UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Host specificity of bacterial communities : natural history, ecology, and conservation Loudon, Andrew Howard 2019

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
24-ubc_2019_september_loudon_andrew.pdf [ 3.49MB ]
Metadata
JSON: 24-1.0380660.json
JSON-LD: 24-1.0380660-ld.json
RDF/XML (Pretty): 24-1.0380660-rdf.xml
RDF/JSON: 24-1.0380660-rdf.json
Turtle: 24-1.0380660-turtle.txt
N-Triples: 24-1.0380660-rdf-ntriples.txt
Original Record: 24-1.0380660-source.json
Full Text
24-1.0380660-fulltext.txt
Citation
24-1.0380660.ris

Full Text

HOST SPECIFICITY OF BACTERIAL COMMUNITIES: NATURAL HISTORY, ECOLOGY AND CONSERVATION by  Andrew Howard Loudon  B.S., University of Mount Union, 2011 M.S., James Madison University, 2013  A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF  DOCTOR OF PHILOSPHY in THE FACULTY OF GRADUATE AND POSTDOCTORAL STUDIES (Zoology)  THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver)   August 2019  © Andrew Howard Loudon, 2019 ii   The following individuals certify that they have read, and recommend to the Faculty of Graduate and Postdoctoral Studies for acceptance, the dissertation entitled:  Host Specificity of Bacterial Communities: Natural History, Ecology and Conservation  submitted by Andrew Howard Loudon in partial fulfillment of the requirements for the degree of Doctor of Philosophy in Zoology  Examining Committee: Laura Wegener Parfrey Supervisor  Steve Perlman Supervisory Committee Member   Supervisory Committee Member Darren Irwin University Examiner William Mohn University Examiner  Additional Supervisory Committee Members: Christopher Harley Supervisory Committee Member Brian Leander Supervisory Committee Member  iii  Abstract Animals live with a symbiotic community of bacteria, and some of these bacteria affect host biology. Generally, these bacterial communities are diverse and complex. A key question in microbiome research is whether bacteria are specific to a host and affect host biology. In my dissertation, I use observational and experimental approaches to 1) understand the factors that contribute to composition of host-microbiota, 2) determine the core bacteria of a host by comparing the abundance of host bacteria to the bacteria in environmental bacterial communities, 3) determine the phylogenetic and environmental distribution of core bacteria and their close relatives to better understand their natural history, and 4) test whether core bacteria are correlated with aspects of host biology. Across three host species, I found bacterial communities that were populated by a few common and prevalent core bacteria that are absent from the environment; in most cases, these core bacteria belong to host-associated clades of bacteria. In chapter 2, I characterize the microbiota of the keystone sea star Pisaster ochraceus, identify core bacteria, and use phylogenetic trees to assess the distribution of their relatives. In chapter 3, I study the interactions between the skin microbiota and host innate immunity of Columbia spotted frogs and to determine whether they explain the occurrence of the amphibian pathogen, Batrachochytrium dendrobatidis (Bd). In chapter 4, I characterize the microbiota of wild and captive endangered Oregon spotted frogs (OSF) and whether their skin bacteria are associated with Bd intensity. Lastly, I experimentally test the hypothesis environmental reservoirs of bacteria influence the bacteria on the skins of frogs. Understanding the factors that structure captive communities has substantial conservation implications for captive breeding and head start programs. My results differ from previous studies by comparing the host to their environment and focusing on specific core bacteria.       iv  Lay Summary Animals often have complex bacterial communities with tens to hundreds of species and some influence host biology. Determining which bacteria to study within a complex community is a challenge that can be met by focusing on bacteria that are frequent across a host species. Bacteria that are frequent across hosts are more likely to affect host biology. I found just a handful of bacteria that are dominant and frequently associated with sea stars and amphibians; the closest relatives of these bacteria often associate with related hosts. My work suggests that host-associated bacteria are associated with multiple animal species, rather than each animal possessing a unique species of bacteria. Additionally, I explore factors that shape bacterial community composition, such as the immune defenses of frogs, captivity, or access to bacteria in the environment. Understanding which factors shape bacterial communities has implications to captive rearing programs and disease mitigation. v  Preface This thesis is based on original work by the author (AHL).  In chapter 2: AHL and Laura Wegener Parfrey (LWP) conceived and designed the study. AHL did the analyses. AHL and LWP wrote a manuscript that is near ready to be submitted. Dr. Chris Harley facilitated fieldwork and provided permits for sampling. Dr. Matt Lemay and Dr. Sam Starko assisted in the field.  In chapter 3: This chapter was in collaboration with Dr. Brandon Sheafor (Carroll College), Dr. Kevin Minbiole and Dr. Thomas Umile (University of Villanova).  Metabolite processing was done by KM and TU. Sample Collection, AMP processing and DNA extraction were done by BS and undergraduate student Alex Kurtz. Library prep for sequencing and qPCR was done by AHL. AHL performed analyses with guidance from LWP.   In chapter 4: This project partnered with the Precious Frog group, which consists of members from the Ministry of Forests, Lands Natural Resource Operations and Rural Development, Vancouver Aquarium and Greater Vancouver Zoo. Sample Collection in the field was done in coordination with a team led by Kendra Morgan and Aleesha Switzer.  The experiment was done with at the Greater Vancouver Zoo in collaboration with Andrea Gielens.  Sampling at the Vancouver Aquarium with Darren Smy. UBC students Coreen Forbes, Dr. Sam Starko, and Remi Matthey-Doret, Melissa Chen and Dr. Stilian Louca aided in sampling.  Lab work was done by AHL and Analyzes were done by AHL with guidance from LWP.  vi  Table of Contents Abstract	.................................................................................................................................	iii	Lay Summary	........................................................................................................................	iv	Preface	...................................................................................................................................	v	Table of Contents	..................................................................................................................	vi	List of Abbreviations	.............................................................................................................	xi	Glossary	...............................................................................................................................	xii	Acknowledgements	..............................................................................................................	xiii	Dedication	............................................................................................................................	xv		Introduction	.....................................................................................................................	1	Chapter 1:1.1	 Objectives	..................................................................................................................................	2	1.2	 General	Approach	......................................................................................................................	3	1.3	 Summary	of	questions	for	each	chapter	....................................................................................	5		The	sea	star,	Pisaster	ochraceus,	has	unique	and	dominant	bacteria	that	are	related	to	Chapter 2:symbionts	of	other	marine	invertebrates	.........................................................................................	7	2.1	 Abstract:	.....................................................................................................................................	7	2.2	 Introduction:	...............................................................................................................................	7	2.3	 Methods	....................................................................................................................................	11	2.4	 Results	......................................................................................................................................	18	2.5	 Discussion:	...............................................................................................................................	31		Elucidating	the	skin	microbial	ecology	of	Columbia	spotted	frogs	...................................	39	Chapter 3:vii  3.1	 Abstract:	...................................................................................................................................	39	3.2	 Introduction:	............................................................................................................................	40	3.3	 Methods:	...................................................................................................................................	43	3.4	 Results:	.....................................................................................................................................	53	3.5	 Discussion:	...............................................................................................................................	66		Establishing	the	core	microbiota	of	Oregon	spotted	frogs:	distribution,	potential	function	Chapter 4:and	stability	...................................................................................................................................	76	4.1	 Abstract:	...................................................................................................................................	76	4.2	 Introduction:	.............................................................................................................................	77	4.3	 Methods:	..................................................................................................................................	81	4.4	 Results	......................................................................................................................................	88	4.4.1	 Survey	..............................................................................................................................................	88	4.4.2	 Experiment:	......................................................................................................................................	99	4.5	 Discussion:	.............................................................................................................................	105		Concluding	remarks	.....................................................................................................	119	Chapter 5:5.1	 Limitations of work	................................................................................................................	127	5.2	 Final remarks	.........................................................................................................................	129	Bibliography	......................................................................................................................	131	Appendix A	................................................................................................................................................	145	Appendix B	.................................................................................................................................................	152	   viii  List of Tables Table 2.1 Samples that are included in the Pisaster study. ........................................................... 12	Table 2.2 Adonis results for community composition comparisons. ............................................ 21	Table 3.1 Samples that were processed and successfully analyzed by site and sample type. ....... 44	Table 4.1 Samples that are included in the OSF survey. ............................................................... 82	Table 4.2 Pairwise Adonis results comparing experimental treatments. ..................................... 102	Table 4.3 Statistical results of predicted function and alpha diversity by treatment and time. ... 105	Appendix Table A.1 Adonis result for testing the effect of filtering .......................................... 147 Appendix Table A.2 Adonis results for testing the effect of filtering for location ..................... 148	Appendix Table A.3 Common bacteria that are differentially abundant on aboral and aboral sides of Pisaster .................................................................................................................................... 148	Appendix Table B.1 Environmental characteristic for MT sampling locations.. ........................ 152	Appendix Table  B.3  Summary of protein concentration and MIC ........................................... 152	Appendix Table B.4 Shared OTUs between different locations ................................................. 153	Appendix Table B.5 Statistics and antifungal status of core bacteria ......................................... 153	Appendix Table B.6 Prevalence of core bacteria on Columbia spotted frogs. ............................ 153	Appendix Table B.7 Kendall’s rank Correlations between Bd intensity and core bacteria ........ 153	Appendix Table B.8 Logistic regression between Bd status and core bacterial abundance ........ 154	Appendix Table B.9 Correlations between AMP MIC against Bd and core bacteria. ................ 154	Appendix Table B.10 Summary of results from correlations across the entire dataset. .............. 154	Appendix Table B.11 Summary of results from mixed effect logistic regressions. .................... 155	Appendix Table B.12 Rare bacteria that are significantly more abundant on CSF than water. .. 155	 ix  List of Figures Figure 2.1 Ordination plots of Pisaster bacterial community compositions ................................ 20	Figure 2.2 Heat map of the proportion of shared OTUs between Pisaster and other bacterial communities .................................................................................................................................. 22	Figure 2.3 Stacked bar plot of core Pisaster bacteria .................................................................... 24	Figure 2.4 Phylogenetic tree of the dominant Spirochaete. ........................................................... 26	Figure 2.5 Phylogenetic tree of Salinispira ................................................................................... 27	Figure 2.6 Phylogenetic tree of Hepatoplasma ............................................................................. 30	Figure 3.1 Bd intensity for all four sampling locations ................................................................. 53	Figure 3.2 Bacterial community composition of Columbia spotted frogs .................................... 56	Figure 3.3 Alpha diversity and proportion of bacteria shared with water. .................................... 56	Figure 3.4 The core bacteria of Columbia spotted frogs ............................................................... 62	Figure 3.5 The relative abundance of bacteria that match the antifungal database by location. ... 63	Figure 3.6 Correlations between core bacteria and AMP MIC against Bd ................................... 63	Figure 3.7 Phylogenetic tree of Rhizobacter. ................................................................................ 64	Figure 3.8 Phylogenetic tree of Chryseobacterium ....................................................................... 65	Figure 4.1 Ordination plots of Oregon spotted frog bacterial community compositions .............. 90	Figure 4.2 Comparison of predicted function and diversity across sample types.. ....................... 92	Figure 4.3 Predicted function and alpha diversity for Oregon spotted frogs ................................ 93	Figure 4.4 Core bacteria on Oregon spotted frogs ........................................................................ 95	Figure 4.5 The relative abundance and occurrence core bacteria associated with OSF eggs. ...... 95	Figure 4.6 The frequency and intensity of the fungal pathogen, Bd. ............................................ 96	Figure 4.7 Relationship between bacterial communities and Bd .................................................. 98	x      	  	  	  	  	   	      	  	  	  	   	    	  	   	   	  	Figure 4.8 Relationships between the predictive function of bacteria, alpha diversity, corebacteria and the intensity of Bd. D-F show the relative abundance of the core bacteria..............99 Figure 4.9 Reduction of bacteria from the water reservoir..........................................................100 Figure 4.10 Ordination plots of community composition during the experiment.....................101 Figure 4.11 Predicted antifungal function and alpha diversity of frogs during the experiment..103 Figure 4.12 Dot plot of common, abundant bacteria in the experiment......................................104 Figure5.1 Schematic of core bacteria.........................................................................................121 Appendix Figure A.1 Alpha diversity of bacterial communities of oral and aboral surfaces and ceca of Pisaster........................................................................................................................145Appendix Figure A.2 Three species of sympatric sea stars have different bacterial communitycompositions................................................................................................................................146 Appendix Figure A.3 Filtering low abundance reads from individual samples has no affect on beta diversity ...............................................................................................................................146 Appendix Figure A.4 A comparison of the shared bacteria between sea star surfaces and their respective environments..........................................................................................................147 Appendix Figure A.5 Phylogenetic tree of Spirochaete OTU 10720.....................................149 Appendix Figure A.6 Phylogenetic tree of Reichenbachiella.....................................................150 Appendix FigureA.7 Phylogenetic tree of Peregrinibacteria ....................................................151 Appendix Figure B.1 Relationship between peptide concentration and MIC of peptides..........156 Appendix Figure B.2 Relationship between AMP MIC/Surface are and Log Bd ......................156 Appendix Figure B.3 Metabolite richness differs between locations......................................157 Appendix Figure B.4 Metabolite composition differs by geographic location..........................158xi  List of Abbreviations AMPs: Antimicrobial peptides ASV: Amplicon Sequence Variant  Bd:  The fungal pathogen, Batrachochytrium dendrobatidis CSF: Columbia spotted frogs LOR: Low observed reads MIC: Minimum inhibitory concentration, the lowest concentration of crude peptides at which no growth is detectable   OSF: Oregon spotted frogs OTU: Operational taxonomic units   xii  Glossary Bd intensity: Bd load, as measured by the number of genetic equivalents determined by qPCR.   Bd status: Presence or absence of Bd.  Common bacteria: Bacteria that are present within a community at an average relative abundance of greater than 1% across a sample type (here most often individuals of a given host species).     Core bacteria: Bacteria that are more abundant on hosts compared to environmental bacterial  communities across all sampled natural sites. This definition does not inform which type of symbiosis (mutualism, commensalism, parasitism) is occurring.     Dominant: The most relative abundant bacterium within a community.  Low observed reads (LOR): Reads present at low abundance per individual sample (less than ten reads/sample). These are filtered out of single samples to decrease the impact of sequencing errors within our dataset.   Rare bacteria: Bacteria that are present within a community at an average relative abundance across samples of less than 1%.     Symbiosis or symbiont: Host-associated microbe without regard for their fitness impact on the host to broadly encompass the spectrum of host-bacterial interactions (mutualism, commensalism, parasitism).   Transient bacteria: Bacteria detected in or on a host that appear to be stochastically assembled from environmental species pools. These are generally rare and infrequent.           xiii  Acknowledgements  I thank Dr. Laura Parfrey for taking me on as her first Ph.D. student. I thank her for her unwavering support, patience, insights, and incredible hard work.   I thank Dr. Chris Harley for teaching me the intertidal ropes and great feedback, and Dr. Brian Leander and Dr. Steve Perlman for all of their insights and great feedback.   I thank Dr. Brandon Sheafor for being a great collaborator.    Thank you to Coreen Forbes for all of support in life and in the field.  I thank Eric Peterson and Christina Munck and the Hakai Institute for supporting my sea star work on Calvert Island. Special thanks to Rhea Sanders-Smith for all that you have done to support this work.   I would like to thank past and current members of the Parfrey lab for all of their help: Dr. Matt Lemay, Melissa Chen, Jordan Lin, Cody Foley, Katy Davis, Dr. Flo Mazel, Vince Billy. Special thanks to Evan Morien for all of his help in bioinformatics.  Thank you to Dr. Sam Starko for your assistance in the field and Rémi Matthey-Doret for your assistance in fieldwork and statistics.  xiv   I would like to thank Oregon spotted frog recovery team. Special thanks to Kendra Morgan, Kristina Robbins and Aleesha Switzer, and for allowing me to research Oregon spotted frogs. Thank you to Andrea Gielens and the Greater Vancouver Zoo. Thank you to Darren Smy and the Vancouver Aquarium. And Monica Pearson for connecting me to the Oregon spotted frog community in BC.  I thank Dr. Reid Harris, Dr. Molly Bletz and Dr. Eria Rebollar for fielding my many amphibian  microbiome questions.   I thank Dr. Pam Dennis for all of her insights and the opportunity to apply what I have learned about microbiomes to animals at the Cleveland Metroparks Zoo.  Thank you to Dr. Kat Krynak for recruiting me to the zoo.  Thank you to The University of British Columbia for providing me with a 4-year fellowship.  Special thanks to my parents, Howard and Kathy, who are always supportive. Thank you to my sister, Sarah, for always having my back and always being supportive. Thank you to my Ursu’s and my Oshaben’s.  Thank you to my rabbit, Bunbun, for sometimes allowing me to pet him after a long day of working on this Ph.D. xv  Dedication  I dedicate this to my family. 1   Introduction Chapter 1:All organisms live with bacteria, and many bacteria affect the biology of their hosts. For example, bacteria aid in digestion (Dillon and Dillon 2004b), provide beneficial by-products from fermentation in the gut (Koh et al. 2016), make inorganic materials available to hosts in deep-sea thermal vent ecosystems (Cavanaugh 1983), educate the immune system (Gensollen et al. 2016), affect morphological development (Egan et al. 2013), and affect disease susceptibility (Woodhams et al. 2007b). Bacteria associated with a host can even affect mental health (Foster and McVey Neufeld 2013), social interactions with conspecifics (Theis et al. 2013), and the behaviour of disease vectors (Verhulst et al. 2011). With more research on host-associated bacterial communities, compelling connections between hosts and their symbionts become common.  In this dissertation, I use the term symbiont without regard for their fitness impact on the host to broadly encompass the spectrum of host-associated bacteria from mutualist to commensal to parasitic (Leung and Poulin 2008). Host-associated bacterial communities often consist of tens to hundreds of types of different bacteria and species vary in their complexity (O’Brien et al. 2019). Hosts tend to harbour species-specific bacterial communities that are distinct from the environment. These bacteria can be acquired through the environment, from con-specifics, or vertically from a parent (Bletz et al. 2013, Koskella et al. 2017). The level of host specificity of communities varies between host species, where some hosts have a few symbionts with a high specificity, and other have communities that have symbionts that are also in the environment (i.e., greater influence of environmental bacteria (Lemieux-Labonté et al. 2016, Lemay et al. 2018, Roth‐Schulze et al. 2018). Establishing which bacteria may be relevant to their host within a complex community can be challenging; one solution is to identify core bacteria that are 2  frequent across a study species (Apprill et al. 2017, Hernandez-Agreda et al. 2017, Hernandez-Agreda et al. 2018).  Core bacteria are hypothesized to play a key role in host ecosystem function (Shade and Handelsman 2012); however, this method cannot differentiate the type of symbiosis is occurring (i.e., the core could include pathogens or commensals). In previous studies determining the core bacteria for a host species has been done by setting frequency thresholds across a host population; these thresholds are often high, for example, 80% (Hernandez-Agreda et al. 2018), or 90% frequency across populations (Loudon et al. 2014b). High thresholds exclude bacteria that are abundant on some individuals, but not at a high frequency across individuals. For example, a bacterium abundant on half of the individuals within a sampled population may be biologically relevant, (e.g., to a particular phenotype) but excluded with high-frequency thresholds. Furthermore, these studies only focus on the microbiota of the target host species, which provides no information on where bacteria originate (e.g., are they filtered from an environmental source or transmitted between hosts?). Without also collecting environmental samples, it not possible to differentiate bacteria that are specific to a host from those predominantly found in the environment (i.e., transients). In my dissertation, I use extensive sampling of hosts and their environment across distinct sampling locations to examine how populated hosts bacterial communities are by particular bacteria that are adapted to particular hosts.            1.1 Objectives    In this dissertation, I seek to:   3  1) Define core as bacteria that are more abundant on hosts compared to environmental bacterial communities across all sampled natural sites.  This definition does not imply which type of symbiosis (mutualism, commensalism, parasitism) is occurring.  2) Determine the phylogenetic position of core bacteria and the environmental distribution of their close relatives to understand their natural history better.  3) Understand the factors that contribute to the composition of host microbiota.  4) Determine whether parameters of core bacteria and aspects of host biology are correlated.     1.2 General Approach    For my dissertation, I have focused on two host-bacteria systems. Studying two systems allowed me to establish generalities across systems. First, I characterize the bacterial communities that are associated with the purple ochre star, Pisaster ochraceus. I sampled the surfaces of sea stars, but also the ceca, where digestion and nutrient absorption occurs. This species is the iconic keystone species (Paine 1966) and sea stars in the Pacific Northwest have experienced mass die-offs disease (Schrope 2014, Miner et al. 2018).  Microbes can affect disease dynamics in other systems (Harris et al. 2009, Cheng et al. 2017), so as a marker of disease, establishing the signal of a healthy microbiota is a first step in detecting diseased individuals. In Chapters 3 and 4, I focused on amphibian systems in which it has been established that skin bacterial communities can protect against the chytrid fungus Batrachochytrium dendrobatidis (Harris et al. 2006). This system allowed me to examine whether there are any relationships between host microbiota and a pathogen. I studied two frog 4  species, Columbia spotted frogs (Rana luteiventris), and Oregon spotted frogs (Rana pretiosa). These two species are closely related and are in the spotted frog complex (Green et al. 1996). Furthermore, both species are resistant to Bd, and therefore, they serve as a model to detect characteristics of host disease resistance that are effective against Bd. Columbia spotted frogs are common in Montana, where my study took place, making them convenient species to study the interactions between host innate immunity, symbiotic bacteria, and Bd. Oregon Spotted frogs are endangered in British Columbia and are currently in captive breeding, and head start programs (Team 2014). This provides me the opportunity to study the effects of captivity on OSF microbiota, since captivity is known to affect the microbiota of some amphibians (Becker et al. 2014, Loudon et al. 2014b). In general disturbed or unnatural microbiota can have long-term health consequences for the host (Funkhouser and Bordenstein 2013, Zeissig and Blumberg 2014, Knutie et al. 2017).   In this dissertation, I use amplicon sequencing to profile bacterial communities that are associated with hosts and environmental bacterial communities. I use different tools to compare the distributions of bacteria in the environment to that of the host, such as the Sloan neutral model, to examine the abundance of bacteria in a source (environment), compared to the host to determine bacteria that are over-represented, neutral, or under-represented (Sloan et al. 2007). I also use the statistical program DESeq, which tests for differences in the distribution of data between two groups (host vs. environment) and corrects for the number of bacteria in the comparison (Love et al. 2014). Both of these methods allowed me to determine which bacteria are more abundant on the host compared to the environment.  I use phylogeny to infer elements of the natural history and to some extent, the environmental distribution of core bacteria by comparing them to their closest relatives. This was done by placing core bacteria into a 5  phylogenetic tree of bacteria from the SILVA database (Quast et al. 2013), which is an immense database of high-quality DNA sequences, to first determine their placement among curated bacteria. Then I made fine-resolution trees by acquiring the closest relatives from NCBI and curating the metadata for those sequences with EukRef (del Campo et al. 2018). This allowed me to determine whether the close relatives of core bacteria are also detected on hosts, or in the environment. If the clade contains other host-associated bacteria, I then also assess the taxonomy of other host organisms and their phylogenetic relationship to sea stars and amphibians.   1.3 Summary of questions for each chapter   In chapter 2, I ask whether Pisaster ochraceus have core bacteria across three distinct populations and a large geographic area. Furthermore, I ask whether the closest relatives of core bacteria are on other hosts. These analyses identified core bacteria that are in bacterial clades that are associated with other marine invertebrate hosts and are candidates to have long evolutionary histories of symbiosis.  In chapter 3, I ask whether Columbia Spotted frogs have core microbiota and if they correlate with properties of antimicrobial peptides (AMPs; a form of innate immunity); this question examines a mechanism of environmental filtering caused by the host. I ask whether metabolite profiles, which are produced by bacteria and the agents that inhibit Bd, correlate with bacterial communities, and Bd prevalence. I also ask whether bacteria on Columbia Spotted frogs are found in an antifungal database (Woodhams et al. 2015) and whether their relative abundances correlate with Bd; these questions address whether specific bacteria are associated with Bd in these frogs. Lastly, I ask whether the closest relatives of core bacteria are found on other hosts, to determine the extent of which these bacteria are found on other 6  amphibians and if they have a potential evolutionary history with amphibians.  In chapter 4, I first ask whether Oregon Spotted frogs (Rana pretiosa) have core microbiota and whether they correlate with Bd. I then test whether captivity affects bacterial community composition; this applies to conservation programs since frogs are in captive breeding and head start programs. In addition, to learn at what stage Oregon spotted frogs begin their relationship with core bacteria, I asked whether wild and captive egg masses have core bacteria. Lastly, I experimentally reduced the bacterial diversity in the water that frogs were housed in to test the effect of environmental bacteria on the stability and residence of bacterial diversity on frogs.  The resilience and stability of bacteria on Oregon spotted frogs has implications for housing frogs in captivity and the potential function of their skin microbiota.  My dissertation research reveals bacterial communities that are dominated by a few bacteria that have relatives that are on similar hosts. All of the hosts that I sampled had bacteria with a high fidelity to their host across the geographical range that I sampled. Furthermore, I reveal that these core bacteria tend to belong to clades of bacteria that are associated with closely related host organisms, indicating that these bacteria are either pre-adapted to live with that host or have an evolutionary history entwined with the host. My methods can be used to advance the field of host-microbial ecology to detect the abundant and frequent bacteria that may be important to host biology or serve as a baseline of health. My research also sheds light on patterns of host-symbiont relationships and corresponds to a study that argues that many bacterial symbionts are found on multiple hosts, rather than each host having unique symbionts (Louca et al. 2019).  7   The sea star, Pisaster ochraceus, has unique and dominant Chapter 2:bacteria that are related to symbionts of other marine invertebrates 2.1 Abstract:  All animals live in close association with bacteria, but for most hosts little is known about the distribution and host-specificity of these bacteria. I used amplicon sequencing to survey the bacterial communities living on and in the sea star Pisaster ochraceus, as well as in their environment and on sympatric marine hosts, across three geographically distinct populations. Overall, the bacterial communities on Pisaster are distinct from their environment and differ by both body region and geography.  I detected core bacteria that are more abundant on hosts compared to environmental bacterial communities across all sampled natural sites. The core taxa fall within Spirochaetes (genus Salinispira and two uncultured clades), Mollicutes (genus Hepatoplasma), Peregrinibacteria, and Bacteroidetes (genus Reichenbachiella). Using phylogenetic trees, I show that some of the core bacteria have close relatives that are found on other sea stars or other marine animals, suggesting that these clades are adapted to an animal-associated lifestyle.   2.2 Introduction:  All animals live closely associated with bacteria that collectively can have an impact on host health and biology (Egan et al. 2013, McFall-Ngai et al. 2013, Apprill 2017). While an explosion of recent research has highlighted the universality and general importance of host-associated microbiota, we still know little about how to best identify the key players within a host-associated bacterial community. This is a challenging problem since host-associated 8  bacterial communities are complex and often comprised of tens to hundreds of bacterial types on a given individual with a large amount of intraspecific variation across host populations (O’Brien et al. 2019).   Core bacteria are bacteria that are frequent across microbiomes and are hypothesized to play a key role in ecosystem function (Shade and Handelsman 2012). Determining core bacteria is done by establishing frequency thresholds; these thresholds are often high, for example, 80% (Hernandez-Agreda et al. 2018), or 90% frequency across host populations of a host (Loudon et al. 2014b). High thresholds exclude bacteria that are abundant on some individuals, but not at a high frequency across a high threshold of individuals; a bacterium abundant on half of the individuals sampled may be biologically important but excluded with those criteria. Alternately, there are transient bacteria that are infrequent across microbiomes. Efforts to partition the community into core and transient bacteria in sponges (Ramsby et al. 2018), and corals (Hernandez-Agreda et al. 2018), and amphibians (Loudon et al. 2014b) generally find a small number of core taxa and a large number of transient bacteria. Alternatively, hosts may not have core bacteria, such as the green alga, Ulva spp, (Roth‐Schulze et al. 2018). Establishing which bacteria are members of a host’s core bacteria is a good start in identifying bacteria that may provide a function for the host. However, there should be some caution since commensal, and pathogenic bacteria are identified with this definition. In this study, I define core bacteria as bacteria that are more abundant on hosts compared to environmental bacterial communities across all natural sites that are sampled. This definition does not imply which type of symbiosis (mutualism, commensalism, parasitism) that is occurring.  Host-associated bacterial communities tend to be host species-specific and distinct from environmental microbiota; these patterns have now been documented for a broad range of marine 9  organisms including sponges (Thomas et al. 2016a), anemones (Mortzfeld et al. 2015), corals (Ainsworth et al. 2015), and kelp (Lemay et al. 2018). Within a host species, bacterial communities also differ across body region (e.g. internal, external) (Ainsworth et al. 2015), likely reflecting the fact that differences in chemistry and morphology across body regions result in different habitats that filter for bacteria. At the same time, bacteria from the environment can influence microbiota composition on marine organisms with many microbes being acquired from surrounding seawater and biofilms (Fahimipour et al. 2017, Lemay et al. 2018). Microbiota composition tends to vary across geographical space, reflecting background variation in environmental microbiota and the contribution of environmental microbes to the microbiota of marine organisms (Rodriguez‐Lanetty et al. 2013, Mortzfeld et al. 2015). Widespread host-specificity of bacterial communities, in addition to the variation of transient bacteria, influenced by background environmental conditions, are likely driven by a few core bacteria and transient bacteria from environmental bacterial communities.   Extensive characterization of host-associated microbiota spurred by amplicon sequencing and the emergence of a sea star wasting disease (Hewson et al. 2014, Schrope 2014) have fueled recent interest in the microbiome of sea stars, though the associated bacteria were detected first in the 1970s with microscopy (Holland and Nealson 1978). Sea stars are a diverse clade of marine echinoderms and many, including Pisaster ochraceus (hereafter referred to as Pisaster), are predators that have an impact on macro-community structure (Paine 1966). Culture-based and microscopy studies reported dense populations of subcuticular bacterial symbionts in many echinoderms, including sea stars, and showed that different species harbour morphologically different bacterial communities, including bacteria with a spiral morphology (Lesser and Blakemore 1990, Kelly et al. 1995, Lawrence et al. 2010). More recently, the microbiota across 10  body regions has been characterized for several species of sea stars (Nakagawa et al. 2017, Hoj et al. 2018, Jackson et al. 2018, Lloyd and Pespeni 2018). A survey of the bacteria associated with surfaces, pyloric ceca, gonads, and coelomic fluid of twelve healthy sea star species from Washington, USA, and Queensland, Australia found that microbiota differed compared to seawater samples and specific bacterial taxonomic groups were common across sea star species (Jackson et al. 2018). A similar study that also sampled sea star surfaces, pyloric ceca, gonads, and coelomic fluid of Ancanthaster cf. solaris found differences between body regions and found Proteobacteria, Spirochaetes, and Tenericutes (Hoj et al. 2018). Altogether this work provides an initial image of the sea star microbiome, but because of the sample size and limited sampling of the environment that sea stars live in, little can be said about the distribution and host specificity of sea star bacteria.   In this study, I surveyed the bacterial communities from three body regions of Pisaster ochraceus at three geographical locations along with communities in their environment and on sympatric host species. I used my broad sampling to 1) test whether bacterial communities on Pisaster are unique compared to other hosts and the environment, 2) identify core bacteria. I also built phylogenetic trees of core bacteria and their closest relatives to determine whether they are generally host-associated or environment-associated. Lastly, I tested how removing low abundance sequence reads (LOR) from individual samples, which may be artifacts of sequencing, affects the analyses of alpha and beta diversity, and the shared number of bacteria between sample types. My data show that Pisaster have bacterial communities that are distinct from the environment, consistent with environmental filtering being a dominant factor in structuring communities. Most bacteria are shared with other host species such as barnacles, mussels, anemones or rockweed in addition to the environment, but a handful of bacteria are core 11  bacteria that were only detected on or in sea stars. I placed these core bacteria in broadly sampled phylogenetic trees, and for some core bacteria, their close relatives are on other sea stars or other marine benthic invertebrates.   2.3 Methods Sampling Pisaster were sampled from Bamfield, BC (n=10), Port Moody, BC (n=10) and the Hakai Institute on Calvert Island, BC (Hakai; n=6), under permits XR 232 2014 and XR 204 2014. Bamfield and Hakai are located on the outer coast, exposed sites of British Columbia with a salinity of ~32ppt. Samples were collected from tide pools or exposed sea stars at low tide. In contrast, Port Moody is a protected, calm inlet with a salinity of ~20 ppt. Port Moody is also different in that sea stars were collected from pilings that are coated in creosote, and sea stars were submerged at the time of sampling. All of these locations are hundreds of kilometers apart. Microbial sampling occurred at the site of collection: sea stars were initially rinsed with sterile seawater (instant ocean; 32 ppt) to remove transient bacteria (Culp et al. 2007) and then oral and aboral surfaces were each swabbed with sterile cotton swabs (Puritan, Guilford, ME). All surface samples were collected by swabbing 10 times, with a back and forth stroke equaling one swab. Ceca samples were collected by making an incision into an arm of a sea star with sterile scissors and pulling out a piece of ceca, which included host tissue, with sterile forceps. The ceca were then cut with sterile scissors and placed in cryotubes taking care to avoid contamination. Seawater and rock/abiotic substrate samples from the surrounding environment and other marine hosts were also sampled for comparison. Seawater samples were collected in triplicate at each site in 1-liter Nalgene bottles and then transported back to the lab on ice for filtering within four 12  hours of collection. Between 500 and 1000 mL of seawater was filtered through 0.22um GV filters (Duapore®Membrane filters, Millipore®- GVWP04700) with the Masterflex ® L/S peristaltic pump (Cole-Parmer- Item#- RK-77913-70) on level 70 until the filter clogged. Surfaces of surrounding rocky substrate and other hosts present at the site were swabbed similar to sea stars and included rocks, settling plates, mussels (Mytilus spp.), barnacles (Belanus glandulus), sea anemones (Anthopleura elegantissima), and the rockweed, Fucus distichus (Table 2.1). Two other sea star species, Dermasterias imbricata, and Henricia leviuscula, were also sampled at Hakai at West Beach. All samples were stored on ice until frozen at -20° C until further processing.   Table 2.1 Samples that are included in the Pisaster study.  Molecular methods: DNA was extracted from swabs, water filters and ceca samples using the MoBio Powersoil-htp 96 well DNA extraction kit (Carlsbad, CA) following the manufacturer's protocol Sample	Type Bamfield	(n) Port	Moody	(n) Hakai	(n)Oral	Pisater 10 10 6Aboral	Pisater 10 10 8Ceca	Pisater 7 6 6Leather	star:	Dermasterias	imbricata* 0 0 9Blood	star: 	Henricia	leviuscula* 0 0 14Mussels:	Mytilis	spp. 0 3 0Sea	Anemone:	Anthopleura	elegantissima 3 3 0Barnacles:	Belanus	glandulus 3 0 0Rockweed:	Fucus	distichus 3 0 0Abiotic	surface 3 3 4Seawater 3 3 5*Includes	oral	and	aboral	samples.13  with modifications according to the Earth Microbiome protocol (http://www.earthmicrobiome.org/) except for plates, which were shaken at 20 shakes per second. To determine the identities and relative abundances of bacteria associated with sea stars we amplified the V4 region of the 16S rRNA gene to detect bacteria and Archaea on using the 515f and 806r primers: 515f: 5'–GTGYCAGCMGCCGCGGTAA–3', 806r: 5'–GGACTACNVGGGTWTCTAAT–3'. These primers included Illumina adapters and the forward included a 12 nucleotide Golay barcode. Each PCR reaction contained 10µl of 5-Prime Master Mix (Bamfield and Port Moody samples), or 10ul of Phusion Mastermix (Hakai samples), 1µl of each primer (final concentration = 0.2µM each), 1µl of DNA, and PCR grade water to a final volume of 25µl. PCR started with an initial denaturation step at 94˚C for 3 minutes, followed by 35 cycles of denaturation at 94˚C for 45 seconds, primer annealing at 50˚C for 60 seconds, and extension at 72˚C for 90 seconds, with a final extension step of 72˚C for 10 minutes. PCR products were quantified using Quant-IT Pico Green® ds DNA Assay Kit (Life Technologies). Equal quantities of DNA (25ng) from each sample were pooled and then purified using the MoBio UltaClean® PCR clean-up kit. Pooled library quantitation and paired-end Illumina MiSeq sequencing (2 x 300bp) were carried out at the Integrated Microbiome Resource (IMR) facility in the Centre for Genomics and Evolutionary Bioinformatics at Dalhousie University (Halifax, Canada). Raw sequences were demultiplexed using Quantitative Insights into Microbial Ecology (QIIME v. 1.9.1) with a pfred score of 20 to ensure quality sequences (Caporaso et al. 2010b). The sequences were then trimmed to 250 base pairs using the FastX toolkit (http://hannonlab.cshl.edu/fastx_toolkit/). Operational taxonomic units (OTUs) were made using 14  the Minimum Entropy Decomposition method (Eren et al. 2015). This was done with the minimum substantive abundance parameter (-M) set at 250 reads, which means that 250 sequences of an OTU must be present for it to be included in the data. All other parameters were at default. Taxonomy was assigned using uclust (Edgar 2010) and the SILVA 128 database clustered at 99%. Chloroplast and mitochondrial sequences, and sequences unidentified at the domain level were removed. One OTU assigned to Achromobacter marplatenisis was also removed as it was found in PCR controls and is a common lab contaminant. Furthermore, OTUs that were only present in one sample was removed. Samples with fewer than 2400 reads were removed from further analysis. The dataset was rarefied to 2400 sequences/sample for alpha and beta diversity analyses. To minimize the impact of cross-well contamination, I also filtered low abundance reads (LOR) on a per-sample basis.  Data and MiMARKS compliant metadata will be accessioned at EBI.  Diversity To investigate the richness of bacteria that are associated with sea stars, I calculated Chao1 index in QIIME. Chao1 index is a richness estimator that adds a correction factor for rare taxa (singletons and doubletons) (Chao 1984). I used a linear model to test for differences between ceca, surface, and environmental samples. I then used Pairwise least-square means tests with a Tukey adjustment based on sample type to test for differences across sample types using the package emmeans (Lenth 2018). Differences in community composition were assessed with the Bray-Curtis dissimilarity and visualized with NMDS plots using Phyloseq in R (McMurdie and Holmes 2013). I used 15  permutational analysis of variance (Adonis) in the Vegan package of R (Oksanen et al. 2015) to test for differences in community composition. I used a one-way Adonis to test whether bacterial communities differed by sample types (i.e., Pisaster, ceca, water, abiotic surfaces, etc.). I used a two-way Adonis to test for differences between sea star geographical location and body regions. Lastly, I used a one-way Adonis to test whether three sympatric sea star species had distinct bacterial communities.  Distribution of bacteria I assessed the distribution of bacterial OTUs found on Pisaster to determine how many bacterial OTUs are shared with environmental sources and the oral side and ceca of Pisaster using shred_phylotypes.py in QIIME (Caporaso et al. 2010b). Furthermore, I tested different filtering parameters to determine the effect of rare bacteria on the distribution of bacteria on sea stars by eliminating sequence reads that were at average less than two or ten reads per sample using an in house python script. Furthermore, I conducted comparisons to determine if bacterial communities from oral and aboral Pisaster shared more OTUs with the environmental sources that they are in the most contact with (seawater or abiotic surface samples). Identifying core Pisaster bacteria I defined core bacteria that are more abundant on hosts compared to environmental bacterial communities across all sampled natural sites. This definition does not imply which type of symbiosis (mutualism, commensalism, parasitism) that is occurring. I used my sampling design that included three distinct populations and samples from surrounding environments at each site to identify core bacteria using the Sloan neutral model (Sloan et al. 2007). The Sloan 16  neutral model uses the frequency at which an OTU is found in a target community (i.e., the surface of Pisaster) compared to the average relative abundance of a source community (i.e., seawater) to determine which OTUs in the target community are over-represented, neutral, or under-represented compared to the target community. The model uses the premise that the probability of detecting an OTU in a target community due to random dispersal and ecological drift is proportional to the relative abundance of that OTU in the source community. I ran this test for the aboral surfaces of Pisaster (target community), compared to seawater, abiotic substrate, and other non-sea star hosts (source communities) separately. These analyses were run separately for each location because microbial communities on Pisaster and in the environment differ significantly by geography; unequal sampling sizes would skew results to favour the location with greater samples. I determined the core as those OTUs over-represented or unique to Pisaster compared to all of the source communities for all three sampling locations. For these analyses, the sympatric sea stars Dermasterias imbricata, and Henricia leviuscula were not used as a source since including them would prevent us from identifying bacteria that are over-represented across multiple sea star species. This process identified 38 OTUs within my dataset that are specific to the aboral side of Pisaster compared to the environment and other non-echinoderm hosts across all geographical sites. I focused on the common core OTUs (overall average relative abundance greater than 1% (Pedros-Alio 2012) since all of the rare core bacteria were close sequence variants of the common core bacteria. Lastly, to understand the differences between the oral and aboral sides of Pisaster I used DESeq2 (Love et al. 2014) to identify differentially abundant bacteria on oral and aboral surfaces.  Phylogenetic trees 17  I placed my OTUs into the overall SILVA 128 phylogenetic tree to investigate their phylogenetic position. I aligned the representative sequences of MED OTUs to SILVA 128 with PyNAST (Caporaso et al. 2010a) then placed them into the SILVA128 tree with EPA algorithm (Berger et al. 2011) using RaxML version 8 (Stamatakis 2014). This tree excludes the 29 OTUs that did not align, but these account for less than 2% of the overall dataset and are not part of the Pisaster core. To examine the distribution of and test whether the core bacteria of Pisaster are within sea star bacterial clades I examined their closest sequence matches in SILVA (Quast et al. 2013) and NCBI using BLAST (Madden 2013). Trees were constructed by selecting neighbors from the overall SILVA tree. Then a new maximum likelihood tree was constructed using a subset of the closest SILVA sequences that are neighbors to the core bacteria in the SILVA tree, sequences from this study, Hoj et al. (2018), Jackson et al. (2018) and NCBI from a BLAST search against NCBI were compiled for the tree. A recent study on Pisaster microbiota found a Spirochaete at high abundances (Lloyd and Pespeni 2018), but this sequence was not included in these analyses. The amount of bacterial diversity that is present in each tree varies since some bacterial groups are specious, where others are species poor. The percent similarity of sequences compared to the core bacteria of interest available in databases varied for each taxon to accommodate taxa with few bacteria that closely related. For example, an 85% threshold for the Spirochaete OTU13197 and OTU13140; a 93% threshold for Salinispira OTU29617 and OTU29659; an 88% threshold for Mollicutes OTU31698. Bootstrapped (n=100) trees were constructed using RAxML v 8.24 (Stamatakis 2014). I annotated trees using metadata pulled from GenBank records by the EukRef pipeline (del Campo et al. 2018) to determine whether sequences were detected on a host, and more likely to be host-associated, or detected from an environmental sample; detecting a bacterium on a host or from the environment is not an 18  absolute indicator of whether a bacterium is host-associated or from the environment. Figtree v 1.4.3 (Rambaut 2012) also, Inkscape v 0.92.3 was used to visualize and annotate the trees.   2.4 Results I sampled the bacterial communities of the sea star Pisaster ochraceus (Pisaster) from three populations in coastal British Columbia, along with comparative samples of the microbiota from the nearby environment and other marine hosts. In total there were 134 samples; 5,075,052 amplicon sequences that fell within 1,796 OTUs. I use these data to characterize the Pisaster microbiota and identify core bacteria.   Bacterial richness:  I investigated the bacterial richness of communities detected on and in Pisaster and their environment using the alpha diversity metric Chao1. I used a linear mixed-effects model to test for differences across sample type with the geographical site as a fixed effect. Bacterial richness was significantly different across sample types (Linear Model: Sample Type F4:86=24.18, p <0.0001) and sampling location (F2:86=3.90, p =0.02). To detect differences across sample types, I used Pairwise least-square means tests with a Tukey adjustment. These pairwise contrasts show that ceca samples consistently had a lower richness than the other sample types (Appendix Figure A.1). There was not a consistent trend in richness for other sample types across sites (Appendix Figure A.1).   19  Community composition:  Pisaster have bacterial communities that cluster separately compared to other hosts, surrounding seawater and abiotic surfaces. The Pisaster ceca are quite distinct and differ from all other samples in this study (Adonis: Pseudo-F(1:132)=11.198, R2 = 0.078, p <0.001; Table 2.2). Since ceca samples greatly differ compared to all other samples, I removed them from further beta diversity analyses. Pisaster surfaces have distinct bacterial communities compared to other sample types (other hosts, water, rocks, abiotic substrates; Adonis: Pseudo-F(5:104)=11.67, R2 = 0.331, p <0.001; Figure 2.1A). Furthermore, communities also differ by geographical location (Pseudo-F(2:104)=6.88, R2 = 0.078, p <0.001; Table 2.2). Focusing on only Pisaster external surfaces I found differences between the oral and aboral sides of the sea stars (Pseudo-F(1:47)=4.25, R2 =0.056, p <0.001) and across geographical locations (Pseudo-F(2:47)=8.55, R2 = 0.225, p <0.001). These two factors had a significant interaction (Pseudo-F(1:47)=3.74, R2 = 0.098, p <0.001;Figure 1B; Table 2.2). Pairwise comparisons show that the oral and aboral surfaces have distinct bacterial communities in the Port Moody location where sea stars were on creosoted pilings, but do not differ at the other two sites (Figure 2.1B).  20   Figure 2.1 Ordination plots of Pisaster bacterial community compositions. Differences in bacterial community composition between Pisaster and other sample types (stress = 0.188). B) Pisaster samples further cluster by oral or aboral surfaces and by geographical location (Stress = 0.149).  −101−1 0 1NMDS1NMDS2SamplesAbiotic SurfacesDermasteriasDermasterias CecaHenreciaHenrecia CecaOther HostsPisasterPisaster CecaSeawaterA: All Samples−0.75−0.50−0.250.000.250.0 0.5NMDS1NMDS2LocationBamfieldHakaiPort Moody RegionaboraloralB: Pisaster Only21    Table 2.2 Adonis results for community composition comparisons.     At the Hakai site, I sampled two sympatric sea star species, Dermasterias imbricata and Henricia leviuscula in addition to Pisaster. Analyzing only sea star surface samples from Hakai (e.g., no environmental and ceca samples), I find significantly different community structure on these three species (Pseudo-F(2:34)= 14.461. R2 = 0.460, p<0.001; Appendix Figure A.2; Table 2.2).  I investigated how removing low abundance sequence reads (LOR) from my dataset from individual samples affect my results. I found no differences in statistical outcomes (Adonis) between different filtering thresholds of low observation sequences (Appendix Figure A.3 and Appendix Tables A.1 and A.2). I then asked which bacteria make Pisaster communities distinct and to what degree bacteria are shared between Pisaster and environment (e.g., found on at least one host and one environmental sample). Here, I found that filtering rare bacteria out of samples had a large impact on my results. When all LOR are included in my shared analyses, there was a substantial overlap in shared OTUs between the aboral side of stars and the oral sides of sea stars, abiotic substrates, and water samples (Figure 2.2). Although, when I removed LOR the number of shared OTUs decreased for these groups. Ceca samples shared the least number of Test Factor	 Pseudo-F R2 df p-valueCeca	vs.	all	other	samples Ceca 11.198 0.078 1:132 <0.001Sample	type	vs.	location Sample	types 11.67 0.331 5:104 <0.001Geographical	location 6.88 0.078 2:104 <0.001Pisaster:	Body	region	vs.	location Body	Region 4.25 0.56 1:47 <0.001Geograpahical	location 8.55 0.225 2:47 <0.001Interaction 3.74 0.098 1:47 <0.001Hakai:	Sea	star	species Sea	star	species 14.461 0.46 2:34 <0.00122  OTUs, and this number also decreased further as rare OTUs were removed (Figure 2.2). I also found that Pisaster surfaces shared more OTUs with the environmental sources that they are more likely to be in contact with; there was little difference when all the rare OTUs were in the analyses, but differences emerged as rare sequence observations were removed (Appendix Figure A.4).   Figure 2.2 Heat map of the proportion of shared OTUs between Pisaster and the oral side of and other bacterial communities. The oral side of Pisaster, abiotic surfaces, seawater, and ceca of Pisaster at different filtering depths. Low observed reads (LOR) greatly how you interpret the proportion of shared OTUs of bacteria between sample types.  Identifying the Pisaster core community:  Core bacteria are bacteria that are more abundant on hosts compared to environmental bacterial communities across all natural sites that are sampled. This definition does not imply which type of symbiosis (mutualism, commensalism, parasitism) that is occurring. I used the Sloan neutral model to identify OTUs that are over-represented in comparison to the environment and other hosts. This was accomplished by comparing the aboral surface of Pisaster against different source bacterial communities (e.g., environment and other hosts) for each field 23  location, to prevent biases from unequal samples sizes that would affect the Sloan neutral model. Furthermore, this was only done for the aboral surfaces, since I saw composition and diversity differences between oral and aboral sides, and between sampling sites. There were between 67 and 127 over-represented OTUs on the Pisaster aboral surface at each geographical location compared to their environment (10 – 15% of the total number of OTUs per site). To establish which bacteria are in the core bacteria of Pisaster I then asked which OTUs are consistently over-represented across sites with different conditions and background communities to identify OTUs that are core bacteria of Pisaster overall. There are 38 consistently over-represented OTUs on the Pisaster surface. I focus further analyses on the 8 core bacteria that are also relatively abundant (greater than 1% relative abundance overall). The other 30 OTUs are nearly all variants (e.g., same distribution and close sequence similarity) of the abundant OTUs.  24  .  Figure 2.3 Stacked bar plot of core Pisaster bacteria that have a relative average abundance of 1% or greater for all Pisaster. Few bacteria make up the majority of bacteria detected on the surfaces of Pisaster; the bacteria that relatively abundant are either over-represented or unique to Pisaster. B) Stacked bar plot of bacteria found in the ceca that have a relative abundance of 1% or greater averages for all geographical locations.  25  Core bacteria Spirochaetes are the most abundant bacteria on Pisaster surfaces, and there are at least three clades that are associated with Pisaster. The dominant core OTUs, 13197 and 13140 that makeup ~14% and 34% of the total sequences from Pisaster, respectively (Figure 2.3B). They are on all Pisaster individuals included in this study. These two OTUs are 99.7 similar to each other and are closely related to a Spirochaetes previously found on the sea stars Linckia leavigata and Pentaceraster spp. (Jackson et al. 2018)( (Figure 2.4). These Spirochaetes differ in their distribution across sites; Spirochete OTU13197 was at a lower abundance at Port Moody compared. This demonstrates that the two OTUs do not co-vary and strongly suggests that these represent distinct organisms despite close 16S similarity. These Spirochaetes fall within a clade containing the genus Borrelia, the causal agent of Lyme disease, and the genus of Cristispira (Kuhn 1981), which is a large Spirochete associated with oysters morphologically similar to Borrelia (Paster et al. 1996, Fernandez-Piquer et al. 2012). All representatives within this clade are associated with hosts, including bivalves and octocorals (Figure 2.4). Although OTUs 13197 and 13140 are related to Borrelia, my sequences are only approximately 77% similar to the sequences in my tree that were associated with ticks or mice.  26   Figure 2.4 Phylogenetic tree of the dominant Spirochaete (OTUs 13140 and 13197) detected on the surfaces of Pisaster. Some of the bacteria in this tree are identified as Borrelia, and one is identified as Cristispira (U42638); the Borrelia are likely misidentified. OTUs 13140 and 13197 are approximately 78% related to the Borrelia that are terrestrial and associated with ticks. This is likely a novel clade of marine Spirochaetes that are associated with invertebrates. This tree was constructed with RAxML with 100 bootstrap replicates. Bootstrap confidence (n=100) that are greater than 50 are shown on nodes.  Spirochaetes related to Salinispira are also core bacteria on Pisaster; OTU29617 and OTU29659 represent ~10% and 1% of the total sequences found on Pisaster surfaces, respectively (Figure 2.3A). These two OTUs were also detected in the ceca at all three sites (Figure 2.3B). These OTUs cluster with sequences that were detected on five species of sea stars 27  (Figure 2.5). Some of the sequences were found associated with multiple sea star species. More broadly this clade contains sequences associated with marine sponges and environmental samples, such as microbial mats.   Figure 2.5 Phylogenetic tree of Salinispira OTU29617 and OTU29659 that are abundant on Pisaster. Similar sequences are found on other species of sea stars, some of which are found on multiple sea star species. Other similar sequences are associated with marine substrates or marine invertebrates. Numbers located at nodes represent bootstrap confidence (n=100) that are greater than 50.  OTU10720 is a Spirochaete that is distinct from the dominant Spirochaetes (OTUs 13197 and 13140) and is most abundant on Pisaster at Port Moody. OTU10720 is also more common 28  on the oral surfaces than aboral surfaces. It is represented by ~6% of all sequences found on the surfaces on the Pisaster (Figure 2.3A). This OTU falls within a cluster that is associated with a sponge and bivalves (Appendix Figure A.5). Those marine hosts fall among sequences that are primarily associated with microbial mats, and they may be associated with larval invertebrates that house that bacterium as a symbiont.  OTU29557 and OTU37708 are classified as Reichenbachiella (Bacteroidetes: Sphingobacteria) and are 95% similar to each other and make up ~4 % and 2% of the sequences found on the surfaces of Pisaster (Figure 2.3A). I found that these OTUs were more abundant on aboral surfaces than oral surfaces. Some closely related OTUs are associated with marine invertebrates and others are associated with environmental sources (Appendix Figure A.6). OTU5961 make up ~1% of the sequences found on the surfaces of Pisaster within the genus Peregrinibacteria (Figure 2.3A), which falls with the Candidate Phyla by the same name. This OTU does not cluster with sequences that were detected with marine hosts (Appendix Figure A.7). Of the sequences in this Peregrinibacteria phylogenetic tree, there is no clear pattern in the distribution between sequences detected with marine hosts and environmental samples.  The other OTUs that were unique or over-represented on Pisaster across all three sites had a relative abundance less than 1% and were almost all close sequence variants of the OTUs that are dominant, other than two Rickettsiales OTUs (Alphaproteobacteria) and one Herbaspiillum huttiense OTU (Betaproteobacteria; Burkholderiales).  29  To better understand whether the core bacteria change in abundances between oral and aboral surfaces I used DESeq2 (Love et al. 2014) to test for bacteria that are differentially represented between the oral and aboral sides of sea stars. 144 OTUs were differently abundant between the oral and aboral sea star; however, only four had an abundance of over 1%. Three core Pisaster bacteria, had a greater abundance on the aboral side of stars (OTU 10702; Spirochaete, OTU29557; Reichenbachiella, and OTU37708; Reichenbachiella. One abundant OTU was more abundant on the oral side of stars (OTU36954; Rubritalea), but it was restricted to the Port Moody site (Appendix Table A.3). 114 other OTUs were found to be differentially abundant but were under 1% in relative abundance.  I also investigated the dominant bacteria in the ceca of Pisaster, OTU31698 and OTU 29599 (Figure 2.3B), which are within the Phylum Tenericutes and in the genus Hepatoplasma. This group consists of bacteria that are associated with marine invertebrates (Figure 2.6). I was able to establish that the two dominant OTUs are indeed two separate groups, although they fall within the same genus. One of these two OTUs was present in every Pisaster, and often, when one was absent, the other made up as much as 90.3% of the relative abundance. OTU31698 was detected in 94% of Pisaster and had a relative abundance of 48.0% across all ceca samples; OTU 29599 was detected in 61% of stars and had a relative abundance of 24.5% relative abundance (Figure 2.3B; n= 18). Both OTUs are within this genus but do not cluster together in the tree. Almost all the sequences in this tree were detected with hosts, and the most similar sequences are associated with other sea star species; some sequences are found with multiple sea star host species.   30                                 Figure 2.6 Phylogenetic tree of Hepatoplasma. OTU29699 and OTU31698 fall within the genus Hepatoplasma. OTU31698 fall among many sequences that are associated with other sea star species and some sequences were detected on multiple species. Sequences that are similar to 31  OTU 29699 are also host-associated, and one is associated with a sea star species. Numbers that are located at nodes represent bootstrap confidence (n=100) that are greater than 50. Branches for sequences that are found on sea stars are in green.  2.5 Discussion: I show that Pisaster has distinct bacterial communities compared to their environment and any other sympatric hosts, even including sympatric sea stars (Figure 2.1A; Appendix Figure A.2). Furthermore, I find differences between our field sites (Figure 2.1B) and different body regions of the stars (Figure 2.1AB). Multiple studies have shown that hosts have unique bacterial community compositions compared to other hosts (McKenzie et al. 2012, Kueneman et al. 2014, Thomas et al. 2016b, Hernandez-Agreda et al. 2017), can vary across geographical space (Rodriguez‐Lanetty et al. 2013, Mortzfeld et al. 2015), and vary between morphological regions of an individual (Ainsworth et al. 2015, Hoj et al. 2018, Jackson et al. 2018). All these patterns are results of environmental filtering.  I break bacteria into two categories that are on or in Pisaster: transient bacteria and core bacteria. Transient bacteria are generally rare or infrequent across individuals and are consistent with stochastic assembly from local species pools. Core bacteria (Figure 2.3A) are frequent across populations and abundant, and I expect that there are specific mechanisms by which they reliably colonize Pisaster, such as vertical inheritance or specific mechanisms (Koskella et al. 2017) for environmental acquisition similar to the squid-Vibrio system (Nyholm and McFall-Ngai 2004). I concede that there is a spectrum of association between hosts and bacteria but focus on the extremes to more clearly examine the Pisaster bacteria. Others have classified bacteria into categories in the past. For example, bacteria have been described as ‘environmental responsive' (i.e., transient bacteria), ‘resident microbiota': based on an individual host species, 32  and ‘core' found across species of a taxonomic host group (Hernandez-Agreda et al. 2018). I have categorized bacteria into two simple categories that detect bacteria at the extremes of host specificity. My methods differ by comparing the distribution of bacteria present on a host compared to environmental samples: this allowed us to find bacteria that were over-represented and abundant on hosts. I also only focus on one host species, although my further analyses asked whether my core bacteria are on other closely related species.  The majority of bacterial diversity that I detected consists of transient bacteria. I used extensive sampling of the environment and other hosts in addition to using the neutral model to determine which bacteria are transient. Other studies have found transient bacteria that are infrequent and low abundance, although transient bacteria make up a large amount of the detected diversity (Loudon et al. 2014b, Hernandez-Agreda et al. 2018). These bacteria are often immigrating and emigrating or undergoing local extinction (Loudon et al. 2016). The transient bacteria detected on sea stars may be filtered by host defenses or are outcompeted by the core bacteria, or these bacteria could be from the water column, and an artifact of my sampling. Core bacteria are bacteria that are more abundant on hosts compared to environmental bacterial communities across all sampled natural sites. This definition does not imply the type of symbiosis (mutualism, commensalism, parasitism) that is occurring. Within the category of core, there is also varying strength of fidelity. For example, some of the core bacteria are differentially abundant across field sites (OTU10720; Marine Spirochaete, OTU29557, and OTU37708; Reichenbachiella). In addition, these bacteria have not been detected in other sea stars species, and some relatives were detected from environmental samples (Appendix Figures A5-7).  33  Other core bacteria have greater fidelity to Pisaster and are frequently detected and abundant on Pisaster regardless of geography or surface of the star. A prime example if this is dominant Spirochaete (OTU OTU29557 and OTU37708) and Salinispira. The closest relatives to these core bacteria are detected with other sea star species and other marine invertebrates, and some sequences are in multiple species of sea stars leading to an emergent hypothesis that these bacteria may have a long evolutionary history and play an important function to sea stars (Figures 2.4, 2.5).  Some relatives to the Pisaster core bacteria (i.e., dominant Mollicutes and Salinispira clades, and Spirochaete clade to a less extent) that are within my sea star clade are also found in multiple sea star hosts (Hoj et al. 2018, Jackson et al. 2018). Indeed, multiple species of sea stars have multiple closely related OTUs that are not sister taxa, i.e., the distribution of related OTUs to related sea stars is not indicative of co-evolution where you would expect concurrent phylogenies between hosts and symbionts (Moran et al. 2008). Importantly, some of these sequences were on multiple sea star species. This pattern is more indicative of host switching, where symbionts are not restricted to one host but can live on similar hosts. Host switching can be explained by a host being a niche space that a symbionts can fit within; closely related bacteria have similar niche requirements and be able to fit into similar niche spaces; however, their fitness may not be equal, depending on how well they fit within that niche space (Araujo et al. 2015). Host switching has been observed with marine worms and their chemoautotrophic bacteria (Zimmermann et al. 2016) and vesicomid clams and chemosynthetic bacteria (Ozawa et al. 2017). Furthermore, there is a clade of Endozoicomonas that are coral generalists (Pollock et al. 2018). There are examples of symbionts that have co-evolved, such as Buchnera and aphids 34  (Baumann et al. 1995, Clark et al. 2000), symbionts of halichondrid sponges (Erpenbeck et al. 2002) and a clade of Endozoicomonas and with corals (Pollock et al. 2018), but I expect that this is less frequent than host switching. Host specificity of symbionts is rare in macro-symbiosis. For example, feather mites do not have one specific host but are often found on multiple closely related hosts (Dona et al. 2018). Having host clades of bacteria that can live with multiple hosts is consistent with the results of a massive census estimating the global biodiversity of bacteria that shows finite bacterial diversity that is not hugely inflated by the incorporation of host-associated communities (Louca et al. 2019). Globally, this demonstrates that many symbiotic bacteria are associated with multiple hosts, as I found with Pisaster. The core bacteria Peregrinibacteria and Reichenbachiella, do not cluster with sequences that were detected with marine hosts and fall amongst sequences from environmental samples (Appendix Figure A6 and A7). Peregrinibacteria is part of the candidate phyla radiation and little is known about this group other than from some recent genomic studies that focused metabolic capabilities of Peregrinibacteria from groundwater (Anantharaman et al. 2016). Genomic studies from marine host-associated members of Peregrinibacteria would be useful for predicting the function of the Peregrinibacteria that are associated with Pisaster. Reichenbachiella has been associated with seaweeds (Huggett et al. 2018) and the specialized cells of sponges (Godefroy et al. 2019). I had access to few bacteria that closely match these Pisaster core bacteria, and as more data becomes available; host clades may become apparent.  I was unable to determine the function of the bacteria associated with Pisaster, and further studies are needed to determine specific functions. For surface communities, I predict that these bacteria are playing a role in preventing fouling. Surface bacteria producing anti-fouling 35  chemicals have shown to be defensive in marine organisms (Holmstrom et al. 2002, Ma 2009). Bacteria on their surfaces could also provide disease defense. For example in one study a high prevalence of bacterial strains cultured from sea star surfaces were able to inhibit pathogenic bacterial strains (N=38; 21%), which was much greater than the bacteria cultured from sea slugs (N=19; 0%) or seaweed (N=21; 5%) (Burgess et al. 1999). If bacterially derived chemicals are the mechanism for disease defense or anti-fouling, this is likely the result of interference competition, which occurs when bacteria compete against each other by chemical warfare and produce a biologically relevant compound as a by-product (Scheuring and Yu 2012). A study found that the density of bacteria on the surfaces of tropical sea stars was between 104 and 105 cells/cm2 and was an order of magnitude higher than on neutral surfaces (e.g., small stones), which may make interference competition more likely. Two Mollicutes OTUs dominate the ceca (Figure 3C) and had substantially less diversity than surface communities. It is unclear whether these bacteria exclude each other or whether one dominating the community is an artifact of the data being compositional (Gloor et al. 2017). Low diversity communities are a trend in some insects (Jones et al. 2013). The diversity that I detected in Pisaster differs from other recent sea star studies. Jackson et al. (2018) did not find a low diversity of bacteria across ceca samples of multiple species of sea stars compared to seawater. In addition, Hoj et al. (2018), found a relatively "high" diversity in the ceca compared to gonads, body tissue and tube feet, however, did not sample seawater as a comparison. For both studies, sampling techniques and filtering of rare data (e.g., OTUs) may greatly impact the interpretation of these results. I predict that bacteria in the ceca may facilitate the breakdown nutrients that sea stars cannot breakdown on their own (e.g., fermentation) or could protect 36  against pathogens. Mollicutes that were detected on sea stars in the Jackson et al. (2018) study were closely related to the Mollicutes found in the ceca of Pisaster. This indicates that these bacteria may have an evolutionary relationship with sea stars or fit the sea star niche well, and due to this, more research on this group is warranted.  I am unable to determine how Pisaster acquire their core bacteria. Since these bacteria are rare in the environment, they are less likely to be acquired by environmental transmission; however, rare species in the environment with strong environmental filtering could make this possible. Due to the distribution and abundance of these bacteria with Pisaster, they are candidates to be vertically transmitted. Whether vertical transmission is the mode of colonization could be established by testing the gonads of reproducing adults and larvae of Pisaster for the presence of the core bacteria. Bacteria have been detected in the gonads of sea stars (Hoj et al. 2018, Jackson et al. 2018) and detected in larval sea stars (Bosch 1992, Galac et al. 2016). Furthermore, vertical transmission has been observed in broadcast spawning corals, where bacteria are associated with gametes (Leite et al. 2017), so it is possible that sea stars, who reproduce by broadcast spawning, are capable of acquiring symbionts through vertical transmission. There are other close relationships between a host and symbiont that are not vertically transmitted and are horizontally transmitted, such as bobtail squid and deep-sea bivalves (Nyholm and Nishiguchi 2008, Roeselers and Newton 2012). Knowing whether these bacteria are vertically transmitted would provide insight into the potential host dependence of a symbiont; a meta-analysis across 38 host-symbiont pairs found that losing vertically transmitted symbionts results in a greater fitness loss than losing horizontally transferred symbionts (Fisher et al. 2017). 37  I have taken a closer examination of the bacterial communities associated with Pisaster by comparing larger sample sizes and environmental samples than most studies. This allowed me to compare the bacterial communities of Pisaster with the environment and other hosts to determine which bacteria are Pisaster-specific rather than a generalist and also found in other bacterial communities. In addition, I tested how filtering low observation sequence reads from individual samples affect my analyses and found that filtering rarely-detected sequences has no effect on observing differences in bacterial composition (Appendix Figure A.3; Appendix Tables A.1 and A.2), but does affect the number of OTUs that are shared between sample sites (Figure 2.2A; Appendix Figure A.4). Removing rarely detected sequences not affecting composition is explained by my method of determining community similarity, Bray-Curtis, which accounts for abundance, so rare community members have little impact when computing a similarity matrix. Indeed, since transient bacteria are numerous, rare, and infrequent, I suggest using a weighted similarity metric when focusing on bacteria that are host specialists. The large influence that is removing LOR for comparing shared bacteria is explained by the reality of 16s amplicon sequencing data having a long tail of low abundant reads. Low abundance sequences should be accounted for with hesitation since sequence errors may indicate their presence in a sample even when they are not present. Studies that account for all the data when computing shared bacteria are inflating the actual shared results (Lemay et al. 2018). This study highlights that sea stars have a few specific bacteria are frequent across geographical locations that are hundreds of kilometers apart. The bacteria that are most abundant and frequent (Spirochaete, Salinispira, Mollicutes), belong to clades, where little of the diversity is known. From my phylogenetic analyses, I show that these bacteria are found to be associated 38  with other sea stars. This pattern is striking in some cases, such as with Salinispira, where similar sequences cluster together and consist of a multitude of sea star hosts that live across the globe (Figure 5). These results indicate that sea stars may have an evolutionary history with these bacteria; further studies should focus on characterizing the bacterial communities across sea star taxa and across the globe to determine if these patterns of host specificity strengthen. Furthermore, future studies should focus on determining the function of these bacteria and whether they are obligate or facultative. Altogether, I have identified a procedure to better characterize host communities by extensively sampling potential sources of bacteria, I characterized the Pisaster microbiome in-depth, and I identified clades that may have long evolutionary histories with Asteroidea.   39   Elucidating the skin microbial ecology of Columbia spotted frogs Chapter 3: 3.1 Abstract:  Global amphibian declines due to the fungal pathogen Batrachochytrium dendrobatidis (hereafter Bd) have led to questions about how amphibians defend against skin diseases. Studying amphibian species that are resistant to Bd can inform understanding of disease defense mechanisms that are effective against Bd and could lead to mitigation strategies. Both antimicrobial peptides (AMPs), a component of amphibian innate immune defense, and symbiotic skin bacteria can inhibit Bd. These disease defense factors can synergistically inhibit Bd, but their interactions are not well understood. Here I characterize each of these factors in four populations of Columbia spotted frogs (CSFs), a species that is resistant to Bd. We measured the ability of AMPs to inhibit Bd, the composition and predicted antifungal function of symbiotic bacteria, the bacterial metabolite profiles from the frogs' skin and the presence and intensity of Bd in four populations of CSFs with variable Bd prevalence and intensity. We found that AMPs from CSFs inhibit Bd, but AMPs ability to inhibit Bd did not correlate with Bd status or intensity. The bacterial community is composed of prevalent core bacteria: Rhizobacter and Chryseobacterium are dominant, present on all frogs, and negatively correlated with each other. The abundance of Rhizobacter negatively correlates with AMPs ability to inhibit Bd, and negatively correlates with Bd intensity. Bacterial composition and AMPs ability to inhibit Bd did not correlate with bacterial metabolite composition. The four populations have differences in Bd presence and the compositions of Rhizobacter and Chryseobacterium, which may drive some of these findings. Further studies with more sampling locations are needed to determine if AMPs affect the relative abundance of Rhizobacter, which in turn affects Bd status. Future studies 40  should determine the direct role of AMPs on these bacteria and whether Rhizobacter and Chryseobacterium are competitors. My results suggest that the AMPs from CSFs may affect CSF skin bacterial communities.   3.2 Introduction: Some amphibians are susceptible to a fungal pathogen, Batrachochytrium dendrobatidis (hereafter Bd), which is the causal agent of chytridiomycosis, a disease linked to global amphibian declines (Skerratt et al. 2007); some species are more susceptible to the disease than others and, subsequently, are more at risk of extinction (Woodhams et al. 2007a, Gahl et al. 2012).  Due to the threat of extinction and rapid decline of some species, much attention has been focused on identifying how amphibians defend against Bd infection and whether mitigation is feasible (Woodhams et al. 2011, Bletz et al. 2013). In this study, we focus on Columbia spotted frogs (Rana luteiventris), which appear to be resistant to Bd, and maintain consistent but low-level infections (Russell et al. 2010). Using a resistant species allows us to study the mechanisms of disease defense that are effective against Bd, or the magnitude of which Bd affects their microbiota (Jani and Briggs 2014); this knowledge may promote the development of mitigation strategies for susceptible species.  Amphibians defend against pathogens by utilizing adaptive immunity (targeting specific pathogens), innate immunity (nonspecific defense) and symbiosis with bacteria. Although amphibians have adaptive immunity capabilities, these defenses are rather ineffective against Bd, as Bd produces chemicals that mask itself from the amphibian immune system (Fites et al. 2013, Fites et al. 2014, Rollins-Smith et al. 2015). Some amphibians produce antimicrobial peptides, 41  but not all, e.g., Atelopus varius, Hyla, Pseudacris, Bufo, Scaphiopus (Conlon et al. 2004, Woodhams et al. 2006b), a component of the innate immune system that inhibits Bd (Rollins-Smith et al. 2002, Rollins-Smith et al. 2005). Another mechanism of disease defense for amphibians is the bacteria on their skins and the antimicrobial metabolites they produce (Harris et al. 2006, Harris et al. 2009).  Antimicrobial peptides, a component of the innate immune system, are small proteins that are produced by the granular gland of many amphibians (Rollins-Smith et al. 2005). Numerous species have been shown to produce AMPs that can inhibit Bd, although their effectiveness can vary (Woodhams et al. 2006a, Woodhams et al. 2006b). Some common peptides are shared among species, but there is generally a suite of species-specific peptides (Conlon et al. 2004, Rollins-Smith et al. 2005). Species that have more effective AMPs against Bd have a greater survival rate when challenged with Bd (Woodhams et al. 2007a). A mitigation strategy using AMPs against Bd would potentially include selectively breeding of amphibians possessing AMPs that are more effective against Bd (Woodhams et al. 2011). Bacterial communities associated with amphibian skins are species-specific (McKenzie et al. 2012), differ from the environmental bacterial communities (Loudon et al. 2014b), across geographical space (Kueneman et al. 2013), by host ecology (Bletz et al. 2017a) and along environmental gradients, such as elevation (Hughey et al. 2017b, Muletz Wolz et al. 2018). Recent work highlights a strong influence of the abiotic environment on amphibian microbiota, with communities differing due to temperature (Kueneman et al. 2019) water conductivity (Krynak et al. 2016), and land use (Krynak et al. 2016, Hughey et al. 2017b). These patterns likely result from a combination of environmental filtering by the abiotic environment (Kraft et al. 2015) and by the host habitat (e.g., cutaneous mucous layer, host AMPs), where the 42  fundamental niche of some bacteria fit within the amphibian habitat, and then non-host environmental factors influence biotic interaction of the bacteria on the amphibian.   Bacteria that protect against Bd do so by producing metabolites that inhibit Bd growth (Brucker et al. 2008a, Brucker et al. 2008b, Becker et al. 2009) or cause negative chemotaxis of Bd zoospores (Lam et al. 2011); bacteria may also compete for space with Bd; however, this has yet to be tested. Woodhams et al. (2015) have compiled a database of bacteria that can inhibit Bd, and this can be used to predict the antifungal function of skin bacterial communities (Bletz et al. 2017c). The production of these metabolites may be the result of interference competition (Loudon et al. 2014a), following the model for the genesis of beneficial symbiosis put forth by Scheuring and Yu (2012). According to this model, bacteria compete over a nutrient source (e.g., cutaneous mucus) and produce chemicals to inhibit their competitor, which indirectly results in protection against a pathogen (Scheuring and Yu 2012). The metabolites produced by symbiotic bacteria can be detected noninvasively on the skins of amphibians (Umile et al. 2014), and have experimentally been shown to differ with Bd infection (Walke et al. 2015).  Symbiotic bacteria and AMPs interact with each other, which can influence Bd growth. For example, symbiotic bacteria and AMPs produced by amphibians synergistically inhibit Bd in vitro (Myers et al. 2012).   Additionally, there is evidence that AMPs can stabilize bacterial communities.  Kung et al. (2014) demonstrated that bacterial bioaugmentation did not affect cutaneous bacterial community composition, but induced a specific AMP to be up-regulated, indicating that the AMPs were stimulated to stabilize community composition.  Indeed, specific AMPs have been correlated with specific bacteria, supporting that they have a role in shaping bacterial community structure (Davis et al. 2017). Furthermore, bacteria can affect AMPs: the addition of a bacterial probiotic cocktail down-regulated a specific AMP that is anti-Bd 43  (Woodhams et al. 2019). Therefore, knowing how AMPs and bacterial communities interact across amphibian species is important to understanding disease dynamics and in developing mitigation strategies. I investigated the interactions between a component of the amphibian innate immune system and symbiotic bacteria of the Bd resistant Columbia spotted frog (CSF) to understand their influence on the ecology of amphibian microbiota and Bd presence and intensity. I surveyed 40 CSF across four locations and tested them for the presence of Bd, characterized their bacterial communities via 16s rRNA sequencing, surveyed their metabolite profiles through HPLC-MS, and collected and tested the efficacy of their AMPs against Bd using in vitro assays. I hypothesize that 1) frogs with more inhibitory AMPs have less Bd, 2) AMP effectiveness associates with specific bacteria, 3) metabolite profiles correlate with bacterial communities, and 4) greater abundances of antifungal bacteria associate with lower Bd prevalence and intensity.  3.3 Methods:  Sampling location Columbia spotted frogs (CSF, Rana luteiventris) were collected from four lakes in central Montana. They were Gipsy Lake, Doney Lake, Park Lake, and Jones Pond at the Lubrecht Experimental Forest. Park Lake and Gipsy Lake are cooler, and Doney Lake and Jones Pond are warmer, which is informative because Bd thrives at cooler temperatures (Piotrowski et al. 2004). Environmental parameters are in Appendix Table B.1. These locations were chosen because they have CSF that have tested positive for Bd in previous seasons, and range in altitude (Global Bd mapping project, Sheafor, unpublished data). Ten frogs were collected by hand (wearing sterile 44  nitrile gloves) at each of these locations. Each frog was then sampled to collect metabolites, AMPs, and microbial DNA for bacterial community analysis and Bd quantification (Table 3.1). This study followed Carroll Colleges Animal Care Policy and was approved by the IACUC committee (CC0002). Since frogs were not removed from the site of capture and sampling was invasive, a permit from the state of Montana was not necessary in accordance with their laws.  Table 3.1 Samples that were processed and successfully analyzed by site and sample type.   Sample collection After capture, each frog was swabbed ten times with a sterile polyurethane swab (14-960-3J, Fisher Scientific) on a randomly chosen side of the ventral surface (left or right side) to collect metabolites produced by skin bacteria. Before collection, swabs were cleaned in methanol to remove any methanol-soluble impurities (Umile et al. 2014). Metabolite swabs were placed in sterile tubes and stored on dry ice until they were stored at -80°C.  Following the collection of metabolites, amphibians were rinsed with sterile Provasoli medium (Wyngaard and Chinnappa 1982) to remove transient bacteria (Culp et al. 2007) Frogs were swabbed ten times on the previously un-swabbed ventral side using rayon swabs to collect Location Bd	analyses Bacterial	analyses Metabolite	analyses AMP	analysesDoney	CSF 10	(10) 10	(8) 10	(10) 2	(2)Gipsey	CSF 10	(10) 10	(7) 10	(9) 7	(7)Jones	CSF 10	(10) 10	(10) 10	(10) 4	(4)Park	CSF 10	(10) 10	(9) 10	(9) 6	(6)Doney	Water 3	(3) 3	(1) - -Gipsey	Water 3	(3) 3	(3) - -Jones	Water 3	(3) 3	(3) - -Park	Water	 3	(3) 3	(3) - -()	Indicates	samples	that	were	successfully	analysed45  the microbiota and Bd cells for DNA extraction. Swabs were placed in sterile tubes and stored on ice until they could be stored at -80°C until DNA extraction. Disposable nitrile gloves were changed between each amphibian. To determine the bacterial composition of the water the frogs were caught in, three 50 mL water samples were obtained at each site and filtered using Whatman 0.22 micron filters (GE Life Sciences, Pittsburgh, Pennsylvania) to collect DNA to determine bacteria communities. Frogs were placed in 50 mL of sterile collection buffer (50 mM NaCl, 25 mM CH3COONa, pH=7.0) containing 500 µl of 20 mM norepinephrine hydrochloride for 15 minutes to obtain skin secretions (Sheafor et al. 2008). After the frog was removed from the bath, the collection buffer was acidified with 500 µl of trifluoroacetic acid (TFA) and passed over a C-18 Sep-Pak cartridge (Waters Corporation, Milford, Massachusetts, USA). Both the Sep-Pak and the collection buffer were kept on ice until they were stored in a refrigerator. After a final rinse with sterile Provasoli medium, all animals were released at the site of capture to remove AMP collection buffers. DNA Extraction: DNA was extracted using a MoBio Power Soil DNA isolation kit according to the manufacturer's protocol with an adaptation from McKenzie et al. (2012). The DNA was used to quantify Bd infection status (positive or negative) and intensity (how infected: genetic equivalents (ge)) and to determine bacterial community composition.  Testing for Bd  46  To test for the presence and intensity of the fungal pathogen, Bd, on the skins of Columbia Spotted frogs, I used quantitative Polymerase Chain Reaction using PerfeCTa qPCR Fastmix II (QuantaBio, Beverly MA) in 20 ul reactions using the reaction times and primers from (Boyle et al. 2004). I used amplicon standards (Pisces Molecular, Boulder CO.) to establish the copy number of the ITS gene (a marker for Bd) for each frog. Tests were performed in triplicate, and samples had to have a positive value in 2/3 of the replicates to be considered positive. One sample only had two replicates, and another only one replicate; however, they were left in the analyses to account for all frogs. To test for differences in Bd levels, I used the Kruskal-Wallis analysis (Kruskal and Wallis 1952) in R (Team 2015) and Dunn test in FSA (Ogle 2015). Bd status refers to whether an individual is infected, and Bd intensity will refer to how infected an individual is, and is measured as genomic equivalents.  Antimicrobial peptides Peptides were eluted from C-18 Sep-Paks using a solution of 70% acetonitrile, 29.9% water, and 0.1% TFA (v/v/v). Eluted peptides were concentrated to approximately 1 mL using a CentriVap benchtop centrifugal concentrator system (Labconco, Kansas City Mo). The concentration of eluted peptides was determined via a micro BCS assay using bradykinin as a standard. The effectiveness of the secretions against Bd was determined using a variation of the bioassay from (Sheafor et al. 2008). In short, zoospores from a cultured plate of Bd (strain PTH02, collected initially from British Columbia, and obtained from Dr. Joyce Longcore, University of Maine) were collected and diluted with 2% tryptone broth to a concentration between 5x104 and 5x105 zoospores in a volume of 50 µl. Concentrations were determined using a hemocytometer. Three replicates of each zoospore dilution were plated in a 96-well 47  microtiter plate and mixed with 50 µl of peptide dilutions ranging from 5 to 200 µg protein/mL for each animal. Growth over 10 days was measured daily as the change in absorbance at 492 nm using a Titertek Multiskan plate reader (Titertek Instruments Inc., Huntsville, Alabama). A minimal inhibitory concentration (MIC), the lowest concentration of crude peptides at which no growth is detectable (Woodhams et al. 2006a, Sheafor et al. 2008) was calculated to determine how active peptides from a specific frog were in combatting Bd infection. All Bd assays were performed at Carroll College in Helena Montana.  For this study, MIC analyses are normalized to ug protein/ml per cm2 of frog skin to control for different sized frogs. Surface areas of frogs were calculated using an equation for Ranids: Surface Area=1.107*Mass(g)0.606 (Hutchinson et al. 1968). To test for differences in the abundance of peptides collected and for MIC, I used Kruskal-Wallis analyses with a Monte Cristo distribution with 10,000 permutations in the R package Coin (Hothorn et al. 2008). This test was used since there were small sample sizes and uneven sampling between locations. To test for relationships between AMPs and Bd for all frogs in this study, I used a logistic regression and Kendall's rank correlation test. I also used mixed models to account the sampling location of the frogs to test if AMP MIC or protein concentration /surface area explain Bd status. I used a mixed-effects logistic regression with the location as the random effect (Bates et al. 2014). I also tested whether protein concentration is related to MIC by using a linear mixed model with the location as a random effect with the package lme4 (Bates et al. 2014).  Identifying Bacterial communities To determine the identities and relative abundances of bacteria associated with CSFs I amplified the V4 region of the 16S rRNA gene to detect bacteria and archaea using a redesigned 48  version of the 515f and 806r primers: 515f: 5'–GTGYCAGCMGCCGCGGTAA–3', 806r: 5'–GGACTACNVGGGTWTCTAAT–3'. These primers have been modified to include a 12 bp Golay barcode on the forward primer and degeneracies were added to improve taxonomic coverage (http://www.earthmicrobiome.org/). Each PCR reaction contained 10µl of phusion Master Mix, 1µl of each primer (final concentration = 0.2µM each), 1µl of DNA, and PCR grade water to a final volume of 25µl. PCR started with an initial denaturation step at 94˚C for 3 minutes, followed by 35 cycles of denaturation at 94˚C for 45 seconds, primer annealing at 50˚C for 60 seconds, and extension at 72˚C for 90 seconds, with a final extension step of 72˚C for 10 minutes. PCR products were quantified using Quant-IT Pico Green® ds DNA Assay Kit (Life Technologies). Equal quantities of DNA (25ng) from each sample were pooled and then purified using the MoBio UltraClean® PCR clean-up kit. Pooled library quantitation and paired-end Illumina MiSeq sequencing (2 x 300bp) were carried out at the Integrated Microbiome Resource (IMR) facility in the Centre for Genomics and Evolutionary Bioinformatics at Dalhousie University (Halifax, Canada). Raw sequences were demultiplexed using Quantitative Insights into Microbial Ecology (QIIME v. 1.9.1) with a pfred score of 20 to ensure quality sequence (Caporaso et al. 2010b). The sequences were then trimmed to 250 bp using the FastX toolkit (http://hannonlab.cshl.edu/fastx_toolkit/). Operational taxonomic units were made using the Minimum Entropy Decomposition method (MEDs) (Eren et al. 2015). This was performed with the minimum substantive abundance parameter (-M) set at 250 reads (i.e., 250 sequences of an OTU must be present for it to be included in the data). All other parameters were at default settings. Taxonomy was assigned using uclust (Edgar 2010) and the SILVA 128 99% sequence 49  similarity database. Chloroplast, mitochondrial, and sequences unidentified at the domain level were removed. Taxonomy was assigned using uclust and the Silva 132 99% database. Two OTUs (Achromobacter and Pseudomonas) were present in negative controls and every sample and were removed as likely lab or reagent contaminants. OTUs that had fewer than five reads per sample were removed on a per sample basis to minimize the impact of barcode switching and more accurately compare shared OTUs (See chapter 2). Samples were rarified to each have 1,150 sequences, hence removing any samples with fewer than 1,150 sequences; this number was chosen to include the greatest number of samples (Table 3.1).  I calculated Chao1, which is the predicted richness of a community based on the data, to test for differences in bacterial richness across sites and differences with Bd status. To test for differences in bacterial community compositions, I calculated Bray-Curtis dissimilarity in Phyloseq (McMurdie and Holmes 2013) then performed PERMANOVA tests using the Adonis function in Vegan (Oksanen et al. 2015). A one-way PERMANOVA was used to test for differences between CSFs and water. A two-way PERMANOVA was performed to test for differences in the composition of CSF communities by site and Bd status. For posthoc comparisons, I used pairwise.adonis (Arbizu 2017). Furthermore, I examined how many bacteria were shared between CSFs and water, as well as between the frogs based on location, using shared phylotypes in QIIME.  Antifungal matches I used a database of 16S sequences from bacterial isolates taken from amphibian skin and experimentally shown to inhibit Bd (Woodhams et al. 2015) to determine the prevalence of bacteria that match known antifungal isolates. This allows us to infer which OTUs might have 50  the ability to inhibit Bd. I clustered my data against the Woodhams antifungal database with closed reference OTU picking at 97% similarity. This sequence similarity was used since 97% similarity binning is often used for making OTUs. To test for a relationship between antifungal bacteria and Bd, I used Kendall's rank correlation tests (intensity) and logistic regressions (status). To account for location, I performed a Mixed Effects Logistic model with the location as the random effect.  Core bacteria  I defined core bacteria on CSFs as bacteria that are more abundant on hosts compared to environmental bacterial communities across all sampled natural sites. This definition does not imply which type of symbiosis (mutualism, commensalism, parasitism) that is occurring. Core OTUs were determined to be over-represented on CSF compared to water at all four locations using DeSeq (Love et al. 2014). I focused on OTUs that are considered common, meaning they consist of greater than 1% average relative abundance, in addition to having a having greater than 1% relative abundance on more than 5 % of the frogs. This allowed me to focus on bacteria consistently associated with CSF and excludes OTUs dominant only in one sample. I tested for correlations between each core bacterium and Bd for all frogs using Kendall's rank correlation (Bd intensity) and logistic regressions (Bd presence). There are substantial differences in Bd status by site, so I also ran a Mixed Effects Logistic regression of relative abundance of core bacteria and Bd status that included location as a random effect. To test for a relationship between the two most abundant bacteria, I used a Linear Mixed Effects model with Chryseobacterium as the fixed effect and location as the random effect. Furthermore, I tested whether the dominant antifungal bacteria were correlated with the relative abundance of other 51  bacteria (antifungal other than Rhizobacter) that match the antifungal database within my data using a Linear Mixed Effects model with the location as the random effect.  Taxonomy of OTU87682 The taxonomy acquired for OTU87682 using the described methods was identified as Rhizobacter, and in Blast, OTU87682 was identified as Piscinobacter (Madden 2013); however, in a taxonomy assignment with SILVA 132 and using DADA2 (Callahan et al. 2016), OTU87682 was only identified to the level of family: Burkholderiaceae. To determine the proper taxonomy of OTU87682, I extracted all of the 99% Rhizobacter and Piscinobacter sequences from SILVA 132 (Quast et al. 2013) (410 sequences) and combined them with ASV13 and aligned them with SINA (Pruesse et al. 2012). Alignments were filtered using QIIME with a gap fraction of 0.99 and an entropy threshold of 0.05. A tree was made using RAxML with a 100 bootstraps (Stamatakis 2014). This resulted in OTU87682 falling within the Rhizobacter clade.  Phylogenetic tree of close relatives of Rhizobacter and Chryseobacterium I made phylogenetic trees of OTU87682 (Rhizobacter) and the closest SILVA sequences. I also made a tree for the OTU39630 (Chryseobacterium), both bacteria were common and found on Oregon Spotted frogs in Chapter 4. For both trees, sequences were acquired from SILVA and NCBI and aligned with SINA (Pruesse et al. 2012). Alignments were filtered with QIIME using a gap fraction of 0.99 and an entropy threshold of 0.05. A maximum-likelihood tree was constructed using RAxML with a 100 bootstraps (Stamatakis 2014). Metadata was retrieved from Genbank database using EukRef (del Campo et al. 2018). Metabolites 52  Metabolites were extracted with methanol, and HMPC-MS was performed with naphthalene used as a positive control following the methods described in Umile et al. (2014). This is a non-invasive assay that detects the presence of small methanol soluble organic compounds. This non-targeted approach does not identify the structure or name of the metabolites. This method results in knowing whether a metabolite, identified by retention time, is present on a frog and allows for the analyses based on the richness of metabolites for a specific frog, and the ability to use non-weighted ‘community' similarity matrices between individual frogs. The richness of metabolites is comparable between samples since this is a non-targeted approach. A Kendall's rank correlation test and logistic regressions were used to determine relationships between metabolite richness and Bd intensity and Bd status, respectively. Mixed-Effects logistic regressions were used to determine relationships between metabolite richness and Bd load, antifungal bacteria prevalence, and AMP MIC. Pearson correlations were performed to test for associations between metabolite richness, antifungal bacteria prevalence, and AMP MIC for all samples. The Holm p-value correction was used to account for multiple comparisons.  To examine metabolite composition, I used the Adonis test in Vegan, using the Sorensen index, to determine differences in metabolite composition between location and Bd status. The Sorensen index was used since only the profile of metabolites rather than abundances of the metabolites were identified. As a posthoc test to determine differences between the four locations, I used pairwise.adonis (Arbizu 2017). Lastly, I used a Mantel test in Vegan (Oksanen et al. 2015) to test for a correlation between bacterial community composition (Bray-Curtis) and metabolite composition (Sorensen).   53  3.4 Results:  Batrachochytriun dendrobatidis status and intensity  I determined the status (whether Bd is present or absent on a frog) and intensity (Bd load measured as genetic equivalents (ge) by qPCR) of Bd in 40 frogs from four field locations. Twenty frogs were positive for Bd, and 20 were negative. Bd status differs by location: all frogs at Gipsy were positive for Bd while all at Jones were negative, and the other two locations had both positive and negative individuals (Figure 1). Bd intensity ranged from 0 to 24,692 genetic equivalents (ge). I found differences in Bd intensity between locations using a Kruskal-Wallis multiple comparisons (chi-squared = 18.615, df = 3, p = 0.0003); a Dunn test was used to test for differences between groups (Figure 1).   Figure 3.1 Bd intensity (log base 10) for all four sampling locations. Red points represent Bd negative individuals, and blue points represent Bd positive individuals. Bd intensity varies by site; Letters denote differences established by the Dunn test with alpha <0.05. Points are jittered to avoid overlap and do not signify any biological factor.   54  Inhibition of Bd by Antimicrobial peptides    AMPs were from frog skin by bathing them in norepinephrine to elicit a stress response and then tested the ability of these AMPs to inhibit Bd for a subset of frogs. AMPs from CSF were randomly selected for 19 frogs, and challenge assays were run to test whether AMPs inhibited Bd. Proteins were detected in all samples, and the concentration ranged between 204 and 2,556 ug protein/cm2 frog skin. These secretions had MIC against Bd values from 0.12 to 1.11 ug protein/ml per cm2; this is the lowest quantity of AMPs needed to inhibit Bd in an in vitro assay. Samples with a low MIC have a greater Bd inhibition. All comparisons of protein concentration and MIC against Bd are normalized for surface area. There was no difference in protein concentration or MIC across locations (Appendix Table B.2); Doney Lake was excluded from cross-site comparisons because of low sample size. Protein concentration is negatively correlated with MIC overall, taking location into account (linear mixed model, t1.400= -3.260, p = 0.0057, cor = -0.837, with location as a random effect, Appendix Figure B.1). I then tested if protein concentration or MIC explain Bd status with logistic regression; they do not (Appendix Table B.10). Similarly, MIC is not correlated with Bd intensity (Kendall test, Appendix Table B.9). Taking location into account using mixed-effects logistic regression did not change results (Appendix Figure B.2).   Bacterial composition Amplicon sequencing of the 16S rRNA gene yielded 440,654 sequences binned into 873 OTUs through minimum entropy decomposition after quality filtering across 44 samples (34 55  frogs and 10 water samples; Table 3.1); sequencing for 6 frog samples failed. Colombia spotted frogs had different bacterial community compositions than water samples (PERMANOVA: F1:42= 17.677, R2=0.2962;p = 0.001; Figure 3.2A). In addition, communities were different by location, but not different by Bd status when nested into location (i.e.- there was no interaction; PERMANOVA: location: F3:28= 4.8955, R2=0.33187; p = 0.001. Bd: F1:28= 0.7635, R2=0.0.01725; p = 0.51; Interaction: F1:28= 0.8042, R2=0.01817; p = 0.485, Figure 3.2B). Bd status affected community composition when location was not accounted for (PERMANOVA Bd alone: F1:33= 5.6968, R2=0.1511; p = 0.003), but this is likely driven by strong differences in Bd load across sites. Estimated bacterial richness (Chao1 index) did not differ across sites for CSF, but richness was significantly higher in water compared to frogs (ANOVA, F= 21.67, df= 4:39 p < 0.00001; HSD test alpha =<0.05, Figure 3.3A).  Approximately 74% of the bacterial OTUs found on CSFs were shared (i.e., found on at least one frog and one water sample) with the water (Figure 3.3B), although these bacteria were at low average relative abundance (less than 1%). There were 39 bacteria on CSFs that were shared across all locations, regardless of being core or transients. 56   Figure 3.2 Bacterial community composition of CSFs. Communities differed between CSFs and water, (stress = 0.142) B: Compositions differed between the location of sampling, but not based on Bd status (stress = 0.206). Bd intensity affected composition when the location is not in the model.  .  Figure 3.3 Alpha diversity and proportion of bacteria shared with water. A) Chao1 bacterial diversity of CSF and water (pooled). Water had a greater diversity than CSFs. B) Approximately ¾ of the bacteria detected on CSFs were also in the water.  57   Core bacteria   I then defined the core and transient components of the CSF microbiota by comparing distributions across sites and with water from which the frogs were collected. I defined core bacteria as bacteria that are more abundant on hosts compared to environmental bacterial communities across all sampled natural sites. In contrast, transients are defined as OTUs with similar abundance on CSF and in the water column, or more common in the water. Using DESeq, I identified 15 OTUs that were significantly more abundant on CSFs than in water. For all the bacteria detected on CSF, the 15 core bacteria make up ~78% of the total sequence reads on CSFs. I find 333 transient OTUs on CSF: 133 OTUs were significantly less abundant on frogs compared to water (comprise ~2% total sequence reads), and 200 OTUs were at similar abundance on frogs and water (~17% total sequence reads of CSF microbiota; all of these are considered transient bacteria). To focus on bacteria that might be important to CSF, I established a core group of bacteria that were significantly more abundant on CSF and used a 1% average relative abundance threshold across frogs: In addition they had to be greater than 1% relative abundance on more than 5% of the frogs to allow us to focus on the dominant bacteria that are not outliers (e.g., high abundance in one sample) (Figure 3.4A; Appendix Tables B.4 and B.5). The other 10 bacteria that were significantly greater on frogs than in water had an average relative abundance of less than 1%; all of these OTUs were variants of the dominant five core OTUs (Appendix Table B.11).  Rhizobacter (OTU87682) was the most abundant core bacterium (33% of total sequence reads) and was present on all frogs. The average relative abundance of Rhizobacter is greater in 58  Jones Pond where Bd is absent than at Gipsy Lake where Bd is the most abundant and frequent (Kruskal-Wallis multiple comparisons with Holm correction: z= -3.4, p=0.004) (Figure 3.4b). Rhizobacter was positively associated with AMP MIC against Bd overall (Pearson: t15=2.5069, cor = 0.54337, p=0.024) and when modeling the effect of location as a random effect in a Linear Mixed Model (t12.7851=3.123, cor = -0.668, p=0.008; Figure 3.8A). This means that there is more Rhizobacter when AMPs were less effective against Bd. Rhizobacter trends toward a negative association with Bd prevalence (Logistic regression: z=-2.656, p=0.0113, adjusted-p = 0.055) and intensity (Kendall: z = -2.5793, tau= -0.3282, p= 0.0099, adjusted-p =0.0396) overall. However, when accounting for location, the relationship is not significant (Mixed Effects Logistic regression: z=-1.53, p=0.878).  Chryseobacterium (OTU39630) was the second most abundant bacterium and was present on 94% of CSFs and constituted 29.7% of the sequences on frogs. It is not correlated with MIC with or without accounting for location (Linear Mixed Model: t12.95=-0.5934, cor = -0.705, p=0.3295). Chryseobacterium trends toward positive association with Bd status overall (Logistic regression: z=-2.329, p=0.0198, adjusted-p = 0.0792; and intensity (Kendall: z = -2. 7039, tau= 0.3443, p= 0.0069, adjusted-p =0.03427), but again this effect disappears when including location as a random effect (Mixed Effects Logistic regression: z=-0.784, p=0.433). Rhizobacter and Chryseobacterium negatively correlate with each other, both overall and accounting for location (Pearson correlation: t32:-5.9263, cor=-0.7224, p<0.00001; Linear Mixed Model: t32:-5.224, cor=-0.832, p<0.0001) (Figure 3.6C); this pattern may be a result of amplicon sequence data being compositional (Gloor et al. 2017). The average relative abundance of Chryseobacterium differs between locations (Kruskal-Wallis: chi-squared= 14.834, df= 3, 59  p=0.002) and is greater at Gipsy Lake, where Bd is the most frequent and abundant while at Jones Pond, Bd was not detected (Figure 3.4b). An OTU in the candidate phylum Gracilibacteria (OTU4446) was found on 85% of CSFs and did not predict Bd status (Mixed Effects Logistic regression: z=0.095, p=0.925). An OTU in the family Oligoflexaceae (OTU5105) was present on 62% of CSFs and did not predict Bd status (Mixed Effects Logistic regression: z=0.763, p=0.446). Oligoflexaceae did not differ by location (Figure 4b; Kruskal-Wallis: chi-squared= 2.37, df= 3, p=0.5). An OTU in the genus Luteolibacter (OTU7414) was present on 68% of CSFs and did not predict Bd status (Mixed Effects Logistic regression: z=-0.944, p=0.345). Luteolibacter also did not vary by location (Figure 4b; Kruskal-Wallis: chi-squared= 5.625, df= 3, p=0.13). In addition, all three of these core bacteria did not associate with MIC against Bd (Appendix Table B.8), or with Bd prevalence (number frogs that are Bd positive)(Appendix Table B.7) and intensity (number of genetic equivalent (ge)) when location is not considered (Appendix Table B.6).  Phylogeny of Rhizobacter and Chryseobacterium  To learn more about OTU87682 (Rhizobacter) and OTU39630 (Chryseobacterium), I made phylogenetic trees that included their closest relatives. I detected the same Rhizobacter sequence on Oregon spotted frogs, rough-skinned newts and northwest salamanders (Chapter 4), and the sequence had a 100% match to an antifungal isolate from mountain yellow-legged frogs (Woodhams et al. 2015). Most of the closely related sequences to the core Rhizobacter were previously found on other amphibians, such as cricket frogs (Krynak et al. 2016), American bullfrogs and eastern newts (Walke et al. 2014) and the Perez's frog (Costa et al. 2016); these mostly clustered together within a clade that is generally amphibian associated (Figure 3.7). I 60  also detected Chryseobacterium on Oregon spotted frogs, rough-skinned newts, and northwest salamanders (Chapter 4). The closest matches in Genbank of Chryseobacterium have approximately 97% similarity, and there is no clear amphibian clade or clear clade that is primarily host-associated. Of the sequences detected on hosts, one sequence was found on red-backed salamanders (Lauer et al. 2008) and another on a frog from Madagascar (Bletz et al. 2017b). The sequences that were associated with amphibians do not cluster within the core Chryseobacterium. Another closely related Chryseobacterium sequence from a host was associated with little brown bats (NCBI KT951707.1). Although most sequences are detected in environmental sources, there was no apparent clustering based on sample affiliation (Figure 3.8).  Predictive antifungal function I matched my bacterial DNA sequence data to the Woodhams antifungal database (Woodhams et al. 2015) and found that, out of 873 OTUs, 83 matched inhibitory isolates, five matched bacterial isolates that were isolated from amphibians but not tested against Bd, one OTU enhanced Bd growth, 38 matched isolates that had no effect on Bd growth, and 746 did not match the database (Figure 3.5). The relative abundance of bacteria that match antifungal isolates is positively associated with AMP MIC /Surface Area against Bd with the location as a random effect (Linear Mixed-effects model; t12.73=2.92, p =0.012; Figure 6B). When not accounting for location, I found a relationship between the presence of Bd and the average relative abundance of antifungal bacteria, where it is less likely to have Bd detected if there is a greater relative abundance of bacteria that match the antifungal database (Logistic regression, z= -2656, p=0.00791). The same is true when the intensity of Bd is the response variable (Kendall, z=-27347, p= 0.006244, tau =-0.33479756). However, with the location included as a random effect, 61  no correlation was found between the presence of Bd and the average relative abundance of bacteria that match the antifungal isolates (Mixed logistic regression, z= -1.088, p=0.277). The most dominant bacterium on CSFs, Rhizobacter, which matched the antifungal database, was positively correlated with the relative abundance of other bacteria that match the antifungal database (Linear Mixed Model; t30.7=-2.745, p =0.01; Figure 3.6D). Chryseobacterium and antifungal bacteria negatively correlate (Linear Mixed Model: t= -5.941, p <0.000001), which is not surprising since it is also associated with Rhizobacter, and Rhizobacter is the major contributor to the relative abundance of bacteria that match antifungal bacteria. However, Chryseobacterium was also strongly negatively associated with other antifungal bacteria that do not include Rhizobacter (Linear Mixed Model; t30.72=-4.984, p <0.0002). The average relative abundance of antifungal matches differed by location (Kruskal-Wallis: chi-squared= 14.5, df= 3, p=0.002), with Jones Pond having a greater average relative abundance than Gipsy Lake (Kruskal-Wallis multiple comparisons with Holm correction: Z= -3.41, p=0.003). Jones Pond also had a greater average relative abundance than Doney Lake (Kruskal-Wallis multiple comparisons with Holm correction: z= -2.54, p=0.044; Figure 3.5).  62   Figure 3.4 The core bacteria of CSF (bacteria that are significantly different than water and >1% of all bacteria found on > 5% of frogs) make up a large portion of bacteria found on CSFs. Other bacteria refer to all other bacteria in the dataset. B: The relative abundance of the core bacteria, in addition to whether the individual was Bd positive or negative. The y-axis is not fixed.   63   Figure 3.5 The relative abundance of bacteria that match the antifungal database on CSF by location.   Figure 3.6 Correlations between core bacteria and AMP MIC against Bd A: Relationship between AMP MIC against Bd and Rhizobacter are positively correlated B: Relationship between AMP MIC against Bd and antifungal bacteria are positively correlated. C: Relationship between Rhizobacter and Chryseobacterium. D: Rhizobacter and the relative abundance of bacteria that match antifungal isolates are positively correlated. Regressions are coloured by location, and the red dotted line is the mean for all frogs; p-values are the result of an LMM with the location as a random effect. 64   Figure 3.7 Phylogenetic tree of the closest relatives to OTU87682, Rhizobacter. Core bacteria are in bold. Bacteria associated with a host are in red and bacteria associated with the environment are in blue. CSF= Columbia spotted frog, RLF = Red-legged frogs, OSF = Oregon spotted frog. MYLF= Mountain yellow-legged frog. 0.03CSF_89640: amphibian CSFJN178856: soilEU907887: waterKM187584: amphibian Notophthalmus viridescens87836: amphibian OSF (ASV69), CSFCSF_87684: amphibian CSFAM936015: soilAM157297: rhizosphereAY212562: waterLN794432: amphibian Acris blanchardiKM187626: amphibian Notophthalmus viridescensAM157292: rhizosphereKY611677: amphibian Pelophylax pereziFJ849067: waterKT720394: amphibian Pelophylax pereziAB245357: soilAM934713: soil87632: amphibian OSF (ASV975), CSFJQ919617: soilFM872852: dustAM935558: soilKC286726: snowASV49: amphibian OSFKC286788: snowASV48: amphibian OSFJQ919600: soilAM935634: soilEF522266: rockJF417757: soilJF199040: skinLN794508: amphibian Acris blanchardiJQ977182: rhizoplaneDQ837237: waterCSF_87687: amphibian CSFJX559248: airCSF_87683: amphibian CSFASV1735: amphibian OSFAM935181: soilFJ849071: waterJQ919613: soilJQ675368: waterJQ919603: soilKM187100: amphibian Rana catesbeianaLN794471: amphibian Acris blanchardiLMDS01000004: plantHQ327181: snowHM270718: skinJQ919537: soilEF516518: soilMF366369: waterAB991186: snowFJ849257: waterKM253106: plantKU477679: hot springsKT720386: amphibian Pelophylax pereziCSF_87685: amphibian CSFLN794503: amphibian Acris blanchardiFJ849100: waterFJ464985: waterAB862896: hot springsEF580947: waterJQ977338: rhizosphereEF580981: waterKF533788: snowJQ318959: plantHQ327219: snowASV29: amphibian OSF (Experiment)HQ327204: snowLN794519: amphibian Acris blanchardiKU462123: hot springsJQ977109: rhizoplaneASV425: amphibian OSFHQ327249: snowHQ327188: snowKU713256: waterFM872725: dustKC286835: snowKC255351: waterLGRD01000030: plantAB245356: soilASV91: amphibian OSF87682: amphibian OSF (ASV13), CSF, , RLF (Core)JQ919569: soilAM935737: soilHQ327157: snowFM872569: dust507488100898387901008796959410068765594696881878968561009956775510052 10098546559100949010010088MYLF65    Figure 3.8 Phylogenetic tree of the closest relatives to OTU39630, Chryseobacterium (denoted with an asterisk). Core bacteria are in bold. OSF= Oregon spotted frog, Newt = Rough-Skinned Newt, NWS= North West Salamander. Bacteria associated with a host are in red and bacteria associated with the environment are in blue.  Metabolites:  Metabolite composition differed by geographic location (PERMANOVA: F3:31= 3.1513, R2=0.21762; p = 0.001), but was not associated with Bd status F1:31= 1.5707, R2=0.03616; p = 0.16); there was no interaction between metabolites and Bd status (F1:31= 1.4175, R2=0.03263; p 0.03JX141782: soilDQ156146: amphibian   Hemidactylium scutatumKC560018: rhizosphereGU385856: soilAF538774: waterJQ977178: rhizosphereJUGG01000004: unknownASV1263: amphibian   OSFCP015199: waterLN811705: sludgeKT766029: soilHQ377322: insectKU924005: rhizosphereHM263400: skinEU057843: amphibian   Plethodon cinereus87682: amphibian  OSF (ASV2), CSF, Newts, NWS (Core)39632: amphibian   OSF (ASV2524), CSFKC306431: amphibian   Leiopelma archeyiNR_159892: soilKF672598: rhizosphereJQ977309: rhizosphereJQ976992: soilKP899178: soilEF540483: soilJQ864376: insect39631: amphibian   OSF (ASV492), CSFHG934369: milkJPRH01000001: soilFJ535184: waterJQ977175: rhizosphereGU377122: -FR682717: soilGU138380: soilJQ977154: rhizosphereEF601823: soilJX287891: fish39634: amphibian   OSF (ASV28), CSF, NWSKJ482752: soilHM204917: soilKM114952: amphibian   Anaxyrus boreas; CSFJQ977693: rhizosphereKM187399: amphibian   Pseudacris cruciferKR856232: batKC618502: rhizosphereJQ977391: rhizosphereJQ977188: rhizosphereKM187232: amphibian   Bufo americanusKM187408: amphibian   Pseudacris cruciferJQ697106: plantMH929966: soilKC560017: plantKF704082: oilHQ911369: milkHM049695: soilJQ511867: soilJX287893: fishAY599654: fungiEF591302: soilKT767785: milkGQ274324: insectJF833804: soilJQ977140: rhizosphereKR106668: insectHF563551: mudKT951707: batAM988901: waterMF524689: amphibian   Boophis quasiboehmeiJF915334: fishJQ977061: rhizosphereAB681439: -EF093133: biofilm73171: amphibian   OSF (ASV21), CSFJX287902: fishJQ977174: rhizosphereJF496404: soil39633: amphibian   CSFKF641258: soilHG738135: plant10010052100869972999581761001008497100100988710010010010010083100100100100100100999710010099100811001008066  = 0.197). I also determined that there was no relationship between bacterial community composition and metabolite composition using a Mantel test with 999 permutations between similarity matrix data of bacterial compositions (Bray-Curtis) and metabolite similarity (Sorensen); rho-0.1241, p=0.944). I used pairwise.adonis (Arbizu 2017) to determine that Jones Pond was different compared to the other three locations concerning metabolites, however, the three other locations do not differ from each other (Appendix Figure B.3). Likewise, metabolite richness differs between locations (Kruskal-Wallis, chi-squared=14.119, df = 3, p = 0.0027), with Jones Pond having a lower richness than the other locations (Dunn test, alpha<0.05; Appendix Figure B.4). Metabolite richness did not explain the presence of Bd without or with location included as a random effect (Kendall's rank correlation, t=0.568, tau= 0.075, p =0.5699; Logistic regression: z= 0.616, p =0.538; Mixed Logistic Regression: z = -1.1355, p = 0.175).  3.5 Discussion: In this study, I compared AMPs, cutaneous bacterial communities, and bacterial metabolites to Bd status and intensity (measured as genetic equivalents) per individual to determine if specific components of disease defense explain Bd status in CSFs. This study is unique in that it surveyed multiple aspects of disease defense in Columbia spotted frogs and quantified the presence of the disease agent, Bd, across wild populations. Most studies have focused on one aspect of disease defense at a time, showing that both host-derived antimicrobial peptides and the metabolites produced by bacterial communities protect amphibians from the disease. We are just beginning to understand the interactions between amphibian antimicrobial peptides and bacterial community composition.  67  Half of the frogs sampled in this study were positive for Bd (Figure 3.1), although none had visible signs of morbidity at the time of sampling. Bd presence and intensity varied across locations and was absent in one site; however, Bd was previously detected at all locations on CSFs (Sheafor, personal correspondence). Bd presence is known to vary both spatially and seasonally on CSFs without clinical signs of infection (Russell et al. 2010); for the population that is Bd negative (Jones pond), either Bd was cleared by frogs or the environmental conditions associated with seasonality were not conducive to Bd at the time of sampling. It is also possible that with a sample size of 10 frogs that Bd was not detected. Furthermore, the sister taxa of CSF, Oregon spotted frogs (Rana pretiosa), have been experimentally shown to be resistant to Bd infection; all of these factors suggest that CSF are generally resistant to Bd.  Although the frogs in this study showed no signs of morbidity, in other amphibian species the presence of low levels of Bd has been shown to have negative effects on individuals or populations (Chatfield et al. 2013, Brannelly et al. 2018, Campbell et al. 2019). A mark and recapture study with northern cricket frogs showed that frogs that are Bd positive are more likely to be recaptured, which could mean they are more easily preyed upon (Brannelly et al. 2018). Furthermore, jumping abilities are reduced in northern leopard frogs that are Bd positive (Chatfield et al. 2013). An experiment on golden bell frogs showed that being inoculated and clearing Bd still resulted in a higher metabolism, less fat, and males with reduced testes compared to frogs that were not exposed to Bd (Campbell et al. 2019). I did not measure aspects of body condition, which I would expect to be worse if frogs are negatively affected by Bd. The long-term impact of Bd has not been studied in this species.  68  I found that AMPs from CSFs inhibited Bd, but my hypothesis that frogs with more potent AMPs are less likely to have Bd was not supported (Appendix Figure B.2). Having more potent AMPs (low MIC) did not correlate with lower Bd intensity or Bd absence. The MIC values detected (0.1 to 1.0 µg protein/ml per cm2) fall within the range of Australian frogs (Woodhams et al. 2006a) and Tiger salamanders (Sheafor et al. 2008), suggesting there were no methodological problems. My results suggest that AMPs are not directly affecting Bd status in CSFs. For my study, frogs were induced to release AMPs, and therefore, my results do not capture the constitutive quantities at which peptides are released. Constitutive AMPs may give a more accurate picture the role of AMPs in resistance to Bd for CSF.  Columbia spotted frogs have a few common bacteria (>1% average relative abundance) that are amphibian specific, where most bacteria within their communities are rare; the rare bacteria are site-specific and likely from environmental reservoirs. Indeed, only 15 OTUs had a greater abundance on frogs compared to water, and these OTUs consisted of nearly 80% of the bacteria detected on their skin. When comparing the CSF microbiota by geographic location, there were only 39 bacterial OTUs that were present at all locations (not necessarily the core). CSFs share 75% of OTUs with the water species pool, but these are in low abundances (Figure 3.3B), and the CSF microbiota differs across sites, perhaps reflecting differences in abiotic conditions and the environmental species pool. This means that the bacteria that are consistent on frogs are abundant, and the bacteria that are variable on frogs are rare and influenced by species pools of bacteria in the environment. Other studies have found that the environment acts as an environmental reservoir and that many bacteria are shared with the environment (Muletz et al. 2012, Kueneman et al. 2013, Loudon et al. 2014b). Previous work has shown the importance of 69  the species pool in shaping community structure experimentally (Loudon et al. 2016), and environmental conditions have been shown to influence bacterial community compositions in woodland salamanders (Muletz Wolz et al. 2018). The extent to which the environment affects community dynamics of amphibian skin bacterial communities likely differs on a species by species basis. More efforts in understating the structure of communities (i.e., lots of rare vs. a few dominate for communities) and how that relates to host biology, phylogeny, and disease susceptibility should be the focus of future studies.  I identify core bacteria that are more abundant on hosts compared to environmental bacterial communities across all natural sites that are sampled and does not inform understanding of the type of symbiosis (mutualism, commensalism, parasitism) that is occurring. Rhizobacter is the most abundant core bacterium (Figure 3.4) and matches an antifungal isolate that was cultured from a yellow-legged mountain frog (Woodhams et al. 2015). I find a positive relationship between MIC against Bd and the relative abundance of Rhizobacter. This finding is robust across locations and supports the hypothesis that AMPs are affecting bacterial communities on CSF. Previously, specific peptides from toads correlated with specific bacterial OTUs (Davis et al. 2017). Future studies should focus on characterizing the identities, antifungal abilities, and interactions with bacteria of CSF AMPS. Rhizobacter was negatively associated with Bd overall, but the effect disappears after accounting for location, which means there was a large site effect. Jones Pond, where all frogs were Bd negative, had the greatest abundance of Rhizobacter, and Gipsy Lake, where Bd was the most frequent and abundant, had the least amount of Rhizobacter. Jones Pond is also the hottest site, and high temperatures negatively influences Bd. On the contrary, Park lake was the coolest site, and the temperature may be more 70  in line with Bd's optimal temperature (Piotrowski et al. 2004). I am unable to state whether Rhizobacter is protecting CSF against Bd, or whether it just has different environmental requirements to Bd (e.g., thrives in warmer temperatures). Sampling more sites would elucidate these unclear relationships.  To learn more about Rhizobacter, I made a phylogenetic tree of its closest relatives and found close sequence matches on six other amphibian species. This suggests that there may be an amphibian associated Rhizobacter clade (Figure 3.7), and fits a pattern suggested by a meta-analysis that most symbionts are found across different hosts (Louca et al. 2019). As more Rhizobacter sequences are detected on other amphibian species, analysis of congruence between host and symbiont phylogeny can be used as an indicator of co-evolution.  The second most abundant bacterium, present on 94% of CSFs was Chryseobacterium, which is not in the antifungal database and negatively correlates with Rhizobacter. Chryseobacterium did not correlate with MIC and did not have an effect on Bd status or intensity; although Chryseobacterium was most abundant at Gipsy Lake where Bd is the most frequent and abundant and least abundant at Jones Pond where all frogs were Bd negative. An isolate identified as Chryseobacterium has been cultured from amphibians, but this isolate only has a 94% sequence similarity (Muletz-Wolz et al. 2017). I also placed Chryseobacterium into a phylogenetic tree with its closest relatives and found that similar sequences are from amphibians, but there are also environmental samples and no apparent clustering (Figure 3.8). Also, the sequence with the highest similarity was 97%, which is fairly low sequence similarity and means that there is likely missing diversity that would provide more insight into the host-specificity of this Chryseobacterium. Having host symbionts that are similar to environmental strains (i.e., no 71  clear host clade) is not unheard of; for example, the protective symbionts Pseudonocardia and Amycolatopsis of attine ants do not fall in a host or environmental clades (Mueller et al. 2008). The other three core bacteria are identified as being in the candidate phylum Gracilibacteria, the family Oligoflexaceae, and the genus Luteolibacter, and did not match the antifungal database or correlate with Bd presence, Bd intensity, or MIC (Appendix Tables B6-10).  There is a strong negative relationship between the rations of Rhizobacter and Chryseobacterium. This result could be caused by the data being compositional, for example, if only one of these OTUs is changing abundance resulting in this pattern (Gloor et al. 2017). I am unable to decipher whether relationships with compositional data are a result of a biological effect or a mathematical artifact without knowing the actual abundances of bacteria. I found that the relative abundance of Rhizobacter is associated with AMPs, and my results indicate that it may have a negative associated with Bd. Conversely, the relative abundance of Chryseobacterium is not associated with AMPs and is associated positively with Bd. These patterns lead to a new hypothesis that frog AMPs are controlling Rhizobacter populations (Figure 3.6A), and Rhizobacter is competing with Chryseobacterium (Figure 3.6C). Alternatively, the presence of Rhizobacter could be influencing AMPs. Data that accounts for actual abundances of bacteria, or in vitro tests are required to test these hypothesizes. Rhizobacter is also positively correlated with other bacteria that match the antifungal database (Figure 3.6D), which further supports the hypothesis that AMPs influence Rhizobacter, and then Rhizobacter influences other bacteria within the community. An antifungal bacterium on tropical ants, Pseudonocardia, facilitates the horizontal transmission of other antibiotic bacteria, which 72  results in a multi-drug defense against pathogens (Scheuring and Yu 2012). My results indicate that Rhizobacter may be playing a similar role on CSFs, and possibly other amphibians.  I surveyed the metabolites that are found on the skins of CSFs and found that metabolite composition and richness differed by location. Both patterns are driven by one location: Jones pond was distinct from the other three locations (Appendix Figures B4, B5). Metabolite compositions have been shown to differ by sampling site and elevation in a neotropical frog (Medina et al. 2017), which could be the result of variation in bacterial communities or environmental factors. I did not find an association between metabolite assemblages and Bd status, or a correlation between metabolite composition and bacterial community composition, which is similar to what has been observed in Panamanian frogs (Belden et al. 2015). However, this is contrary to an experiment where bullfrogs that were treated with a probiotic (Janthinobacterium lividum), antibiotics, or a placebo before exposure to Bd had different metabolite assemblages that correlate with Bd load. In addition, metabolite composition and bacterial community composition correlate (Walke et al. 2015). The lack of clarity in my results on metabolite compositions highlights the need to understand why, at times, this method is informative, whereas other times it is not. This might be caused by specific metabolites having greater anti-Bd properties compared to others. For examples, some chemicals such as violacein, are known to be strongly inhibitory against Bd (Brucker et al. 2008b), and its abundance has been shown to drive patterns in mortality caused by Bd (Becker et al. 2009). Therefore, if specific metabolites are driving patterns in disease dynamics caused by Bd, then metabolite richness is a poor predictor for Bd. We also did not detect clear patterns when comparing metabolite compositions, and most sites appeared to cluster randomly, other than Jones pond. 73  This pattern is likely since interspecific bacterial interactions lead to the production of different suites of metabolites (Loudon et al. 2014a); metabolites on the skins of amphibians may be inheritably variable. Jones pond had no detectable Bd (Figure 3.1) and also had a trend of less bacterial richness, significantly fewer metabolites (Appendix, Figure B.3), and distinct metabolite assemblages (Appendix, Figure B.4) and bacterial compositions (Figure 3.2B) with low dispersion. Furthermore, Jones Pond was distinct in having the greatest abundance of Rhizobacter, and the least amount if Chryseobacterium. However, we cannot say that these characteristics resulted in low Bd, especially since Jones pond was the warmest location and temperature could be driving all of these patterns. Future studies could demonstrate the role of Rhizobacter and Chryseobacterium in Bd resistance by increasing the number of sample locations.  In this study, I examined the ability of AMPs to inhibit Bd and the profiles of metabolites in comparison to Bd. There are multiple things to consider, 1) it is unknown whether the frogs produce all of the detected AMPs and that skin bacteria produce all of the detected metabolites. 2) For my study, it is unclear how the combination of AMPs and metabolites affects Bd. Not all amphibians produce AMPs, and different species produce different suites of peptides (Rollins-Smith et al. 2005). This means that there are different strengths of environmental filtering if AMPs are essential in such a mechanism. Columbia spotted frogs and Oregon Spotted frogs (Chapter 4), are closely related (Green et al. 1996), and I have shown they have similar bacterial symbionts. I have shown that Columbia spotted frogs have peptides that can inhibit Bd, and Oregon spotted frogs are known to have peptides that are strong enough to protect against Bd 74  (Conlon et al. 2013). Determining whether AMPs explain bacterial community composition may explain bacterial community patterns across amphibians. An emerging hypothesis is that stronger peptides against pathogens will cause a more potent defense against other microbes as well, and this will result in a stronger environmental filter for bacteria that leads to less complicated communities, and higher host specificity. A cross taxa study that examines the AMPs and bacterial communities at different scales of amphibian relatedness would refute or corroborate this prediction.  Isolating and culturing Rhizobacter and Chryseobacterium in future studies will be vital to test these emerging hypotheses: 1) Isolates that match my core bacteria from CSFs will inhibit Bd, 2) AMPs from CSF will have a dose-dependent effect on the growth of core bacteria in vitro and 3) the abundances of Rhizobacter and Chryseobacterium will be determined by competition and impacted by the starting densities of each core bacterium. Culturing these bacteria will also allow us to determine which metabolites or proteins are being produced by which bacteria and then allow us to determine if they are present in my metabolite analyses. It will be essential to growing these in co-culture, since bacteria produce different suites of metabolites, with variable antifungal abilities when grown with other bacteria (Loudon et al. 2014a). Furthermore, I can expose these bacteria to AMPs from CSF to determine whether they are resistant to AMPs. From my data on the correlation between MIC and Rhizobacter, I can expect that AMPs likely affect Rhizobacter populations. Furthermore, with Chryseobacterium in culture, it can be tested against other amphibian pathogens to determine if it has a role in protecting against diseases other than chytridiomycosis.  75  My results suggest that the AMPs from CSFs may affect CSF skin bacterial communities. I found deterministic bacterial communities associated with CSF, which is likely due to environmental filtering by frog skin and AMPs: in this scenario, some bacteria can live in the environment, which is the frog. I found core bacteria that were abundant and frequent across Columbia spotted frogs from four different populations. I observed a pattern where the site with the most Bd had the least amount of antifungal bacteria, and the site without Bd had the greatest amount of antifungal bacteria, which leads to questions about thresholds of antifungal bacteria and herd immunity. Surveying a greater number of CSF populations and having the most dominant core bacteria in culture will be vital to better understanding this system in the future.   76   Establishing the core microbiota of Oregon spotted frogs: Chapter 4:distribution, potential function and stability  4.1 Abstract:  Amphibians have bacteria on their skins that can protect against the fungal pathogen, Batrachochytrium dendrobatidis (Bd). Oregon spotted frogs (OSF) are an endangered amphibian that are in captive breeding and head start programs in British Columbia. In the wild OSF are resistant to Bd. Skin bacteria can mediate resistance to Bd in some amphibians, but little is known about OSF bacteria and whether they are associated with Bd. Furthermore, captivity changes the microbiome of many animals, including amphibians, which may have implications for reintroducing endangered amphibians into the wild. Here we survey the microbiota of wild and captive OSF. We show that wild OSF have three core bacteria that are absent from the environment but found on other sympatric amphibians. These core bacteria are also on captive frogs, but two are lower in abundance. Two of the core bacteria are also detected at a low relative abundance on some wild egg masses, although they are absent, or rare, on egg masses in captivity. Using qPCR, I detected Bd in wild populations, allowing me to test the hypothesis that Bd alters microbiota structure. Bd intensity weakly correlates with overall microbiota composition, and the core bacterium Rhizobacter positively correlates with Bd. Lastly, I experimentally test the role of environmental reservoirs in maintaining OSF microbiota. Housing frogs in low diversity settings resulted in subtle bacterial community changes, with most community members remaining stable. Furthermore reducing the diversity of the environmental reservoir did not reduce diversity on frogs or change the predicted antifungal function of OSF 77  communities. Overall, we show that OSF have a handful of specific bacteria that are dominant, stable, and associated with a wide range of amphibians.   4.2 Introduction:  All animals have bacteria that live in and on them. The complexity of host-associated microbiota (the entire collection of microbes associated with a host) varies across host species, and the number of symbionts per host species ranges from only one symbiont in a specialized organelle to hundreds in the gut and on the skin (O’Brien et al. 2019). The importance of the microbiota—and individual microbes—for host fitness is a continuum, ranging from no effect on host fitness to obligate for host survival (Hammer et al. 2019). Determining which bacteria may have an impact on the host within these complex communities is a challenge; one solution is focusing on core bacteria that are frequent across a host species (Apprill et al. 2017, Hernandez-Agreda et al. 2017, Hernandez-Agreda et al. 2018). These core bacteria are hypothesized to play a key role for the host (Shade and Handelsman 2012). The presence of core bacteria, which by definition are stable and widespread across populations, may indicate the stable provision of beneficial functions by the symbiont(s) (Lozupone et al. 2012, Relman 2012, Ainsworth et al. 2015). Core bacteria are not necessarily performing beneficial functions for the host; for example, they could be commensal or even pathogenic. There are also cases where provisioning of beneficial functions do not require specific microbes, and key functions occur by functionally redundant taxa (Louca et al. 2018, Kruger 2019). Examining core bacteria allows me to examine further bacteria that are characteristic of my host species. I define core bacteria as the bacteria that are more abundant on hosts compared to environmental bacterial communities across all 78  sampled natural sites; this definition does not capture which type of symbiosis (mutualism, commensalism, parasitism) that is occurring. Host-associated bacteria can be acquired from the environment, vertically from parents, or horizontally from other hosts (Koskella et al. 2017). The specificity of bacterial taxa to a host species also varies considerably, ranging from unique to a given host, shared across multiple hosts, to generalists (Roth‐Schulze et al. 2018, O’Brien et al. 2019). Comparing host microbiota to potential bacterial source pools (i.e., water, soil, sympatric animals) can generate hypotheses about whether bacteria are acquired from the environment or other sources (Venkataraman et al. 2015, Loudon et al. 2016). Sampling additional host species from the same environments can suggest taxa that are host specialists but not unique to the focal species. Sampling eggs can elucidate the possibility of vertical transmission, or at least early horizontal transmission (Dominguez-Bello et al. 2010, Trevelline et al. 2018). Understanding transmission mode can provide insight into the potential host dependence on a symbiont; a meta-analysis found that the loss of vertically transmitted symbionts causes a greater decrease in host fitness than losing horizontally transferred symbionts (Fisher et al. 2017). Symbionts acquired via specific horizontal mechanisms that bypass environmental species pool are also more likely functionally important or indicators of host dependence (Shukla et al. 2018).  Chytridiomycosis, a disease caused by the fungus Batrachochytrium dendrobatidis (Bd) (Longcore et al. 1999, Piotrowski et al. 2004), has caused global amphibian declines (Skerratt et al. 2007). Defense against Bd is primarily mediated by antimicrobial peptides, which are part of the amphibian innate immune system (Rollins-Smith et al. 2002), and the bacteria that live on the skins of amphibians (Harris et al. 2006, Woodhams et al. 2007b, Harris et al. 2009). Many 79  bacterial isolates from amphibians have been shown to inhibit Bd in vitro (Woodhams et al. 2015) and do so by producing antifungal metabolites (Harris 1966, Brucker et al. 2008a, Brucker et al. 2008b, Becker et al. 2009). These metabolites likely affect susceptibility to other pathogens as well, and mediate competition between bacteria from amphibian skin (Loudon et al. 2014a). Protection from Bd that is mediated by bacteria can vary with bacterial community composition, and relies on an amphibian harboring one or more bacterial types that can persist and inhibit Bd (Harris et al. 2009). Bacterial community composition on amphibians, and the portion of the community producing antifungal metabolites are affected by bacteria reservoirs (i.e., sources of immigration from natural bacterial pools) (Loudon et al. 2016). Many factors impact amphibian microbiota structure overall, including environmental conditions (Krynak et al. 2016, Kueneman et al. 2019), developmental stage (Kueneman et al. 2013), and host species identity (McKenzie et al. 2012, Bletz et al. 2017a).  Oregon spotted frogs (Rana pretiosa; OSF) are endangered and historically had a range from the southwest corner of British Columbia to the northwest corner of California. This species has had a 70-90% reduction of its historical range (Hayes 1997, Team 2014). As of 2012, OSF was Canada's most endangered frog, with an estimated population of 316 individuals at four locations (Team 2014); declines are mostly attributed to habitat destruction (Watson et al. 2003, Team 2014). To ameliorate the population declines in Canada, captive breeding and head-starting programs for OSF have been established at The Greater Vancouver Aquarium, Toronto Zoo, and Greater Vancouver Zoo (Team 2014, Olive and Jansen 2017). The prevalence of Bd on Oregon spotted frogs is high in the wild (Pearl et al. 2009). In inoculation trials, Bd colonized OSF skin but did not cause mortality (Padgett-Flohr and Hayes 2011). Bd resistance in OSFs is 80  attributed to antimicrobial peptides (AMP), part of amphibian innate immunity (Conlon et al. 2011, Conlon et al. 2013).  The role of the bacteria on OSF skin in protecting against Bd and the influence of captivity on skin bacteria is unknown. Captivity generally changes both environmental conditions and the availability of bacterial reservoirs (captive individuals are often reared in microbially depauperate environments to reduce the risk of disease introduction). Indeed, comparisons of wild and captive conspecifics have shown altered community compositions in red-backed salamanders, Panamanian golden frogs, boreal toads, fire-bellied toads, Japanese fire-bellied newts, red-eyes tree frogs, and giant Japanese salamanders (Antwis et al. 2014, Becker et al. 2014, Loudon et al. 2014b, Bataille et al. 2016b, Kueneman et al. 2016, Sabino-Pinto et al. 2016, Bletz et al. 2017d). The resulting unnatural bacterial communities associated with captivity can affect health. For example, boreal toads in captivity that lost their ‘wild' microbiota were more susceptible to Bd when not given a bacterial probiotic (Kueneman et al. 2016). In addition, having altered bacterial communities early in development caused Cuban tree frogs to be more susceptible to parasitic worms as adults (Knutie et al. 2017). Here I characterize the bacterial community structure of wild and captive OSF and include samples from egg masses, adults, sympatric amphibians, and environmental samples. We use these samples to infer the core community and identify potential modes of transmission. We measure Bd load across individuals and test for associations between Bd and the OSF microbiota. We then compare the bacterial communities of wild and captive OSF populations to determine the impact of captivity. As microbial diversity of the environment is generally lower in captivity, we experimentally investigated the stability and resilience of OSF bacteria, and the importance 81  of environmental bacteria by manipulating the diversity of bacterial reservoirs of frogs. For the experiment, I expected that the lack of an environmental bacterial reservoir would result in different bacterial communities that have less diversity and bacteria that match antifungal isolates.  4.3 Methods: Survey of wild and captive OSF  We surveyed the bacterial communities, and Bd status of wild Oregon spotted frogs from two locations in 2015 and 2016. Permitting for surveys was through the Oregon Spotted From Recovery Team, of the Ministry of Forests, Lands and Natural Resource Operations: permit SU15-216280. Frogs were collected from traps that were checked daily at the Maria Slough and Morris Valley in the Fraser River Valley near Agassiz, BC. Maria Slough is a wetland that maintains water year-round; Morris Valley location is a horse pasture that floods in the spring during OSF breeding season. II also surveyed the bacterial communities on captive Oregon spotted frogs at the Vancouver Aquarium in 2015 and 2016 at the Greater Vancouver Zoo in 2016 (Table 4.1). Frogs were handled with sterile nitrile gloves and rinsed with Provasoli media (Wyngaard and Chinnappa 1982, Culp et al. 2007) before sampling to remove transient bacteria. Frogs were swabbed 10 times with a sterile swab on a randomly chosen side of the ventral surface (left or right side) to collect DNA for bacterial community and Bd analyses. We also opportunistically sampled Northern red-legged frogs (Rana aurora), rough-skinned newts (Taricha granulosa) and adult and juvenile Northwest salamanders (Ambystoma gracile) at Maria Slough. To determine if any bacteria associated with adult OSF are candidates to be vertically transmitted, we sampled eggs of wild and captive egg masses (Table 4.1). While 82  wearing sterile nitrile gloves eggs were gently raised out of the water and spayed with Provasoli media to remove transient bacteria and debris since eggs are in contact with environmental bacteria. Eggs were then gently swabbed using the above protocol. In addition, seven individual OSF eggs were collected and surface sterilized by emerging in a 1% hydrogen peroxide bath twice for 60 seconds. Eggs were rinsed with sterilized water to remove the hydrogen peroxide and then frozen at -80° C.  Table 4.1 Samples that are included in the OSF survey.  SS= Surface Sterilized NWS= Northwest salamander Newt = Rough-skinned newt RLF= Red-legged Frog   Bacteria were also sampled from the water, sediment, and vegetation to understand the bacterial community assembly and maintenance of OSF. Bacteria were collected from water by filtering 500 – 150 mL (until filter clogged) of water through a 0.22 µm filter using a Sterivex syringe. Sediment was collected in a 50 mL falcon tube, and 0.1 g was used for DNA extraction. Microbiota from vegetation was collected by swabbing for 10 seconds. All samples were frozen at -80° C until DNA extraction.  Experiment: To determine the stability, resilience, and the importance of environmental bacterial reservoirs on OSF community composition, we experimentally reduced the bacterial diversity of Survey OSF OSF	Eggs SS	OSF	Eggs Water Sediment Vegetation RLF Newt	 NWS	Adult NWS	JuvinileMaria	Slough 32 20 7 8 21 13 3 2 5 6Morris	Valley 15 15 	- 10 5 5 	- 	- 	- 	-Vancouver	Aquarium 15 6 	- 12 	- 5 	- 	- 	- 	-Greater	Vancouver	Zoo 7 	- 	- 4 	- 	- 	- 	- 	- 	-83  the water in which the frogs were housed. This experiment met the animal welfare standards of the Greater Vancouver Zoo and was under their permit, SU19-524756. Twenty-one Oregon spotted Frogs (OSF) from the Greater Vancouver Zoo that were approximately 1.5 years old were placed in tilted, Rubbermaid containers (14.12(D) X 7.75 (W) X 3.88 (H) inches), such that they had access to water and dry areas, and randomly placed into one of three treatments: Reservoir, Reduced Diversity or Crossover to Reservoir. The Reservoir treatment received water from the frog enclosure at the zoo. The Reduced Diversity treatment received frog enclosure water that was filtered with 0.22 µm filters using a peristaltic pump and then passed through a UV filter. Approximately 0.7 liters of water was added to each container and changed every three days. The Crossover to Reservoir treatment initially received the Reduced Diversity water, and then was crossed over to the enclosure water after 7 days. This was done to determine whether bacterial communities are resilient and would become more like those of frogs in the Reservoir treatment if the Reduced Diversity treatment had an impact. Frogs lived in cattle tanks that were open-air and had live vegetation before the experiment; some of the frogs were from the same cattle tank. Frogs were swabbed for bacteria before being placed into treatments and then swabbed on days four, seven eleven and fourteen of the experiment. The swabbing procedure was the same as above. Filtered water and unfiltered water were sampled to characterize reservoir communities and ensure that diversity was reduced in the Reduced Diversity treatment. In addition to sampling frogs, bacteria from water and from the biofilm that grew on the containers in contact with water were swabbed on days 7 and 14 (the middle and end) of the experiment for each frog. Water was changed every two to three days. Frogs were fed crickets every 3 days; the bacterial communities associated with crickets were sequenced, and we 84  determined that they did not impact the experiment (e.g., bacteria from crickets did not colonize the frogs).  DNA Extraction: DNA was extracted using a MoBio PowerSoil kit according to the manufacturer's protocol. This DNA was used to determine bacterial community composition and to test for Bd and determine bacterial community composition. To better assess whether bacteria are inside of OSF eggs, DNA was extracted from whole eggs using the same protocol as above; our methods likely did not remove all surface bacteria. Identifying Bacterial communities The V4-V5 16s rRNA gene was targeted for this study. Library prep for samples from 2016 was conducted as described in Chapter 3. Library prep for samples from 2017 was prepared by the Integrated Microbiome Resource (IMR) facility in the Centre for Genomics and Evolutionary Bioinformatics at Dalhousie University (Halifax, Canada) using their standard procedure (Comeau et al. 2017).  I used DADA2 (Callahan et al. 2016) to form Amplicon Sequence Variants (ASVs) from the forward read of the 16s amplicon sequencing data. ASVs are exact sequence variants that take account of sequencing errors and should correspond to discrete bacteria at the highest possible resolution. The quality of the reverse reads was poor, and they were excluded from further analysis. This method collapses sequences into exact amplicon sequence variants (ASVs) and accounts for sequencing errors (Callahan et al. 2016). This produced over 40,000 ASVs. As with all clustering methods, DADA2 still inflates bacterial diversity, and sequencing errors likely 85  cause many ASV. Thus, we removed any ASVs that had fewer than 600 reads, which resulted in 2,145 of ASVs. The number of ASVs is likely inflated; however, we sampled multiple environmental sources that are expected to have widely different bacterial compositions. Before filtering sequences, two contaminants (Achromobacter and Pseudomonas), and any chloroplast and mitochondria matches were removed from the data. Taxonomy was assigned to these sequences by using vsearch against the SILVA 132 database (Quast et al. 2013). ASV13 is a 100% sequence match to OTU87682 found on Columbia spotted frogs; in chapter 3 I found that this sequence falls within the genus Rhizobacter. Antifungal matches I used an antifungal database (Woodhams et al. 2015) to determine the prevalence of my bacteria that match known antifungal isolates. We did this by doing closed reference OTU picking at 97% similarity and the antifungal database as the reference.  Survey Analyses  Comparing diversity  To test whether bacterial compositions differ between frogs and the environment, I used Bray-Curtis similarity and an Adonis test. I also tested whether bacterial communities differ by location, site, and captivity with Adonis. I used Chao1 to test whether alpha diversity differs between amphibians and the environment, and then between captive and wild OSF. Differences in Chao1 diversity between locations were tested with the Kruskal-Wallis test, and then the Dunn test. Furthermore, I tested whether predated antifungal function, using the Woodhams antifungal 86  database, differed between wild and captive OSF using a Kruskal-Wallis test, and then the Dunn test.  Identifying Core bacteria I define core bacteria as the bacteria that are more abundant on hosts compared to environmental bacterial communities across all natural sites that are sampled and does not inform understanding of the type of symbiosis (mutualism, commensalism, parasitism) that is occurring. One of my main goals was to determine if OSF have a core bacterial community that differs from the environment and then determine whether these bacteria are present on captive frogs and in eggs. To find the core bacteria of wild OSF, I compared OSF to Water, sediment, and vegetation samples using DESeq2 for Maria Slough and Morris separately (Love et al. 2014). I tested the field sites separately since there were uneven sampling sizes between the sites, which would have skewed the results to favour bacteria on frogs at Maria Slough. I focused on the common bacteria (greater average relative abundance of 1%). Testing OSF for Bd  To test for the presence and intensity of Bd on the skins of OSF, I used quantitative Polymerase Chain Reaction using PerfeCTa qPCR Fastmix II (QuantaBio, Beverly MA) in 20µl reactions using the reaction times and primers from (Boyle et al. 2004). I used amplicon standards (Pisces Molecular, Boulder CO.) to establish the copy number of the ITS gene for each frog. Tests were performed in triplicate, and samples had to have a positive value in 2/3 of the replicates to be considered positive. To test for differences between locations, I used the Kruskal -Wallis test (Kruskal and Wallis 1952) in R (Team 2015). I also used a Kruskal-Wallis test to test 87  whether Bd differed between male and female OSF. My data was skewed towards males (46 males vs. 9 females), which aligns with their natural history: males wait in the wetland for the females during the breeding season, whereas the females arrive, lay eggs, and then leave.  Furthermore, I tested whether Bd intensity explained beta diversity of bacterial communities using the Adonis test in vegan (Oksanen et al. 2015). I then extracted the NMDS coordinates using vegan and performed Kendall's rank correlations with NMDS1 and NMDS2 against Bd intensity, as similarly done by Jani and Briggs (2014). I then tested whether Bd intensity correlated with matches to the antifungal database, alpha diversity metrics, and the core bacteria with Kendall's rank correlation. Lastly, I tested whether body condition explained Bd intensity with Kendall's rank correlation for only males since some females were gravid while others had already laid their eggs. Body condition was determined by using the residuals from a linear model (R2=0.589) of Log weight (g) and Log SVL (mm) (Schulte-Hostedde et al. 2005). I also tested whether body condition correlated with ASV13 Rhizobacter and ASV2 Chryseobacterium.  Experiment Analyses Filtering water for the experiment To reduce bacterial diversity for the experiment, I initially collected water from the Greater Vancouver Zoo and then autoclaved it for 1 hour. I found that this increased the pH by one pH unit and increased the conductivity of the water ~ 50% (µS/cm). I found this to be too great of a confounding variable and had concerns whether these changes could affect the health of the frogs. I instead filtered the water with a peristaltic pump using a 0.22 µm filter and then 88  ran the water through a UV filter that was designed to reduce algal populations in ponds. The pH and conductivity were not greatly affected by this treatment.  I initially tested whether frog bacterial communities differ from water and substrates, which I expected to be different, using the Adonis test with the Bray-Curtis similarity metric. To determine whether changing bacterial reservoirs affect the bacterial communities of OSF, I first established that I was able to manipulate the reservoir source successfully. This was done with the Adonis function in vegan (Oksanen et al. 2015) using the Bray-Curtis similarity metric to test for community similarity differences and alpha diversity metrics with a mixed-effects model with the day as a random effect. I also used DESeq to determine how many bacterial ASVs were altered from the manipulation. To test whether manipulating bacterial reservoirs changed the bacterial communities of OSFs, I excluded the baseline sample day and performed an Adonis test with Bray-Curtis similarity metrics with treatment and day as explanatory variables.  To understand how a reduced diversity reservoir affects specific bacteria, I used DESeq to determine which bacteria differ between the Reservoir and Reduced Diversity treatments. To better understand the dynamics of OSF bacterial communities, I also focused on the bacteria that are common (greater than 1% average relative abundance), compared to water sources and environmental samples from their housing.  4.4  Results 4.4.1 Survey Bacterial communities  89  I surveyed the bacterial communities of 272 samples of wild and captive Oregon spotted frogs along with their environment and sympatric amphibians (Table 4.1) to assess the similarity of OSF microbiota across sites and with the environment. This resulted in over 12.9 million sequences that fall into 2,145 ASVs. OSF communities are distinct from environmental samples (Adonis on Bray-Curtis dissimilarities: F3:144=17.76, R2= 0.239, p<0.00;1 Figure 4.1A). When examining only OSF, frogs cluster based on the location of sampling (Adonis: F3:63=14.205, R2= 0.372, p<0.001), with a clear separation between wild and captive populations (Figure 4.1). The year of sampling is also a significant determinant of microbiota structure (Adonis: F1:63=3.229, R2= 0.0282, p=0.005) and there is an interaction between sampling location and year (Adonis: F2:63=12.190, R2= 0.049, p<0.001), indicating that year of sampling explains more variation at some locations, particularly Morris (Figure 4.1B).  90   Figure 4.1 Ordination plots of Oregon spotted frog bacterial community compositions A: NMDS of Bacterial community compositions based on Bray-Curtis dissimilarity of wild and captive OSFs and their environment (Stress= 0.196). B: Bacterial compositions of only OSFs (Stress=0.219).  -1012-1 0 1 2NMDS1NMDS2Sample_TypeOSFSedimentVegetationWaterstatusCaptiveWildA: All Samples-1.5-1.0-0.50.00.51.0-1 0 1NMDS1NMDS2LocationAquariumMariaMorrisZooYear2016_year2017_yearB: Only OSF91  Environmental bacterial communities tend to be more diverse compared to hosts. Bacterial richness, as measured by Chao1, was lower on all of the amphibians compared to their environment (Figure 4.2B). Environmental sources in captivity (water and sediment) have lower diversity than wild source pools (Figure 4.2B). If frogs acquire bacteria from their environment, this would be expected to lead to depauperate bacterial communities on frogs in captivity. I do not see lower diversity on captive frogs generally, though frogs from the aquarium have a greater richness than frogs from Morris and the zoo (Kruskal-Wallis: chi-squared = 20.376, df = 3, p= 0.0001; Dunn test: α= 0.05; Figure 4.3B). The aquarium is a more controlled environment, while the zoo is more similar to an open system. Other amphibians (red-legged frogs, adult, and juvenile northwest salamanders and rough-skinned newts) appear to have similar richness as Oregon spotted frogs (Figure 4.2B), though low sample sizes preclude statistical tests.        92     Figure 4.2 Comparison of predicted function and diversity across sample types. A) The relative abundance of bacteria that match the Woodhams antifungal database for captive and wild OSF, other amphibians, and for the environment. B: Chao1 diversity, (estimated richness) of bacteria for captive and wild OSF, other amphibians, and for the environment. Samples from Maria and Morris are from the wild, and samples from the Aquarium and the Zoo are captive.  0.000.250.500.751.00Wild OSFCaptive OSFWild RLFWild NewtWild NWS AdultWild NWS LarvalWild WaterCaptive WaterWild VegetationCaptive VegetationWild SedimentMatches to Inhibitory IsolatesLocationMariaMorrisAquariumZooA0200400600Wild OSFCaptive OSFWild RLFWild NewtWild NWS AdultWild NWS LarvalWild WaterCaptive WaterWild VegetationCaptive VegetationWild SedimentChao1LocationMariaMorrisAquariumZooB93    Figure 4.3 Predicted function and alpha diversity for only OSF. The relative abundance of bacteria that match the Woodhams antifungal database for OSF by location. B: Chao1 diversity (estimated richness) of bacteria for OSF by location. Samples from Maria and Morris are from the wild, and samples from the Aquarium and the Zoo are captive. Letters indicate significant differences using the Dunn test, with α= 0.05.  I then identified core bacteria that are more abundant on hosts compared to environmental bacterial communities across all sampled natural sites. For my analyses, I focused on core bacteria that are considered to be common. I found three core ASVs that are always statistically more abundant on OSF than in water, on vegetation, and in sediment at both wild AB BABACBCACD94  sites (Figure 4.4). Two of the core bacteria (ASV2: Chryseobacterium and ASV13: Rhizobacter) are also core bacteria in Columbia spotted frogs with identical sequences (Chapter 3). One of these, ASV13: Rhizobacter, matched the antifungal database. All three core bacteria are on captive OSF, with ASV13: Rhizobacter at highest abundance (Figure 4.4). Overall, there is a negative relationship between the relative abundance of Rhizobacter and Chryseobacterium (Kendall's tau: z = -2.46, tau= -0.24, p =0.014). To better understand how OSF acquire their bacteria, we sampled egg masses and extracted DNA from single eggs that were surface sterilized. The core bacteria, Rhizobacter and Chryseobacterium, are also detected on wild eggs, but at a lower frequency and abundance than on the skins of OSF. Furthermore, both were with at least one of the surface sterilized eggs. Chryseobacterium was not detected on captive eggs, but Rhizobacter was detected in low abundances on captive eggs (Figure 4.5). The core bacterium, Luteolibacter, was only detected on one egg mass from Maria Slough.   95   Figure 4.4 Core bacteria on wild OSF. These ASVs are significantly greater on frogs than their environment for both sampling locations. Many are also found on other amphibians sampled from the same sites. ASVs in the red text match the antifungal database and inhibit Bd.   Figure 4.5 The relative abundance and occurrence of adult wild OSF core bacteria associated with OSF eggs. Chryseobacterium and Rhizobacter are associated with eggs at a low frequency and abundance compared to OSF skins. The core bacterium Luteolibacter only occurred on one wild egg mass from Maria and is not shown here.  I predicted the antifungal function of bacterial communities by matching my sequences to an antifungal database (Woodhams et al. 2015). Frogs have a higher relative abundance of bacteria that match the Woodhams antifungal database than environmental samples (Figure 4.2A); this is inevitable since the antifungal database was assembled from bacteria from frogs, and is therefore skewed towards finding antifungal bacteria on frogs. The average relative abundance of taxa with antifungal matches varies across locations, and is highest on Morris and Chryseobacterium RhizobacterFrequency   12/21     1/6      3/15     0/7                   2/21     2/6       0/15     3/7 96  Aquarium frogs; Zoo frogs do not differ compared to the other locations (Kruskal-Wallis: chi-squared = 27.527, df = 3, p <0.0001; Dunn test: α= 0.05; Figure 4.3A). Although I did not have adequate sample sizes of other amphibians to perform statistics, Northwest salamanders and rough-skinned newts appear to have a lower relative abundance of bacteria that match the antifungal database compared to Oregon spotted frogs, and red-legged frogs appear to have similar abundances of bacteria that match the antifungal database (Figure 4.2A).  I tested 58 wild OSF from two locations over two years for Bd. Fifty of the 58 OSF tested positive for Bd. There was no difference in Bd intensity between the sites and two sampling years 2017 (Kruskal-Wallis, chi-squared= 6.9291, df = 4, p = 0.1397). Bd intensity ranged from 0 to 97,390 ge (Figure 4.6). There were no differences in Bd intensity between males and females (Kruskal-Wallis, chi-squared= 1.245, df =1, p = 0.264).  Figure 4.6 The frequency and intensity of the fungal pathogen, Bd, on OSFs. Each point represents one frog; blue represents Bd positive frogs, where red represents Bd negative frogs.   012345Maria 2016Maria east 2017Maria west 2017Morris 2016Morris 2017Collection siteLog Bd intensity (ge)StatusBd negativeBd positive97  Bacterial community compositions are associated with Bd in other species of amphibians (Jani and Briggs 2014, Becker et al. 2015). I tested whether bacterial communities corresponded with Bd intensity and year and location of capture. Both Bd intensity (Adonis: F1:37=3.6, R2= 0.05, p= 0.008) and Location and year (Adonis: F3:37=2.1, R2= 0.346, p= 0.001) of capture explained bacterial community composition, and there was no interaction between these variables (Adonis: F3:37=1.2, R2= 0.053, p= 0.239; Figure 4.7A). In addition, I tested the NMDS axes against Bd intensity for all samples for both axes and found that NMDS 1 did not correlate with Bd intensity (Kendall's rank correlation: z=0.598, tau=0.062, p=0.550); however NMDS 2 approached significance (Kendall's rank correlation: z= 1.891, tau=0.196, p=0.059; Figure 4.7B). I also performed Adonis tests independently by location. Either Bd was not significant (Maria 2017: Adonis: F1:21=1.898, R2= 0.0829, p= 0.137: Figure 4.7C), or there were not enough replicates per site to perform the Adonis test. Frogs with a greater relative abundance of bacteria that match the antifungal database had a greater Bd intensity (Kendall's tau: z=2.018, tau=0.207 p=0.044). This correlation is driven by the antifungal matching bacteria, ASV13 Rhizobacter, which is also positively associated with Bd; it is the most abundant and bacteria and matches the antifungal database (Kendall's tau: z=3.059, tau=0.297 p=0.002). Alpha diversity does not correlate with Bd intensity (Kendall's tau: Chao1, z= -1.1657, tau=-0.12 p=0.244; Simpson's evenness, z=-1.28, tau=-0.131 p=0.2). The other two core bacteria are not associated with Bd intensity (Kendall's tau: ASV2, z=-0.448, tau=--0.044, p=0.654, ASV26, z=-1.448 tau=-0.141 p=0.145; Figure 4.8. Furthermore, body condition did not explain Bd intensity for males (Linear model: F= 1.02, df = 1:44, p =0.318 (Figure 4.10). This was not tested for females since some were gravid, where others had already 98  laid their eggs. I also found that body condition does not correlate with Bd intensity, Rhizobacter, or Chryseobacterium; although there is a trend that Bd and Rhizobacter are negatively associated with body condition and that Chryseobacterium is positively associated with body condition.  Figure 4.7 Relationship between bacterial communities and Bd A: NMDS of Bacterial community compositions using Bray-Curtis similarity of OSF and their Bd intensity for all wild 99  frogs tested for Bd. B. Correlation between NMDS2 and Bd log of all wild frogs that were tested for Bd. C: A subset of the samples focusing on frogs from Morris in 2017.    Figure 4.8 Relationships between the predictive function of bacteria, alpha diversity, core bacteria and the intensity of Bd. D-F show the relative abundance of the core bacteria.  4.4.2 Experiment: The bacterial communities on some amphibians are influenced by bacteria from environmental species pools (Loudon et al. 2014b). I experimentally reduced the bacterial species pool and sampled 20 frogs and their water and substrate over time, for a total of five sampling times. For this experiment, 147 samples successfully amplified and had 2,145 ASVs total. I successfully reduced the bacterial richness (LM: F1:20=23.75, p<0.0001; Day: F2:20=17.46, p<0.0001) and altered the bacterial composition of the OSF's reservoir (Adonis: Reduction: F1:20= 31.47, p<0.001; Day: F2:20=7.751, p<0.001(Figure 4.9). Fifty-three bacteria were 100  significantly differently represented in the Reduced Diversity Reservoir compared to the Reservoir source, as measured by DESeq2.   Figure 4.9 Reduction of bacteria from the water reservoir. A) NMDS of bacterial communities using Bray-Curtis similarity showed that filtering the bacterial source changed the community composition (stress= 0.0679). B) Filtering reduced the richness of bacteria within the communities.  101   Figure 4.10 Ordination plots of community composition during the experiment A) NMDS of bacterial communities using Bray-Curtis similarity showed that filtering the bacterial source changed the community composition (stress= 0.21). B) NMDS of OSF communities during the experiment (stress = 0.23).  -101-1.0 -0.5 0.0 0.5 1.0NMDS1NMDS2SampleFrogSubstratewaterTreatmentBaselineCrossover to ReservoirReduced DiversityReduced Diversity SourceReservoirResevoir SourceA: All Samples012-1 0 1NMDS1NMDS2TreatmentBaselineCrossover to ReservoirReduced DiversityReservoirDayElevenFourFourteenSevenZeroB: OSFs102  When examining all the samples from the experiment, bacterial communities differ based on sample type: frogs, water sources, water housed with frogs, and biofilms in frog containers (Adonis: F2:127=29.311, R2= 0.267, p<0.001). These communities also differ by treatment (Adonis: F5:127=6.763, R2= 0.154, p<0.001; Figure 4.10A). When only examining frogs, communities differ based on treatment (Adonis: F2:56=4.189, R2= 0.098, p<0.001) and day of experiment (Adonis: F3:56=5.526, R2= 0.195, p<0.001), but there was no interaction (Adonis: F4:56=1.044, R2= 0.049, p<0.355; Figure 4.10B). Reduced diversity treatment differs from the Reservoir treatment, but not Crossover to Reservoir treatment. The Reservoir treatment differs from the Crossover to Reservoir treatment (Table 4.2). I did not observe a difference in the relative abundance of bacteria that match the antifungal database between the treatments, but there was a difference in the day; there was not an interaction between treatment and day (Table 3; Figure 4.11A). There was also no difference in Chao1 for the treatment, day, or interaction (Table3: Figure 4.11B).   Table 4.2 Pairwise Adonis results comparing experimental treatments.  Stability of microbiota may be necessary for a host since it may correspond to the continual function provided by stable community members (Lozupone et al. 2012, Relman 2012, Ainsworth et al. 2015). 14 bacteria were differentially abundant between the Reduced Diversity treatment and Reservoir treatment. Two ASVs were greater on frogs in the Reservoir treatment, and 12 were greater on frogs in the Reduced Diversity treatment. Of these bacteria, only three Comparison	 SumsOfSqs F.Model R2 p-value p.adjustedReduced	Diversity	vs	Reservoir 1.205 5.043 0.088 0.001 0.003Reduced	Diversity	vs	Crossover	to	Reservoir 0.583 2.468 0.056 0.02 0.06Reservoir	vs	Crossover	to	Reservoir 0.558 2.161 0.063 0.013 0.039103  ASVs are considered common, by having a greater average relative abundance of 1%, across frogs. Of all the bacteria on the frogs, 19 ASVs had an average abundance of greater than one percent in the experiment. One bacterium, ASV1002, was abundant in the baseline frogs and plummeted once the experiment started. Four bacteria are on OSF for all treatments, but not in any water or substrate samples; core ASV13 and ASV29 are 99.2% similar in sequence similarity (Rhizobacter) and the same clade as ASV13 (Figure 3.7). Six bacteria are on the frogs before the experiment, but not in the Water Reservoir source, that appears in the environmental sources. Five bacteria are on the initial frogs and are in the Water Reservoir Source that are then present in the environmental sources. Two of those are not present or in low abundance in the substrate samples (Figure 4.12).  Figure 4.11 Predicted antifungal function and alpha diversity of frogs during the experiment. Statistic results are reported in Table 4.3 104      Figure 4.12 Dot plot of common, abundant bacteria in the experiment. 19 bacteria had an average relative abundance of greater than 1% on OSF through this study. Of these bacteria, only three were significantly different from between the Reservoir and Reduced Diversity Treatment. ASVs in red matched the antifungal database and inhibited Bd. The ASV with an asterisk matched the ASV but did not affect Bd.   105  Table 4.3 Statistical results of predicted function and alpha diversity by treatment through time.   4.5 Discussion: In this study, I focus on Oregon spotted frogs (OSF) and determine that they have core bacteria, such as Rhizobacter and Chryseobacterium, which correlate with each other and are on other amphibian hosts. These core bacteria are on some wild OSF egg masses, and in low occurrence detected with surface-sterilized eggs, which indicates early development of the relationship, likely from horizontal transmission. Furthermore, these core bacteria are on captive OSF; however, Chryseobacterium is at lower relative abundances. Bacterial communities are weakly associated with Bd intensity, and Rhizobacter drives this pattern. Lastly, I experimentally reduced bacterial reservoirs that frogs were housed with and found stable bacterial communities that were dominated by Rhizobacter.  Oregon spotted frogs have bacterial communities that are dominated by a few bacteria that are not found in the environment, while environmental transients are a minor component of OSF microbiota. These results are consistent with previous work showing that a handful of relatively abundant taxa similarly populates the skin microbiota of many amphibians. For example, a comprehensive study of 89 species of amphibians across 30 locations in Madagascar found that amphibian microbiota sharply differs from their environment and many amphibian species harbour skin microbiota dominated by a few bacteria (uneven composition) (Bletz et al. Linear	models:	OSF F-value	 Sum	Sq. Mean	Sq. F-value	 p-valueInhibitory	Matches 2 0.165 0.083 2.04 0.139Day	 3 0.453 0.151 3.73 0.016Interaction	 4 0.021 0.005 0.13 0.97Chao1 2 64.9 32.4 0.209 0.812Day	 3 430.4 143.5 0.925 0.435Interaction	 4 433.3 108.3 0.699 0.596106  2017a). Other amphibians, such as red-backed salamanders, have skin microbiota that are more strongly influenced by environmental bacteria (Loudon et al. 2014b); however, their core bacteria are maintained by deterministic processes, e.g., core bacteria are over-represented on salamanders (Loudon et al. 2016). Red-backed salamander communities were also more even with few bacteria having a relative abundance greater than 1%. The bacterial communities of other amphibians are also affected by environmental bacteria, such as yellow-legged frogs (Jani et al. 2017, Jani and Briggs 2018). Whether relative abundance of a bacterium links to greater function is not clear and is likely a case-by-case situation. There are examples of bacteria that are keystone species within amphibian bacterial communities, such as Janthinobacterium lividum (Bletz et al. 2013), which has a high impact for being relatively in low abundances. This bacterium has been in low relative abundances (in some cases less than 1% relative abundance) (Loudon et al. 2014b) and likely has a large impact on host resistance to Bd since it produces a metabolite that is considered to be highly antifungal (Brucker et al. 2008b). Here I present data showing that OSF have highly deterministic bacterial communities that are dominated by only a few bacteria.  Why some amphibian species bacterial communities that are dominated by a few bacteria, and communities are maintained by more deterministic processes, while other species have diverse bacterial communities that are more influenced by the environmental microbial species pool (Loudon et al. 2014b) is unclear. Multiple factors may play a role in this spectrum of determinism between amphibian species such as host ecology, the presence, abundance, composition, and effectiveness of antimicrobial peptides (AMPs) and mucus production. Host ecology is a factor since the extent of contact a host has with environmental bacteria affects 107  bacterial colonization. For example, a fossorial salamander is in contact with soil microbes, whereas arboreal frog is in contact with bacteria associated with leaves and branches, and the densities of bacteria available to colonize the host are different in those environments. There is some evidence for this: host ecology plays a greater role than phylogeny in structuring bacterial communities in Madagascar frogs (Bletz et al. 2017a).  Amphibian skins are habitats for bacteria to colonize and have features that make for a unique environment. These features likely cause environmental filtering, where colonizers are incapable of establishing on the amphibian (Kraft et al. 2015). For example, host immune factors such antimicrobial peptides (AMPs) (Conlon et al. 2011, Conlon et al. 2013), a form of innate immunity, are likely a major factor shaping OSF bacterial communities. AMPs from amphibians have been shown to have a dose effect on a bacterium isolated from a frog, where low doses of AMPs do not inhibit the isolate, while high dose inhibits the isolate (Myers et al. 2012). Furthermore, AMPs are proposed to cause the stability of bacterial communities in a bioaugmentation study: adding bacteria to an amphibian did not affect community composition in an experiment, but specific AMPs were up-regulated (Kung et al. 2014). Indeed, specific AMPs correlate with specific bacteria, suggesting that AMPs have a role in structuring bacterial communities (Davis et al. 2017). The ability of Columbia spotted frog AMPs to inhibit Bd correlated with the dominant bacterium Rhizobacter in chapter 3. Their close relatedness and likely similar AMPs may explain similar bacterial communities and Bd tolerance in CSF and OSF. Lastly, mucus production, which is understudied in amphibian microbiome work, is suggested to limit bacterial density in fire salamanders, which have a minimal layer of mucus compared to frogs (Bletz et al. 2018); the quantity of mucus and other chemicals that are species-108  specific likely vary across amphibians, and may be associated with host ecology (e.g., aquatic vs. arboreal). My work suggests that core OSF bacteria are acquired horizontally or from the environment. The core OSF bacteria are on OSF egg masses, but they were in low relative abundance and had a low frequency (Figure 4.5). Some of the core bacteria are also detected on surface sterilized eggs, but in less than half of the eggs that were sampled. This method likely does not fully sterilize the surface of eggs, but likely reduces the amount of bacterial DNA that would be extracted from the surface of the egg masses and should increase the ability to detect bacteria that are within the egg. These results are more supportive of environmental transmission since I would expect all of the eggs to have core bacteria if they are vertically transmitted. Vertical transmission occurs in other amphibian species; however, this has only been described in species that exhibit parental care, in the form of guarding eggs (Banning et al. 2008, Walke 2011, Hughey et al. 2017a). Since OSF do not exhibit parental care, vertically transmission would have to happen when the eggs are initially laid; there is some evidence that bird and lizards vertically transmit some bacteria inside eggs (Trevelline et al. 2018); so vertical transmission is not impossible. The core bacteria could also be independently colonizing both the eggs and frogs. If the niche space of eggs were similar to the skins of frogs, then the likelihood for egg masses to be colonized by bacteria that do well in a frog niche through environmental transmission would be increased. This would create a scenario where bacteria that do well on frogs surround newly hatched larvae. Understanding how organisms acquire bacteria has health implications since there is evidence that early bacterial colonization influences the developing immune system. For example, rearing Cuban tree frog tadpoles in sterile water translated into a 109  longer devolvement time, lower survival to metamorphosis, and a decreased resistance to parasitic worms as adults (Knutie et al. 2017).  I found that many core bacteria of OSF are also found on sympatric amphibian hosts, even though they are absent or nearly absent from diverse environmental samples (Figure 4.4). This is not surprising since amphibians are likely similar habitats so environmental filtering of bacteria will likely be similar between amphibians. I hypothesize that these bacteria either are adapted to live on amphibian hosts and not exclusively may be participating in host switching (Araujo et al. 2015). In this study, I sampled three other amphibian species from the same habitat (adult northern red-legged frogs, adult rough-skinned newts, and adults and juvenile northwest salamanders). I do not have a large host sample size to assess generalities about taxonomic patterns within this dataset. My results also indicate that some these bacteria, such as Rhizobacter, fall within clades where the bacteria are amphibian associated (Figure 3.7).  I investigated whether the two dominant core bacteria, Rhizobacter and Chryseobacterium, fell within clades that are associated with amphibians. Both of these are also core bacteria on Colombia spotted frogs in Chapter 3. Rhizobacter has a 100% sequence similarity match to an antifungal isolate that was cultured from a mountain yellow-legged frog (Woodhams et al. 2015). Rhizobacter falls within a clade that is overwhelmingly amphibian associated (Figure 3.7). Some bacteria fall within the genus Chryseobacterium that are associated with amphibians, but all of these are at least a 97% sequence similarity match in NCBI (Benson et al. 2005), many are in environmental samples, and none fall within the same clade as the core Chryseobacterium (Figure 3.8). The phylogenetic tree of Rhizobacter and Chryseobacterium 110  only consist of bacteria that are in GenBank and the SILVA database and does not include amplicon sequence data from other studies. The relative abundances of Rhizobacter and Chryseobacterium are negatively correlated on OSF and in Columbia Spotted frogs. This relationship could reflect the compositional nature of the data. For example, this pattern would occur if the absolute abundance of Rhizobacter is constant across samples, and the absolute abundance of Chryseobacterium changes (Gloor et al. 2017). Therefore, caution must be in place when interpreting relationships between bacteria within the dataset, or between bacteria and Bd. It is also worth noting that this correlative pattern is not surprising since these bacteria are the two most dominant bacteria and often make up a majority of the bacteria on OSF in the wild, hence having a high relative abundance of one necessarily means the other is lower. A biological relationship could mean that Rhizobacter and Chryseobacterium are competitors and trade-off in abundances, or there is something different about the frogs (e.g., selective AMPs) that are leading to different bacterial distributions. I also tested whether the body condition of OSF correlates with Rhizobacter or Chryseobacterium. Neither relationship was significant, but Rhizobacter approached significance with a negative relationship and Chryseobacterium approached significance with a positive relationship. Even if this were a significant relationship, there would be no way to tell if this was causative relationship; a pattern between a bacterium and body condition could be a reflection of the factor causing the condition (e.g., Bd).  I determined the presence and intensity of Bd on wild OSF to determine if Bd is associated with characteristics of their bacterial communities. Most OSF were Bd positive (Figure 4.6), which corresponds with a large survey that found that Bd was common and 111  widespread across the range of OSF (Pearl et al. 2009). Although there was a negative trend, Bd did not correlate with the body condition of the frogs. OSF have previously been shown to be tolerant of Bd (Padgett-Flohr and Hayes 2011), so this is not surprising. In my data, bacterial communities weakly differ based on Bd intensity. The dominant bacterium, Rhizobacter, positively correlates with Bd intensity (Figure 4.8D). This is the opposite relationship that I found between Rhizobacter (100% sequence identity), and Bd in Columbia Spotted frogs. The positive relationship is also surprising because Rhizobacter has a 100% match to an antifungal isolate (Woodhams et al. 2015); I expected that antifungal activity would lead to a negative correlation. Other studies have found differences in bacterial community composition, or relationships between specific bacteria and Bd (Jani and Briggs 2014, 2018). A major question is whether differences in bacterial community composition are the cause or the result of Bd. Multiple studies have demonstrated that Bd can drive bacterial community patterns (Jani and Briggs 2014, Jani et al. 2017, Longo and Zamudio 2017, Jani and Briggs 2018). Alternatively, the initial bacterial communities pre-Bd exposure predicted survival in Panamanian golden frogs implying that the pre-exposure bacterial communities equated to Bd survival (Becker et al. 2015). The presence of antifungal bacterial isolates at a herd immunity threshold (at least one antifungal isolate on greater than 80% of frogs) predicted the survival of a population of mountain yellow-legged frogs (Lam et al. 2010). Others have found that the bacterial communities of frogs that are co-existing with Bd differ from bacterial communities of frogs that are naïve to Bd. Frogs that are co-existing with Bd are enriched in genera that are known to be antifungal (Rebollar et al. 2016). Whether bacteria play a role in OSF resistance to Bd should be further investigated.  112  Captive OSF have different bacterial community compositions compared to wild OSF and each location clusters together (Figure 4.1B). This suggests that factors related to the sampling locations are playing a role in bacterial community composition. Captivity is known to affect bacterial compositions in red-backed salamanders, Panamanian golden frogs, boreal toads, fire-bellied toads, Japanese fire-bellied newts, red-eyes tree frogs, and giant Japanese salamanders (Antwis et al. 2014, Becker et al. 2014, Loudon et al. 2014b, Bataille et al. 2016b, Kueneman et al. 2016, Sabino-Pinto et al. 2016, Bletz et al. 2017d). In this study, there was no clear pattern in alpha diversity between captive and wild frogs (Figure 4.3A). Other studies have found that captivity causes a decrease in alpha diversity for red-backed salamanders, boreal toads, fire-bellied toads, Japanese fire-bellied newts, red-eyes tree frogs (Antwis et al. 2014, Loudon et al. 2014b, Bataille et al. 2016a, Kueneman et al. 2016, Sabino-Pinto et al. 2016); however, alpha diversity increased for captive Panamanian golden frogs and giant Japanese salamanders (Becker et al. 2014, Bletz et al. 2017d). Of these studies, only one tested for functional consequences of frogs having ‘wild' like bacteria; captive toads that lost wild-like bacteria were less susceptible to Bd if they were given a probiotic (Kueneman et al. 2016). I used the antifungal database and found that there was no difference in predicted antifungal function. I also examined whether the core bacteria are also on captive frogs. Captive frogs have similar relative abundances of Rhizobacter compared to wild frogs, but they have a lower abundance of Chryseobacterium as well as the third core bacterium, Luteolibacter. Furthermore, captive eggs lacked Chryseobacterium and had low relative abundances of Rhizobacter. This further suggests that early colonization of core bacteria on the eggs is not through vertical transmission, but rather from the environmental or horizontal transmission. The development of the host-microbiome throughout amphibian development from egg to frog is a subject of importance to captive rearing 113  programs for any species of conservation significance. There is a growing body of work that indicates the importance of early colonization of biologically relevant bacteria (Funkhouser and Bordenstein 2013, Zeissig and Blumberg 2014, Knutie et al. 2017). I used an antifungal database (Woodhams et al. 2015) to predict the function of the bacteria in my dataset. In the survey, I found that the relative abundance of bacteria that are predicted to be antifungal increase when frogs have a greater Bd intensity. This is in direct contradiction to my results in Chapter 3, where the relative abundance of bacteria that are predicted to be antifungal is greater in frogs that do not have Bd or have a lower Bd intensity. Furthermore, Rhizobacter, which has a 100% match to the antifungal database, is the driver of this pattern. It is positively associated with Bd on OSF and negatively correlated with Bd on Columbia spotted frogs. These contradicting results suggest caution with regards to the utility of the antifungal database in making ecological predictions. Inconsistent results from the antifungal database may be due to the context-dependency of the amphibian microbiota- Bd system (Daskin and Alford 2012) and functional variation of amphibian-associated bacteria at the strain-level. For example, isolates that fall within the same 97% OTU can have different antifungal functions against Bd (Kruger 2019), and this is also the case for isolates with 100 % sequence similarity of the 16srRNA gene (Molly Bletz, personal communication). In addition, different genotypes of Bd have variation in resistance to the same bacterial isolate (Bletz et al. 2017b, Antwis and Harrison 2018) (Bletz et al. 2017b, Antwis and Harrison 2018). To complicate this further, the ability of known antifungal bacteria to inhibit Bd also changes with temperature (Daskin et al. 2014, Woodhams et al. 2014). This database is useful for identifying whether core bacteria are found on other amphibians and may have the ability to inhibit Bd.  114  There are additional limitations to using the Woodhams antifungal database. For example, the bacteria in the database are biased: early studies only sequenced antifungal isolates, so the database is missing many isolates that did not affect Bd. Furthermore, the database only contains bacteria that were isolated from amphibians, which skews results when comparing the relative abundance of antifungal matches between amphibians and environmental samples. In addition, the database is incomplete and will miss some inhibitory bacteria. Lastly, we are counting all bacteria that match antifungal isolates equally, whereas it is known that different types of bacteria have varying strengths of inhibition against Bd (Bell et al. 2013). For example, the impact of keystone species, such as Janthinobacterium lividum, which is often in a low relative abundance but has a large impact against Bd, can not be discerned with this method (Bletz et al. 2013). Future studies should use better indicators of disease susceptibility, such as the mucosome assay, which challenges amphibian secretions against Bd (Woodhams et al. 2014). The weak correlation between bacteria in this study and Bd and the conflicting results that I have observed while using the antifungal database suggest that bacteria on OSF may be doing little to affect Bd. OSF are resistant to Bd (Woodhams et al. 2014) (Padgett-Flohr and Hayes 2011), with their AMPs as a suggested mechanism of resistance (Conlon et al. 2013). The deterministic bacterial communities that I observe on OSF may be a result of environmental filtering of bacteria by strong AMPs, rather than any selection for their antifungal functions.  Experiment  I experimentally reduced the bacterial reservoirs that are in direct contact with OSF by filtering and UV treating water that came from a naturalized OSF habitat (Figure 4.9). I found 115  that community composition changed (Figure 4.10), but the difference in community change was less than expected compared to previous studies. Communities differed between the Reduced Diversity treatment and the Reservoir treatment and differed between the Reservoir treatment and Crossover to Reservoir treatment, but there was no difference between the Reduced Diversity and Crossover to Reservoir treatment. Reducing the diversity of bacterial reservoirs has been shown to alter the community compositions of amphibians in previous studies (Loudon et al. 2014b, Knutie et al. 2017, Jani and Briggs 2018). Priority effects could explain a difference between the Reservoir treatment and Crossover to Reservoir treatment since communities on frogs did not converge once they had the same reservoir community. Furthermore, the Crossover to Diversity was not different from the Reduced diversity treatment, suggesting that the communities that developed with Reduced Diversity were slow to change when exposed to the Reservoir treatment.  Alpha diversity and the relative abundance of bacteria that match the antifungal database also did not change (Figure 4.11). Other studies have seen more dramatic shifts based on bacterial reservoirs. Previously I removed the bacterial reservoir from red-backed salamanders, which are terrestrial and fossorial, by housing salamanders in containers that had either soil from their habitat or sterile media rather than water and saw a large decrease in diversity and change in composition (Loudon et al. 2014b). A study on Sierra Nevada mountain yellow-legged frogs tested whether water sources from ponds where frogs persisted with Bd or were extirpated due to Bd affected bacterial communities and frog susceptibility using two source populations of frogs (Jani and Briggs 2018); they also included sterile water in their experiment. Bacterial communities differed based on which population frogs were from, which indicates that host 116  factors are playing a role in community composition. Bacterial communities also differed based on the water sources that the frogs were housed in, and the frogs in sterile water also had different bacterial compositions. Once these communities were established, there was a Bd inoculation treatment; frogs from different source populations had different survivorship curves, but survivorship curves did not differ between the water source treatments. Frogs in the sterile water had a steeper survivorship curve (they died more quickly). This study demonstrated that host genetics and the environmental bacteria affect bacterial communities on frogs and the rate at which death occurs from Bd (Jani and Briggs 2018). Another experiment using Cuban tree frogs exposed tadpoles to 1) natural water, 2) autoclaved water, 3) autoclaved water with short-term doses of antibiotics and 4) autoclaved water with long-term antibiotics (Knutie et al. 2017). Tadpoles in all autoclaved water treatments had lower skin and gut bacterial diversity. As adults, frogs that were in any of the autoclaved water treatments were more susceptible to the nematode Aplectana hamatospicula, than control frogs (Knutie et al. 2017). As in other studies, there are differences in bacterial community composition when frogs are housed in different bacterial reservoirs; other studies have found short term (Jani and Briggs 2018) and long-term (Knutie et al. 2017) negative consequences to being exposed to unnatural bacterial reservoirs.  In my experiment, there are 19 bacteria with an average abundance of over 1% on OSF (Figure 4.12). Most of these were consistent between treatments, with only three bacteria having a different abundance between the Reservoir treatment and the Reduced Diversity treatment; this indicates a stable and resilient community, which is not what I expected, and is not what is seen in red-backed salamanders and mountain yellow-legged frogs (Loudon et al. 2014b, Jani and Briggs 2018). Multiple patterns emerged when focusing on these abundant bacteria. The bacteria 117  that increased in relative abundance in the Reduced Diversity treatment match antifungal isolates. Their increase may be a result of 1) their ability to produce metabolites that might have the ability to kill bacteria and 2) having fewer bacteria to compete within the Reduced Diversity treatment.  Some bacteria are only on frogs and not the environment (Figure 4.12). This includes Rhizobacter, ASV13, which is in abundance at all of my sampling sites and on wild frog eggs. There is another Rhizobacter, ASV (29), which is 99.2% similar to ASV13 and is also abundant. However, Chryseobacterium (ASV2) was not on frogs during the experiment. Some bacteria were on the baseline frogs, but not in the Reservoir source (i.e., the water the baseline frogs came from) that maintained similar relative abundances during the experiment. This indicates that the water is becoming more frog-like, which may be a result of bacteria sloughing into the environment and persisting as a potentially biologically irrelevant by-product of the experiment. Some bacteria are on Baseline frogs, frogs throughout the experiment, the reservoir water source, and water throughout the experiment, but not in the substrate samples. One ASV (1002; Flavobacterium), was abundant before the experiment, and then disappeared once the experiment began; frogs lived in a natural habitat before the experiment (Figure 13). Although some of these patterns are unlikely in the wild, since frogs do not live in small enclosures with a relatively small volume of water, these results highlight that there is a range of host specificity of bacteria, where some are only frog specific, and others can persist in the environment when conditions are permissive. I expect that the stability of communities will be species-specific and be associated with host ecology. Summary 118  In this study, I used environmental sampling to find core bacteria on frogs that are frequent and abundant on a host, but not the environment. OSF have core bacteria that dominate their community; two of the core bacteria are on other amphibians, and one may represent amphibian clades. Further work is necessary to understand why some amphibians have communities that are more similar to their environments, whereas others, like Oregon spotted frogs and Colombia spotted frogs in chapter 3, have highly deterministic communities. A highly deterministic community, as seen here, may be an indicator of a stable mutualism, whether for disease or not. In addition, some of the core bacteria are associated with OSF eggs masses; suggesting that these relationships start early. Further studies should focus on the development of bacterial communities associated with eggs, with a focus on eggs in captivity. This is worth investigating since the early colonization of bacteria can have long-term impacts (Knutie et al. 2017). Lastly, I find that the bacterial communities on OSF are relatively stable and resilient when held in environmental reservoirs with reduced diversity; this suggests that environmental reservoirs have a minimal effect when OSF communities are established and is not surprising since I found that OSF bacterial communities are deterministic. Incorporating the natural history of symbiotic bacteria in captive breeding programs may have substantial health implications. Understanding how core bacteria contribute to the composition of a community will allow us to understand the interspecific variation that is frequently observed in host bacterial communities (Bletz et al. 2017a, Bletz et al. 2017c, Sabino-Pinto et al. 2017), and how stressors or captivity (Kueneman et al. 2016) might affect community compositions and potential function.  119   Concluding remarks Chapter 5:  All organisms live with bacteria, and many bacteria affect the biology of their hosts. Examples include bacteria aiding in digestion (Dillon and Dillon 2004a), educating the immune system (Gensollen et al. 2016), and disease resistance (Woodhams et al. 2007b). Host-associated bacterial communities range in complexity, ranging from tens to hundreds of types of different bacteria (O’Brien et al. 2019). Furthermore, hosts tend to have species-specific bacterial communities that are distinct from the environment.  Establishing which bacteria may affect their host within a complex community is a challenge. One solution is to focus on core bacteria that are hypothesized to play a key role in host ecosystem function (Shade and Handelsman 2012). One problem with this approach is that there is not an agreed-upon way to define the core. There is general agreement that core taxa are frequent across a population and specific to a host. Previous studies identify core bacteria by setting a frequency threshold across a host population; these thresholds are often high, for example, 80% (Hernandez-Agreda et al. 2018), or 90% frequency across populations (Loudon et al. 2014b). High thresholds exclude bacteria that are abundant on some individuals, but not at a high frequency across individuals. For example, a bacterium that is emerging as a biological relevant symbiont and is abundant on half of the individuals within a sampled population may be biologically impactful but excluded with those criteria. Furthermore, these studies only focus on the microbiota of the target host species, and this provides no information on where bacteria originate (e.g., are bacteria acquired from an environmental source or another host?), and cannot differentiate bacteria specific to a host from those predominantly found in the environment (i.e., transient bacteria). In my dissertation, I define core bacteria as bacteria that are more abundant 120  on hosts compared to environmental bacterial communities across all sampled natural sites. This was done by testing the distributions of individual bacteria within bacterial communities of a population against environmental reservoirs. This was done with either the Sloan neutral model (Sloan et al. 2007), or with DESeq (Love et al. 2014) and avoided using thresholds to establish which bacteria are of focus. Furthermore, I focused on common bacteria, using a threshold at 1% average relative abundance (Li et al. 2012, Pedros-Alio 2012).  Across three study systems, I find core bacteria make up most of the relative abundance of bacteria within a community. In general, for the host species that I studied, there are only a few core bacteria, where there are many more bacterial species found in environmental species pools. Furthermore, core bacteria have a greater relative abundance than transient bacteria and more frequent across the population of hosts. These core bacteria are more likely to impact host biology than transient bacteria because these specific and consistent associations make it more likely that hosts have come to depend upon the functions of these core bacteria (Figure 5.1). They could also be a parasite or pathogen (e.g., Bd is frequent across OSF). Alternatively these bacteria just fit into a narrow niche space within the host (e.g., environmental filtering, Kraft et al. 2015), and are a signature of health, since the host niche likely changes when diseased, and disease has been shown to shift bacterial communities (Jani and Briggs 2014, Lloyd and Pespeni 2018, Ange-Stark et al. 2019). I identify core bacteria by extensively sampling multiple individuals across multiple host populations and their environment; the core bacteria are more abundant on the host than in the environment and consistently present across individuals and populations. I argue that focusing on specific bacteria within a host bacterial community yields 121  more information about host-associated communities than comparisons of community metrics, such as beta and alpha diversity between hosts and the environment.    Figure 5.1 Schematic of core bacteria.  Core bacteria are bacteria that are more abundant on hosts compared to environmental bacterial communities across all natural sites that are sampled and does not inform understanding of the type of symbiosis (mutualism, commensalism, parasitism) that is occurring. They generally have a high relative abundance, are less diverse than the environmental bacterial communities and frequent across the host population.  They are hypothesized to impact host function, whereas transient bacteria are hypothesized to not impact hosts.   Host	Popula+on	Core	bacteria																				High																			Low																		High																High	Transient	bacteria												Low																			High																		Low																	Low	Rela+ve		Abundance	Environmental	Bacteria	Hypothesized		impact	on	host	Diversity		Frequency	in		Popula+on	122  I investigated the surface and ceca communities of the sea star, Pisaster ochraceus, and find them dominated by a handful of core bacteria, including clades of Mollicutes and Salinispira (Spirochetes). Relatives of these Mollicutes and Salinispira are also found on multiple additional sea star species in the Northern and Southern hemisphere, suggesting that these specific relationships have a long history. I do not know what, if any, functional role these bacteria play in sea star biology. However, in other marine invertebrates with intensely studied host-specific bacteria, there is a strong functional role for bacteria and communities, and one or a few bacterial taxa are dominant (McFall-Ngai 1991, Ozawa et al. 2017, O’Brien et al. 2019). The microbiota of well-studied marine invertebrates ranges from diverse and highly variable for corals and sponges (Ainsworth et al. 2015, Thomas et al. 2016a) to communities of only a handful of core taxa in squid and deep-sea mussels (McFall-Ngai 1991, Ozawa et al. 2017). The microbiota structure of most marine animals has yet to be determined.  The two amphibian species that I studied, Columbia spotted frogs and Oregon spotted frogs, are recently diverged sister species (Green et al. 1996). Each of these frogs has two core bacteria on their skin, Rhizobacter, and Chryseobacterium, that alternately are highest in relative abundance across populations in both species (i.e., one of these tends to be most abundant). Both Rhizobacter and Chryseobacterium are identical across host species in terms of ribosomal DNA sequence even though these frogs were hundreds of kilometers apart (Montana vs. British Columbia) and these species are reproductively isolated. Rhizobacter, the most dominant of core bacteria, is consistently found on wild and captive frogs and wild egg masses and falls within a clade that consists mostly of amphibian host-associated taxa. Chryseobacterium does not fall in a clade of host-specific bacteria, and closely related Chryseobacterium sequences are detected in 123  environmental samples. Chryseobacterium is less abundant and frequent on captive frogs but found on wild egg masses.  The hosts studied in my dissertation have a few dominant bacteria that are also found on related host species. What can these results tell us about the global diversity of host-associated bacteria? One line of thought proposes that animal hosts are a rich and largely untapped source of novel bacterial lineages, and that sampling the microbiota of these hosts will massively expand the number of bacterial lineages (‘species') known to science. In other words, it has been proposed that each host species unique bacterial taxa, which result in inflated immense diversity (Larsen et al. 2017). However, if most animals have a similar microbiota structure to the focal species (dominated by a few host-specific bacteria that are on related hosts, with much of the community transients), we should expect fewer bacterial lineages overall. My data support a massive census estimating the global biodiversity of bacteria that shows finite bacterial diversity that is not greatly inflated by the incorporation of host-associated communities (Louca et al. 2019). Globally, this suggests that many symbiotic bacteria are found associated with multiple hosts, as I found in a few target species. Studying the distribution of symbionts across hosts will be a tool into the future, for example, enabling researchers to decipher between host switching and co-evolution between hosts and symbionts. Core bacteria are good candidates for further study across systems because they are more likely to shape host biology, including performing beneficial functions for the host. In applied research, knowing which clades of bacteria are specifically associated with an animal of conservation concern may aid in the development of probiotic therapy as a potential treatment for disease. My dissertation research shows a clade of Rhizobacter is a core taxon on diverse 124  amphibians and correlates with Bd load. It is, therefore, a good candidate to investigate as a probiotic against amphibian fungal disease. Furthermore, consistent and stable symbionts may mean their animal hosts are more resilient in the face of climate change or other stressors such as pollution if the symbiont performs a function on which a host relies. Inventories of microbiota may help to identify populations that are at particularly high risk, and changes in microbiota could be the proverbial canary in the coal mine when it comes to forecasting changes to host populations. Bacteria can be acquired through the environment, from con-species, or vertically from a parent (Bletz et al. 2013, Koskella et al. 2017). Sampling design for an experiment can be informative for determining how hosts acquire their bacteria. For example, comparing host microbiota to potential bacterial pools (i.e., environmental bacteria) can be informative on whether bacteria are acquired from environment (Venkataraman et al. 2015, Loudon et al. 2016). Including eggs or the potential sources of vertical transmission as well as environmental samples as sources of environmental acquisition enable us to posit the most likely mode of transmission (Dominguez-Bello et al. 2010, Trevelline et al. 2018). In this study, I heavily sampled environmental samples to determine whether the bacteria that are associated with my hosts are also found in the environment and therefore are potentially acquired through the environment. For my Oregon spotted frog study, I also sampled egg masses and whole eggs to determine if the core bacteria on adult frogs are also found on or in eggs; some of the core bacteria were found on some eggs masses or associated with surface sterilized whole eggs: since these bacteria are not on all egg masses or eggs suggests that these bacteria are likely not vertically transmitted, but are likely transmitted from the environment or horizontally transmitted from other amphibians. 125  Knowing whether a bacterium is vertically acquired or acquired through horizontal or through the environment can provide insight into the potential host dependence on a symbiont; a meta-analysis found that the loss of vertically transmitted symbionts causes a greater decrease in fitness than losing a horizontally transferred symbiont (Fisher et al. 2017). Furthermore, bacteria that are acquired through specific horizontal mechanisms rather than an environmental species pool are more likely to play a role in host biology. Understanding how host-associated bacteria are acquired may have a high conservation importance when rearing animals of conservation value in captivity since environmental conditions and bacteria can be different access to environmental bacteria can be limited; we know that having the biologically relevant bacteria during early development can have long-lasting impacts on health (Sjogren et al. 2009, Gensollen et al. 2016, Knutie et al. 2017). To best characterize host-associated bacterial communities and identify core bacteria, I make the following suggestions. 1) Intensely sample the surrounding environment to distinguish between host-specific bacteria and those transients coming in from the environment. 2) Sample an individual host species across distinct populations with different abiotic conditions and to determine the range and extent of host specificity across a host species. 3) Sample related host species within a clade to uncover clades of bacteria that are host clade-specific and potentially have a long evolutionary history with the host-species of interest. 4) Sample the same individual repeatedly through time to determine if the bacteria are persistent through time and different seasons.  126  In my dissertation, I focused on common bacteria that are defined as being present in a greater than 1% average relative abundance. This criterion has previously been used to define common bacteria (Li et al. 2012, Pedros-Alio 2012). These methods can be beneficial since it simplifies the data, allowing the researcher to examine the most abundant bacteria; however, there are also problems with focusing on common core bacteria. For example, since I used relative abundance the actual abundance of bacteria are within one of these communities is not known, and there may be bacteria that meet my criteria for core bacteria that are less than 1% average relative abundance, but have a great impact on the function of the bacteria community (i.e., keystone species). I examined the core bacteria that have a relative abundance of bacteria less than 1% for Chapters 2 (Pisaster) and Chapter 3 (Columbia spotted frogs) and found that the rare core bacteria were all sequence variants of the common core bacteria. These sequence variants of the common core bacteria are likely sequencing errors and the products of our methods for binning bacteria into discrete units; this is even more likely since the recent methods (i.e., MED nodes, DADA2) bin sequences at high resolution. Therefore, focusing on the common core bacteria allowed me to examine bacteria within my communities that are likely real (not sequencing artifacts) and resulted in not inflating the diversity of the core bacteria. The pattern of the rare core bacteria being close sequence variants to the common core bacteria is likely due to the core bacteria being highly representative during sequencings. Having greater representation of a sequence will result in a higher number of artifact sequences for that particular sequence that is not filtered in bioinformatics steps that are in place to remove sequencing errors. Care should be practiced to ensure that there are not core bacteria that are rare in relative abundance that may be keystone bacteria while accounting for sequences that are 127  likely artifacts. Therefore, I recommend determining whether any bacteria meet core criteria that are not close sequence variants of the common core bacteria.   5.1 Limitations of work To better understand the distribution of core bacteria in other species and across geographical space, more data needs to be collected and accessible. Although more surveys are essential, there is already a massive amount of data that is collected and is currently not easily accessible. Greater accessibility to amplicon sequencing data would allow better characterization of the distribution of symbionts. The data generated from hundreds of amplicon sequencing studies each year is not difficult in principle to incorporate since they are publicly available as raw sequences. In practice, meta-analyses with many studies are time-consuming and computationally demanding. This is because metadata need to be compiled and cleaned, often manually, then data need to be re-processed together to make robust comparisons, and so on. The sequences that were in my phylogenetic trees are from my study, NCBI, SILVA, and from other researchers that I contacted directly. Large datasets that comprise many amplicon sequencing studies are being complied for amphibians (Kueneman et al. 2019), for as many sample types as possible (Thompson et al. 2017, Louca et al. 2019). Work remains to be done curating metadata for specific questions, but these efforts will open up more avenues for describing symbiont distribution into the future.  I tested whether core bacteria on frogs correlated with a pathogen in chapters 3 and 4. For chapter 3, studying Columbia spotted frogs, I found that Rhizobacter negatively correlated with 128  the amphibian fungal pathogen, Bd; my functional interpretation of this finding was strengthened since Rhizobacter has a 100% match to an antifungal isolate (Woodhams et al. 2015).. However, the same sequence of Rhizobacter positively correlated with Bd in Oregon spotted frogs. These results are in direct contradiction and are puzzling. It may be that Rhizobacter is only antifungal under specific abiotic and biotic conditions. The antifungal function of bacteria can be affected by temperature (Woodhams et al. 2014) and interactions with other bacteria (Loudon et al. 2014a). Another explanation for inconsistencies between Rhizobacter and Bd correlations could be due to strain-level differences that cannot be discriminated by our sequencing method, which only captures a 250 bp region; bacteria isolates from frogs with up to 97% 16s rRNA gene similarity have been shown to have different antifungal properties against Bd (Kruger 2019). Strain-level differences could be due to horizontal gene transfer (Thomas and Nielsen 2005), if genes for antifungal metabolites are on a plasmid (Bletz et al. 2013). These conflicting results may demonstrate that the bacteria on these frogs' skins are not a significant factor in Bd resistance on these species; indeed, antimicrobial peptides were previously hypothesized to mediate Bd resistance in this species. Future studies must test the function of bacteria of these frogs rather than surveying the communities as I have done to understand whether spotted frog bacterial are related to Bd; this could be done by direct Bd challenges or the possible using the mucosome assay (Woodhams et al. 2014). The bacterial community data that I generated for my dissertation are compositional rather than being absolute. Unfortunately, this is unavoidable due to the methods that generate amplicon sequencing data (Gloor et al. 2017). This leads to problems when interpreting correlations between the relative abundances of bacteria within a dataset (e.g., Rhizobacter and 129  Chryseobacterium), or between the relative abundance of a bacterium and the abundance of some other variable (e.g., Rhizobacter and Bd), due to the mathematical consequences of the data being compositional. For example, the relative abundances of a bacterium can greatly change between samples, even if it is present at the same absolute abundance if the absolute abundance of a second bacterium changes. Due to this, there must be caution when interpreting relationships between bacteria within the dataset, or between bacteria and Bd. Although biotic interactions, such as competition, are possible between Rhizobacter and Chryseobacterium, it is also possible that this relationship is a mathematical consequence of compositional data. These results, however, do serve as a way to generate hypotheses that can be later tested using methods that capture biotic interactions. Therefore, future studies that focus on these bacteria in vitro are needed to decipher whether there are biotic interactions between core bacteria and between core bacteria and Bd.  5.2 Final remarks In sum, I found bacterial communities that were dominated by a few abundant and prevalent core bacteria that are not detected in the environment; in most cases, these core bacteria belong to host-associated clades of bacteria. My results differ from previous studies by comparing host to their environment and focusing on specific core bacteria; these core bacteria are likely relevant to the host and should be further studied. More efforts should be made in testing whether most host-associated bacteria are on multiple hosts and the function of those bacteria. Furthermore, I found a pattern where similar symbionts are found across multiple host species that are in line with host switching and not co-evolution. The pattern detected in this dissertation corresponds with a study that concludes that there is finite bacterial diversity that is 130  not greatly inflated by bacteria that are associated with hosts. I have highlighted methods to establish core bacteria that are frequent, dominant, and more likely to influence a host in complex datasets.  131  Bibliography Ainsworth, T. D., L. Krause, T. Bridge, G. Torda, J.-B. Raina, M. Zakrzewski, R. D. Gates, J. L. Padilla-Gamiño, H. L. Spalding, and C. Smith. 2015. The coral core microbiome identifies rare bacterial taxa as ubiquitous endosymbionts. The ISME journal 9:2261. Anantharaman, K., C. T. Brown, D. Burstein, C. J. Castelle, A. J. Probst, B. C. Thomas, K. H. Williams, and J. F. Banfield. 2016. Analysis of five complete genome sequences for members of the class Peribacteria in the recently recognized Peregrinibacteria bacterial phylum. PeerJ 4:e1607. Ange-Stark, M. A., T. L. Cheng, J. R. Hoyt, K. E. Langwig, K. L. Parise, W. F. Frick, A. M. Kilpatrick, M. D. MacManes, and J. T. Foster. 2019. White-nose syndrome restructures bat skin microbiomes. bioRxiv:614842. Antwis, R. E., and X. A. Harrison. 2018. Probiotic consortia are not uniformly effective against different amphibian chytrid pathogen isolates. Mol Ecol 27:577-589. Antwis, R. E., R. L. Haworth, D. J. P. Engelmoer, V. Ogilvy, A. L. Fidgett, and R. F. Preziosi. 2014. Ex situ Diet Influences the Bacterial Community Associated with the Skin of Red-Eyed Tree Frogs (Agalychnis callidryas). PLoS One 9. Apprill, A. 2017. Marine animal microbiomes: toward understanding host–microbiome interactions in a changing ocean. Frontiers in Marine Science 4:222. Apprill, A., C. A. Miller, M. J. Moore, J. W. Durban, H. Fearnbach, and L. G. Barrett-Lennard. 2017. Extensive Core Microbiome in Drone-Captured Whale Blow Supports a Framework for Health Monitoring. mSystems 2. Araujo, S. B., M. P. Braga, D. R. Brooks, S. J. Agosta, E. P. Hoberg, F. W. von Hartenthal, and W. A. Boeger. 2015. Understanding Host-Switching by Ecological Fitting. PLoS One 10:e0139225. Arbizu, M. 2017. Pairwiseadonis: Pairwise multilevel comparison using adonis. R Package Version 0.0 1. Banning, J. L., A. L. Weddle, G. W. Wahl, M. A. Simon, A. Lauer, R. L. Walters, and R. N. Harris. 2008. Antifungal skin bacteria, embryonic survival, and communal nesting in four-toed salamanders, Hemidactylium scutatum. Oecologia 156:423-429. Bataille, A., L. Lee-Cruz, B. Tripathi, H. Kim, and B. Waldman. 2016a. Microbiome diversity differs between amphibian skin regions: Implications for chytridiomycosis mitigation efforts. Microbial ecology. Bataille, A., L. Lee-Cruz, B. Tripathi, H. Kim, and B. Waldman. 2016b. Microbiome Variation Across Amphibian Skin Regions: Implications for Chytridiomycosis Mitigation Efforts. Microbial ecology 71:221-232. Bates, D., M. Mächler, B. Bolker, and S. Walker. 2014. Fitting linear mixed-effects models using lme4. arXiv preprint arXiv:1406.5823. Baumann, P., L. Baumann, C. Y. Lai, D. Rouhbakhsh, N. A. Moran, and M. A. Clark. 1995. Genetics, physiology, and evolutionary relationships of the genus Buchnera: intracellular symbionts of aphids. Annu Rev Microbiol 49:55-94. Becker, M. H., R. M. Brucker, C. R. Schwantes, R. N. Harris, and K. P. Minbiole. 2009. The bacterially produced metabolite violacein is associated with survival of amphibians infected with a lethal fungus. Appl Environ Microbiol 75:6635-6638. 132  Becker, M. H., C. L. Richards-Zawacki, B. Gratwicke, and L. K. Belden. 2014. The effect of captivity on the cutaneous bacterial community of the critically endangered Panamanian golden frog (Atelopus zeteki). Biological Conservation 176:199-206. Becker, M. H., J. B. Walke, S. Cikanek, A. E. Savage, N. Mattheus, C. N. Santiago, K. P. Minbiole, R. N. Harris, L. K. Belden, and B. Gratwicke. 2015. Composition of symbiotic bacteria predicts survival in Panamanian golden frogs infected with a lethal fungus. Proc Biol Sci 282. Belden, L. K., M. C. Hughey, E. A. Rebollar, T. P. Umile, S. C. Loftus, E. A. Burzynski, K. P. Minbiole, L. L. House, R. V. Jensen, M. H. Becker, J. B. Walke, D. Medina, R. Ibanez, and R. N. Harris. 2015. Panamanian frog species host unique skin bacterial communities. Front Microbiol 6:1171. Bell, S. C., R. A. Alford, S. Garland, G. Padilla, and A. D. Thomas. 2013. Screening bacterial metabolites for inhibitory effects against Batrachochytrium dendrobatidis using a spectrophotometric assay. Dis Aquat Organ 103:77-85. Benson, D. A., I. Karsch-Mizrachi, D. J. Lipman, J. Ostell, and D. L. Wheeler. 2005. GenBank. Nucleic Acids Res 33:D34-38. Berger, S. A., D. Krompass, and A. Stamatakis. 2011. Performance, accuracy, and Web server for evolutionary placement of short sequence reads under maximum likelihood. Systematic Biology 60:291-302. Bletz, M. C., H. Archer, R. N. Harris, V. J. McKenzie, F. C. E. Rabemananjara, A. Rakotoarison, and M. Vences. 2017a. Host Ecology Rather Than Host Phylogeny Drives Amphibian Skin Microbial Community Structure in the Biodiversity Hotspot of Madagascar. Front Microbiol 8:1530. Bletz, M. C., M. Kelly, J. Sabino-Pinto, E. Bales, S. Van Praet, W. Bert, F. Boyen, M. Vences, S. Steinfartz, F. Pasmans, and A. Martel. 2018. Disruption of skin microbiota contributes to salamander disease. Proc Biol Sci 285. Bletz, M. C., A. H. Loudon, M. H. Becker, S. C. Bell, D. C. Woodhams, K. P. Minbiole, and R. N. Harris. 2013. Mitigating amphibian chytridiomycosis with bioaugmentation: characteristics of effective probiotics and strategies for their selection and use. Ecol Lett 16:807-820. Bletz, M. C., J. Myers, D. C. Woodhams, F. C. E. Rabemananjara, A. Rakotonirina, C. Weldon, D. Edmonds, M. Vences, and R. N. Harris. 2017b. Estimating Herd Immunity to Amphibian Chytridiomycosis in Madagascar Based on the Defensive Function of Amphibian Skin Bacteria. Front Microbiol 8:1751. Bletz, M. C., R. G. B. Perl, B. T. Bobowski, L. M. Japke, C. C. Tebbe, A. B. Dohrmann, S. Bhuju, R. Geffers, M. Jarek, and M. Vences. 2017c. Amphibian skin microbiota exhibits temporal variation in community structure but stability of predicted Bd-inhibitory function. ISME J 11:1521-1534. Bletz, M. C., M. Vences, J. Sabino-Pinto, Y. Taguchi, N. Shimizu, K. Nishikawa, and A. Kurabayashi. 2017d. Cutaneous microbiota of the Japanese giant salamander (Andrias japonicus), a representative of an ancient amphibian clade. Hydrobiologia:1-15. Bosch, I. 1992. Symbiosis between bacteria and oceanic clonal sea star larvae in the western North Atlantic Ocean. Marine Biology 114:495-502. 133  Boyle, D. G., D. B. Boyle, V. Olsen, J. A. Morgan, and A. D. Hyatt. 2004. Rapid quantitative detection of chytridiomycosis (Batrachochytrium dendrobatidis) in amphibian samples using real-time Taqman PCR assay. Dis Aquat Organ 60:141-148. Brannelly, L. A., M. W. Chatfield, J. Sonn, M. Robak, and C. L. Richards-Zawacki. 2018. Fungal infection has sublethal effects in a lowland subtropical amphibian population. BMC ecology 18:34. Brucker, R. M., C. M. Baylor, R. L. Walters, A. Lauer, R. N. Harris, and K. P. Minbiole. 2008a. The identification of 2,4-diacetylphloroglucinol as an antifungal metabolite produced by cutaneous bacteria of the salamander Plethodon cinereus. J Chem Ecol 34:39-43. Brucker, R. M., R. N. Harris, C. R. Schwantes, T. N. Gallaher, D. C. Flaherty, B. A. Lam, and K. P. Minbiole. 2008b. Amphibian chemical defense: antifungal metabolites of the microsymbiont Janthinobacterium lividum on the salamander Plethodon cinereus. J Chem Ecol 34:1422-1429. Burgess, J. G., E. M. Jordan, M. Bregu, A. Mearns-Spragg, and K. G. Boyd. 1999. Microbial antagonism: a neglected avenue of natural products research. J Biotechnol 70:27-32. Callahan, B. J., P. J. McMurdie, M. J. Rosen, A. W. Han, A. J. A. Johnson, and S. P. Holmes. 2016. DADA2: high-resolution sample inference from Illumina amplicon data. Nature Methods 13:581. Campbell, L., D. S. Bower, S. Clulow, M. Stockwell, J. Clulow, and M. Mahony. 2019. Interaction between temperature and sublethal infection with the amphibian chytrid fungus impacts a susceptible frog species. Scientific Reports 9:83. Caporaso, J. G., K. Bittinger, F. D. Bushman, T. Z. DeSantis, G. L. Andersen, and R. Knight. 2010a. PyNAST: a flexible tool for aligning sequences to a template alignment. Bioinformatics 26:266-267. Caporaso, J. G., J. Kuczynski, J. Stombaugh, K. Bittinger, F. D. Bushman, E. K. Costello, N. Fierer, A. G. Pena, J. K. Goodrich, J. I. Gordon, G. A. Huttley, S. T. Kelley, D. Knights, J. E. Koenig, R. E. Ley, C. A. Lozupone, D. McDonald, B. D. Muegge, M. Pirrung, J. Reeder, J. R. Sevinsky, P. J. Tumbaugh, W. A. Walters, J. Widmann, T. Yatsunenko, J. Zaneveld, and R. Knight. 2010b. QIIME allows analysis of high-throughput community sequencing data. Nature Methods 7:335-336. Cavanaugh, C. M. 1983. Symbiotic chemoautotrophic bacteria in marine invertebrates from sulphide-rich habitats. Nature 302:58. Chao, A. 1984. Non-parametric estimation of the number of classes in a population. Scand. J. Stat. 11:265-270. Chatfield, M. W., L. A. Brannelly, M. J. Robak, L. Freeborn, S. P. Lailvaux, and C. L. Richards-Zawacki. 2013. Fitness consequences of infection by Batrachochytrium dendrobatidis in northern leopard frogs (Lithobates pipiens). Ecohealth 10:90-98. Cheng, T. L., H. Mayberry, L. P. McGuire, J. R. Hoyt, K. E. Langwig, H. Nguyen, K. L. Parise, J. T. Foster, C. K. Willis, and A. M. Kilpatrick. 2017. Efficacy of a probiotic bacterium to treat bats affected by the disease white‐nose syndrome. Journal of Applied Ecology 54:701-708. Clark, M. A., N. A. Moran, P. Baumann, and J. J. Wernegreen. 2000. Cospeciation between bacterial endosymbionts (Buchnera) and a recent radiation of aphids (Uroleucon) and pitfalls of testing for phylogenetic congruence. Evolution 54:517-525. 134  Comeau, A. M., G. M. Douglas, and M. G. Langille. 2017. Microbiome helper: a custom and streamlined workflow for microbiome research. mSystems 2:e00127-00116. Conlon, J. M., J. Kolodziejek, and N. Nowotny. 2004. Antimicrobial peptides from ranid frogs: taxonomic and phylogenetic markers and a potential source of new therapeutic agents. Biochim Biophys Acta 1696:1-14. Conlon, J. M., M. Mechkarska, E. Ahmed, L. Coquet, T. Jouenne, J. Leprince, H. Vaudry, M. P. Hayes, and G. Padgett-Flohr. 2011. Host defense peptides in skin secretions of the Oregon spotted frog Rana pretiosa: implications for species resistance to chytridiomycosis. Dev Comp Immunol 35:644-649. Conlon, J. M., L. K. Reinert, M. Mechkarska, M. Prajeep, M. A. Meetani, L. Coquet, T. Jouenne, M. P. Hayes, G. Padgett-Flohr, and L. A. Rollins-Smith. 2013. Evaluation of the skin peptide defenses of the Oregon spotted frog Rana pretiosa against infection by the chytrid fungus Batrachochytrium dendrobatidis. J Chem Ecol 39:797-805. Costa, S., I. Lopes, D. N. Proença, R. Ribeiro, and P. V. Morais. 2016. Diversity of cutaneous microbiome of Pelophylax perezi populations inhabiting different environments. Science of the total environment 572:995-1004. Culp, C. E., J. O. Falkinham, and L. K. Belden. 2007. Identification of the natural bacterial microflora on the skin of eastern newts, bullfrog tadpoles and redback salamanders. Herpetologica 63:66-71. Daskin, J. H., and R. A. Alford. 2012. Context-dependent symbioses and their potential roles in wildlife diseases. Proc Biol Sci 279:1457-1465. Daskin, J. H., S. C. Bell, L. Schwarzkopf, and R. A. Alford. 2014. Cool temperatures reduce antifungal activity of symbiotic bacteria of threatened amphibians--implications for disease management and patterns of decline. PLoS One 9:e100378. Davis, L. R., L. Bigler, and D. C. Woodhams. 2017. Developmental trajectories of amphibian microbiota: response to bacterial therapy depends on initial community structure. Environ Microbiol 19:1502-1517. del Campo, J., M. Kolisko, V. Boscaro, L. F. Santoferrara, R. Massana, L. Guillou, A. Simpson, C. Berney, C. de Vargas, and M. Brown. 2018. EukRef: phylogenetic curation of ribosomal RNA to enhance understanding of eukaryotic diversity and distribution. bioRxiv:278085. Dillon, R., and V. Dillon. 2004a. The gut bacteria of insects: nonpathogenic interactions. Annual Reviews in Entomology 49:71-92. Dillon, R. J., and V. M. Dillon. 2004b. The gut bacteria of insects: nonpathogenic interactions. Annu Rev Entomol 49:71-92. Dominguez-Bello, M. G., E. K. Costello, M. Contreras, M. Magris, G. Hidalgo, N. Fierer, and R. Knight. 2010. Delivery mode shapes the acquisition and structure of the initial microbiota across multiple body habitats in newborns. Proceedings Of The National Academy Of Sciences Of The United States Of America 107:11971-11975. Dona, J., H. Proctor, S. Mironov, D. Serrano, and R. Jovani. 2018. Host specificity, infrequent major host switching and the diversification of highly host‐specific symbionts: The case of vane‐dwelling feather mites. Global Ecology and Biogeography 27:188-198. Edgar, R. C. 2010. Search and clustering orders of magnitude faster than BLAST. Bioinformatics 26:2460-2461. 135  Egan, S., T. Harder, C. Burke, P. Steinberg, S. Kjelleberg, and T. Thomas. 2013. The seaweed holobiont: understanding seaweed-bacteria interactions. FEMS Microbiol Rev 37:462-476. Eren, A. M., H. G. Morrison, P. J. Lescault, J. Reveillaud, J. H. Vineis, and M. L. Sogin. 2015. Minimum entropy decomposition: unsupervised oligotyping for sensitive partitioning of high-throughput marker gene sequences. ISME J 9:968-979. Erpenbeck, D., J. Breeuwer, H. van der Velde, and R. van Soest. 2002. Unravelling host and symbiont phylogenies of halichondrid sponges (Demospongiae, Porifera) using a mitochondrial marker. Marine Biology 141:377-386. Fahimipour, A. K., M. R. Kardish, J. M. Lang, J. L. Green, J. A. Eisen, and J. J. Stachowicz. 2017. Global-scale structure of the eelgrass microbiome. Applied and Environmental Microbiology 83:e03391-03316. Fernandez-Piquer, J., J. P. Bowman, T. Ross, and M. L. Tamplin. 2012. Molecular analysis of the bacterial communities in the live Pacific oyster (Crassostrea gigas) and the influence of postharvest temperature on its structure. J Appl Microbiol 112:1134-1143. Fisher, R. M., L. M. Henry, C. K. Cornwallis, E. T. Kiers, and S. A. West. 2017. The evolution of host-symbiont dependence. Nature communications 8:15973. Fites, J. S., J. P. Ramsey, W. M. Holden, S. P. Collier, D. M. Sutherland, L. K. Reinert, A. S. Gayek, T. S. Dermody, T. M. Aune, K. Oswald-Richter, and L. A. Rollins-Smith. 2013. The invasive chytrid fungus of amphibians paralyzes lymphocyte responses. SCIENCE 342:366-369. Fites, J. S., L. K. Reinert, T. M. Chappell, and L. A. Rollins-Smith. 2014. Inhibition of local immune responses by the frog-killing fungus Batrachochytrium dendrobatidis. Infection and immunity 82:4698-4706. Foster, J. A., and K. A. McVey Neufeld. 2013. Gut-brain axis: how the microbiome influences anxiety and depression. Trends Neurosci 36:305-312. Funkhouser, L. J., and S. R. Bordenstein. 2013. Mom knows best: the universality of maternal microbial transmission. PLoS Biol 11:e1001631. Gahl, M. K., J. E. Longcore, and J. E. Houlahan. 2012. Varying responses of northeastern North American amphibians to the chytrid pathogen Batrachochytrium dendrobatidis. Conserv Biol 26:135-141. Galac, M. R., I. Bosch, and D. A. Janies. 2016. Bacterial communities of oceanic sea star (Asteroidea: Echinodermata) larvae. Marine Biology 163:162. Gensollen, T., S. S. Iyer, D. L. Kasper, and R. S. Blumberg. 2016. How colonization by microbiota in early life shapes the immune system. SCIENCE 352:539-544. Gloor, G. B., J. M. Macklaim, V. Pawlowsky-Glahn, and J. J. Egozcue. 2017. Microbiome datasets are compositional: and this is not optional. Frontiers in microbiology 8:2224. Godefroy, N., E. Le Goff, C. Martinand-Mari, K. Belkhir, J. Vacelet, and S. Baghdiguian. 2019. Sponge digestive system diversity and evolution: filter feeding to carnivory. Cell and tissue research:1-11. Green, D. M., T. F. Sharbel, J. Kearsley, and H. Kaiser. 1996. Postglacial range fluctuation, genetic subdivision and speciation in the western North American spotted frog complex, Rana pretiosa. Evolution 50:374-390. Hammer, T. J., J. G. Sanders, and N. Fierer. 2019. Not all animals need a microbiome. FEMS Microbiology Letters. 136  Harris, H. 1966. Enzyme polymorphisms in man. Proc. Roy. Soc. Lond. B 164:298-310. Harris, R. N., R. M. Brucker, J. B. Walke, M. H. Becker, C. R. Schwantes, D. C. Flaherty, B. A. Lam, D. C. Woodhams, C. J. Briggs, V. T. Vredenburg, and K. P. Minbiole. 2009. Skin microbes on frogs prevent morbidity and mortality caused by a lethal skin fungus. ISME J 3:818-824. Harris, R. N., T. Y. James, A. Lauer, M. A. Simon, and A. Patel. 2006. Amphibian pathogen Batrachochytrium dendrobatidis is inhibited by the cutaneous bacteria of Amphibian species. Ecohealth 3:53-56. Hayes, M. P. 1997. Status of the Oregon spotted frog (Rana pretiosa sensu stricto) in the Deschutes Basin and selected other systems in Oregon and northeastern California with a rangewide synopsis of the species' status. Hernandez-Agreda, A., R. D. Gates, and T. D. Ainsworth. 2017. Defining the Core Microbiome in Corals' Microbial Soup. Trends Microbiol 25:125-140. Hernandez-Agreda, A., W. Leggat, P. Bongaerts, C. Herrera, and T. D. Ainsworth. 2018. Rethinking the Coral Microbiome: Simplicity Exists within a Diverse Microbial Biosphere. mBio 9. Hewson, I., J. B. Button, B. M. Gudenkauf, B. Miner, A. L. Newton, J. K. Gaydos, J. Wynne, C. L. Groves, G. Hendler, M. Murray, S. Fradkin, M. Breitbart, E. Fahsbender, K. D. Lafferty, A. M. Kilpatrick, C. M. Miner, P. Raimondi, L. Lahner, C. S. Friedman, S. Daniels, M. Haulena, J. Marliave, C. A. Burge, M. E. Eisenlord, and C. D. Harvell. 2014. Densovirus associated with sea-star wasting disease and mass mortality. Proceedings Of The National Academy Of Sciences Of The United States Of America 111:17278-17283. Hoj, L., N. Levy, B. K. Baillie, P. L. Clode, R. C. Strohmaier, N. Siboni, N. S. Webster, S. Uthicke, and D. G. Bourne. 2018. Crown-of-thorns sea star, Acanthaster cf. solaris, have tissue-characteristic microbiomes with potential roles in health and reproduction. Appl Environ Microbiol. Holland, N. D., and K. H. Nealson. 1978. The fine structure of the echinoderm cuticle and the subcuticular bacteria of echinoderms. Acta zoologica 59:169-185. Holmstrom, C., S. Egan, A. Franks, S. McCloy, and S. Kjelleberg. 2002. Antifouling activities expressed by marine surface associated Pseudoalteromonas species. FEMS Microbiol Ecol 41:47-58. Hothorn, T., K. Hornik, M. A. Van De Wiel, and A. Zeileis. 2008. Implementing a class of permutation pests: the coin package. Huggett, M. J., K. McMahon, and R. Bernasconi. 2018. Future warming and acidification result in multiple ecological impacts to a temperate coralline alga. Environmental Microbiology 20:2769-2782. Hughey, M. C., J. Delia, and L. K. Belden. 2017a. Diversity and stability of egg‐bacterial assemblages: The role of paternal care in the glassfrog Hyalinobatrachium colymbiphyllum. Biotropica 49:792-802. Hughey, M. C., J. A. Pena, R. Reyes, D. Medina, L. K. Belden, and P. A. Burrowes. 2017b. Skin bacterial microbiome of a generalist Puerto Rican frog varies along elevation and land use gradients. PeerJ 5:e3688. Hutchinson, V., W. Whitford, and M. Kohl. 1968. Relation of body size and surface area to gas exchange in anuran amphibians. J. Comp. Physiol 109:199-207. 137  Jackson, E. W., C. Pepe-Ranney, S. J. Debenport, D. H. Buckley, and I. Hewson. 2018. The microbial landscape of sea stars and the anatomical and interspecies variability of their microbiome. Frontiers in microbiology 9:1829. Jani, A. J., and C. J. Briggs. 2014. The pathogen Batrachochytrium dendrobatidis disturbs the frog skin microbiome during a natural epidemic and experimental infection. Proc Natl Acad Sci U S A 111:E5049-5058. Jani, A. J., and C. J. Briggs. 2018. Host and Aquatic Environment Shape the Amphibian Skin Microbiome but Effects on Downstream Resistance to the Pathogen Batrachochytrium dendrobatidis Are Variable. Front Microbiol 9:487. Jani, A. J., R. A. Knapp, and C. J. Briggs. 2017. Epidemic and endemic pathogen dynamics correspond to distinct host population microbiomes at a landscape scale. Page 20170944 in Proc. R. Soc. B. The Royal Society. Jones, R. T., L. G. Sanchez, and N. Fierer. 2013. A cross-taxon analysis of insect-associated bacterial diversity. PLoS One 8:e61218. Kelly, M. S., M. Barker, J. D. McKenzie, and J. Powell. 1995. The incidence and morphology of subcuticular bacteria in the echinoderm fauna of New Zealand. The Biological Bulletin 189:91-105. Knutie, S. A., C. L. Wilkinson, K. D. Kohl, and J. R. Rohr. 2017. Early-life disruption of amphibian microbiota decreases later-life resistance to parasites. Nature communications 8:86. Koh, A., F. De Vadder, P. Kovatcheva-Datchary, and F. Bäckhed. 2016. From dietary fiber to host physiology: short-chain fatty acids as key bacterial metabolites. Cell 165:1332-1345. Koskella, B., L. J. Hall, and C. J. E. Metcalf. 2017. The microbiome beyond the horizon of ecological and evolutionary theory. Nature ecology & evolution 1:1606. Kraft, N. J., P. B. Adler, O. Godoy, E. C. James, S. Fuller, and J. M. Levine. 2015. Community assembly, coexistence and the environmental filtering metaphor. Functional Ecology 29:592-599. Kruger, A. 2019. Functional Redundancy of Batrachochytrium dendrobatidis Inhibition in Bacterial Communities Isolated from Lithobates clamitans Skin. Microb Ecol. Kruskal, W. H., and W. A. Wallis. 1952. Use of ranks in one-criterion variance analysis. Journal of the American statistical Association 47:583-621. Krynak, K. L., D. J. Burke, and M. F. Benard. 2016. Landscape and water characteristics correlate with immune defense traits across Blanchard's cricket frog (Acris blanchardi) populations. Biological Conservation 193:153-167. Kueneman, J. G., M. C. Bletz, V. J. McKenzie, C. G. Becker, M. B. Joseph, J. G. Abarca, H. Archer, A. L. Arellano, A. Bataille, M. Becker, L. K. Belden, A. Crottini, R. Geffers, C. F. B. Haddad, R. N. Harris, W. M. Holden, M. Hughey, M. Jarek, P. J. Kearns, J. L. Kerby, J. Kielgast, A. Kurabayashi, A. V. Longo, A. Loudon, D. Medina, J. J. Nunez, R. G. B. Perl, A. Pinto-Tomas, F. C. E. Rabemananjara, E. A. Rebollar, A. Rodriguez, L. Rollins-Smith, R. Stevenson, C. C. Tebbe, G. Vargas Asensio, B. Waldman, J. B. Walke, S. M. Whitfield, K. R. Zamudio, I. Zuniga Chaves, D. C. Woodhams, and M. Vences. 2019. Community richness of amphibian skin bacteria correlates with bioclimate at the global scale. Nat Ecol Evol 3:381-389. 138  Kueneman, J. G., L. W. Parfrey, D. C. Woodhams, H. M. Archer, R. Knight, and V. J. McKenzie. 2014. The amphibian skin-associated microbiome across species, space and life history stages. Mol Ecol 23:1238-1250. Kueneman, J. G., D. C. Woodhams, R. Harris, H. M. Archer, R. Knight, and V. J. McKenzie. 2016. Probiotic treatment restores protection against lethal fungal infection lost during amphibian captivity. Proc Biol Sci 283. Kuhn, D. A. 1981. The genus Cristispira. Pages 555-563  The prokaryotes. Springer. Kung, D., L. Bigler, L. R. Davis, B. Gratwicke, E. Griffith, and D. C. Woodhams. 2014. Stability of microbiota facilitated by host immune regulation: informing probiotic strategies to manage amphibian disease. PLoS One 9:e87101. Lam, B. A., J. B. Walke, V. T. Vredenburg, and R. N. Harris. 2010. Proportion of individuals with anti-Batrachochytrium dendrobatidis skin bacteria is associated with population persistence in the frog Rana muscosa. Biological Conservation 143:529-531. Lam, B. A., D. B. Walton, and R. N. Harris. 2011. Motile Zoospores of Batrachochytrium dendrobatidis Move Away from Antifungal Metabolites Produced by Amphibian Skin Bacteria. Ecohealth 8:36-45. Lauer, A., M. A. Simon, J. L. Banning, B. A. Lam, and R. N. Harris. 2008. Diversity of cutaneous bacteria with antifungal activity isolated from female four-toed salamanders. Isme Journal 2:145-157. Lawrence, S. A., R. O'Toole, M. W. Taylor, and S. K. Davy. 2010. Subcuticular Bacteria Associated With Two Common New Zealand Echinoderms: Characterization Using 16S rRNA Sequence Analysis and Fluorescence in situ Hybridization. Biological Bulletin 218:95-104. Leite, D. C., P. Leão, A. G. Garrido, U. Lins, H. F. Santos, D. O. Pires, C. B. Castro, J. D. van Elsas, C. Zilberberg, and A. S. Rosado. 2017. Broadcast spawning coral Mussismilia hispida can vertically transfer its associated bacterial core. Frontiers in microbiology 8:176. Lemay, M. A., P. T. Martone, P. J. Keeling, J. M. Burt, K. A. Krumhansl, R. D. Sanders, and L. Wegener Parfrey. 2018. Sympatric kelp species share a large portion of their surface bacterial communities. Environ Microbiol 20:658-670. Lemieux-Labonté, V., N. Tromas, B. J. Shapiro, and F.-J. Lapointe. 2016. Environment and host species shape the skin microbiome of captive neotropical bats. PeerJ 4:e2430. Lenth, R. 2018. Emmeans: Estimated marginal means, aka least-squares means. R package version 1. Lesser, M. P., and R. P. Blakemore. 1990. Description of a Novel Symbiotic Bacterium from the Brittle Star, Amphipholis squamata. Appl Environ Microbiol 56:2436-2440. Leung, T., and R. Poulin. 2008. Parasitism, commensalism, and mutualism: exploring the many shades of symbioses. Vie et Milieu 58:107. Li, K., M. Bihan, S. Yooseph, and B. A. Methe. 2012. Analyses of the microbial diversity across the human microbiome. PLoS One 7:e32118. Lloyd, M. M., and M. H. Pespeni. 2018. Microbiome shifts with onset and progression of Sea Star Wasting Disease revealed through time course sampling. Scientific Reports 8:16476. Longcore, J. E., A. P. Pessier, and D. K. Nichols. 1999. Batrachochytrium dendrobatidis gen. et sp. nov., a chytrid pathogenic to amphibians. Mycologia 91:219-227. 139  Longo, A. V., and K. R. Zamudio. 2017. Environmental fluctuations and host skin bacteria shift survival advantage between frogs and their fungal pathogen. ISME J 11:349-361. Louca, S., F. Mazel, M. Doebeli, and L. W. Parfrey. 2019. A census-based estimate of Earth's bacterial and archaeal diversity. PLoS Biology 17:e3000106. Louca, S., M. F. Polz, F. Mazel, M. B. Albright, J. A. Huber, M. I. O’Connor, M. Ackermann, A. S. Hahn, D. S. Srivastava, and S. A. Crowe. 2018. Function and functional redundancy in microbial systems. Nature ecology & evolution:1. Loudon, A. H., J. A. Holland, T. P. Umile, E. A. Burzynski, K. P. Minbiole, and R. N. Harris. 2014a. Interactions between amphibians' symbiotic bacteria cause the production of emergent anti-fungal metabolites. Front Microbiol 5:441. Loudon, A. H., A. Venkataraman, W. Van Treuren, D. C. Woodhams, L. W. Parfrey, V. J. McKenzie, R. Knight, T. M. Schmidt, and R. N. Harris. 2016. Vertebrate Hosts as Islands: Dynamics of Selection, Immigration, Loss, Persistence, and Potential Function of Bacteria on Salamander Skin. Front Microbiol 7:333. Loudon, A. H., D. C. Woodhams, L. W. Parfrey, H. Archer, R. Knight, V. McKenzie, and R. N. Harris. 2014b. Microbial community dynamics and effect of environmental microbial reservoirs on red-backed salamanders (Plethodon cinereus). ISME J 8:830-840. Love, M. I., W. Huber, and S. Anders. 2014. Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2. Genome Biology 15:550. Lozupone, C. A., J. I. Stombaugh, J. I. Gordon, J. K. Jansson, and R. Knight. 2012. Diversity, stability and resilience of the human gut microbiota. Nature 489:220-230. Ma, Y. L., P; Yu, S;  Li, D; Cao S. 2009. Inhibition of common fouling organisms in mariculture by epiphytic bacteria from the surfaces of seaweeds and invertebrates. Acta Ecologica Sinica 29:222–226. Madden, T. 2013. The BLAST sequence analysis tool. The NCBI Handbook [Internet]. 2nd edition. National Center for Biotechnology Information (US). McFall-Ngai, M., M. G. Hadfield, T. C. Bosch, H. V. Carey, T. Domazet-Loso, A. E. Douglas, N. Dubilier, G. Eberl, T. Fukami, S. F. Gilbert, U. Hentschel, N. King, S. Kjelleberg, A. H. Knoll, N. Kremer, S. K. Mazmanian, J. L. Metcalf, K. Nealson, N. E. Pierce, J. F. Rawls, A. Reid, E. G. Ruby, M. Rumpho, J. G. Sanders, D. Tautz, and J. J. Wernegreen. 2013. Animals in a bacterial world, a new imperative for the life sciences. Proceedings Of The National Academy Of Sciences Of The United States Of America 110:3229-3236. McFall-Ngai, M. J. a. E. G. R. 1991. Symbiont recognition and subsequent morphogenesis as early events in an animal-bacterial mutualism. SCIENCE 254:191-194. McKenzie, V. J., R. M. Bowers, N. Fierer, R. Knight, and C. L. Lauber. 2012. Co-habiting amphibian species harbor unique skin bacterial communities in wild populations. ISME J 6:588-596. McMurdie, P. J., and S. Holmes. 2013. phyloseq: an R package for reproducible interactive analysis and graphics of microbiome census data. PLoS One 8:e61217. Medina, D., M. C. Hughey, M. H. Becker, J. B. Walke, T. P. Umile, E. A. Burzynski, A. Iannetta, K. P. Minbiole, and L. K. Belden. 2017. Variation in Metabolite Profiles of Amphibian Skin Bacterial Communities Across Elevations in the Neotropics. Microbial ecology 74:227-238. 140  Miner, C. M., J. L. Burnaford, R. F. Ambrose, L. Antrim, H. Bohlmann, C. A. Blanchette, J. M. Engle, S. C. Fradkin, R. Gaddam, and C. D. Harley. 2018. Large-scale impacts of sea star wasting disease (SSWD) on intertidal sea stars and implications for recovery. PLoS One 13:e0192870. Moran, N. A., J. P. McCutcheon, and A. Nakabachi. 2008. Genomics and Evolution of Heritable Bacterial Symbionts. Annual Review Of Genetics 42:165-190. Mortzfeld, B. M., S. Urbanski, A. M. Reitzel, S. Kunzel, U. Technau, and S. Fraune. 2015. Response of bacterial colonization in Nematostella vectensis to development, environment and biogeography. Environ Microbiol. Mueller, U. G., D. Dash, C. Rabeling, and A. Rodrigues. 2008. Coevolution between attine ants and actinomycete bacteria: a reevaluation. Evolution: International Journal of Organic Evolution 62:2894-2912. Muletz, C. R., J. M. Myers, R. J. Domangue, J. B. Herrick, and R. N. Harris. 2012. Soil bioaugmentation with amphibian cutaneous bacteria protects amphibian hosts from infection by Batrachochytrium dendrobatidis. Biological Conservation 152:119-126. Muletz Wolz, C. R., S. A. Yarwood, E. H. Campbell Grant, R. C. Fleischer, and K. R. Lips. 2018. Effects of host species and environment on the skin microbiome of Plethodontid salamanders. J Anim Ecol 87:341-353. Muletz-Wolz, C. R., J. G. Almario, S. E. Barnett, G. V. DiRenzo, A. Martel, F. Pasmans, K. R. Zamudio, L. F. Toledo, and K. R. Lips. 2017. Inhibition of Fungal Pathogens across Genotypes and Temperatures by Amphibian Skin Bacteria. Front Microbiol 8:1551. Myers, J. M., J. P. Ramsey, A. L. Blackman, A. E. Nichols, K. P. Minbiole, and R. N. Harris. 2012. Synergistic inhibition of the lethal fungal pathogen Batrachochytrium dendrobatidis: the combined effect of symbiotic bacterial metabolites and antimicrobial peptides of the frog Rana muscosa. Journal of chemical ecology 38:958-965. Nakagawa, S., H. Saito, A. Tame, M. Hirai, H. Yamaguchi, T. Sunata, M. Aida, H. Muto, S. Sawayama, and Y. Takaki. 2017. Microbiota in the coelomic fluid of two common coastal starfish species and characterization of an abundant Helicobacter-related taxon. Scientific Reports 7. Nyholm, S. V., and M. J. McFall-Ngai. 2004. The winnowing: establishing the squid-vibrio symbiosis. Nat Rev Microbiol 2:632-642. Nyholm, S. V., and M. K. Nishiguchi. 2008. The Evolutionary Ecology of a Sepiolid Squid-Vibrio Association: From Cell to Environment. Vie Milieu Paris 58:175-184. O’Brien, P. A., N. S. Webster, D. J. Miller, and D. G. Bourne. 2019. Host-Microbe Coevolution: Applying Evidence from Model Systems to Complex Marine Invertebrate Holobionts. mBio 10:e02241-02218. Ogle, D. H. 2015. FSA: fisheries stock analysis. R package version 0.6 13. Oksanen, J., F. G. Blanchet, R. Kindt, P. Legendre, P. R. Minchin, R. B. O'Hara, G. L. Simpson, P. Solymos, M. H. H. Stevens, and H. Wagner. 2015. Package vegan: Community Ecology Package. R package version 2.3-0. Olive, A., and K. Jansen. 2017. The contribution of zoos and aquaria to Aichi Biodiversity Target 12: A case study of Canadian zoos. Global ecology and conservation 10:103-113. Ozawa, G., S. Shimamura, Y. Takaki, K. Takishita, T. Ikuta, J. P. Barry, T. Maruyama, K. Fujikura, and T. Yoshida. 2017. Ancient occasional host switching of maternally 141  transmitted bacterial symbionts of chemosynthetic vesicomyid clams. Genome Biology and Evolution 9:2226-2236. Padgett-Flohr, G. E., and M. P. Hayes. 2011. Assessment of the vulnerability of the Oregon spottedfrog (Rana pretiosa) to the amphibian chytrid fungus (Batrachochytrium dendrobatidis). Herpetological Conservation and Biology 6:99-106. Paine, R. T. 1966. Food Web Complexity and Species Diversity. The American Naturalist 100:65-75. Paster, B. J., D. A. Pelletier, F. E. Dewhirst, W. G. Weisburg, V. Fussing, L. K. Poulsen, S. Dannenberg, and I. Schroeder. 1996. Phylogenetic position of the spirochetal genus Cristispira. Appl Environ Microbiol 62:942-946. Pearl, C. A., J. Bowerman, M. J. Adams, and N. D. Chelgren. 2009. Widespread occurrence of the chytrid fungus Batrachochytrium dendrobatidis on Oregon spotted frogs (Rana pretiosa). Ecohealth 6:209-218. Pedros-Alio, C. 2012. The rare bacterial biosphere. Annual Review of Marine Science 4:449-466. Piotrowski, J. S., S. L. Annis, and J. E. Longcore. 2004. Physiology of Batrachochytrium dendrobatidis, a chytrid pathogen of amphibians. Mycologia 96:9-15. Pollock, F. J., R. McMinds, S. Smith, D. G. Bourne, B. L. Willis, M. Medina, R. V. Thurber, and J. R. Zaneveld. 2018. Coral-associated bacteria demonstrate phylosymbiosis and cophylogeny. Nature communications 9:4921. Pruesse, E., J. Peplies, and F. O. Glöckner. 2012. SINA: accurate high-throughput multiple sequence alignment of ribosomal RNA genes. Bioinformatics 28:1823-1829. Quast, C., E. Pruesse, P. Yilmaz, J. Gerken, T. Schweer, P. Yarza, J. Peplies, and F. O. Glockner. 2013. The SILVA ribosomal RNA gene database project: improved data processing and web-based tools. Nucleic Acids Research 41:D590-596. Rambaut, A. 2012. FigTree v1. 4. Ramsby, B. D., M. O. Hoogenboom, S. Whalan, and N. S. Webster. 2018. Elevated seawater temperature disrupts the microbiome of an ecologically important bioeroding sponge. Molecular Ecology 27:2124-2137. Rebollar, E. A., M. C. Hughey, D. Medina, R. N. Harris, R. Ibanez, and L. K. Belden. 2016. Skin bacterial diversity of Panamanian frogs is associated with host susceptibility and presence of Batrachochytrium dendrobatidis. ISME J 10:1682-1695. Relman, D. A. 2012. The human microbiome: ecosystem resilience and health. Nutrition reviews 70:S2-S9. Rodriguez‐Lanetty, M., C. Granados‐Cifuentes, A. Barberan, A. J. Bellantuono, and C. Bastidas. 2013. Ecological Inferences from a deep screening of the C omplex B acterial C onsortia associated with the coral, Porites astreoides. Molecular Ecology 22:4349-4362. Roeselers, G., and I. L. Newton. 2012. On the evolutionary ecology of symbioses between chemosynthetic bacteria and bivalves. Applied microbiology and biotechnology 94:1-10. Rollins-Smith, L. A., C. Carey, J. Longcore, J. K. Doersam, A. Boutte, J. E. Bruzgal, and J. M. Conlon. 2002. Activity of antimicrobial skin peptides from ranid frogs against Batrachochytrium dendrobatidis, the chytrid fungus associated with global amphibian declines. Dev Comp Immunol 26:471-479. 142  Rollins-Smith, L. A., J. S. Fites, L. K. Reinert, A. R. Shiakolas, T. P. Umile, and K. P. Minbiole. 2015. Immunomodulatory metabolites released by the frog-killing fungus Batrachochytrium dendrobatidis. Infection and immunity 83:4565-4570. Rollins-Smith, L. A., L. K. Reinert, C. J. O'Leary, L. E. Houston, and D. C. Woodhams. 2005. Antimicrobial Peptide defenses in amphibian skin. Integrative and comparative biology 45:137-142. Roth‐Schulze, A. J., J. Pintado, E. Zozaya‐Valdés, J. Cremades, P. Ruiz, S. Kjelleberg, and T. Thomas. 2018. Functional biogeography and host specificity of bacterial communities associated with the Marine Green Alga Ulva spp. Molecular Ecology 27:1952-1965. Russell, D. M., C. S. Goldberg, L. P. Waits, and E. B. Rosenblum. 2010. Batrachochytrium dendrobatidis infection dynamics in the Columbia spotted frog Rana luteiventris in north Idaho, USA. Diseases of Aquatic Organisms 92:223-230. Sabino-Pinto, J., M. C. Bletz, M. M. Islam, N. Shimizu, S. Bhuju, R. Geffers, M. Jarek, A. Kurabayashi, and M. Vences. 2016. Composition of the Cutaneous Bacterial Community in Japanese Amphibians: Effects of Captivity, Host Species, and Body Region. Microbial ecology 72:460-469. Sabino-Pinto, J., P. Galan, S. Rodriguez, M. C. Bletz, S. Bhuju, R. Geffers, M. Jarek, and M. Vences. 2017. Temporal changes in cutaneous bacterial communities of terrestrial- and aquatic-phase newts (Amphibia). Environ Microbiol 19:3025-3038. Scheuring, I., and D. W. Yu. 2012. How to assemble a beneficial microbiome in three easy steps. Ecol Lett 15:1300-1307. Schrope, M. 2014. Sea star wasting. Proc Natl Acad Sci U S A 111:6855. Schulte-Hostedde, A. I., B. Zinner, J. S. Millar, and G. J. Hickling. 2005. Restitution of mass–size residuals: validating body condition indices. Ecology 86:155-163. Shade, A., and J. Handelsman. 2012. Beyond the Venn diagram: the hunt for a core microbiome. Environ Microbiol 14:4-12. Sheafor, B., E. W. Davidson, L. Parr, and L. Rollins-Smith. 2008. Antimicrobial peptide defenses in the salamander, Ambystoma tigrinum, against emerging amphibian pathogens. Journal of Wildlife Diseases 44:226-236. Shukla, S. P., H. Vogel, D. G. Heckel, A. Vilcinskas, and M. Kaltenpoth. 2018. Burying beetles regulate the microbiome of carcasses and use it to transmit a core microbiota to their offspring. Molecular Ecology 27:1980-1991. Sjogren, Y. M., M. C. Jenmalm, M. F. Bottcher, B. Bjorksten, and E. Sverremark-Ekstrom. 2009. Altered early infant gut microbiota in children developing allergy up to 5 years of age. Clinical and experimental allergy : journal of the British Society for Allergy and Clinical Immunology 39:518-526. Skerratt, L. F., L. Berger, R. Speare, S. Cashins, K. R. McDonald, A. D. Phillott, H. B. Hines, and N. Kenyon. 2007. Spread of chytridiomycosis has caused the rapid global decline and extinction of frogs. Ecohealth 4:125-134. Sloan, W. T., S. Woodcock, M. Lunn, I. M. Head, and T. P. Curtis. 2007. Modeling taxa-abundance distributions in microbial communities using environmental sequence data. Microb Ecol 53:443-455. Stamatakis, A. 2014. RAxML version 8: a tool for phylogenetic analysis and post-analysis of large phylogenies. Bioinformatics 30:1312-1313. 143  Team, C. O. S. F. R. 2014. Recovery Strategy for the Oregon Spotted Frog (Rana pretiosa) in British Columbia. Page    Ministry of the Environment, Victoria: B.C. Team, R. C. 2015. R: a language and environment for statistical computing. R Foundation for Statistical Computing. R version 3.2. 1 Vienna, Austria. Theis, K. R., A. Venkataraman, J. A. Dycus, K. D. Koonter, E. N. Schmitt-Matzen, A. P. Wagner, K. E. Holekamp, and T. M. Schmidt. 2013. Symbiotic bacteria appear to mediate hyena social odors. Proc Natl Acad Sci U S A 110:19832-19837. Thomas, C. M., and K. M. Nielsen. 2005. Mechanisms of, and barriers to, horizontal gene transfer between bacteria. Nature Reviews Microbiology 3:711. Thomas, T., L. Moitinho-Silva, M. Lurgi, J. R. Bjork, C. Easson, C. Astudillo-Garcia, J. B. Olson, P. M. Erwin, S. Lopez-Legentil, H. Luter, A. Chaves-Fonnegra, R. Costa, P. J. Schupp, L. Steindler, D. Erpenbeck, J. Gilbert, R. Knight, G. Ackermann, J. V. Lopez, M. W. Taylor, R. W. Thacker, J. M. Montoya, U. Hentschel, and N. S. Webster. 2016a. Diversity, structure and convergent evolution of the global sponge microbiome. Nature communications 7. Thomas, T., L. Moitinho-Silva, M. Lurgi, J. R. Bjork, C. Easson, C. Astudillo-Garcia, J. B. Olson, P. M. Erwin, S. Lopez-Legentil, H. Luter, A. Chaves-Fonnegra, R. Costa, P. J. Schupp, L. Steindler, D. Erpenbeck, J. Gilbert, R. Knight, G. Ackermann, J. Victor Lopez, M. W. Taylor, R. W. Thacker, J. M. Montoya, U. Hentschel, and N. S. Webster. 2016b. Diversity, structure and convergent evolution of the global sponge microbiome. Nature communications 7:11870. Thompson, L. R., J. G. Sanders, D. McDonald, A. Amir, J. Ladau, K. J. Locey, R. J. Prill, A. Tripathi, S. M. Gibbons, and G. Ackermann. 2017. A communal catalogue reveals Earth’s multiscale microbial diversity. Nature 551:457. Trevelline, B. K., K. J. MacLeod, S. A. Knutie, T. Langkilde, and K. D. Kohl. 2018. In ovo microbial communities: a potential mechanism for the initial acquisition of gut microbiota among oviparous birds and lizards. Biol Lett 14. Umile, T. P., P. J. McLaughlin, K. R. Johnson, S. Honarvar, A. L. Blackman, E. A. Burzynski, R. W. Davis, T. L. Teotonio, G. W. Hearn, and C. A. Hughey. 2014. Nonlethal amphibian skin swabbing of cutaneous natural products for HPLC fingerprinting. Analytical Methods 6:3277-3284. Venkataraman, A., C. M. Bassis, J. M. Beck, V. B. Young, J. L. Curtis, G. B. Huffnagle, and T. M. Schmidt. 2015. Application of a neutral community model to assess structuring of the human lung microbiome. mBio 6. Verhulst, N. O., Y. T. Qiu, H. Beijleveld, C. Maliepaard, D. Knights, S. Schulz, D. Berg-Lyons, C. L. Lauber, W. Verduijn, G. W. Haasnoot, R. Mumm, H. J. Bouwmeester, F. H. Claas, M. Dicke, J. J. van Loon, W. Takken, R. Knight, and R. C. Smallegange. 2011. Composition of human skin microbiota affects attractiveness to malaria mosquitoes. PLoS One 6:e28991. Walke, J. B., M. H. Becker, S. C. Loftus, L. L. House, G. Cormier, R. V. Jensen, and L. K. Belden. 2014. Amphibian skin may select for rare environmental microbes. ISME J 8:2207-2217. 144  Walke, J. B., M. H. Becker, S. C. Loftus, L. L. House, T. L. Teotonio, K. P. Minbiole, and L. K. Belden. 2015. Community Structure and Function of Amphibian Skin Microbes: An Experiment with Bullfrogs Exposed to a Chytrid Fungus. PLoS One 10:e0139848. Walke, J. B., Reid N. Harris, Laura K. Reinert, Louise A. Rollins‐Smith,  Douglas C. Woodhams. . 2011. Social immunity in amphibians: evidence for vertical transmission of innate defenses. Biotropica 43:4. Watson, J. W., K. R. McAllister, and D. J. Pierce. 2003. Home ranges, movements, and habitat selection of Oregon Spotted Frogs (Rana pretiosa). Journal of herpetology 37:292-301. Woodhams, D., K. Ardipradja, R. Alford, G. Marantelli, L. Reinert, and L. Rollins‐Smith. 2007a. Resistance to chytridiomycosis varies among amphibian species and is correlated with skin peptide defenses. Animal Conservation 10:409-417. Woodhams, D. C., R. A. Alford, R. E. Antwis, H. Archer, M. H. Becker, L. K. Belden, S. C. Bell, M. Bletz, J. H. Daskin, and L. R. Davis. 2015. Antifungal isolates database of amphibian skin-associated bacteria and function against emerging fungal pathogens: Ecological Archives E096-059. Ecology 96:595-595. Woodhams, D. C., J. Bosch, C. J. Briggs, S. Cashins, L. R. Davis, A. Lauer, E. Muths, R. Puschendorf, B. R. Schmidt, B. Sheafor, and J. Voyles. 2011. Mitigating amphibian disease: strategies to maintain wild populations and control chytridiomycosis. Frontiers in zoology 8:8. Woodhams, D. C., H. Brandt, S. Baumgartner, J. Kielgast, E. Kupfer, U. Tobler, L. R. Davis, B. R. Schmidt, C. Bel, S. Hodel, R. Knight, and V. McKenzie. 2014. Interacting Symbionts and Immunity in the Amphibian Skin Mucosome Predict Disease Risk and Probiotic Effectiveness. PLoS One 9. Woodhams, D. C., L. A. Rollins-Smith, C. Carey, L. Reinert, M. J. Tyler, and R. A. Alford. 2006a. Population trends associated with skin peptide defenses against chytridiomycosis in Australian frogs. Oecologia 146:531-540. Woodhams, D. C., L. A. Rollins-Smith, L. K. Reinert, B. A. Lam, R. N. Harris, C. J. Briggs, V. T. Vredenburg, B. T. Patel, R. M. Caprioli, and P. Chaurand. 2019. Probiotics Modulate a Novel Amphibian Skin Defense Peptide That Is Antifungal and Facilitates Growth of Antifungal Bacteria. Microbial ecology:1-11. Woodhams, D. C., J. Voyles, K. R. Lips, C. Carey, and L. A. Rollins-Smith. 2006b. Predicted disease susceptibility in a Panamanian amphibian assemblage based on skin peptide defenses. J Wildl Dis 42:207-218. Woodhams, D. C., V. T. Vredenburg, M. A. Simon, D. Billheimer, B. Shakhtour, Y. Shyr, C. J. Briggs, L. A. Rollins-Smith, and R. N. Harris. 2007b. Symbiotic bacteria contribute to innate immune defenses of the threatened mountain yellow-legged frog, Rana muscosa. Biological Conservation 138:390-398. Wyngaard, G., and C. Chinnappa. 1982. General biology and cytology of cyclopoids. Developmental biology of freshwater invertebrates:485-533. Zeissig, S., and R. S. Blumberg. 2014. Life at the beginning: perturbation of the microbiota by antibiotics in early life and its role in health and disease. Nature immunology 15:307. Zimmermann, J., C. Wentrup, M. Sadowski, A. Blazejak, H. R. Gruber-Vodicka, M. Kleiner, J. A. Ott, B. Cronholm, P. De Wit, C. Erseus, and N. Dubilier. 2016. Closely coupled evolutionary history of ecto- and endosymbionts from two distantly related animal phyla. Mol Ecol 25:3203-3223. 145  Appendices Appendix A   All figures and tables in this Appendix correspond to Chapter 2.       Appendix Figure A.1 Alpha diversity of bacterial communities of oral and aboral surfaces and ceca of Pisaster compared to Abiotic surfaces and seawater. Letters represent statistically significant differences between sample type determined by least square means with a Tukey adjustment (𝝰 = 0.05).      146   Appendix Figure A.2 Three species of sympatric sea stars have different bacterial community compositions.   Stress = 0.148. df= 2:34 R2 = 0.459, Pseudo F= 14.461. p<0.001.      Appendix FigureA.3 Filtering low abundance reads from individual samples has no affect on beta diversity. Testing the effect of filtering sequences that are in samples in low abundances on beta diversity.  Stress: No filtering in all surface samples: 0.191; Filtered 2, all samples 0.185; 147  Filtered 10, all samples 0.188.  No filtering just Pisaster: 0.143, Filtered 2, just Pisaster: 0.140, Filtered 10 just Pisaster: 0.148.      Appendix Figure A.4 A comparison of the shared bacteria between sea star surfaces and their respective environments at different filtering depths of LOR.  All the Pisaster and environmental samples are pooled in this figure.      Appendix Table A.1 Adonis result for testing the effect of filtering on beta diversity for sample type.     df:	8:125R2	 F	value	 P	value	All 0.3703 9.1876 0.0012 0.3701 9.1801 0.00110 0.3694 9.1543 0.001148  Appendix Table A.2 Adonis results for testing the effect of filtering LOR on beta diversity when looking at geographical location and side of sea star surface.      Appendix Table A.3 Abundant bacteria that are differentially abundant on aboral and aboral sides of Pisaster (DESeq2) and are present in an average relative abundance of greater than 1%.      PisasterFiltering Test df R2	 F-	value	 P	value	Sample	side	 1:47 0.0554 4.194 0.001All location 2:47 0.2262 8.562 0.001Interaction 2:47 0.0974 3.687 0.001Sample	side	 1:47 0.0562 4.28 0.0012 location 2:47 0.2268 8.643 0.001Interaction 2:47 0.1005 3.83 0.001Sample	side	 1:47 0.0558 4.224 0.00210 location 2:47 0.2224 8.466 0.001Interaction 2:47 0.099 3.746 0.001ID Aboral	Rel.	abund Oral	Rel.	abund stat padj10720 0.035 0.007 4.292 0.002 Marine	Spirochaete29557 0.032 0.003 -2.662 0.039 p:	Bacteroidetes,	g:	Reichenbachiella37708 0.015 0.008 2.603 0.040 p:	Bacteroidetes,	g:	Reichenbachiella36954 0.001 0.018 4.791 0.000 p:	Verrucomicrobia,	g:	Rubritalea149   Appendix Figure A.5 Phylogenetic tree of Spirochaete OTU 10720. Spirochaete OTU 10720 was abundant on Pisaster and its most similar sequences were detected on bivalves and a sponge. Other similar sequences were detected in marine mats. Numbers that are located at nodes represent bootstrap confidence intervals (n=100) that are greater than 50.    150   Appendix Figure A.6 Phylogenetic tree of Reichenbachiella. OTU29557 and OTU37708 fall within the genus Reichenbachiella (Bacteroidetes; Cytophagia) and make up ~4 % and 2% of the sequences found on the surfaces of Pisaster. Some closely related OTUs are associated with marine invertebrates and others are associated with environmental sources. Numbers that are located at nodes represent bootstrap confidence intervals (n=100) that are greater than 50.   151   Appendix Figure A.7 Phylogenetic tree of Peregrinibacteria. OTU5961 make up ~1% of the sequences found on the surfaces of Pisaster within the genus Peregrinibacteria.  This OTU does not cluster with sequences that were detected with marine hosts. There is no clear pattern between host and environmental sequences. Numbers that are located at nodes represent bootstrap confidence intervals (n=100) that are greater than 50.     152  Appendix B   All tables and graphs in this section correspond to Chapter 3.  Appendix Table B.1 Environmental characteristic for MT sampling locations. At each sampling location, three water samples from geographically separate areas of the lakes were collected in 50 mL Falcon tubes.  Samples were acidified with 500 µL of trace metal grade HNO3 for preservation. These samples were used to determine the concentrations of several elements important to bacterial growth (Na, Mg, K, Ca, Mn, Fe) at each site using a PerkinElmer Optima 2000DV atomic emission spectrometer.  Total nitrogen, phosphorus, and sulfur were determined by Energy Labs, Inc (Helena, Montana).          Appendix Table B.2  Summary of protein concentration and AMP MIC by site.      Environmental	characters Doney Park	 Jones	 GipseyMean	Total	N	(mg/L) 1.20 0.20 1.60 0.20Mean	Total	P	(mg/L) 0.04 0.02 0.08 0.00Mean	Total	S	(mg/L) 0.70 2.10 6.70 3.30Mean	Total	Fe	(mg/L) 0.15 4.21 0.25 0.02Mean	Total	Mg	(mg/L) 3.10 0.74 1.27 3.31Mean	Total	Mn	(mg/L) 0.03 0.32 0.01 0.00Mean	Total	Na	(mg/L) 1.10 2.59 11.20 1.72Mean	Total	Ca	(mg/L) 20.83 0.22 0.40 4.37Mean	Total	K	(mg/L) 0.29 1.16 4.12 7.82Mean	Water	Temp	(C) 21.60 10.07 25.57 12.00Mean	DO	(mg/L) 7.55 8.88 12.30 9.14Mean	pH 7.65 6.95 10.84 8.08Mean	Conductivity	(mg/L) 109.07 33.30 88.60 126.87153  Appendix Table B.3 Shared OTUs between different locations. Bold numbers indicate the total number of OTUs for the frogs at their respective location. Gipsy 178    Jones 74 130   Park 81 56 230  Doney 83 62 69 148   Gipsy Jones Park Doney  Appendix Table B.4 Statistics and antifungal status of core bacteria on Columbia spotted frogs.    Appendix Table B.5 Prevalence of core bacteria on Columbia spotted frogs.       Appendix Table B.6 Kendall’s rank Correlations between Bd intensity and core bacterial abundance; the Holm adjustment was used to take account of multiple comparisons           OTU log2FoldChange stat pvalue padj CSF Water Antifungal	Status taxonomy87682 12.923 14.81 1.30E-49 4.51E-47 0.329 0.000 Inhibitory Proteobacteria;	Burkholderiaceae;	Rhizobacter39630 13.171 10.11 5.19E-24 2.58E-22 0.297 0.000 No	Match Bacteroidetes;	Weeksellaceae;	Chryseobacterium44464 11.310 9.202 3.51E-20 1.02E-18 0.072 0.000 No	Match Patescibacteria;	Gracilibacteria74141 8.710 5.82 5.89E-09 5.69E-08 0.028 0.000 No	Match Proteobacteria;	Oligoflexaceae51056 24.676 14.46 2.24E-47 3.90E-45 0.014 0.000 No	Match Verrucomicrobia;	Rubritaleaceae;	LuteolibacterDoney Gipsey Jones Park All All	%OTU87682:Proteobacteria;	Burkholderiaceae; 	Rhizobacter 8/8 10/10 7/7 9/9 34/34 100OTU39630:Bacteroidetes;	Weeksellaceae;	Chryseobacterium 8/8 10/10 6/7 8/9 32/34 94OTU44464:Patescibacteria;	Gracilibacteria 8/8 10/10 4/7 7/9 29/34 85OTU5105:Proteobacteria	;Oligoflexaceae 6/8 8/10 3/7 4/9 21/34 62OTU7414:Verrucomicrobia;	Rubritaleaceae;	Luteolibacter 6/8 10/10 5/7 2/9 23/34 68z	 tau	 p p-adjBd	X		OTU39630.Chryseobacterium 2.70 0.34 0.01 0.03Bd	X	OTU87682.Rhizobacter -2.58 -0.33 0.01 0.04Bd	X	OTU7414.Luteolibacter 1.94 0.26 0.05 0.16Bd	X	OTU5105.Oligoflexaceae 1.70 0.23 0.09 0.18Bd	X		OTU44464.Gracilibacteria 1.18 0.15 0.24 0.24154  Appendix Table B.7 Logistic regression between Bd status and core bacterial abundance; the Holm adjustment was used to take account of multiple comparisons.    Appendix Table B.8 Correlations between AMP MIC against Bd and core bacteria. Holm adjustment was used to take account of multiple comparisons; df = 15.     Appendix Table B.9 Summary of results from correlations across the entire dataset. Bd is intensity for these results.      n z p	 adjusted	p-value	Bd	X	OTU87682.Rhizobacter 34 -2.534 0.0113 0.0565Bd	X	OTU39630.Chryseobacterium 34 2.329 0.0198 0.0792Bd	X	OTU44464.Gracilibacteria 34 1.161 0.246 0.738Bd	X	OTU5105.Oligoflexaceae 34 0.989 0.322 0.644Bd	X	OTU7414.Luteolibacter 34 0.743 0.457 0.457t cor p-value adjusted	p-valueMIC	X	OTU87682.Rhizobacter 2.51 0.54 0.024 0.121MIC	X	OTU7414.Luteolibacter -1.33 -0.33 0.203 0.81MIC	X	OTU44464.Gracilibacteria -1.22 -0.3 0.24 0.72MIC	X	OTU39630.Chryseobacterium	 -0.59 -0.15 0.562 1.12MIC	X	OTU5105.Oligoflexaceae 0.17 0.05 0.864 0.864Test df z	or	t	 tau	or	cor pBd	X	MIC Kendall 16 -0.41937 -0.07883 0.6749Bd	X	[Protein] Kendall 16 0.16725 0.031117 0.8672MIC	X	[Protein] Pearson 16 -3.6091 -0.0669898 0.002354MIC	X	Antifungal Pearson 15 2.1913 0.4924 0.04463Bd	X	Antifungal	 Kendall 33 -2.7347 -0.3479756 0.006244OTU87682.Rhizobacter	X	OTU39630.ChryseobacteriumPearson 32 -5.9263 -0.7233613 1.344E-06MIC	X	Metabolite	Richness Pearson 15 -1.14511 -0.3508565 0.1673[Protein]	X	Metabolite	Richness Pearson 15 3.0422 0.6177182 0.00822Bd	X	Metabilite	Richness Kendall 32 0.56818 0.075026 0.5699MIC	X	Chao1 Pearson 15 -1.7084 -0.40359 0.1082MIC	X	Simpson_e Pearson 15 -0.85322 -0.215141 0.407[Protein]	X	Chao1 Pearson 15 1.612 0.3842633 0.1278[Protein]	X	Simpson_e Pearson 15 -0.20571 -0.05304 0.8398Chao1	X	Bd Kendall 33 -0.54403 -0.06939 0.5864Simpson_e	X	Bd Kendall 33 0.93228 0.11862 0.3512Chao1	X	Metabolites	 Pearson 31 1.18244 0.31138 0.07774Simpson	X	Metabolites Pearson 31 1.1449 0.201407 0.261155  Appendix Table B.10 Summary of results from mixed effect logistic regressions, which take location into account.     Appendix Table B.11 Bacteria that are significantly more abundant on CSF (DeSeq) that are less than 1% average relative abundance.   Logistic	regression n z p	Bd	X	Antifungal	 34 -2.656 0.00791Simpson_e	X	Bd 34 1.549 0.121Bd	X	[Protein] 17 1.159 0.247Chao1	X	Bd 34 -0.724 0.469Bd	X	Metabilite	Richness 33 0.616 0.538Bd	X	MIC 17 -0.307 0.759OTU stat pvalue padj CSF	* Water* taxonomy39632 6.04 2E-09 2E-08 0.009 0 Chryseobacterium39633 6.44 1E-10 1E-09 0.008 0 Chryseobacterium87687 9.35 9E-21 3E-19 0.009 0 Rhizobacter44467 4.68 3E-06 2E-05 0.001 0 Gracilibacteria39631 2.57 1E-02 3E-02 0.009 0 Chryseobacterium87684 6.19 6E-10 6E-09 0.002 0 Rhizobacter44465 4.53 6E-06 4E-05 0.001 0 Gracilibacteria87683 6.26 4E-10 4E-09 0.001 0 Rhizobacter87685 5.55 3E-08 2E-07 0.001 0 Rhizobacter74143 2.31 2E-02 5E-02 0.001 0 Verrucomicrobia;	Roseibacillus	156   Appendix Figure B.1 Relationship between peptide concentration and MIC of peptides. There is a relationship between the concentration of skin proteins per surface area and MIC per surface area (Pearson, t= -3.4576, df = 15, p = 0.0035, cor = -0.6659).   Appendix Figure B.2 Relationship between AMP MIC/Surface are and Log Bd, showing that AMP MIC does not correspond to Bd intensity for CSF.  Mean for all samples in red dotted line. 157   Appendix Figure B.3 Metabolite richness differs between locations  (Kruskal-Wallis, chi-squared=14.119, df = 3, p = 0.0027), with Jones having a lower richness than the other locations (Dunn test, alpha<0.05).  158   Appendix Figure B.4 Metabolite composition differs by geographic location (F3:31= 3.1513, R2=0.21762; p = 0.001), there was no difference with Bd status F1:31= 1.5707, R2=0.03616; p = 0.16) and there was no interaction (F1:31= 1.4175, R2=0.03263; p = 0.197). We used pairwise.adonis to determine that Jones is different from the other three locations (Letters next to location names).    

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
https://iiif.library.ubc.ca/presentation/dsp.24.1-0380660/manifest

Comment

Related Items