56 Notes: Loading values for principal component analysis (PCA) formed from three phagocytosis measures. Whole-blood phagocytosis in cane toads was assessed via mean luminescence across time points, maximum luminescence, and time to reach maximum luminescence. These response variables were natural log-transformed, and then run through a PCA.

Phagocytosis assay We used a phagocytosis assay in which whole blood samples were stimulated by zymosan in the presence of luminol, generating luminescence (in relative light units, RLU) as a measure of phagocytosis (Martinez & Lynch, 2013). Whole blood was first diluted 1:20 in Amphibian Ringers solution, and 240 mL of the mixture was added to duplicate wells in a 96-well plate along with 30 mL luminol (Sigma-Aldrich, Castle Hill, Australia) and 10 mL zymosan (Sigma-Aldrich, Castle Hill, Australia). Another 240 mL of the sample was added to a control well along with 30 mL luminol and 10 mL PBS instead of zymosan. After the addition of zymosan, the 96-well plate was immediately inserted into a luminometer. Light emissions were recorded every 5 min for 200 min. The luminescence value in the control PBS well of each sample was subtracted from the luminescence values in the two corresponding zymosan wells to control for variations in light emissions between samples unrelated to the addition of zymosan; duplicates were then averaged together. Because there are multiple facets to the strength of an immune response (duration, maximum, and speed), phagocytosis was assessed via three response variables: mean luminescence across time points, maximum luminescence, and time to reach maximum luminescence. These three response variables were natural log-transformed for data normalization, then run through a principal component analysis to determine the best-fitting vector to represent all of the data in a single measure, called principal component 1 (PC1; Table 1). High PC1 values indicate high average luminescence, high maximum luminescence, and short time to maximum luminescence.

Urine sampling To obtain urine for corticosterone analysis, toads were lifted gently from their boxes and held over a plastic cup for up to 3 min. Urine was not collected from toads that did not urinate within this period. Urine was immediately transferred to a 2 mL snap-cap tube and stored at -20 C. Urine sampling was conducted at seven time points during the experiment: three days (10:00 and 22:00 h) and two days (10:00 h) prior to injection, as well as six hours (22:00 h), one day (22:00 h), seven days (22:00 h), and 12 days (10:00 h) after injection. Samples were collected at two different times of day to incorporate periods of activity (22:00 h), when corticosterone levels are relatively high, and periods of inactivity (10:00 h), when corticosterone levels are lower (Jessop et al., 2014).

Selechnik et al. (2017), PeerJ, DOI 10.7717/peerj.3856 6/19 Creatinine assay Creatinine concentration (mg/mL) was measured in every urine sample to standardise corticosterone levels by controlling for concentration of urine. Creatinine quantification was based on the Jaffe reaction in which creatinine turns orange in the presence of alkaline picrate. Briefly, 100 mL neat urine was mixed with 50 mL 0.75M sodium hydroxide (NaOH) and 50 mL 0.04N picric acid in duplicate on a 96-well plate. The plate was then incubated at room temperature for 15 min. Absorbance was measured with a plate reader (Biochrom Anthos 2010, Biochrom Ltd., Cambourne, UK) using a 405 nm measuring filter and a 620 nm reference filter.

Corticosterone assay Urinary corticosterone metabolites were analysed using an enzyme-immunoassay that has been previously validated for cane toads (Narayan et al., 2012). Urinary corticosterone has been shown to lag behind plasma corticosterone by only 1 h (Narayan, Cockrem & Hero, 2013). The polyclonal corticosterone antibody (CJM06) and corresponding label (corticosterone conjugated with horseradish peroxidase [HRP]) were supplied by Smithsonian National Zoo (Washington, DC, USA). Briefly, high binding 96-well plates (Costar) were coated with 150 ml of coating buffer containing goat anti-rabbit IgG (GARG; 2 mg/mL). After 24 h, the coating solution was discarded and 200 mLof Trizma buffer solution rich in bovine serum albumin was added to each well and incubated for at least 4 h. Plates were washed five times and immediately loaded with 50 mL of standard, control, or neat urine sample, 50 mL of corticosterone-HRP (working dilution = 1:80,000), and 50 mL of corticosterone antibody (working dilution = 1:100,000). After incubating for 2 h at room temperature, plates were washed and 150 mL of substrate solution (1.6 mM hydrogen peroxide, 0.4 mM azino-bis (3-ethylbenzthiazoline-6-sulphonic acid) in 0.05 M citrate buffer, pH 4.0) was added to each well. The plate was incubated at room temperature for 45 min, and absorbance was quantified using a 450 nm measuring filter and a 620 nm reference filter. All samples were analysed in duplicate and hormone concentration is expressed as ng/mg creatinine. Urine corticosterone metabolite concentrations were natural log-transformed to meet model assumptions.

Statistical analyses All statistical analyses were performed using JMP Pro 11.0 (SAS Institute, Cary, NC, USA). We first subtracted each individual’s pre-injection PC1 score from its post-injection PC1 score, and then analysed the differences using a linear mixed model containing red blood cell concentration, population (QLD vs WA), and treatment (LPS injection vs PBS injection) as fixed effects. The interactive effects of population and treatment were also included in the model. We used this model to assess variation in PC1 scores, as well as variation in neutrophil percentages. Changes in PC1 were also tested for correlation with changes in neutrophil percentage due to the putative role of neutrophils in phagocytosis. A p-value less than 0.05 was used as our criterion for statistical significance of an effect or relationship.

Selechnik et al. (2017), PeerJ, DOI 10.7717/peerj.3856 7/19 Hormone data were analysed using a linear mixed model containing population, treatment, and days post-injection (-3.25, -2.75, -2.25, 0.25, 1.25, 7.25, and 12.25 DPI) as fixed effects. The interactive effects of population, treatment, and DPI were also included in the model. Because several repeated measures were taken for each toad, we also included individual ID as a random effect in the model. A p-value less than 0.05 was required to call significance of an effect or relationship.

Ethics Research was conducted in accordance with rules set for by the University of Sydney Animal Ethics Committee. Ethics application was approved under the project number 2016/1003. RESULTS Phagocytosis We found an effect of LPS challenge treatment on phagocytosis PC1 (treatment effect p = 0.02). LPS-injected toads exhibited a greater increase in phagocytosis than did their PBS-injected counterparts after injection (Fig. 2). RBC count also had a significant effect; toads whose blood samples were more concentrated with RBCs exhibited higher levels of phagocytosis. However, population had no significant effect, indicating similarity in the two populations’ phagocytic response to LPS challenge (Table 2).

White blood cell differentials Similar to PC1, a strong treatment effect was seen on change in neutrophil percentage (p = 0.0052). Across both populations, LPS-injected toads exhibited a greater increase in the percentages of neutrophils in their blood than did their PBS-injected counterparts after injection (Fig. 3A). Across treatments, the change in neutrophil percentage before vs after injection was positively correlated with the change in PC1 before vs after injection (p = 0.037, R2 = 0.23; Fig. 3B).

Urinary corticosterone metabolites A strong treatment  days post-injection (DPI) effect was observed on urinary corticosterone levels (Table 3). Corticosterone increased over time after injection in PBS-injected toads, but decreased over time after injection in LPS-injected toads (Fig. 4). However, we found no significant difference between populations. DISCUSSION Injection with LPS evoked up-regulation of immune responses in cane toads from both expanding (invasion front) and established (range core) populations. Two weeks after injection with LPS, toads had increased circulating levels of neutrophils, as well as elevated phagocytic ability. Control toads injected with PBS showed no significant changes in these immune measures. A positive correlation between phagocytosis levels and neutrophils was expected, as neutrophils are the most abundant phagocytic cells in the blood (Summers et al., 2010; Wilgus, Roy & McDaniel, 2013; Wright, 2001). Indeed, toads that increased neutrophil production the most also increased phagocytic ability the most, indicating an

Selechnik et al. (2017), PeerJ, DOI 10.7717/peerj.3856 8/19 Figure 2 Phagocytic responses of cane toads to lipopolysaccharide (LPS). Phagocytosis curves of cane toads (Rhinella marina) from (A) Queensland (QLD) and (B) Western Australia (WA) both before and after injection with either lipopolysaccharide (LPS) or phosphate-buffered saline (PBS). The average of all samples within a treatment group at each time point (N = 5) was calculated to produce the points on the graph. (C) Difference in mean luminescence values between pre-injection and post-injection readings of each population and treatment group.

Selechnik et al. (2017), PeerJ, DOI 10.7717/peerj.3856 9/19 Table 2 Cane toad phagocytosis linear mixed model. Source Estimate Standard error 95% CI DF F ratio p RBC 0.01 0.004 [0.004, 0.02] 1,15 8.70 0.01 Population 0.63 0.33 [-0.07, 1.34] 1,15 3.61 0.08 Treatment -0.87 0.34 [-1.58, -0.15] 1,15 6.64 0.02 Population  treatment -0.06 0.33 [-0.77, 0.65] 1,15 0.03 0.85 Notes: Effect of source population and treatment (LPS or PBS) on phagocytic activity in blood of the cane toad, Rhinella marina. Each individual’s pre-injection PC1 score was subtracted from its post-injection PC1 score, and differences were analysed using a linear mixed model containing red blood cell concentration (RBC), population (Queensland vs Western Australia), and treatment (LPS injection vs PBS injection) as fixed effects. Significant effects are in bold.

increased stimulation of neutrophil-mediated phagocytosis by LPS. Thus, LPS is an effective stimulant of the phagocytic response in cane toads via their neutrophils. Blood samples that were more concentrated with RBCs exhibited higher levels of phagocytosis. However, RBCs had no effect on neutrophil abundance (p = 0.55). Thus, the effect could have potentially arisen through luminescent properties of the RBCs themselves (autofluorescence; Emmelkamp et al., 2003). Contrary to our predictions, toads from the invasion front did not exhibit stronger phagocytic responses to LPS exposure than did toads from the range core; there was no significant difference between populations. Our predictions were based upon the possibility that neutrophil-mediated phagocytosis poses a lower energetic cost than other immune responses, and hence may be favoured in toads at the invasion front undergoing enemy release. Brown et al. (2015) reported higher baseline levels of neutrophil-mediated phagocytosis in common garden-raised toads from WA, consistent with the hypothesis that neutrophil-mediated phagocytosis provides an inexpensive way for WA toads to retain some immunocompetence without expending much energy. Thus, we expected phagocytosis to be favoured in the wild-collected adult toads from WA as well, even with the change in methodology (injection with LPS vs measurement of baseline levels). However, LPS is a common bacterial antigen, and the lack of difference may reflect an equal importance of combating severe bacterial infections across all populations. There may be more than a billion species of bacteria worldwide, and they are found across all environments (Dykhuizen, 1998). Because bacteria are able to tolerate many types of abiotic extremes (Dykhuizen, 1998), it is likely that the cane toad’s entire Australian range is home to rich bacterial communities. Although microbial species richness follows an aridity gradient (Yabas, Elliott & Hoyne, 2015) and the particular species of bacteria present at opposite ends of the range may differ, standing innate immune defences such as neutrophils are unspecialised (Janeway, Travers & Walport, 2001), and thus may not differ across populations based on changes in the bacterial species encountered. However, it is also possible that toads from invasion front vs range core populations differ in phagocytic activity (as reported by Brown et al. (2015)), but that those differences were not apparent in our study. Our sample sizes were limited to N = 5 per treatment per population. When a null hypothesis is not rejected (such as in our study, in which a population-level difference was not found), confidence intervals for the effect size are

Selechnik et al. (2017), PeerJ, DOI 10.7717/peerj.3856 10/19 Figure 3 Neutrophil responses of cane toads to lipopolysaccharide (LPS). (A) Changes in the per- centages of neutrophils (as induced by injection of either lipopolysaccharide [LPS] or phosphate- buffered saline [PBS]) across two treatment groups of cane toads (Rhinella marina)fromtwopopulations (invasion front site in Western Australia [WA] and range core site in Queensland [QLD]). The per- centage of neutrophils in the toad’s blood pre-injection was subtracted from the percentage of neu- trophils in the same toad’s blood post-injection. The average of the difference between pre-injection and post-injection of all samples within a treatment group at each time point (N = 5) was calculated to produce the points on the graph. Error bars indicate standard error. (B) Positive correlation between the changes in neutrophil percentage and PC1 before vs after injection.

recommended in place of retrospective power analysis to check for validity (Steidl, Hayes & Schauber, 1997). Our 95% CI for population as a predictor include zero within their range (Table 2); this indicates that effect could be zero, and thus rejection of the null hypothesis is justified. The same is seen in the 95% CI for the interactive effect of population and treatment. Sample sizes were low and it is possible that this may not have

Selechnik et al. (2017), PeerJ, DOI 10.7717/peerj.3856 11/19 Table 3 Cane toad urinary corticosterone linear mixed model. Source Estimate Standard error DF F ratio p Population -0.14 0.17 1,14 0.67 0.4253 Treatment -0.04 0.17 1,14 0.05 0.8311 DPI -0.02 0.02 1,41 1.83 0.1839 Population  treatment 0.02 0.17 1,14 0.02 0.8909 Population  DPI -0.02 0.02 1,41 1.12 0.2964 Treatment  DPI -0.05 0.02 1,41 10.53 0.0023 Population  treatment  DPI 0.02 0.02 1,41 0.89 0.3516 Notes: Effects of each explanatory variable on corticosterone levels of cane toads after injection with either lipopolysaccharide (LPS) or control (phosphate-buffered saline, PBS). Natural log-transformed corticosterone values were analysed in a linear mixed model with population (Queensland vs Western Australia), treatment (LPS injection vs PBS injection), and days post-injection (DPI) as fixed effects; their interactive effects were also assessed. Significant effects are in bold.

Figure 4 Urinary corticosterone responses of cane toads to lipopolysaccharide (LPS). Changes in urinary corticosterone levels across two treatment groups of cane toads (Rhinella marina) from two populations (invasion front site in Western Australia [WA] and range core site in Queensland [QLD]). Toads were injected with either lipopolysaccharide (LPS) or phosphate-buffered saline (PBS). Error bars indicate standard error, and lines are fitted by linear regression to data from each treatment group.

Selechnik et al. (2017), PeerJ, DOI 10.7717/peerj.3856 12/19 been sufficient to uncover a population-level difference; however, there are also several biological explanations for the lack of population difference. Maintaining toads from both populations in a common captive setting, and feeding them the same food, may have eliminated differences in their gut microbiota (Riddell et al., 2014), which in turn influences immune function (Carpenter et al., 2014). Additionally, toads from WA are constantly dispersing into novel environments (Urban et al., 2008) where they will encounter unfamiliar pathogens and parasites. Our study utilised wild-caught toads; prior to our collection, those from WA may have expended more of the energy allocated to neutrophil production and activity than did those from QLD (Brown et al., 2015), potentially resulting in a diminution of the intrinsically stronger phagocytic response. This idea is supported by a previous study conducted on wild-caught cane toads from the Northern Territory, where the toads were radio-tracked to quantify their movement distances before their immune responses were surveyed (Brown & Shine, 2014). Toads that travelled longer distances exhibited decreased standing innate defences, such as neutrophils, compared to less mobile toads (Brown & Shine, 2014). Although these toads were not collected from the same areas as those in our experiment, ‘more mobile’ may be a reasonable proxy for WA, and ‘less mobile’ may represent QLD. If this is the case, a lifetime use of neutrophils and other standing innate effectors may explain why WA toads do not retain the stronger neutrophil responses with which they are born. Prior energy expenditure is not the only variable that could have obscured a potential difference in phagocytosis levels between populations in our study. Our data suggest that the toads’ immune responses may have also been dampened by abiotic conditions. On average, the PBS-injection groups from both populations exhibited a decrease in levels of phagocytosis after injection. This decrease may be due to temperature; our experiments were conducted in mid-July to early August, when nocturnal temperatures fall as low as 14 C. Low temperatures have immunosuppressive effects on amphibians and other ectotherms (Maniero & Carey, 1997; Pxytycz & Jozkowicz, 1994; Raffel et al., 2006). Neutrophils and phagocytic activity decrease initially during winter, but return to baseline levels as amphibians acclimate to seasonal temperatures (Raffel et al., 2006). Such seasonal effects in our toads may have masked population differences. Stress may have also influenced our results. Corticosterone, the primary stress glucocorticoid hormone in amphibians, can induce differential suppression and activation of immune components (Stier et al., 2009). Cane toads exhibit an acute stress response to capture in which their corticosterone levels initially increase (Graham et al., 2012), but decline back to baseline after seven days of confinement (Narayan, Cockrem & Hero, 2011); our toads were held in captivity for one month prior to our study. Although our study showed that PBS-injected toads increased in corticosterone levels after injection, we observed the opposite effect in LPS-injected toads. This finding was unexpected; in mammals, glucocorticoids increase during infection, possibly to regulate the immune response and prevent excessive inflammation (Hawes et al., 1992; Ruzek et al., 1999; Stewart et al., 1988; Webster & Sternberg, 2004). However, one month in captivity, frequent handling, injection with antigens, and cardiac puncture may have

Selechnik et al. (2017), PeerJ, DOI 10.7717/peerj.3856 13/19 induced a state of chronic stress in the toads (Wingfield & Romero, 2001). Adrenal activity is less predictable during chronic stress, and thus the direction of the corticosterone response differs across taxa and stressor types (Dickens & Romero, 2013). Because the amount and nature of handling was the same for toads across all treatments, their stress states were likely similar; however, infection apparently made a difference for the LPS-injected toads. Some chronically stressed animals exhibit a decrease in corticosterone levels after infection (Cyr et al., 2007), thereby avoiding the suppressive effects of chronically high corticosterone on immune function. Amphibian immune defences are particularly sensitive to glucocorticoids, and increases in these hormones can raise amphibians’ susceptibility to disease (Rollins-Smith, 2001). Additionally, corticosterone is not the only mechanism by which stress can suppress immunity. Stress may also cause leakage of gut microbiota into the bloodstream, triggering immune responses; this results in a lower number of unoccupied effectors (Lambert, 2014; Saunders et al., 1994). Studies on phagocytosis in cane toads have ignored differences in the speed, rather than in the strength, of this immune response. Although phagocytosis levels did not differ significantly between populations, the speed at which the observed up-regulation occurred might have differed. A previous study found that LPS injection triggered a larger metabolic increase after 24 h in QLD toads than in WA toads (Llewellyn et al., 2012). Conceivably, part of this increased metabolic activity seen among QLD toads could have involved increased production of immune components. Our post-injection immune assays were conducted two weeks after LPS injection, by which time toads from both populations had up-regulated immune responses to the same extent. Future studies could explore this question by measuring phagocytosis 24 h after LPS exposure and monitoring changes in phagocytosis across a shorter time frame.

CONCLUSION In this study, we tested phagocytosis in cane toads using the same cell quantification methods and activity assay as those of Brown et al. (2015), but we took repeated measurements before and after injecting LPS in vivo. Our study confirmed that LPS stimulates phagocytosis; however, we did not detect a population-level difference in phagocytosis levels (as had been found in the previous study). Each experiment introduces its own unique confounds; the previous study did not account for inter- individual variation, and ours could not account for differences in environmental effects prior to collection. To definitively compare levels of phagocytosis between individuals from invasion front vs range core populations, a more robust experimental design would employ the experimental antigen methodology simultaneously on wild-caught toads, and on captive-bred toads raised in a common setting from each population.

ADDITIONAL INFORMATION AND DECLARATIONS

Funding This work was supported by the Australian Research Council [FL12010:0074, DE150101393] and the Equity Trustees Charitable Foundation [Holsworth Wildlife

Selechnik et al. (2017), PeerJ, DOI 10.7717/peerj.3856 14/19 Research Endowment]. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Grant Disclosures The following grant information was disclosed by the authors: Australian Research Council: FL12010:0074 and DE150101393. Equity Trustees Charitable Foundation [Holsworth Wildlife Research Endowment].

Competing Interests Lee A. Rollins is an Academic Editor for PeerJ.

Author Contributions  Daniel Selechnik conceived and designed the experiments, performed the experiments, analysed the data, contributed reagents/materials/analysis tools, wrote the paper, prepared figures and/or tables.  Andrea J. West conceived and designed the experiments, performed the experiments, reviewed drafts of the paper.  Gregory P. Brown conceived and designed the experiments, performed the experiments, analysed the data, reviewed drafts of the paper.  Kerry V. Fanson conceived and designed the experiments, reviewed drafts of the paper.  BriAnne Addison conceived and designed the experiments, reviewed drafts of the paper.  Lee A. Rollins conceived and designed the experiments, contributed reagents/materials/ analysis tools, reviewed drafts of the paper.  Richard Shine conceived and designed the experiments, contributed reagents/materials/ analysis tools, reviewed drafts of the paper.

Animal Ethics The following information was supplied relating to ethical approvals (i.e. approving body and any reference numbers): Research was conducted in accordance with rules set for by the University of Sydney Animal Ethics Committee (Approval number: 2016/1003).

Data Availability The following information was supplied regarding data availability: The raw data has been provided as Supplemental Dataset Files.

Supplemental Information Supplemental information for this article can be found online at http://dx.doi.org/ 10.7717/peerj.3856#supplemental-information.

REFERENCES Alberts B, Johnson A, Lewis J. 2002. The Generation of Antibody Diversity. Fourth Edition. New York: Garland Science. Allendorf FW. 2003. Introduction: population biology, evolution, and control of invasive species. Conservation Biology 17(1):24–30 DOI 10.1046/j.1523-1739.2003.02365.x.

Selechnik et al. (2017), PeerJ, DOI 10.7717/peerj.3856 15/19 Borregaard N, Herlin T. 1982. Energy metabolism of human neutrophils during phagocytosis. Journal of Clinical Investigation 70(3):550–557 DOI 10.1172/jci110647. Brown GP, Phillips BL, Dubey S, Shine R. 2015. Invader immunology: invasion history alters immune system function in cane toads (Rhinella marina) in tropical Australia. Ecology Letters 18(1):57–65 DOI 10.1111/ele.12390. Brown GP, Shine R. 2014. Immune response varies with rate of dispersal in invasive cane toads (Rhinella marina). PLOS ONE 9(6):e99734 DOI 10.1371/journal.pone.0099734. Carpenter S, Ricci EP, Mercier BC, Moore MJ, Fitzgerald KA. 2014. Post-transcriptional regulation of gene expression in innate immunity. Nature Reviews Immunology 14:361–376 DOI 10.1038/nri3682. Chapple DG, Simmonds SM, Wong BB. 2012. Can behavioral and personality traits influence the success of unintentional species introductions? Trends in Ecology & Evolution 27:57–64 DOI 10.1016/j.tree.2011.09.010. Colautti RI, Ricciardi A, Grigorovich IA, MacIsaac HJ. 2004. Is invasion success explained by the enemy release hypothesis? Ecology Letters 7:721–733 DOI 10.1111/j.1461-0248.2004.00616.x. Cooper GM. 2000. DNA Rearrangements. The Cell: A Molecular Approach. Second Edition. Sunderland: Sinauer Associates. Cornet S, Brouat C, Diagne C, Charbonnel N. 2016. Eco-immunology and bioinvasion: revisiting the evolution of increased competitive ability hypotheses. Evolutionary Applications 9:952–962 DOI 10.1111/eva.12406. Cornet S, Sorci G, Moret Y. 2010. Biological invasion and parasitism: invaders do not suffer from physiological alterations of the acanthocephalan Pomphorhynchus laevis. Parasitology 137(1):137–147 DOI 10.1017/s0031182009991077. Cote J, Fogarty S, Weinersmith K, Brodin T, Sih A. 2010. Personality traits and dispersal tendency in the invasive mosquitofish (Gambusia affinis). Proceedings of the Royal Society B: Biological Sciences 277:1571–1579 DOI 10.1098/rspb.2009.2128. Cyr NE, Earle K, Tam C, Romero LM. 2007. The effect of chronic psychological stress on corticosterone, plasma metabolites, and immune responsiveness in European starlings. General and Comparative Endocrinology 154(1–3):59–66 DOI 10.1016/j.ygcen.2007.06.016. Dickens MJ, Romero LM. 2013. A consensus endocrine profile for chronically stressed wild animals does not exist. General and Comparative Endocrinology 191:177–189 DOI 10.1016/j.ygcen.2013.06.014. Diener E, Marchalonis J. 1970. Cellular and humoral aspects of the primary immune response of the toad, Bufo marinus. Immunology 18:279–293. Dunn AJ, Powell ML, Meitin C, Small PA. 1989. Virus infection as a stressor: influenza virus elevates plasma concentrations of corticosterone, and brain concentrations of MHPG and tryptophan. Physiology & Behavior 45(3):591–594 DOI 10.1016/0031-9384(89)90078-4. Dunn AJ, Vickers SL. 1994. Neurochemical and neuroendocrine responses to Newcastle disease irus administration in mice. Brain Research 645:103–112 DOI 10.1016/0006-8993(94)91643-8. Dykhuizen DE. 1998. Santa Rosalia revisited: why are there so many species of bacteria? Antonie van Leeuwenhoek Journal of Microbiology 73:25–33. Easteal S. 1981. The history of introductions of Bufo marinus; a natural experiment in evolution. Biological Journal of the Linnean Society 16(2):93–113 DOI 10.1111/j.1095-8312.1981.tb01645.x. Emmelkamp J, DaCosta R, Andersson H, van der Berg A. 2003. Intrinsic autofluorescence of single living cells for label-free cell sorting in a microfluidic system. Transducers Research Foundation 1:85–87.

Selechnik et al. (2017), PeerJ, DOI 10.7717/peerj.3856 16/19 Fanning L, Bertr FE, Steinberg C, Wu GE. 1998. Molecular mechanisms involved in receptor editing at the Ig heavy chain locus. International Immunology 10:241–246 DOI 10.1093/intimm/10.2.241. Gould MJ. 2008. Filtration and purification in the biopharmaceutical industry. In: Jornitz MJ, Jornitz MW, Meltzer TH, eds. Limulus Amebocyte Lysate Assays and Filter Applications. Second Edition. Boca Raton: CRC Press, 425–426. Graham SP, Kelehear C, Brown GP, Shine R. 2012. Corticosterone–immune interactions during captive stress in invading Australian cane toads (Rhinella marina). Hormones Behavior 62:146–153 DOI 10.1016/j.yhbeh.2012.06.001. Gruber J, Brown G, Whiting MJ, Shine R. 2017. Geographic divergence in dispersal-related behaviour in cane toads from range-front vs range-core populations in Australia. Behavioral Ecology and Sociobiology 71(2):38 DOI 10.1007/s00265-017-2266-8. Hamrick JL, Godt MJW, Sherman-Broyles SL. 1992. Factors influencing levels of genetic diversity in woody plant species. New Forests 6(1–4):95–124. Hart BL. 1988. Biological basis of the behavior of sick animals. Neuroscience & Biobehavioral Reviews 12(2):123–127 DOI 10.1016/s0149-7634(88)80004-6. Hawes AS, Rock CS, Keogh CV, Lowry SF, Calvano SE. 1992. In vivo effects of the antiglucocorticoid RU 486 on glucocorticoid and cytokine responses to Escherichia coli endotoxin. Infection and Immunity 60(7):2641–2647. Hudson CM, Brown GP, Shine R. 2016. It is lonely at the front: contrasting evolutionary trajectories in male and female invaders. Royal Society Open Science 3(12):160687 DOI 10.1098/rsos.160687. Janeway CA, Travers P, Walport M. 2001. Immunobiology: The Immune System in Health and Disease. Fifth Edition. New York: Garland Science. Jessop TS, Dempster T, Letnic M, Webb JK. 2014. Interplay among nocturnal activity, melatonin, corticosterone and performance in the invasive cane toad (Rhinella marinus). General and Comparative Endocrinology 206:43–50 DOI 10.1016/j.ygcen.2014.07.013. Klasing KC, Leshchinsky TV. 1999. Functions, costs, and benefits of the immune system during development and growth. International Ornitology Congress, Proceedings 69:2817–2832. Klein SL, Nelson RJ. 1999. Activation of the immune–endocrine system with lipopolysaccharide reduces affiliative behaviors in voles. Behavior Neuroscience 113:1042–1048. Lambert GP. 2014. Stress-induced gastrointestinal barrier dysfunction and its inflammatory effects. Journal of Animal Science 87:E101–E108 DOI 10.2527/jas.2008-1339. Lee KA, Klasing KC. 2004. A role for immunology in invasion biology. Trends in Ecology & Evolution 19:523–529 DOI 10.1016/j.tree.2004.07.012. Lee KA, Martin LB, Wikelski MC. 2005. Responding to inflammatory challenges is less costly for a successful avian invader, the house sparrow (Passer domesticus), than its less-invasive congener. Oecologia 145:244–251 DOI 10.1007/s00442-005-0113-5. Llewellyn D, Thompson MB, Brown GP, Phillips BL, Shine R. 2012. Reduced investment in immune function in invasion-front populations of the cane toad (Rhinella marina) in Australia. Biological Invasions 14(5):999–1008 DOI 10.1007/s10530-011-0135-3. Llewelyn J, Phillips BL, Alford RA, Schwarzkopf L, Shine R. 2010. Locomotor performance in an invasive species: cane toads from the invasion front have greater endurance, but not speed, compared to conspecifics from a long-colonised area. Oecologia 162:343–348 DOI 10.1007/s00442-009-1471-1.

Selechnik et al. (2017), PeerJ, DOI 10.7717/peerj.3856 17/19 Maniero GD, Carey C. 1997. Changes in selected aspects of immune function in the leopard frog, Rana pipiens, associated with exposure to cold. Journal of Comparative Physiology B: Biochemical, Systemic, and Environmental Physiology 167(4):256–263 DOI 10.1007/s003600050072. Martinez NM, Lynch KW. 2013. Control of alternative splicing in immune responses: many regulators, many predictions, much still to learn. Immunological Reviews 253:216–236 DOI 10.1111/imr.12047. McCahon D, Slade WR, King AMQ, Saunders K, Pullen L, Lake JR, Priston RAJ. 1981. Effect of mutation on the virulence in mice of a strain of foot-and-mouth disease virus. Journal of General Virology 54(2):263–272 DOI 10.1099/0022-1317-54-2-263. McDade TW, Georgiev AV, Kuzawa CW. 2016. Trade-offs between acquired and innate immune defenses in humans. Evolution, Medicine, and Public Health 2016:1–16 DOI 10.1093/emph/eov033. Monzon-Arguello C, de Leaniz CG, Gajardo G, Consuegra S. 2014. Eco-immunology of fish invasions: the role of MHC variation. Immunogenetics 66(6):393–402 DOI 10.1007/s00251-014-0771-8. Narayan EJ, Cockrem J, Hero JM. 2013. Changes in serum and urinary corticosterone and testosterone during short-term capture and handling in the cane toad (Rhinella marina). General and Comparative Endocrinology 191:225–230 DOI 10.1016/j.ygcen.2013.06.018. Narayan EJ, Cockrem JF, Hero JM. 2011. Urinary corticosterone metabolite responses to capture and captivity in the cane toad (Rhinella marina). General and Comparative Endocrinology 173:371–377 DOI 10.1016/j.ygcen.2011.06.015. Narayan EJ, Molinia FC, Cockrem JF, Hero JM. 2012. Individual variation and repeatability in urinary corticosterone metabolite responses to capture in the cane toad (Rhinella marina). General and Comparative Endocrinology 175:284–289 DOI 10.1016/j.ygcen.2011.11.023. Pxytycz B, Jozkowicz A. 1994. Differential effects of temperature on macrophages of ectothermic vertebrates. Journal of Leukocyte Biology 56(6):729–731. Que´me´re´ E, Galan M, Cosson JF, Klein F, Aulagnier S, Gilot-Fromont E, Merlet J, Bonhomme M, Mark Hewison AJ, Charbonnel N. 2015. Immunogenetic heterogeneity in a widespread ungulate: the European roe deer (Capreolus capreolus). Molecular Ecology 24:3873–3887 DOI 10.1111/mec.13292. Raffel TR, Rohr JR, Kiesecker JM, Hudson PJ. 2006. Negative effects of changing temperature on amphibian immunity under field conditions. Functional Ecology 20:819–828 DOI 10.1111/j.1365-2435.2006.01159.x. Riddell CE, Lobaton Garces JD, Adams S, Barribeau SM, Twell D, Mallon EB. 2014. Differential gene expression and alternative splicing in insect immune specificity. BMC Genomics 15:1031 DOI 10.1186/1471-2164-15-1031. Rollins-Smith LA. 2001. Neuroendocrine-immune system interactions in Amphibians: implications for understanding global amphibian declines. Immunological Research 23(2–3):273–280 DOI 10.1385/ir:23:2-3:273. Rollins LA, Richardson MF, Shine R. 2015. A genetic perspective on rapid evolution in cane toads (Rhinella marina). Molecular Ecology 24:2264–2276 DOI 10.1111/mec.13184. Ruzek MC, Pearce BD, Miller AH, Biron CA. 1999. Endogenous glucocorticoids protect against cytokine-mediated lethality during viral infection. Journal of Immunology 162:3527–3533. Saunders PR, Kosecka U, McKay DM, Perdue MH. 1994. Acute stressors stimulate ion secretion and increase epithelial permeability in rat intestine. American Journal of Physiology 267:G794–G799.

Selechnik et al. (2017), PeerJ, DOI 10.7717/peerj.3856 18/19 Steidl RJ, Hayes JP, Schauber E. 1997. Statistical power analysis in wildlife research. Journal of Wildlife Management 61(2):270 DOI 10.2307/3802582. Stewart GL, Mann MA, Ubelaker JE, McCarthy JL, Wood BG. 1988. A role for elevated plasma corticosterone in modulation of host response during infection with Trichinella pseudospiralis. Parasite Immunology 10:139–150 DOI 10.1111/j.1365-3024.1988.tb00210.x. Stier KS, Almasi B, Gasparini J, Piault R, Roulin A, Jenni L. 2009. Effects of corticosterone on innate and humoral immune functions and oxidative stress in barn owl nestlings. Journal of Experimental Biology 212:2085–2091 DOI 10.1242/jeb.024406. Summers C, Rankin SM, Condliffe AM, Singh N, Peters AM, Chilvers ER. 2010. Neutrophil kinetics in health and disease. Trends in Immunology 31(8):318–324 DOI 10.1016/j.it.2010.05.006. Torchin ME, Lafferty KD, Kuris AM. 2001. Release from parasites as natural enemies: increased performance of a globally introduced marine crab. Biological Invasions 3:333–345 DOI 10.1023/A:1015855019360. Turvey N. 2009. A toad’s tale. Hot Topics from the Tropics 1:1–10. Turvey N. 2013. Cane Toads: A Tale of Sugar, Politics and Flawed Science. Australia: Sydney University Press. Urban MC, Phillips BL, Skelly DK, Shine R. 2008. A toad more traveled: the heterogeneous invasion dynamics of cane toads in Australia. American Naturalist 171(3):E134–E148 DOI 10.1086/527494. Webster JI, Sternberg EM. 2004. Role of the hypothalamic–pituitary–adrenal axis, glucocorticoids and glucocorticoid receptors in toxic sequelae of exposure to bacterial and viral products. Journal of Endocrinology 181:207–221 DOI 10.1677/joe.0.1810207. White TA, Perkins SE, Dunn A. 2012. The ecoimmunology of invasive species. Functional Ecology 26:1313–1323 DOI 10.1111/1365-2435.12012. Wilgus TA, Roy S, McDaniel JC. 2013. Neutrophils and wound repair: positive actions and negative reactions. Advances in Wound Care 2(7):379–388 DOI 10.1089/wound.2012.0383. Wilson-Rich N, Starks PT. 2010. The polistes war: weak immune function in the invasive P. dominulus relative to the native P. fuscatus. Insectes Sociaux 57:47–52 DOI 10.1007/s00040-009-0049-6. Wingfield JC, Romero LM. 2001. Adrenocortical responses to stress and their modulation in free-living vertebrates. In: McEwen BS, Goodman HM, eds. Handbook of Physiology. New York: Oxford University Press, 211–234. Wright KM. 2001. Amphibian medicine and captive husbandry. In: Wright KM, Whitaker BR, eds. Amphibian Hematology. Malabar: Krieger. Yabas M, Elliott H, Hoyne GF. 2015. The role of alternative splicing in the control of immune homeostasis and cellular differentiation. International Journal of Molecular Sciences 17(1):3 DOI 10.3390/ijms17010003.

Selechnik et al. (2017), PeerJ, DOI 10.7717/peerj.3856 19/19