Variation in Marine Threespine Stickleback (Gasterosteus Aculeatus) and Its Implications for Adaptive Divergence

University of Calgary PRISM: University of Calgary's Digital Repository

Graduate Studies The Vault: Electronic Theses and Dissertations

2017 Contemporary ancestor? Variation in marine threespine stickleback (Gasterosteus aculeatus) and its implications for adaptive divergence

Morris, Matthew

Morris, M. (2017). Contemporary ancestor? Variation in marine threespine stickleback (Gasterosteus aculeatus) and its implications for adaptive divergence (Unpublished doctoral thesis). University of Calgary, Calgary, AB. doi:10.11575/PRISM/25434 http://hdl.handle.net/11023/3910 doctoral thesis

University of Calgary graduate students retain copyright ownership and moral rights for their thesis. You may use this material in any way that is permitted by the Copyright Act or through licensing that has been assigned to the document. For uses that are not allowable under copyright legislation or licensing, you are required to seek permission. Downloaded from PRISM: https://prism.ucalgary.ca UNIVERSITY OF CALGARY

Contemporary ancestor? Variation in marine threespine stickleback (Gasterosteus

aculeatus) and its implications for adaptive divergence

Matthew Richard John Morris

A THESIS

SUBMITTED TO THE FACULTY OF GRADUATE STUDIES

IN PARTIAL FULFILMENT OF THE REQUIREMENTS FOR THE

DEGREE OF DOCTOR OF PHILOSOPHY

GRADUATE PROGRAM IN BIOLOGICAL SCIENCES

CALGARY, ALBERTA

JUNE, 2017

© Matthew Richard John Morris 2017 Abstract Standing genetic variation (SGV) can affect the incidence and pace of adaptation and parallel evolution. The role of SGV versus de novo mutation can be tested in ancestral-derived comparisons when the “contemporary ancestor” is extant. Assumptions about SGV in these contemporary ancestors require formal testing. The threespine stickleback is an icon of adaptive divergence, with multiple freshwater forms having evolved in parallel from a presumably panmictic, evolutionarily static marine population – in part from SGV at Ectodysplasin. Variation among marine stickleback would therefore have consequences for understanding adaptive divergence. I collected marine stickleback from eight locations between Alaska and California. Marine populations varied according to ecogeographic rules. Genotype-by- Sequencing of over 380 000 loci and 5700 SNPs revealed five genetic clusters, including one extending north from Washington to Alaska. Pairwise estimates of genetic differentiation (FST) ranged from 0.02 to 0.18. Tests of phenotypic divergence (PST-FST) for plate counts and body shape fell outside neutral evolutionary expectations, suggesting adaptive divergence may be maintaining this quantitative phenotypic variation among marine populations. Since SGV differed between populations, estimates of candidate loci exhibiting potential selection in response to freshwater colonisation varied depending on the marine population chosen as “ancestral”. It has been theorized that genome-wide heterozygosity improves fitness by buffering against asymmetry. If so, SGV could be maintained if it canalizes plate number. Although heterozygosity and asymmetry varied independently, SGV at Ectodysplasin acted as a genetic stressor that increased asymmetry. Critical thermal minima may have evolved from SGV. Contrary to expectations, marine and freshwater stickleback exhibited the same reaction norm for mitochondrial biogenesis, suggesting that biogenesis has not evolved but has retained an ancestrally adaptive phenotype. Collectively, these results reinforce that SGV is a complex and important factor in the evolution of “contemporary ancestors”, and that failure to take these complexities into account can lead to spurious interpretations of adaptation in derived populations.

ii Preface

Although this thesis represents my work, no research is conducted in a vacuum. Below is a description of my contribution and the contributions of co-authors (for submitted papers) or research assistants for each of Chapters Three through Six.

Morris MRJ, Petrovitch E, Bowles E, Jamniczky HA, Sogers SM (second revised version awaiting approval). Exploring Jordan’s Rule in Pacific threespine stickleback Gasterosteus aculeatus. Journal of Fish Biology. MS 16-580R2.

This paper presents Chapter Three in slightly modified form and has been accepted for publication pending minor revisions in the Journal of Fish Biology. Permissions from co-authors can be found in Appendix E. At this point no contract has been signed giving copyright ownership to the journal. The authors contributed as follows: I collected all but the Alaskan stickleback. E. Petrovitch was an undergraduate student who X-rayed all of the fish and did initial vertebral counts and measures of standard length. She submitted this work as an ECOL 507 paper entitled “Assessing Jordan’s Rule in coastal Pacific threespine stickleback.” Her work was the foundation for my work in this chapter; I re- scored all vertebral counts, partitioning them into abdominal and caudal vertebrae, genotyped most individuals at the idh locus to determine sex, counted all other meristic traits, did all of the statistics reported in this chapter, and wrote the manuscript. An additional undergraduate, R. Kaufman, did some of the idh genotyping. E. Bowles collected the Alaskan stickleback, which were vital to the success of this paper. H. Jamniczky and S. Rogers co-supervised E. Petrovitch, and funded this project.

Morris MRJ, Bowles E, Allen B, Jamniczky HA, Rogers SM (submitted) Contemporary ancestor? Adaptive divergence from standing genetic variation in Pacific marine threespine stickleback. Evolution. 17-0292.

This manuscript presents Chapter Four and has been submitted to Evolution. The authors contributed as follows: E. Bowles collected the Alaskan stickleback. E. Bowles,

iii B. Allen and I collaborated to develop the Genotype-by-Sequencing protocol used in this thesis, which included modifications to the Stacks pipeline freely provided by Eric Normandeau, and R-code for using hierfstat and Adegenet. In particular, E. Bowles wrote the script used for GSnap and qqman used in this thesis. H. Jamniczky provided access to the μCT scanner for 3D morphometrics and valuable feedback on the morphology section of this paper. S. Rogers provided funding, feedback, and support throughout this project. Three undergraduates (E. Ellefson, V. Heather, R. Kaufman), not credited as co-authors, provided some assistance in PCR and gel electrophoresis for both idh and Eda, although most of this work was mine. One of these undergraduates also did initial plate counts, although for various reasons I did all of the counts again. I did the following: collected stickleback, extracted DNA, ran most of the PCR, scanned and landmarked all fish, did all morphological analyses, did the entire GBS pipeline and data analysis after initial development with E. Bowles and B. Allen, did all statistical analyses, and wrote the document. Chapter Five has not yet been submitted for publication. The GBS and Eda contributions have already been described. I further assessed plate position on both the left and right side of each stickleback twice to calculate measurement error. I did all statistical analyses and wrote the document. S. Rogers provided funding, feedback, and support. R. Kaufman was an undergraduate student who provided initial plate count information and did initial PCR for Eda and idh for half of the marine stickleback. For various reasons, I redid all of her work, but it is important to note her initial contribution. She published her results for ECOL 528 as “Standing genetic variation and latitudinal clines in threespine stickleback.” I have not sought permission to use her data for this thesis, as I redid the data collection. Chapter Six has not yet been submitted for publication. S. Smith, A. Pistore, and T. Barry collected all stickleback used in this project and did initial fish care at the Bamfield Marine Sciences Center. Once fish were shipped to the University of Calgary, I cared for them with the help of an undergraduate student, N. Hehar, and the occasional help of F. Malik. I designed the experiment. N. Hehar assisted with dissections. All cardiac tissue was processed by N. Tahbaz at the University of Alberta; he took images on the transmission electron microscope and I counted mitochondria from those images.

iv W. Dong processed pectoral tissue at the University of Calgary and I took all images on the transmission electron microscope. Pectoral mitochondrial counts were divided between me and an undergraduate student, J. Rosebush. I wrote the manuscript and did all statistics. S. Rogers provided funding, feedback, and support.

Publications during my PhD tenure not included in this thesis Morris MRJ, Rogers SM (2013) Overcoming maladaptive plasticity through plastic compensation. Current Zoology. 59: 526-536. Originally written for BIOL 607 – Special topics in biology: Plastic compensation.

Morris MRJ (2014) We know in part: James McCosh on evolution and Christian faith. Journal of the History of Biology. 47: 363-410. Originally written for BIOL 607 – Special topics in biology: Darwin’s Origin of Species.

Morris MRJ, Rogers SM (2014) Integrating phenotypic plasticity within an ecological genomics framework: recent insights from the genomics, evolution, ecology and fitness of plasticity. In: Ecological Genomics. Eds. CR Landry, N Aubin-Horth. Springer: UK. Originally written as the first chapter of this thesis, before this thesis transformed into something different.

Morris MRJ, Richard R, Leder EH, Barrett RDH, Aubin-Horth N, Rogers SM (2014) Gene expression plasticity evolves in respone to colonization of freshwater lakes in threespine stickleback. Molecular Ecology. 23: 3226-3240. Adapted from my MSc thesis.

Morris MRJ (2014) Plasticity-mediated persistence in new and changing environments. International Journal of Evolutionary Biology. 2014: 416497. Greatly modified from one of my PhD candidacy papers.

Rogers SM, Morris MRJ (2014) Alberta fishes: a folding pocket guide to all known native and introduced species. Waterford Press.

v Morris MRJ (2015) Review: “Process and providence: The evolution question at Princeton, 1845-1929” by Bradley J. Gundlach. Reports of the National Center for Science Education. 35: 12.1-12.3. This was an invited review based on my McCosh paper.

Hodgson R, Waytuck T, Morris MRJ, Rogers SM (2015) DNA barcoding in the classroom: investigating fish labeling. The Barcode Bulletin. 6: 6-7. This was a result of an undergraduate project for the molecular ecology class.

Somers C, Morris MRJ (2016) Saskatchewan fishes: a folding pocket guide to all known native and introduced species. Waterford Press.

vi Acknowledgements When I started this thesis in January of 2012 I was living the life of a bachelor in a basement apartment. The second week into this PhD I asked Danielle Mackinnon to marry me and she said yes. We were married that summer, and have celebrated four incredible years together. I now have a two (nearly three!)-year-old girl (Celeste Meredith) and a one-year-old boy (Lochlan Royce), and just completed my first year teaching in a tenure-track biology position at Ambrose University. I am incredibly grateful to my supervisor, Dr. Sean Rogers, who has granted me the space to pursue the family life while also spurring me on to complete this thesis in a timely manner – not to mention the encouragement (and funding opportunities) he has provided along the way. I see how some supervisors treat their students, and understand that my life could have taken a different track under a different supervisor. I also look at his incredible output as a researcher while being a husband and father to three young boys, and cannot help but be in awe of his ability to handle stress with a smile and an encouraging word. I have learned a great deal from Sean about work-life balance, and I hope this is not the end of our collaborative research. This thesis is substantially different from what I had initially envisioned. I was going to focus on the evolution of phenotypic plasticity in stickleback spread along a latitudinal gradient. In the summer of 2013 I collected and preserved stickleback from along the coast, with the intention of collecting live specimens the following year. But there were unforeseen regulatory changes, specifically with the Canadian Food Inspection Agency, that stalled all possibilities of bringing back live fish. Unfortunately, the CFIA constantly offered hopeless hope that resulted in untold hours put in by myself, Sean, the university’s health and safety team, the LESARC staff, and numerous others to make this become a reality. I would like to thank all of those who did everything within their power to fulfill the new regulations, even if ultimately this approach had to be abandoned. This left me with quite a predicament – what was I going to do for my thesis? Thank you to my committee for their understanding and their feedback as the focus of this thesis shifted over time – thanks to Dr. Gordon Chua for his conversations on sequencing, and especially thanks to Dr. Lawrence Harder. During my Masters I made some claims about marine stickleback that Dr. Harder challenged. I had a wealth of

vii literature to rely on that made the same claims – but it quickly became apparent that those claims were themselves unfounded. It was Dr. Harder’s challenges, with Dr. Rogers’ support, that led to my desire to sequence marine stickleback along the coast, and ultimately led to the thesis you see before you. I had never dreamed that this subject would become so interesting - in fact, I was relying on the plasticity experiments so that I could have something novel in this thesis, rather than a #honestthesis line about how “I tested to see if a single population of fish was actually a single population, and it was.” Instead, there was an embarrassment of riches to be harvested from these dead fish that I brought back in the summer of 2013. There are many people who were involved in helping me successfully complete this PhD. First, without samples I would have had nothing. Thanks to Ella Bowles for collecting the Alaskan population used in this thesis, and Sara Smith, Alex Pistore, and Tegan Barry for acquiring and initially caring for the stickleback that make up the mitochondrial plasticity study (Chapter Six), and to Tegan and Alex for all of their assistance in working around Amira and MorphoJ (and for giving up some of their scanner time for my project). Danielle Morris helped me collect stickleback from Vancouver to California. It was, perhaps surprisingly, a highlight of our marriage – from chasing the low tides at two in the morning, to seeing a wall of white fog instead of Mount St. Helen’s, to discovering that Schnitzel Schnandwiches are actually really delicious (Coos Bay is an ugly town with some great restaurants). After several days of frustratingly capturing everything in Coos Bay except stickleback, my wife suggested I contact the Estuarine Research Reserve. This was, beyond her companionship (and incredible Tetris-style packing abilities), her greatest contribution to the trip, as I met some wonderfully helpful and knowledgeable people not only in Coos Bay, but across California. I could not have done this without the help of those researchers: Larry Basch and Tom Gaskill of the South Slough National Estuarine Research Reserve (Coos Bay, Oregon), Kerstin Wasson of the Elkhorn Slough Foundation (Elkhorn Slough, California), Denise Homer of Friends of Arcata Marsh (Arcata, California), the Bamfield Marine Sciences Center (Vancouver Island, BC), and the National Park Service in King Salmon, Alaska. I am also grateful to the numerous government employees at federal and state/provincial levels who approved my permits in a timely fashion. I learned during this

viii project how ineffective government agents can impede research, and am grateful for all of you who excelled at your jobs and were pleasant to interact with in the process. In particular, many thanks to the border agents when an accident on the highway caused me to arrive at the US border after Fish and Wildlife had closed their office – and allowed me to slip the forms under the door and cross the border without so much as an inspection. Thank you to several undergraduate students: Ekaterina Petrovitch, who X-rayed all of my stickleback; Rebecca Kaufman for her initial DNA extractions, plate counts, and PCRing; Emily Ellefson and Viviana Heather for their great work PCRing and gelling some particularly difficult fish; Navneet Hehar for fish care and assistance with tissue preservation for the mitochondrial experiment; Faizan Malik for fish care; and Jonathan Rosebush for taking on some of the tedious mitochondrial counts. My thesis is stronger for your work, and I hope you found it valuable experience. Thank you also to many amazing researchers who provided support along the way. In particular: Dr. Heather Jamniczky and Dr. Benedikt Hallgrimsson for letting me use their CT scanner – and particularly to Dr. Jamniczky for her encouragement and feedback throughout this process; Dr. Anthony Russell for the use of his X-ray; to the incredible people at UCDNA Services for training and assistance on several machines, including the NextSeq 500 used in this thesis; Dr. Campbell Rolian for his support during my candidacy and since; Dr. Nasser Tahbaz of the University of Alberta and Dr. Kristin O’Brien at the Institute of Arctic Biology for initial feedback on the mitochondrial project; and Dr. Wei Dong for processing pectoral tissue. The support staff at the University of Calgary was also incredible. In particular, Wayne Jansen, Stefanie Anderson, Rob Hampton and the rest of the LESARC team have been a huge support throughout my Masters and PhD. Karen Barron, David Beninda, Sophia George, Warren Fitch, Christine Goodwin, Laureen Clement, and Hayley Harris have been invaluable in more ways than I likely even know. None of the sequencing work would have been possible if it wasn’t for the incredible support of WestGrid (www.westgrid.ca) and Compute Canada Calcul Canada (www.computecanada.ca). Many people invested in paths not taken. Thank you to Dr. Patricia Schulte, Dr. Jane Shearer, and Dillon Chung for training on how to use the Oroboros to measure

ix mitochondrial respiration. This is a project I would still love to do, even if it fell outside the scope of this thesis. Thanks as well to Bree Yednock of the South Slough National Estuarine Research Reserve who strove to find a space for me in Coos Bay to work with American-caught stickleback. The timing simply did not work on my end, but I am grateful for her enthusiasm and hope to get back to Coos Bay in the near future. None of this would have been possible without an incredible lab. Ella Bowles, Stevi Vanderzwan, Jon Mee, Jobran Chebib, Sara Smith, Tegan Barry, Alex Pistore, Jori Harrison and honorary lab mates Emma Carroll and Mason Kulbaba contributed a great deal to this project, from participating in fish care to helping establish the protocols that resulted in the Genotype-by-Sequencing run. In particular, Mason and Jon began the GBS protocol, and Ella and Brandon helped refine it. I could not have done this without you. It has been a privilege to work alongside you, and I hope for fruitful collaborations in the years to come. Personal funding for this work came from NSERC Vanier and the Killam pre- doctoral scholarships, for which I am immensely grateful. Grants were also provided by Alberta Innovates Technology Futures top-ups. Funding graduate students is important, and I hope my work reflects the trust you placed in me. While writing this thesis I was also granted a course release during the winter semester from Ambrose University, without which this would have been a far more stressful ordeal. Thanks to Linda Schwartz and Carol Kroeker for their support and encouragement as I divided my time between thesis and work, and I look forward to being able to focus solely on my job. Finally, to my wife Danielle – your support and understanding was crucial, particularly during the final year of this PhD as I worked too many weekends to count. Life happens, even during a PhD, but you worked hard to limit the amount of life that came my way. Without you I would still be in the midst of this thesis rather than towards the end. You are my encouragement and my friend. I look forward to spending significant time with you and Celeste and Lochlan when this is finally over. I love you and am deeply proud to call you family.

x Dedication For Celeste and Lochlan, with love.

xi Table of Contents Abstract ...... ii Preface...... iii Acknowledgements ...... vii Dedication ...... xi Table of Contents ...... xii List of Tables ...... xvi List of Figures and Illustrations ...... xviii List of Symbols, Abbreviations and Nomenclature ...... xxi Epigraph ...... xxvi

STANDING GENETIC VARIATION, CONTEMPORARY ANCESTORS, AND MARINE THREESPINE STICKLEBACK ...... 1 1.1 Introduction ...... 1 1.2 Standing genetic variation ...... 2 1.2.1 Definitions ...... 2 1.2.2 Predictions about SGV during adaptation ...... 4 1.2.3 Assumptions regarding the role of SGV during adaptation ...... 7 1.3 Threespine stickleback ...... 9 1.3.1 Standing genetic variation and ancestral-derived comparisons ...... 9 1.3.2 Contemporary ancestors: assumptions about SGV matter ...... 12 1.4 Marine threespine stickleback ...... 15 1.4.1 Defining “marine” ...... 15 1.4.2 Phenotypic uniformity ...... 19 1.4.3 Genetic variation among marine populations ...... 28 1.4.4 Population size ...... 43 1.4.5 Evolution in marine stickleback ...... 44 1.5 The way forward ...... 45

STICKLEBACK COLLECTION ...... 50 2.1 Sampling effort ...... 50 2.2 The marine environment ...... 54

EXPLORING JORDAN’S RULE IN PACIFIC THREESPINE STICKLEBACK ...... 62 3.1 Introduction ...... 62 3.2 Materials and Methods ...... 64 3.2.1 Collection and sex determination ...... 64 3.2.2 Sample preparation and phenotypic measurements ...... 66 3.2.3 Statistical analysis ...... 66 3.3 Results ...... 68 3.3.1 Sex ...... 68 3.3.2 Vertebral number ...... 69 3.3.3 Standard length ...... 73 3.3.4 Sex-specific differences ...... 73 3.3.5 Fin rays and basals ...... 78

xii 3.4 Discussion ...... 78 3.4.1 Jordan’s Rule ...... 78 3.4.2 Bergmann’s Rule and pleomerism ...... 79 3.4.3 Sexual dimorphism ...... 80 3.4.4 Plasticity and meristic traits ...... 80 3.4.5 Causes of Jordan’s Rule ...... 81 3.5 Future directions ...... 83

CONTEMPORARY ANCESTOR? ADAPTIVE DIVERGENCE FROM STANDING GENETIC VARIATION IN PACIFIC MARINE THREESPINE STICKLEBACK ...... 84 4.1 Introduction ...... 84 4.2 Materials and Methods ...... 87 4.2.1 Sex identification ...... 87 4.2.2 Library preparation and analysis ...... 87 4.2.3 Population genetic structure ...... 89 4.2.4 Effective population size ...... 90 4.2.5 Platedness ...... 90 4.2.6 Morphometrics ...... 91 4.2.7 Selection on phenotypic variation ...... 93 4.2.8 Marine-marine genetic divergence ...... 94 4.2.9 Causes of population structure ...... 95 4.2.10 Marine-freshwater genetic divergence ...... 95 4.3 Results ...... 95 4.3.1 Sequencing results ...... 95 4.3.2 Standing genetic variation ...... 96 4.3.3 Population genetic structure ...... 96 4.3.4 Effective population size ...... 103 4.3.5 Platedness ...... 103 4.3.6 Morphometrics ...... 106 4.3.7 PST-FST and FSTQ-FST comparisons ...... 109 4.3.8 Outlier analysis for marine-marine comparisons ...... 109 4.3.9 The effect of outliers on population structure ...... 114 4.3.10 Outlier analysis for marine-freshwater comparisons ...... 115 4.4 Discussion ...... 119 4.4.1 Marine stickleback exhibit between-population genetic variation ...... 119 4.4.2 Marine stickleback vary in their effective population sizes ...... 121 4.4.3 Marine stickleback exhibit between-population phenotypic variation ...... 121 4.4.4 Causes of a single northern genetic cluster – speculation ...... 123 4.4.5 Contemporary ancestors? ...... 124

HETEROZYGOSITY AND ASYMMETRY: ECTODYSPLASIN AS A FORM OF GENETIC STRESS IN MARINE THREESPINE STICKLEBACK ...... 127 5.1 Introduction ...... 127 5.2 Materials and Methods ...... 130 5.2.1 Sequencing ...... 130

xiii 5.2.2 Plate counts and asymmetry ...... 132 5.2.3 Asymmetry and heterozygosity ...... 135 5.3 Results ...... 136 5.3.1 Genotype, plate morph, and plate counts ...... 136 5.3.2 Structural and non-structural plates ...... 140 5.3.3 Plate position asymmetry ...... 140 5.3.4 Asymmetry and heterozygosity ...... 148 5.4 Discussion ...... 151 5.4.1 Summary ...... 151 5.4.2 Genotype-phenotype mismatch ...... 152 5.4.3 Plate count, asymmetry and selection ...... 152 5.4.4 Asymmetry and heterozygosity ...... 153 5.4.5 Conclusions ...... 155

THE ENVIRONMENT CHANGES, BUT PLASTICITY REMAINS THE SAME: MITOCHONDRIAL BIOGENESIS AND ITS ROLE IN ADAPTIVE DIVERGENCE ...... 157 6.1 Introduction ...... 157 6.2 Materials and Methods ...... 160 6.2.1 Sampling and experimental design ...... 160 6.2.2 Tissue fixation and imaging – cardiac muscle ...... 162 6.2.3 Tissue fixation and imaging – pectoral muscle ...... 164 6.2.4 Statistics ...... 164 6.3 Results ...... 166 6.4 Discussion ...... 168 6.4.1 Mitochondrial biogenesis as an adaptation to the cold ...... 168 6.4.2 Phenotypic plasticity and adaptive divergence ...... 172

CONCLUSION ...... 175

REFERENCES ...... 181

APPENDIX A: EXPLORING ADEGENET ...... 236 A.1. Introduction ...... 236 A.2. Methods ...... 238 A.3. Results and discussion ...... 238 A.3.1. Effect of retaining Principal Components for Discriminant Function Analysis...... 238 A.3.2. The effect of local optima ...... 239 A.3.3. The effect of filtering ...... 239 A.3.4. The effect of removing missing values from the dataset ...... 239 A.3.5. The effect of including freshwater stickleback ...... 245 A.3.6. Summary ...... 245

APPENDIX B: SUPPLEMENTAL TABLES AND FIGURES TO CHAPTER FOUR247

xiv APPENDIX C: RELATIONAL ASYMMETRY (SUPPLEMENT TO CHAPTER FIVE) ...... 280 C.1. Summary ...... 280

APPENDIX D: PROTOCOLS...... 288 D.1. Sex-specific marker: idh PCR protocol ...... 288 D.2. Plate morphs: Stn382 PCR protocol for Eda ...... 289 D.3. SimRAD ...... 290 D.4. Digestion of DNA by restriction enzymes for GBS ...... 291 D.5. Ligation for GBS ...... 292 D.6. PCR amplification of ligated product for GBS ...... 293

APPENDIX E: PERMISSIONS ...... 294

xv List of Tables

Table 1-1 - Proportions of fully plated (FPK), partially plated (PPK + PPNK) and low-plated (LPK + LPNK) morphs of threespine stickleback sampled at polymorphic marine locations across the northern hemisphere...... 22

Table 1-2 - Summary of genetic or genomic analyses of marine threespine stickleback . 29

Table 1-3 - Mean phenotypic measurements (mm) for threespine stickleback from seven populations, and measurement of two fossil specimens ...... 46

Table 2-1 - Collection sites and sampling characteristics ...... 51

Table 3-1 - Average meristic phenotypes and standard length (+ SD) of marine threespine stickleback collected along the Pacific coast from central California to Alaska ...... 71

Table 3-2 - Distributions of total, abdominal, and caudal vertebral numbers for threespine stickleback sampled from each of seven populations ...... 72

Table 3-3 - Spearman correlations for pairs of meristic traits and standard length ...... 74

Table 4-1 - Numbers of stickleback used for plate counts, 3-D morphometrics and included in the sequencing run after process_radtags filtering ...... 88

Table 4-2 - Population genetic statistics for the filtered dataset of marine stickleback for (top) all variant loci and (bottom) all sequenced loci ...... 97

Table 4-3 - Pairwise geographic distances (km: above diagonal) and global pairwise Weir and Cockerham FST (below diagonal) ...... 99

Table 4-4 - Effective population sizes of marine threespine stickleback at sampled sites ...... 104

Table 4-5 - Observed correlations and probabilities expected from sampling error for the null hypothesis (p) for Mantel tests between geographic distance, neutral genetic distance (FST), phenotypic distance (PST – for plates or the first four Principal Components (PCs) of body shape), or genetic distance at Eda (FSTQ) .... 110

Table 4-6 - Genic loci flagged as FST outliers in the marine environment ...... 111

Table 4-7 - Genic loci flagged as FST outliers in at least one marine-freshwater comparison ...... 116

Table 5-1 - The numbers of males and females for which plate counts, Eda genotypes, and standardized multi-locus heterozygosity measures were assessed ...... 131

xvi Table 5-2 - Average numbers of plates, and the number of fish with each possible genotype-phenotype combination per population ...... 137

Table 5-3 - Characteristics and sampling effort per Eda genotype ...... 138

Table 5-4 - Average number of asymmetric myomeres per individual (PPOSA), loci sequenced per individual, and average standardized multi-locus heterozygosity (sMLH) per population ...... 141

Table 5-5 - Results of two-way ANOVA ...... 142

Table 5-6 - Results of Welch’s t-tests for significant deviations from means of 0 for signed positional plate asymmetry (PPOSA) ...... 143

Table 6-1 - Variation in average mitochondrial volume density (Vv) for cardiac and pectoral muscle under different temperature treatments for different ecotypes (marine and freshwater (FW)) of threespine stickleback ...... 161

Table 6-2 - Results of Analysis of Variance comparing mitochondrial volume density in cardiac and pectoral tissue between ecotypes (marine or freshwater) and temperatures (6°C or 21°C) ...... 167

xvii List of Figures and Illustrations

Figure 1-1 - Potential consequences of SGV on adaptive divergence ...... 3

Figure 1-2 - The signal of SGV vs. de novo mutation across a chromosome...... 6

Figure 1-3 - The typical representation of stickleback evolution, showing a diversity of freshwater forms derived from a single ancestral fully-plated marine form ...... 10

Figure 1-4 - The probability of fixation for a de novo mutation (blue), or for standing genetic variants present in the population at frequencies of 2% (red) or 19% or higher (grey, orange), for a range of selection strengths (0 to 0.1) in a population of 10 000 diploid individuals (e.g. 2Ns = 2000 for s of 0.1)...... 14

Figure 1-6 - Approximate distributions of different marine clades determined from mitochondrial DNA ...... 39

Figure 1-7 - PCA of ten phenotypic traits in extant marine stickleback from seven localities and two fossil specimens from California ...... 47

Figure 2-1 - Locations from which threespine stickleback were collected, including eight marine sites (triangles) and one freshwater site (circle) ...... 53

Figure 2-2 - Elkhorn Slough, Monterey Bay, California, U.S.A...... 55

Figure 2-3 - Doran Park, Bodega Bay, California, U.S.A...... 56

Figure 2-4 - Arcata Marsh, Arcata Bay, California, U.S.A...... 58

Figure 2-5 - South Slough, Coos Bay, Oregon, U.S.A...... 59

Figure 2-6 - Tillamook Bay, Oregon, U.S.A...... 60

Figure 2-7 - (Top left) Little Clam Bay, Washington, U.S.A. at low tide. (Top right) Crescent gunnel from seining effort in Little Clam Bay. (Bottom left) Bamfield Marine Sciences Centre, Vancouver Island, Canada. (Bottom right) Cleaning minnow traps and seines at the closest gas station was a common event...... 61

Figure 3-1 - Variation in average sea surface temperature (C), 2010-2014, across the study region ...... 65

Figure 3-2 - Radiograph of a marine threespine stickleback ...... 67

xviii Figure 3-3 - Mean ( SE) of variation among sampling sites, plotted against latitude, for A) total vertebral number, B) number of abdominal vertebrae, C) number of caudal vertebrae, and D) standard length (mm) ...... 70

Figure 3-4 - Relation between the numbers of abdominal and caudal vertebrae ( SE) among populations and sexes ...... 75

Figure 3-5 - Relation between standard length (mm) and total vertebral number ( SE) among populations and sexes ...... 76

Figure 3-6 - Mean (+ SE) of variation among sampling sites, plotted against latitude, for A) number of average pectoral-fin rays, b) number of dorsal-fin rays, C) number of anal-fin rays, D) number of extra dorsal-fin basals, E) number of extra anal-fin basals ...... 77

Figure 4-1 - Positions of 55 landmarks used for the morphometric analysis ...... 92

Figure 4-2 - Adegenet-identified clusters for k = 5...... 100

Figure 4-3 - Result of the phylogenetic analysis using SNPhylo ...... 102

Figure 4-4 - Frequency distributions of Eda genotypes using the Stn382 marker for each sampling site ...... 105

Figure 4-5 – Among-group variation from a 3D morphometric analysis ...... 107

Figure 4-6 - Wireframes of stickleback ...... 108

Figure 4-7 - Expected neutral distributions of PST-FST and observed PST-FST ...... 112

Figure 4-8 - Manhattan plot showing 5% FST outliers for eight marine populations ..... 113

Figure 4-9- Per-locus minimum and maximum FST for eight marine-freshwater pairwise comparisons...... 118

Figure 5-1 - Bony elements of the heads of two marine threespine stickleback ...... 133

Figure 5-2 - Distributions of average plate number ...... 139

Figure 5-3 - Distributions of plate number for each Eda genotype of fully-plated stickleback...... 144

Figure 5-4 - Distributions of signed positional plate asymmetry (PPOSA) ...... 145

Figure 5-5 - Frequencies of stickleback that exhibited symmetrical presence (green), symmetrical absence (red) or asymmetrical absence (blue) of plates per plate position ...... 147

xix Figure 5-6 - Mean (+ SE) for (A) observed heterozygosity and (B) standardized multi-locus heterozygosity (sMLH) ...... 149

Figure 5-7 - Relations of standardized multilocus heterozygosity (sMLH) to the number of myomeres with asymmetrical loss of a plate per fish...... 150

Figure 6-1 - Sample micrographs ...... 163

Figure 6-2 - Micrographs of threespine stickleback pectoral tissue at different magnifications ...... 165

Figure 6-3 - Boxplots of mitochondrial volume density for marine and freshwater threespine stickleback ...... 169

Figure 6-4 - The relation between pectoral and cardiac mitochondrial volume density. 170

xx List of Symbols, Abbreviations and Nomenclature

Symbol Definition Adj. adjusted AFLP amplified fragment length polymorphism AIC Akaike information criterion AK Alaska AK01 Swikshak Lagoon, AK allo allozyme AMOVA Analysis of Molecular Variance ANOVA Analysis of Variance AP ascending process ATP adenosine triphosphate AUP Animal Use Protocol Avg. average BC British Columbia BC01 Bamfield Inlet, BC BIC Bayesian information criterion bp base pair BSA bovine serum albumin C Stn382 allele associated with fully-plated phenotype CA California CA01 Elkhorn Slough, Monterey Bay, CA CA02 Doran Park, Bodega Bay, CA CA03 Arcata Marsh, Arcata Bay, CA CC homozygous genotype for the Stn382 C allele CCAC Canadian Council on Animal Care CGV cryptic genetic variation CI confidence interval CL heterozygous genotype for the Stn382 alleles cog5 Component of oligomeric golgi complex 5 CSNK1G2 Casein kinase 1 gamma 2 CTmin critical thermal minimum/minima CV Canonical Variate CVA Canonical Variate Analysis DAPC Discriminant Analysis of Principal Components DBP dorsal basal plate DF discriminant function °C degrees Celsius Den Denmark d.f. degrees of freedom DFA Discriminant Function Analysis dkk2 Dickkopf WNT signaling pathway inhibitor 2 DNA deoxyribonucleic acid

xxi dNTP deoxynucleoside triphosphate dsDNA double-stranded DNA Eda Ectodysplasin ENA Euro-North American clade F female F fixation index FA fluctuating asymmetry FIS inbreeding coefficient FPK fully-plated with keel morph FST fixation index FSTQ FST for a Quantitative Trait Locus GBS Genotype-by-Sequencing geno genotype Germ Germany GLM generalized linear model GO gene ontology gpc5a Glypican 5a Green Greenland haplo haplotype HE expected homozygosity HetE expected heterozygosity HetO observed heterozygosity h hour h2 narrow-sense heritability HO observed homozygosity hp horsepower HSD Honest Significant Difference HWE Hardy-Weinberg Equilibrium IBD Isolation-by-Distance Ice Iceland idh isocitrate dehydrogenase indel insertion/deletion k number of genetic clusters K keel km kilometre KOH potassium hydroxide kV kilovolt L Stn382 allele associated with low-plated phenotype L left lat latitude LD linkage disequilibrium LG linkage group LKT total number of keel plates on left side LL homozygous genotype for the Stn382 L allele LO Lompoc

xxii long longitude LP1, LP2, etc. plate position 1, 2, etc. on left side LPK low-plated with keel morph LPNK low-plated without keel morph LT total number of plates on left side lzic leucine zipper and CTNNBIP1 domain containing M male M molar m metre maf minor allele frequency max maximum Med Mediterranean Sea mg milligram MgCl2 magnesium chloride μA microampere μCT micro-computed tomography μl microlitre μm micrometre μM micromolar microsat microsatellite mid-Pac mid-Pacific min minimum min minute ml millilitre MLH multi-locus heterozygosity mm millimetre mM millimolar morph morphotype MS-222 tricaine methanesulfonate mt mitochondria(l) N or n number N north NaCl sodium chloride Ne effective population size Neth Netherlands ng nanogram NGS Next-Generation Sequencing nM nanomolar NOAA National Oceanic and Atmospheric Administration Nor Norway Nor Sea Norwegian Sea NRF-1 nuclear respiratory factor-1 NS Nova Scotia NY New York

xxiii OR Oregon OR01 South Slough, Coos Bay, OR OR02 Tillamook Bay, OR P Average frequency of the major allele P1, P2, etc. plate 1, plate 2, etc. PC Principal Component PCA Principal Component Analysis PCR Polymerase Chain Reaction π nucleotide diversity pM picomolar PMP plasticity-mediated persistence PNUMA asymmetry in the total number of plates POLRMT mitochondrial DNA-directed RNA polymerase poly polymorphism PPARAa peroxisome proliferator-activated receptor alpha a PPK partially-plated with keel morph PPNK partially-plated with no keel morph PPOSA positional plate asymmetry ppt parts per thousand PST Phenotypic equivalent of FST PV Palo Verdes QC Quebec QTL Quantitative Trait Locus/Loci R right RAD restriction site associated DNA reps1 RALBP1 associated Eps domain containing 1 RFLP restriction fragment length polymorphism RKT total number of keel plates on right side RP1, RP2, etc. plate position 1, 2, etc. on right side RT total number of plates on right side S100P S100 calcium binding protein P Scot Scotland SD standard deviation SDN selection on de novo mutation sec second seq sequencing SGV standing genetic variation 2 σ B between-population variance component 2 σ W within-population variance component Skag Skagerrak SL standard length sMLH standardized multi-locus heterozygosity SNP single nucleotide polymorphism SS sum of squares SSV selection on standing genetic variation

xxiv TEM transmission electron microscope temp temperature TFB1M mitochondrial transcription factor B1 TFB2M mitochondrial transcription factor B2 TNP Trans-North Pacific clade TUB Tubby bipartite transcription factor UK United Kingdom U.S.A. United States of America vs. versus Vv mitochondrial volume density WA Washington WA01 Little Clam Bay, Puget Sound, WA WNT Wingless-related integration site x times (magnification)

xxv Epigraph

We are a schizophrenic group, believing on the one hand that phenotypes are the real stuff of evolution – the observable qualities that we seek to explain – and yet ultimately treating selection and evolution as something else – a change in gene frequencies – whose study appears always to be obscured by the flexible nature of the phenotype.

Mary Jane West-Eberhard, Developmental plasticity and evolution (2003), p. 18.

It would be premature on the basis of just two fossil records to conclude that marine G. aculeatus has been morphologically static since the Miocene, but geographical homogeneity of extant marine G. aculeatus supports this conclusion.

Micheal A. Bell, The evolutionary biology of threespine stickleback (1994), p. 446.

What, if anything, is wrong with typological thinking?

Tim Lewens, Philosophy of Science (2009), p. 355.

xxvi

Standing genetic variation, contemporary ancestors, and marine threespine stickleback

1.1 Introduction

Contemporary populations may be situated anywhere along a speciation continuum from fully panmictic to reproductively isolated (Hendry 2009; Nosil et al. 2009). The probability that populations evolve into distinct species depends on a number of factors, including the likelihood that different beneficial variants arise in each population and have sufficient effect size to overcome the effects of drift, a probability which itself may be low in complex organisms (Orr 2005). Such de novo mutation has traditionally been assumed in genetic models of adaptation. There are however alternative possibilities – plasticity could move the population to its phenotypic optimum without genetic change (Baldwin 1902; Morris 2014); hybridization could introduce beneficial alleles (Anderson and Stebbins 1954); or variation could be present in the ancestral population at low to moderate frequencies that is adaptive in the new environment (Colosimo et al. 2005). In this chapter I review the latter possibility, standing genetic variation (SGV), and its role in adaptive divergence. I then introduce threespine stickleback as a model species for studying SGV, and specifically discuss their utility from a comparative perspective: the marine ecotype is ancestral to the freshwater ecotype, and is still alive today. This has permitted researchers to use “ancestral-derived” studies to infer the direction of genetic and phenotypic evolution in freshwater populations. However, I will argue that the utility of marine stickleback as contemporary proxies of the ancestral form is spurious if the “contemporary ancestor” is not well characterized. I then review what is known about marine stickleback, and conclude with how this thesis will extend our knowledge of marine stickleback and the role of SGV in adaptive divergence.

1.2 Standing genetic variation

1.2.1 Definitions

SGV encompasses all segregating variants within a population (Orr 2005) – that is, SGV is defined as variants that have escaped initial loss by drift, but have not reached fixation. Generally, the origin and fitness consequences of such variants are not included in the definition (Barrett and Schluter 2008). Thus SGV could represent mutations that arose de novo within the population, as well as variants introduced through gene flow from other populations, including hybridization (Anderson and Stebbins 1954) and horizontal gene transfer (Freeman 1951). Introgressed alleles are largely considered a special case (Hedrick 2013) and will not be discussed here. From a fitness perspective in a stable environment, SGV includes beneficial alleles on their way to fixation by selection (but see Dittmar et al. 2016, who explicitly define SGV as not having beneficial fitness effects), weakly maladaptive alleles that have yet to be purged from the population, neutral mutations fluctuating by drift that are of no adaptive significance, or even alleles that were once adaptive and now are not – and hence had a high frequency before being selected against. Cryptic genetic variation (CGV) is a subset of SGV that has phenotypic and fitness consequences only after environmental change (Gibson and Dworkin 2004; Paaby and Rockman 2014). Although SGV, including CGV, may originate from de novo mutation, the two are treated differently in evolutionary models of adaptation. SGV exists in the ancestral population living in an ancestral environment – and may or may not be adaptive in this environment. When the environment changes, or if the ancestral population, in whole or in part, colonises a new environment, SGV is immediately available for selection to adapt this derived population to its derived environment. De novo mutation, in contrast, arises after environmental change or colonisation. It thus was not available in the ancestral population (Figure 1-1). These distinctions reflect their usage in the published literature (Barrett and Schluter 2008). This definition of SGV lacks some subtlety. For instance, it is largely assumed that an allele with a copy number of one does not constitute SGV because it may still be susceptible to loss by drift, but no threshold above which a minor allele becomes SGV

Figure 1-1 - Potential consequences of SGV on adaptive divergence. Several populations (left) vary in their SGV (different alleles represented by blue, orange, or black). A subset of each population colonises a lake (middle) and subsequently evolves (right). Populations A, B, D, and E undergo parallel evolution, but for populations A, B, and D this is due to selection on SGV, whereas for population E it involves de novo mutation. Population D has higher SGV and so the beneficial allele reaches fixation during the period involved, unlike A and B. Populations C, E and F have low SGV that, due to chance, is not present among colonisers. C goes extinct due to a lack of beneficial variants, including those generated by de novo mutation. F adapts with a different de novo mutation. Populations A, B, and D would show genomic evidence for a soft sweep, whereas population E would show evidence for a hard sweep.

has been provided (perhaps 1/2N?) – and methodological constraints usually require filtering out possibly true low-frequency alleles to avoid including sequencing error. I furthermore disagree with the definition of SGV as presented in one fundamental way, which I will discuss in the “Assumptions of SGV” section below. For simplicity, SGV will be used throughout the paper as defined above. Note that SGV includes point mutations (Single Nucleotide Polymorphisms – SNPs), insertions/deletions (indels) and larger-scale chromosomal changes (Feulner et al. 2013). Because adaptation can occur using two types of variants – SGV or de novo mutations – authors have differentiated between two models of selection: selection on SGV (SSV) and selection on de novo mutations (SDN) (Peter et al. 2012). With SSV, phenotypic selection only begins after the allele exists in the population, allowing immediate genetic response. With SDN, phenotypic selection occurs prior to the existence of the allele, but no genetic response is possible until the variant arises. As will be shown, these two models of selection produce different predictions about the pace and possibility of adaptation to new environments.

1.2.2 Predictions about SGV during adaptation

SGV affects adaptation in several distinct ways (Barrett and Schluter 2008; Rockman 2012; Hedrick 2013; Dittmar et al. 2016). First, SSV begins immediately upon environmental change, whereas SDN requires time for the de novo mutation to rise in frequency rather than be lost by drift. Second, a de novo mutation is initially present as only a single copy within a population. It may still approach fixation if the effective population size and/or its beneficial effect are high (Hermisson and Pennings 2005), but could readily be lost through genetic drift, particularly if recessive (Orr and Betancourt 2001; Barrett and Schluter 2008). SGV, on the other hand, is less likely to be lost by drift. SGV thus permits the beneficial allele to approach fixation sooner than it would if it arose de novo. Third, if SGV has been present for many generations, the alleles will have been tested against other genetic backgrounds, and have likely persisted because they are “compatible” with alleles at other loci. Therefore SGV is less likely to result in negative epistatic interactions or negative pleiotropy than de novo mutation. Fourth, if SGV has 4

been present for many generations, it may have been tested against other environmental backgrounds. This is particularly important if the population has experienced the derived environment at different times during its evolutionary history – the beneficial variant has therefore been pre-tested for the new environment. Collectively, these possibilities suggest that adaptation will occur more rapidly under SGV. Indeed, models suggest that SGV will be more likely than de novo mutation to result in evolutionary rescue when the frequency of SGV is moderately high, and will rescue a population at larger size and return it to its original state faster than de novo mutation (Orr and Unckless 2014). Furthermore, SGV will result in populations moving through larger distances in phenotypic space, thereby keeping pace with rapidly changing environments (Matuszewski et al. 2015). SGV tends to produce a signature on the genome that is distinct from that found under de novo mutation (Figure 1-2), and this difference could promote future evolvability. Specifically, when a de novo mutation appears in an individual, it is physically linked to the other alleles near it. If the mutation is beneficial, selection promoting its increase in frequency towards fixation will tend to increase the frequencies of alleles at neighbouring loci. If loci around the de novo mutation were segregating in the population, this hitchhiking will result in a region of neutral loci with low polymorphism that are highly divergent from the ancestral state. However, recombination during this process will tend to break linkage between loci further from the mutation, causing a gradual decline in divergence and an increase in polymorphism with increasing distance from the de novo mutation. In contrast, SGV would produce such a strong genomic signature only if the variant was recent and/or at low initial frequency. Otherwise, recombination in the ancestral population would cause the conditionally beneficial allele to be located next to most segregating alleles at neighbouring loci. When SSV begins to increase the frequency of the beneficial allele, allele frequency at surrounding loci may show little to no change, although rare variants may be lost (Przeworski et al. 2005). These contrasting scenarios have been referred to as hard and soft sweeps, respectively (Hermisson and Pennings 2005). Soft sweeps maintain genetic variation in the population, providing SGV that may be adaptive during future

Figure 1-2 - The signal of SGV vs. de novo mutation across a chromosome. “Position” refers to a linear sequence of nucleotides across a chromosome. A variant under selection (located on the chromosome at the point of the “V”) shows different patterns of polymorphism with neighbouring loci depending on the source of the variant. Red line – a de novo mutation arises in a single individual and confers a fitness advantage. This mutation is physically linked to the alleles present in that initial individual. As the mutation is favoured, so are the surrounding variants. Once the de novo mutation is fixed, those alleles immediately next to it are also fixed. But as the beneficial mutation rises to fixation, recombination associates the beneficial allele with the other segregating alleles at neighbouring loci – and since recombination is more likely between widely separated loci, polymorphism increases gradually with genetic distance on either side of the beneficial de novo variant. Green line – Similar pattern for standing genetic variation. As this variant had more time to recombine in the ancestral population, loci near the variant are more polymorphic than in the de novo situation. Blue line – For polygenic traits with SGV at numerous loci, the beneficial SGV may never reach fixation. From Barrett and Schluter (2008): used with permission from Elsevier (license number 4077700386030).

environmental change. SGV could also be maintained if the phenotype is affected by many loci of small effect, as small shuffles in multiple allele frequencies could produce the same phenotypic effect as fixation of one large-effect allele, resulting in no loss of SGV (Rajon and Plotkin 2013). SGV is expected be more important for outbred populations (where SGV is present) with relatively long generation times (mutations occur less frequently) (Hendry 2013) or for loci with low mutation rates (Peter et al. 2012). These predictions have yet to be formally tested, although experimental evolution has demonstrated a role for SGV in microbes despite their rapid generation time (Burke 2012). The role of SGV during adaptation to new environments has been documented in a number of species, and evidence is expected to accumulate as Next Generation Sequencing “genomicizes” more non-model species (Stapley et al. 2010). Most recently, the adaptation of resistance to a particular malarial strain in humans, which had long been known but for which no evidence of a hard sweep could be found (Przeworski et al. 2005), was shown to have arisen through SGV (McManus et al. 2017), and Lake Victorian cichlid species that differed in sexual colouration had divergent frequencies for a locus implicated in sexual selection that was ancestrally polymorphic (Brawand et al. 2014). Reported consequences of SSV in nature include warfarin resistance in rats (Pelz et al. 2005), plate loss in stickleback (Colosimo et al. 2005), malathion resistance in blowflies (Hartley et al. 2006), flowering time in Arabidopsis (Scarcelli and Kover 2009), drug resistance in HIV (Pennings 2012), rapid evolution in human height (Turchin et al. 2012), and speciation of cave fishes (Bradic et al. 2013).

1.2.3 Assumptions regarding the role of SGV during adaptation

Barrett and Schluter (2008) listed three main ways that SGV could be implicated in adaptation: (1) through genome scans that identify soft sweeps – although under some situations de novo mutations can also lead to soft sweeps and SGV can lead to hard sweeps (Messer and Petrov 2013); (2) demonstration that an allele fixed in a derived population is present in the ancestral population; or (3) with models providing support that the mutation of interest pre-dates the environment in which it is currently adaptive. 7

Not mentioned by them, the specific role of CGV is evident if phenotypic variance in the ancestral population increases under novel environmental conditions, and a subset of this variation is now the derived character (McGuigan et al. 2011; Jarosz and Lindquist 2010). The first two methods compare sequence variation in contemporary descendants of the ancestors and the derived populations. Apart from technological or methodological constraints that could lead to false positives or negatives (e.g. not sampling enough individuals in the ancestral population to find the rare variant that was selected in the derived population), several assumptions must be met for ancestral-derived comparisons to be of value. (1) SGV in the ancestral population is of no value in the derived population if it does not enter the new environment. This is the missing piece of SGV as defined above – the relevant component of SGV from an adaptive perspective is the subset of ancestral variation that encountered the new environment, and not the total SGV present among the ancestors. If an adaptive variant is rare, it may not encounter the new environment, although it may arise subsequently in the derived population as a de novo mutation (Figure 1-1). (2) The ancestral population may have subsequently evolved or received alleles from the derived population (Schluter and Conte 2009), affecting the interpretation of the identity and origin of ancestral SGV. (3) In their review of parallel evolution, Elmer and Meyer (2011) argued that SGV plays a role in parallel evolution when a single large ancestral population repeatedly colonises similar new environments. But what if the ancestral population is structured? Each derived population could have evolved from different pools of SGV, complicating inferences about the role of SGV in adaptive divergence. In this case, the pool of SGV from which the derived population evolved would no longer be obvious. An important first step in testing these assumptions is to characterize variation within the ancestral population. Threespine stickleback are a powerful model for testing these assumptions.

1.3 Threespine stickleback

1.3.1 Standing genetic variation and ancestral-derived comparisons

The threespine stickleback (Gasterosteus aculeatus) species complex has provided some of the most compelling evidence for the role of natural selection in adapting populations to their respective environments. Several aspects of stickleback biology contribute to their utility for evolutionary biology: they can be readily maintained and bred in captivity; they have rapid generation times compared to larger fish species; they have a relatively small genome that is readily sequenced; and they have undergone replicated natural experiments in numerous freshwater lakes across the northern hemisphere (Wootton 1976, 1984; Bell and Foster 1994). Threespine stickleback have a Holarctic distribution. They are ancestrally marine, with contemporary marine and freshwater forms co-occurring across most of their range (Bell and Foster 1994). After the last glaciation, marine fish colonised new lakes and encountered new conditions, such as lower salinity and mineral content, new predators and parasites, and lower winter temperature. These similar novel environments have led to the evolution of similar adaptive phenotypes in parallel (Figure 1-3). SGV has been implicated in some, but not all, of these adaptations. Intuitively, SGV is the most likely explanation for the origin of the genetic variation on which selection acts to evolve populations in parallel (Schluter et al. 2004). For instance, SGV in stickleback has been implicated as the source of phenotypic variation that resulted in the parallel evolution of body size (McGuigan et al. 2011) and adaptive colour vision (Marques et al. 2017), and possibly colouration (Miller et al. 2007), gill raker number (Glazer et al. 2014), and thermal tolerance (Barrett et al. 2011; Morris et al. 2014). SGV also explains numerous patterns in marine-freshwater genomic divergence (Hohenlohe et al. 2010; Jones et al. 2012ab), although the phenotypes underlying this SGV are unknown. However, selection may operate on different sources of genetic variation to result in parallel evolution: selection for the same allele; selection at different loci on the same gene; and selection for different genes that cause similar phenotypic solutions (Gompel and Prud’homme 2009; Elmer and Meyer 2011; Radwan

IMAGE REMOVED DUE TO COPYRIGHT RESTRICTIONS

Figure 1-3 - The typical representation of stickleback evolution, showing a diversity of freshwater forms derived from a single ancestral fully-plated marine form. From Figure 1.2 of Bell and Foster (1994), p. 6.

and Babik 2012; Stern 2013; Peichel and Marques 2017). SGV seems most necessary when selection consistently favours the same allele, because all derived populations are sourced from one ancestral population and thus share the same pool of SGV (Hendry 2013). Nevertheless, the best-known example of parallel evolution in a vertebrate reveals that all of these possibilities may exist for a single phenotype. Marine stickleback are generally (see below) covered laterally in a series of bony plates (Figure 1-3). For several reasons (e.g. Reimchen 2000; Bell et al. 2004; Vamosi 2006; Marchinko and Schluter 2007; Barrett et al. 2008, 2009a; Marchinko 2009; Myhre and Klepaker 2009; Schluter et al. 2010; Smith et al. 2014; Rennison et al. 2015; Greenwood et al. 2016), selection has consistently but not always favoured low-plated phenotypes in freshwater environments. This loss of all but a few anterior plates is due to a mutation in a regulatory element of the gene Ectodysplasin (Eda) (Colosimo et al. 2005; O’Brown et al. 2015). Experimental and observational studies have demonstrated selection on SGV at Eda (Barrett et al. 2008; Kitano et al. 2008). Surprisingly, selection at Eda results in a large block of linked loci, rather than a “soft sweep” (Hohenlohe et al. 2010; Jones et al. 2012ab). SGV vs. de novo mutation can be tested by sequencing the causative SNP itself. Using this approach O’Brown et al. (2015) identified a suite of “freshwater” loci, including a T  G variant in a regulatory element of Eda, in four freshwater stickleback populations from Alaska, British Columbia, and Washington. In addition, this variant was found in 2.3% of individuals from an Alaskan marine population – SGV in the contemporary ancestor suggests it was present in the original colonists of freshwater lakes and then underwent selection. However, a Japanese freshwater population had typically-“marine” alleles at all other loci but still showed the T  G change, suggesting that de novo mutation can produce identical alleles to that found as ancestral SGV. Intriguingly, fully-plated freshwater populations also exist that illustrate other evolutionary pathways to “low-platedness” – smaller, rather than fewer, plates have evolved (Leinonen et al. 2012; Wiig et al. 2016). This suggests that the SGV available for adaptation has differed with each colonisation event, and that there are limitations to the de novo appearance of the low-plated allele. In such situations, selection leads to alternative solutions.

Complete loss of the pelvis has also evolved in parallel and is found in some, but not all, freshwater populations. This parallel loss suggests shared SGV, but its rarity suggests de novo mutation. Indeed, genetic studies have determined that a tissue-specific regulatory region of the Pitx1 gene has been deleted in multiple populations (Chan et al. 2010), but the extent of the deletion differs among populations. At first glance this seems unlikely, but the regulatory region lies within a fragile section of DNA that is prone to such mutations (Chan et al. 2010). Hence, pelvic loss appears to be a good example of parallel evolution through de novo mutation. These studies suggest that the initial colonisers of freshwater lakes and rivers often carried shared SGV for traits that would be adaptive in freshwater conditions. For some traits all known examples evolved from de novo mutation; in others the trait evolved from SGV when present, from de novo mutation when absent, or did not evolve at all. But how could colonisers have different pools of SGV if they all evolved from a ubiquitous panmictic common ancestor? This question gets to the heart of this thesis – marine stickleback are not in fact one “type” but are phenotypically and genetically diverse, and this diversity matters for the evolution of individual freshwater populations.

1.3.2 Contemporary ancestors: assumptions about SGV matter

If marine stickleback exhibit population structure, it will have five major effects on the evolution of freshwater stickleback (Figure 1-1). (1) Different marine populations may harbour different alleles, and some populations may lack the alleles necessary for adaptation to fresh water. Thus the success of colonisation may differ with different pools of SGV. (2) Different alleles in different marine populations may also impact parallel evolution in freshwater populations. Some populations may adapt in parallel through de novo mutation, whereas others evolve in parallel from shared SGV. Some populations may have yet to evolve in parallel because the adaptive variants are still not present in the population – or they may have evolved other solutions to the problem. (3) Different frequencies of the same allele in marine populations may result in different probabilities of fixation of the adaptive allele in different freshwater populations. For example,

consider the probability of fixation for an allele (p) in a population of 10 000 fish with selection ranging from 0 to 0.1, (1-e-2Nsp)/(1-e-2Ns) (Kimura and Ohta 1971 – following an illustration provided by Hedrick 2013, using Eda allele frequencies from Chapter Four of this thesis). Figure 1-4 shows that the fixation probability is low under all selection strengths for de novo mutation, but is essentially 100% even for very weak selection when the allele frequency is 20% or higher. At 2% frequency the allele is still far more likely to approach fixation than for a de novo mutation, but fixation is not guaranteed when selection is weak. (4) Different frequencies of the same allele in marine populations may affect the likelihood that colonisers possessed that allele. The higher the frequency of a standing allele in the marine population, the more likely it will enter the freshwater environment. (5) Population structuring in marine stickleback could allow local adaptation. Variants will therefore be adaptive in some marine habitats and maladaptive in others. Given some migration, maladaptive alleles could persist in marine populations through gene flow. Thus colonists of fresh water may contain SGV through transport of marine alleles from one marine population to another. Studies of adaptation in freshwater stickleback reveal both parallelism and diversity among freshwater populations. These results are puzzling if all colonising populations contained the same pool of SGV (and encountered similar environments in different freshwater lakes). Does such diversity reflect actual differences in the pools of SGV in the original colonisers of different freshwater systems? Marine stickleback are considered contemporary proxies of the ancestral state. This tends to be justified in four ways: by downplaying life-history variation in the marine ecotype (Bell 1976; Wootton 1984); using genetic variation from a limited distribution of marine populations to show that marine stickleback have no population genetic structure (Hohenlohe et al. 2010); appealing to the large effective populations of marine stickleback to justify assumptions of panmixia (Bell 1976); and drawing on the fossil record to show that marine stickleback are “evolutionarily static” (Foster et al. 2003) (Figure 1-3). Indeed, I used these justifications in Morris et al. (2014), and they are typical of stickleback research in general. However, during the last decade European researchers have begun to appreciate that Atlantic marine stickleback do not conform to these expectations (e.g. DeFaveri et al.

2013ab; Ravinet et al. 2013a). Furthermore, limited data underlies these assumptions on the Pacific coast, and some observations suggest that marine stickleback from Pacific North America are more diverse than usually appreciated (Catchen et al. 2013a; Morris et al. 2014). In particular, “marine stickleback” is likely a collective of populations with different life-history strategies that exhibit phenotypic and genetic variation. Below, I review current knowledge about marine threespine stickleback, and in the process question assumptions about SGV and its role in adaptive divergence.

1.4 Marine threespine stickleback

1.4.1 Defining “marine”

The literature provides little consensus on how to describe contemporary forms of the ancestral stickleback. Bell (1976) argued that only two life-history modes of stickleback exist: freshwater resident, which would include all lake and stream forms, and anadromous. Anadromous stickleback spend “the greater part of the year in marine waters, entering fresh water in the spring where they remain to spawn through the summer. In the late summer or fall, the young return to the sea” (Bell 1976, p. 212). Since then, a greater diversity of life-history modes have been identified in freshwater populations, such as resident lake populations and stream-lake migrants (e.g. Deagle et al. 2012), or benthic versus limnetic ecotypes (e.g. McPhail 1984), all of which are presumed to have evolved from a phenotypically and genetically uniform ancestral population (Figure 1-3). Similarly, Bell and others have come to recognize both “marine” and “anadromous” forms (Bell and Foster 1994). Some authors also recognize an “estuarine” form as distinct from the marine or anadromous form (Kassen et al. 1995). Confusingly, some authors refer to all three as “marine”, whereas others use marine and anadromous interchangeably, and even argue that all three represent the ancestral condition. This is usually justified through the apparently limited genetic and phenotypic divergence among the three marine “types”, although that justification is becoming increasingly tenuous (1.4.2, 1.4.3). What is certainly not disputed is the fact that stickleback can be found in the marine environment, sometimes in incredible numbers,

and sometimes far from shore (Brown and Cheng 1946; Jones and John 1978; Taylor and McPhail 1986; Quinn and Light 1989; Williams and Delbeek 1989; Quinn and Brodeur 1991; Jurvelius et al. 1996). For example, Cowen et al. (1991) collected juveniles that had hatched in May 110 km off the New York coast in June, a migration that coincided with the movement of large predatory fish into their natal estuary. Remarkably, such migration required that they swim three body lengths per second, without assistance by currents or wind. These extreme distances seem to be atypical; although stickleback are fairly well distributed throughout bays they are assumed to stay close to shore in the open ocean. Bell’s (1976) original definition of anadromy most accurately reflects its usage by wetland ecologists today (Simenstad et al. 2002). For a stickleback to be truly anadromous, its juveniles must overwinter in seawater, not estuarine or brackish water, and must move through the brackish waters to enter fresh water to spawn. This stickleback lifestyle is well documented in numerous coastal rivers and streams around the northern hemisphere. Parasite load is a good indicator of anadromy - one study from England reported that the exclusively marine fifteenspine stickleback contained only marine parasite species, whereas threespine stickleback captured in the same location contained a mixture of marine, freshwater, and euryhaline parasites (Dartnall and Walkey 1979). Field surveys have also detected anadromy, in particular noting hybrid zones between anadromous and freshwater populations when they meet during the breeding season (Hagen 1967; Jones et al. 2006; Raeymaekers et al. 2005). Intriguingly, the degree of reproductive isolation differs between rivers. The overwintering habitat and migratory paths of anadromous stickleback are largely unknown (Taylor and McPhail 1986; Snyder and Dingle 1989, 1990; Kitano et al. 2012). Anadromous stickleback breed in a variety of habitats, from fast-flowing rivers (Snyder and Dingle 1989) to lakes (Karve et al. 2008), tidally-influenced waters (Worgan and FitzGerald 1981) to far upstream (Taylor and McPhail 1986). This variety challenges whether the term “anadromous” is always appropriate. Tidal marsh ecologists have described five different uses of marshlands by coastal fishes, which have not yet been applied to stickleback: (1) Anadromous punctuated migration, whereby, after spawning in

fresh water, juveniles move through estuaries in stages before entering the sea. (2) Overwintering, whereby anadromous species remain in the freshwater tidal reaches during periods of high river winter flow, rather than migrating to the ocean. (3) Tidal/event migration regulated by the tides, floods, storms, or winter melt. (4) Straying, in which marine fishes move into estuaries only if a water mass accidentally takes them there, but they do not breed in the estuary. (5) Marine rearing, whereby marine fishes use estuaries for nursery habitat (Simenstad et al. 2002). The degree to which stickleback exemplify strategy (1) versus the other strategies is unclear, and understanding is hampered by the assumption that riverine stickleback that are fully-plated must be fully anadromous (e.g. Withler and McPhail 1985). To further complicate matters, none of these five situations apply for stickleback in the Mediterranean, which are low-plated, breed and overwinter in shallow temporary freshwater marshes, and migrate into nearby saltwater lagoons (5-30 ppt) during the hottest summer months (Crivelli and Britton 1987). The authors considered these, too, to be anadromous stickleback adapted to particular environmental conditions, and they may be typical for the region (Bertin 1925). At the other extreme are St. Lawrence River anadromous stickleback that enter the Rivière de Vases to breed, which is brackish at the mouth (3-20 ppt, depending on tide) but fresh 1.5 km upstream. The highest density of reproductive stickleback is found in saline water, but 43-59% are found at varying locations upstream. Furthermore, a small percentage (~1%) breed in saltwater ponds where salinity varies little and averages 21.7 ppt (Worgan and FitzGerald 1981). These three different breeding locations, within presumably a single anadromous population of stickleback, defy simple categorization. Marine rearing is well documented in stickleback that have been traditionally described as anadromous (e.g. Bay of Fundy stickleback using saltwater marshes, Williams and Delbeek 1989). Such stickleback migrate (presumably) from the ocean and enter shallow estuaries to breed, but they do not specifically seek out fresh water. Salinity may vary from relatively low (e.g. 5 ppt) to marine (e.g. 30 ppt), depending on the location within the estuary. The particular tradeoffs between marine rearing versus true anadromy are unknown but may have to do with the distribution of protective eelgrass beds (Rybkina et al. 2016), physiological constraints (Poulin and FitzGerald 1989), the

costs of hybridizing with freshwater stickleback (Jones et al. 2006), or some other factor. Breeding in saltwater marshes has been described primarily in the Atlantic (Poulin and FitzGerald 1989; Rowland 1989; Whoriskey and FitzGerald 1994), but likely occurs in areas such as Elkhorn Slough, California, and Tillamook Bay, Oregon, two of the populations studied in this thesis. In Elkhorn Slough, where salinity is marine for most of its reach, adult stickleback are the second-most dominant species in eelgrass beds during the spring, but they leave the slough entirely for the winter (Grant 2009) for unknown environments. In Tillamook Bay, stickleback are relatively abundant in all areas throughout the year, across a salinity range of nearly fresh water to 30 ppt, but are most abundant during December and January (Bottom and Forsberg 1978). During the breeding season (June and July) they concentrate in the southern salt marshes, where salinity averages 9-15 ppt (Bottom and Forsberg 1978), but may get as low as 0.3 ppt during heavy river input (Ellis 2002). Threespine stickleback also reproduce in protected saltwater lagoons, or along the shore of bays or open ocean. Saimoto (1993a) and Saimoto (1993b) studied the life history of a truly marine-breeding stickleback population in the Pacific, from Oyster Lagoon, BC. Oyster Lagoon is a relatively small (2 hectare), shallow (2.5 m) lagoon connected to Bargains Bay by a narrow channel. This channel limits tidal flow into the lagoon, although tidal variation does occur. There is no fresh water input beyond rainfall, and salinity varies from 20-30 ppt year-round. Adults enter the lagoon in February and most remain until September, using eelgrass beds as nurseries. Larger juveniles leave the bay in the fall; the smaller ones overwinter in Oyster Lagoon. Interestingly, this life history varied from that in Salt Lagoon, which is located only one km from Oyster Lagoon. Salt Lagoon is larger, deeper, and warms slower in the spring; adults do not enter it until May, and become reproductive later than Oyster Lagoon adults (Saimoto 1993a). Adults removed from Oyster or Salt Lagoon generally returned to their respective lagoon without straying. Furthermore, juveniles marked in the fall were more likely to return to their natal grounds than to neighbouring lagoons, although overall returns were low (Saimoto 1993b).

In the absence of small saltwater lagoons, stickleback tend to nest in other protected areas. In the Baltic Sea, they tend to nest around harbours (Borg 1985). In the White Sea they favour coves and bays where eelgrass is plentiful (Borg 1985; Demchuk 2015). In contrast, stickleback nest along much of Nova Scotia’s coastal shoreline, in salinities from 3-32 ppt in water less than 1 m deep. They tend to nest on exposed substratum, only rarely building nests in protected areas (Blouw and Hagen 1990). Clearly there is much geographic variation waiting to be explored. The diversity of breeding habitats in marine stickleback indicates that “anadromous” and “marine” do not encompass the full range of possibilities - and even homogeneous populations can exhibit multiple strategies (Worgan and FitzGerald 1981). “Migratory” is a poor descriptor, as not all individuals migrate, and even freshwater stickleback exhibit some migration (Harvey et al. 1997). “Marine” or “oceanic” could be used as a catch-all for stickleback that spend some time in saline waters. “Coastal” could be used to refer to those that do not enter estuaries, but rather breed in saltwater bays, lagoons, or unprotected coastline. “Anadromous” should refer to stickleback that leave the sea to spawn in fresh water. “Marine rearing” could be used to refer to stickleback that move from the marine environment into estuaries to reproduce, whereas “estuarine resident” could refer to those that never leave the estuary. “Pelagic” stickleback are known, but their breeding sites are presumed to be coastal. One possible life history for which there is currently no evidence is the movement of freshwater stickleback into brackish or oceanic waters to reproduce. These descriptors are for life histories only, and may not describe a complete population - one population may contain individuals that adopt marine rearing or anadromous strategies (Worgan and FitzGerald 1981). For the rest of this thesis, “marine” will be used as a catch-all term. “Anadromous” will only be used if that was the descriptor in the paper, and is not offered as evidence of anadromy.

1.4.2 Phenotypic uniformity

Freshwater stickleback undoubtedly arose from colonisation of the freshwater environment by marine stickleback; studies have consistently ruled out the possibility of phenotypically similar, but unconnected freshwater populations sharing a common 19

freshwater ancestor (e.g. Ortí et al. 1994; McKinnon et al. 2004; Raeymaekers et al. 2005; Schluter and Conte 2009). Therefore, inferring the direction of evolution is simply a matter of measuring the difference between the ancestral and derived phenotypes. Unfortunately, the ancestral fish themselves no longer exist, and so contemporary marine forms are used. This is easy in part to justify if marine stickleback are phenotypically uniform across their range. According to Bell (1976), “Anadromous populations would tend to be phenotypically uniform and stable because of the greater likelihood of gene flow among populations (even if they home to a stream), absence of genetic drift in large interconnected populations, and the presence of relatively uniform selection regimes in the ocean” (p. 221). Few studies have examined phenotypic variation in marine populations. Most studies examine one or a few marine populations to allow comparison to freshwater populations. Nevertheless, the limited available data reveal that marine stickleback vary phenotypically with respect to egg size (Baker et al. 2008; Oravec and Reimchen 2012), life span and size at maturity (Baker et al. 2008; Yershov and Sukhotin 2015), timing of migration (Kitano et al. 2012), phenotypic modularity (Jamniczky et al. 2015), and body shape (Black 1977; Leinonen et al. 2006; Spoljaric and Reimchen 2007; Aguirre 2009; Aguirre and Bell 2012). A recent study found surprising morphological variation, including plate count, in multiple marine populations around Ireland (Ravinet et al. 2013a), and concluded that The considerable phenotypic variation observed in Irish marine stickleback populations is surprising, since other researchers have typically reported low body shape diversity in anadromous stickleback…It is unlikely, however, that Irish marine stickleback have uniquely evolved greater phenotypic diversity. A more plausible explanation focuses on how habitats are categorized; Ireland, for instance, has a large number of sheltered marine systems and tidal inlets that act as transitional habitats between freshwater and marine environments… (p. 285)

This is precisely the problem – variation in marine stickleback habitat exists, and may be important both for the future evolvability of marine stickleback, and inferences about the pace and form of adaptation in freshwater environments. The idealized marine “type” must be abandoned and data on the distribution and occurrence of variation in the marine

environment be characterized objectively in order to better understand stickleback evolution. Stickleback are well known for heterogeneity in their armature. Low-plated fish with  10 plates are usually considered typical of lake and stream residents, whereas fully-plated fish with > 30 plates and a well-developed keel are usually typical of marine forms. Partially-plated forms, which have a plateless gap before the keel, also occur. This variation is genetically-based, with partially-plated fish tending to be heterozygous at Eda-linked alleles. Typical assertions from published papers include: “Although marine sticklebacks are completely covered with bony armor plates, most freshwater populations have dramatic reductions in plates” (Knecht et al. 2007, p. 141) and “Marine, most anadromous, and some freshwater populations are strongly armored” (Baker et al. 1995, p. 226). The dogged belief that plate morph is associated with ecology has led to the description of Eda-linked alleles as being “typically marine” or “typically freshwater” (e.g. Deagle et al. 2013, as in “We also found typically marine alleles present in a few freshwater lakes”, p. 1917). This is despite substantial evidence that the Eda allele for full-platedness can occur at high frequencies in fresh water (e.g. Hagen and Gilbertson 1972; Kitano et al. 2008; Taugbøl et al. 2014), and the low-plated allele can occur at high frequencies in the marine environment (Table 1-1). Colosimo et al. (2005) discovered numerous SNPs associated with fully-plated and low-plated phenotypes. Remarkably, a similar suite of SNPs were recovered in most low-plated freshwater populations that they examined, despite these locations being scattered across the northern hemisphere. Colosimo et al. (2005) concluded from this parallel genetic evolution that, Despite the shared ancestry of Eda alleles, it is highly unlikely that an ancestral population of low-plated fish migrated through the ocean to found most other low-plated populations in freshwater lakes and streams around the world. Ocean sticklebacks are virtually always completely plated…The most likely interpretation to account for the global sharing of closely related low-plated alleles at the Eda locus is that the alleles controlling the low-plated phenotype are present at some frequency in marine populations (p. 1929-1930).

They tested this by sampling fully-plated stickleback from two river systems along the Pacific coast. From 0.2% to 3.8% of individuals were heterozygous for the low-plated

Table 1-1 - Proportions of fully plated (FPK), partially plated (PPK + PPNK) and low- plated (LPK + LPNK) morphs of threespine stickleback sampled at polymorphic marine locations across the northern hemisphere.

Location % % PPK + % LPK + N Reference FPK PPNK LPNK Swina, Baltic Sea 94.1 5.2 0.7 136 Bańbura 1994 Wapnica, Baltic Sea 98.5 1.5 0 264 Bańbura 1994 Dziwnow, Baltic Sea 97.4 2.6 0 311 Bańbura 1994 Mrzezyno, Baltic Sea 96.2 3.8 0 26 Bańbura 1994 Kolobrzeg, Baltic Sea 97.0 3.0 0 33 Bańbura 1994 Wladyslawowo 1, 97.1 2.9 0 309 Bańbura 1994 Baltic Sea Wladyslawowo 2, 48.6 46.4 5.0 280 Bańbura 1994 Baltic Sea Kuznica, Baltic Sea 70.9 29.1 0 206 Bańbura 1994 Hel, Baltic Sea 72.7 22.7 4.6 22 Bańbura 1994 Gdynia, Baltic Sea 66.7 33.3 0 114 Bańbura 1994 Mikoszewo, Baltic Sea 94.9 5.1 0 39 Bańbura 1994 Båsvik estuary, 78.6 17.9 3.6 Klepaker 1996 Norway Austevoll estuary, 80.2 11.1 8.6 Klepaker 1996 Norway Gjølanger estuary, 51.2 9.8 39.0 Klepaker 1996 Norway Dalevatn estuary, 60 34 6 Klepaker 1996 Norway Botnegård estuary, 34 34 32 Klepaker 1996 Norway Hestvika estuary, 85.4 12.5 2.1 Klepaker 1996

Norway Grunvåg estuary, 72.2 22.2 5.6 Klepaker 1996 Norway Kåfjordbotn, Norway 64 20 16 Klepaker 1996 Smalfjorden, Norway 52.9 29.4 17.7 Klepaker 1996 Guttormsvauen, 83 14.9 2.1 Klepaker 1996 Norway Viksfjord, Norway 68 28 7 Klepaker 1996 Grimstadpollen, 92 6 2 Klepaker 1996 Norway Nordåsvatn, Norway 90 10 0 Klepaker 1996 Vågsbøpollen, Norway 82 16 2 Klepaker 1996 Hardbakke, Norway 86 8 6 Klepaker 1996 Selvågane, Norway 92.9 3.6 3.6 Klepaker 1996 Svanevika, Norway 84 6 10 Klepaker 1996 Vefsnfjord, Norway 54.5 6.8 29.6 Klepaker 1996 Botn, Norway 40.9 22.7 36.4 Klepaker 1996 Bugovik, Norway 98 2 0 Klepaker 1996 Presteid, Norway 76 24 0 Klepaker 1996 Skogvollbukta, 44 32 24 Klepaker 1996 Norway Gisundet, Norway 92 8 0 Klepaker 1996 65°02.7’ 22°27.5’, 51.1 44.5 4.4 45 Lucek et al. 2012 Iceland Doel, Belgium 34 56 10 Raeymaekers et al. 2005 Oudenburg, Belgium 38.8 36.7 24.5 Raeymaekers et al. 2005 Yerseke, Netherlands 10 34 56 Raeymaekers et al.

2005 Mariager Fjord, 80 16 4 Ferchaud and Denmark Hansen 2016 Quilcene, Puget 93 0 7 Kitano et al. 2008 Sound, WA Seabeck, Puget Sound, 90 9 2 Kitano et al. 2008 WA Duwamish, Puget 99 1 0 Kitano et al. 2008 Sound, WA Oyster Lagoon, BC 98.7 1 0.3 Saimoto 1993a Rabbit Slough, AK* 100 0 0 Bell et al. 2010 Oyster Lagoon, BC 99.9 0.01 0 Barrett et al. 2008 Shediac, NB 99 1 0 148 Hagen and Moodie 1981 Oyster Bred Bridge, 99 1 0 105 Hagen and Moodie PE 1981 Covehead Stream, PE 98 2 0 141 Hagen and Moodie 1981 Enmore, PE 99 1 0 230 Hagen and Moodie 1981 Trout, PE (3 locations) 98 2 0 330 Hagen and Moodie 1981 Wheatley, PE 93 7 0 91 Hagen and Moodie 1981 Central Hudson Bay 1 99 0 148 Hagen and Moodie 1981 Central Hudson Bay 0 0 100 2 Hagen and Moodie 1981 Richmond Gulf, QC 0 100 0 3 Hagen and Moodie

1981 Richmond Gulf, QC 11 89 0 35 Hagen and Moodie 1981 Richmond Gulf, 12 88 0 74 Hagen and Moodie Hudson Bay 1981 James Bay 0 100 0 1 Hagen and Moodie 1981 Sylte 53 43 4 1000 Münzing 1963 Elbe estuary 56 40 4 2489 Münzing 1963 Jade Bay 55 32 13 2774 Münzing 1963 Ems estuary 49 47 4 1000 Münzing 1963 Ems 48 48 4 1000 Münzing 1963 Den Helder 38 54 8 1000 Münzing 1963 Flemish Bight 42 49 9 1970 Münzing 1963 Western Baltic 70-90 NP NP Münzing 1963 German Bight 53 NP NP Münzing 1963 Dutch and Belgian 40 NP NP Münzing 1963 coast Western outlet of 20 NP NP Münzing 1963 Channel Comox wharf, BC 93.4 5 1.6 4468 Black 1977 Bonsall Creek, BC 99.8 0.2 0 598 Black 1977 Chase Creek mouth, 80 12.5 7.5 40 Black 1977 BC * included some fully-plated heterozygotes at Eda. NP indicates the data was not provided

allele – a seeming contradiction explained by modifier alleles that restored the fully- plated phenotype. Colosimo et al. proposed that the low-plated allele is “likely maintained by the occasional hybridization that occurs when marine stickleback come into contact with low-plated freshwater populations in coastal streams during the breeding season” (p. 1932) based on their biased sampling of only fully-plated riverine stickleback, and their ardent belief that all marine stickleback are anadromous. Another explanation is possible. The low-plated phenotype has been found, sometimes at appreciable frequencies, in European marine stickleback populations (Table 1-1). Klepaker (1996) sampled 27 marine and estuarine populations from Norway. In contrast to the assumption that marine stickleback are nearly or wholly monomorphic for the fully-plated phenotype, 23 populations were polymorphic, with no association between morph frequency and location (estuarine, inner fjord, outer fjord). Fully-plated morphs varied from 35%-85% in estuarine populations to 41%-100% in fjord populations, and partially-plated (up to 34%), low-plated (up to 37%) and the usually rare low-plated with keel morphs (up to 28%) comprised substantial proportions of some populations. Such polymorphisms were not obviously the result of hybridization with freshwater forms. In fact, low-plated fish were, for other traits, more like marine than freshwater stickleback (Klepaker 1996). In contrast, the high proportion of low-plated fish in brackish creeks of Belgium and the Netherlands is maintained by gene flow from freshwater stickleback, despite selection against the low-plated morph in the estuary (Raeymaekers et al. 2014; Table 1-1). Even in Atlantic and Pacific North America, where researchers are particularly prone to describing marine forms as “fully-plated”, polymorphism in the marine environment is not atypical (Table 1-1). Californian populations exhibit intriguing plate variation. Bell (1979) found a low-plated stickleback population breeding in a saltwater lagoon in California. Multiple monomorphic or nearly-monomorphic low-plated stickleback populations have been reported from coastal rivers south of San Francisco, with increasing frequency of the fully-plated morph to the north (Miller and Hubbs 1969; Hagen and Gilbertson 1972; Baumgartner and Bell 1984). Similarly, clines have been observed within rivers, with fully-plated morphs dominating under high water flow

(Baumgartner and Bell 1984). Unfortunately, sampling locations were not precisely identified, so how many marine as opposed to freshwater resident stickleback were sampled cannot be determined. Baumgartner and Bell (1984) reported morph index results from the lagoons of eight rivers along which they also sampled downstream. In Soquel Creek the lagoon samples had predominately partially or low-plated fish closer to the ocean, with an increase to fully-plated fish further upstream (Baumgartner and Bell 1984). This seems to contradict standard understanding. Furthermore, why the fully- plated allele should be favoured in the marine environment is not understood. The characterization that marine stickleback are fully-plated and likely anadromous has contributed to two hypotheses. The “transporter hypothesis” (Schluter and Conte 2009) proposes that freshwater alleles constantly flow into marine stickleback via hybridization and are made available for future freshwater adaptation. The “cryptic genetic variation hypothesis” proposes that modifier alleles in the marine environment suppress the low-plated allele such that it can persist at low frequencies with little phenotypic consequences (for instance: “EdaL alleles in oceanic populations are usually heterozygous EdaL/EdaC genotypes that express the complete morph”, Bell et al. 2010, p. 190). However, another nonexclusive possibility is also plausible, which I call the “marine transporter” hypothesis. In this scenario, low-plated alleles are favoured under certain marine conditions but not others, being maintained even when selected against via dispersal of low-plated alleles from other marine localities. For instance, with sufficient interchange marine stickleback from Californian lagoons could permit the “freshwater” low-plated allele to persist in Oregon, even without freshwater-marine hybridization. This is not to say that the other two hypotheses are incorrect – clearly there is strong evidence for the transporter hypothesis in Alaska (Hohenlohe et al. 2010). Instead, the marine transporter hypothesis suggests that the low-plated allele persists in marine environments for multiple reasons. Information about the occurrence of Eda alleles in pelagic, coastal, marine rearing, estuarine resident, and truly anadromous populations, and measurements of their connectivity, would help address this possibility. But this is only a question if the concept of a marine “type” is abandoned. Statements such as, “Bony lateral plates on these fish, which are highly variable among freshwater populations, are largely

monomorphic for the completely plated morph throughout their distribution in marine waters of the Atlantic and Pacific oceans” (Reimchen 2000, p. 1082) should be relegated to the history books.

1.4.3 Genetic variation among marine populations

Recognition of phenotypic variation among threespine stickleback ecotypes, particularly in plate morph, has prompted numerous revisions of stickleback classification. For instance, trachurus, semiarmatus, and leiurus are terms that predominate in the early stickleback literature as taxonomically valid categories for the three plate morphs (although semiarmatus, the partially-plated stickleback, were usually considered hybrids of the other two forms) (Wootton 1984). Along the California coast, subspecies were classified based on plate morph, including fully-plated (Gasterosteus aculeatus aculeatus), low or partially plated (G. a. microcephalus), and a rare freshwater unarmoured form (G. a. williamsoni) that is currently recognized as a valid federally protected subspecies (USFWS 2009). As phenotypic variation continued to be studied, heterogeneity in platedness was recognized within freshwater populations, and likely evolved independently and repeatedly from marine ancestors (Bell 1976; Wootton 1976; McKinnon et al. 2004). Consequently, recognition of distinct stickleback subspecies became unsupportable. Gasterosteus aculeatus was considered a “superspecies” (Bell 1976), and only in instances of exceptional variation, such as the benthic and limnetic sympatric species pairs in British Columbian lakes, or the giant stickleback of Haida Gwaii, did the impracticality of naming distinct species come up against different political agendas. For instance, benthic and limnetic sympatric populations have many of the criteria for being truly distinct species (Rundle and Schluter 2004), and in some lakes these species have gone extinct. Whether they are recognized as such depends on the political agency in charge of determining what constitutes a “species” (Figure 1-5). With so much unknown about the numerous independently derived freshwater populations, it is no wonder that little attention has been paid to the genetics of the seemingly more common and less interesting marine populations. Yet some work has been done (Table 1- 2), and has even uncovered the possibility of speciation among marine stickleback.

Table 1-2 - Summary of genetic or genomic analyses of marine threespine stickleback. See List of Symbols, Abbreviations and Nomenclature for meaning of abbreviations.

Ref Mar sites FW sites Mar N Total N Mar N Method N loci Min Max Avg. Mar Mar +

N sites Total or Mar Mar FST FW

sites haplo FST FST FST Withler and BC, WA BC, WA 626 4122 16 56 allo 8 0.046 McPhail 1985 Haglund et Worldwide, but 16 16 allo 18 0.651 al. 1992 ecotype not provided

Ortí et al. Japan, Japan, 15 36 12 26 mtDNA 17 1994 AK, BC, AK, BC, haplo NY, QC, CA, UK, NS, France Scot, Sweden Higuchi and Japan Japan, 25 29 25 29 allo 18 Goto 1996 AK

Taylor and BC BC, 139 493 8 24 mtDNA 42 0.034 McPhail WA, CA RFLP haplo (PhiST) 1999 Johnson and BC AK, BC 229 872* 12 45 mtDNA 2 Taylor 2004 RFLP haplo McKinnon et Japan, Japan, 88 220 4 10 microsat 22 0.085 0.822 0.25 (Nei's 0.804 al. 2004 AK, BC, AK, BC, (Nei's (Nei's D) (Nei's Scot, Scot, D) D) D) Nor Nor, Ice Raeymaekers Belgium, Belgium 199 348 4 7 allo, 13, 5 0.018 0.075 0.044 0.106 et al. 2005 Neth microsat Leinonen et Baltic, Finland 290 559 5 10 microsat 18 al. 2006 Norway Mäkinen et Baltic, Baltic, 460 1274 16 73 microsat 18 0.016 0.09 0.024 0.21 al. 2006 North, North, (coastal), Barents, Barents, 0.036 White, White, (migratory) Atlantic Atlantic, Med, Black

Cano et al. Baltic 257 257 8 8 microsat 18 0 0.0151 0.005 2008 Mäkinen and Europe Europe 33 172 8 49 mtDNA 86 Merilä 2008 and haplo Maine Mäkinen et Baltic, Sweden, 72 168 3 7 microsat, 103, 2 0.06 0.167 al. 2008 North, Finland, indels (0.11 Barents Bosnia neut, 0.40 direct, 0.04 balan) Hohenlohe et AK AK 40 100 2 5 RAD- 45000 0.0076 0.087 al. 2010 seq DeFaveri et Japan, Japan, 144 288 6 12 microsat 50 al. 2011 BC, BC, QC, Maine, Germ, North, Norway, Norway, Finland Barents

Lind and Baltic, 244 244 8 8 AFLP 248 0.01 0.24 0.11 Grahn 2011 Bothnia Shimada et Baltic, Norway, 72 204 3 9 microsat 157 0.034 0.045 0.038 0.104 al. 2011 North, Sweden, Barents Belgium, Germ, Russia Deagle et al. BC, BC 974 82 mtDNA 2 2012 mid-Pac RFLP haplo Jones et al. Japan, Japan, 50 196 9 34 Genome 1159 0.193 2012a AK, BC, AK, BC, seq CA, NS, WA, Iceland, CA, Scot, Iceland, Germ Scot, Norway, Germ Jones et al. Japan, AK, BC, 10 22 10 22 Genome 5.8 2012b AK, BC, WA, seq million CA, NS, CA,

Iceland, Scot, Germ, Nor, Scot Germ Catchen et OR OR 252 578 3 9 RAD- >91 0.003 0.01 0.007 0.092 al. 2013a seq 000 Deagle et al. BC, BC 31 462 11 115 SNP 1170 2013 mid-Pac geno SNPs DeFaveri Baltic 1288 1288 14 14 microsat 20 + 0.006 and Merilä Eda (neut), 2013 0.102 (Eda) DeFaveri et Baltic 1288 1288 38 38 microsat 40 0.008 al. 2013a (nongenic), 0.026 (genic) DeFaveri et Maine, 240 240 10 10 microsat 138 0.033 al. 2013b Barents, (neut), White, 0.097 Nor Sea, (direct), Skag, 0.011

North, (balan) Baltic Drevecky et AK AK 144 480 3 10 microsat 9 0.001 0.017 0.011 0.093 al. 2013 Ravinet et al. Ireland Ireland 149 449 11 38 mtDNA, 9, 3 2013a microsat haplo Ravinet et al. Japan Japan ??? 249 11 19 microsat 10 2014 Roesti et al. BC BC 54 270 2 10 RAD- ??? 0.023 (avg. 0.304 2014 seq, across all Sanger SNPs) seq DeFaveri Baltic, Finland 271 585 4 8 microsat 15 0.004 0.199 and Merilä North (2003), (2003), 2015 0.009 0.209 (2009) (2009) Lescak et al. AK AK ~132 1057 3 20 RAD seq 130 0 0 0 0.051 2015 000 Ferchaud Den Den 40 177 2 9 RAD seq 28888 0.003 0.206 and Hansen

2016 Liu et al. Green, AK, BC, 12 81 3 16 NGS 85698 2016 Den Green, Norway Den, Germ Mazzarella Norway Norway 60 129 3 6 RAD seq 92000 et al. 2016

Figure 1-5 - Numbers of historically extirpated and extinct freshwater fish species reported by the federal government of Canada with successive editions of the General Status of Wildlife in Canada (CESCC 2001, 2002, 2006). The reduction in the reported number of extinct species is largely due to removing protected stickleback populations, such as benthic and limnetic ecotypes, from the species list.

Beginning in the 1960s, genetics techniques were applied to stickleback ecotypes to address basic population genetic questions. Based on allozyme variation, Hagen (1967) first differentiated anadromous from freshwater forms of stickleback in the Little Campbell River, BC. Additional loci were found that differentiated other marine and freshwater ecotypes (e.g. Muramoto et al. 1969; Raunich et al. 1972; Zietara 1989; Rafiński et al. 1989; Taniguchi et al. 1990 - reviewed in Buth and Haglund 1994). The first hint of genetic differentiation among marine stickleback came from an allozyme study of 40 freshwater and 16 anadromous populations in BC and Washington (Withler and McPhail 1985). FST for each allozyme was higher among freshwater populations (avg. 0.293) than marine populations (avg. 0.046). The latter result led Withler and McPhail (1985, p. 532) to characterize the “electrophoretic homogeneity” of the marine stickleback “by virtue of gene flow” (Withler and McPhail 1985). Although FSTs for marine populations were low compared to freshwater populations, they are not 0 and are in sharp contrast to the exceptionally low, but often still biologically relevant, FSTs recently reported using SNPs in marine fish populations (e.g. cod FST = 0.0037, Knutsen et al. 2011). A subsequent study (Taylor and McPhail 1999) revealed that 97% of genetic variation occurred within and 3% among eight marine/anadromous populations, as opposed to 23% within and 77% among 20 freshwater populations. The results from this paper seemingly provided evidence to researchers that there was little of interest going on in the marine stickleback. Their conclusions were, however, based on comparing southern BC marine populations to freshwater populations from BC, Washington, and California. Despite the limited differentiation of these marine populations, the relationships among populations revealed at least two clusters of marine stickleback. Taylor and McPhail (1999) did not remark on this finding, but instead concluded that benthic and limnetic stickleback could not have originated from the double invasion of genetically distinct marine stickleback populations, because this “would require highly structured anadromous/marine populations, but our data suggest this is not so…” (p. 287). Yet SGV clearly differed among marine sites and the sampling was not representative of marine stickleback at large.

Although microgeographic genetic variation was initially ignored, macrogeographic variation among marine stickleback from separate ocean basins was quickly uncovered. Haglund et al. (1992) first suggested that European freshwater stickleback may be derived from an Atlantic marine form long isolated from the Pacific marine form: “Obviously even the anadromous stocks from which they were derived have undergone a long period of potential evolution while isolated” (p. 434). If true, this would mean globally there were distinct marine forms of threespine stickleback, even if regionally they were genetically and phenotypically relatively homogeneous. Allozymes, mitochondrial DNA, microsatellites, and, more recently, genome sequencing have uncovered such large-scale regional differences (e.g. Ortí et al. 1994; Johnson and Taylor 2004; Lescak et al. 2015; Liu et al. 2016), but marine stickleback tend to be underrepresented. For instance, Ortí et al. (1994) did a global survey of threespine stickleback variation but only freshwater stickleback were sampled south of British Columbia, whereas Lescak et al. (2015) exclusively sampled freshwater populations. In short, a large body of work has uncovered three major clades of marine and freshwater stickleback (Figure 1-6): a Japan Sea clade, which has since been described as a distinct species (Higuchi and Goto 1996; Higuchi et al. 2014; Yoshida et al. 2016); a Trans-North Pacific (TNP) clade, which encompasses all Japanese freshwater populations, some Japanese marine populations, and various stickleback at varying frequencies along the Aleutian Islands and as far south as Vancouver Island; and a Euro-North American (ENA) clade, which encompasses both sides of the Atlantic and the eastern Pacific (Haglund et al. 1992; Ortí et al. 1994; Taylor and McPhail 1999; Johnson and Taylor 2004; Lescak et al. 2015). The ENA clade has been further subdivided into an eastern Pacific lineage and a derived Atlantic lineage (Haglund et al. 1992; Thompson et al. 1997; Liu et al. 2016). Finally, the Atlantic lineage has been further subdivided (Mäkinen and Merilä 2008; Ravinet et al. 2013a) (Figure 1-6). Clades were first identified based on mitochondrial haplotypes, for which no adaptive significance could be found. On Haida Gwaii, marine and most freshwater populations belonged to the ENA clade, whereas unusual unarmoured freshwater stickleback were all TNP (Gach and Reimchen 1989; O’Reilly et al. 1993; Ortí et al.

Figure 1-6 - Approximate distributions of different marine clades determined from mitochondrial DNA. Green (1), Japan Sea clade (now Gasterosteus nipponicus). Blue (2), Trans-North Pacific (TNP) clade. Black outline (3-6), Euro-North American (ENA) clade. Red (3), Eastern Pacific lineage of ENA. 4-6, Atlantic lineage of ENA: Brown (4), Trans-Atlantic sublineage; Purple (5), European sublineage; Yellow (6), Baltic Sea sublineage. Not shown are the exclusively freshwater Irish and Mediterranean sublineages of the Atlantic lineage. When clades overlap, thick outlines indicate the more common clade. Adapted from Haglund et al. 1992; Ortí et al. 1994; Higuchi and Goto 1996; Thompson et al. 1997; Taylor and McPhail 1999; Johnson and Taylor 2004; Mäkinen and Merilä 2008; Ravinet et al. 2013a, 2014; Lescak et al. 2015.

1994; Deagle et al. 1996). However, the morphological associations seemed coincidental with the habitat occupied by TNP fish, and subsequent analyses discounted any association between freshwater phenotype and haplotype (Deagle et al. 1996; Johnson and Taylor 2004). Deagle et al. (2013) further found no association between mitochondrial haplotype and variation in nuclear DNA, suggesting limited adaptive relevance of these clades. In contrast, the Japan Sea and TNP clades are adaptively divergent. Japan Sea stickleback had higher mortality in fresh water than TNP marine stickleback (Ishikawa et al. 2016). Hybridization experiments have shown that these two marine clades are reproductively isolated (Yamada et al. 2001; Ishikawa et al. 2006; Kitano et al. 2007), despite past hybridization that introduced TNP mitochondrial haplotypes into the Japan Sea populations (Yamada et al. 2001; but see Yamada et al. 2007). The Japan Sea clade was formally recognized as a distinct species, Gasterosteus nipponicus, in 2014 (Higuchi et al. 2014; Yoshida et al. 2016), indicating that marine stickleback can indeed evolve, and that clades have some adaptive relevance. In this case speciation apparently occurred after threespine stickleback became isolated in the Sea of Japan subsequent to sea level lowering approximately two million years ago (Higuchi et a. 2014). A more recent speciation event may be underway in marine threespine stickleback in the western Atlantic. Multiple behavioural and morphological phenotypes distinguish the “white stickleback” of Nova Scotia (Haglund et al. 1990) from typical marine threespine stickleback. White stickleback nest in distinct habitats, are smaller, display white instead of red nuptial colouration, do not provide parental care, have reversed sexual dimorphism for brain size, and consistently mate assortatively in lab settings (Blouw and Hagen 1990; Jamieson et al. 1992ab; Blouw 1996; McKinnon and Rundle 2002; Samuk et al. 2014). However, allozyme variation did not differentiate them from typical threespine stickleback (Haglund et al. 1990), hybridization can be readily achieved in the lab (McKinnon and Rundle 2002), and mitochondrial haplotypes place them in the ENA clade (Ortí et al. 1994). Deep sequencing is needed to determine the genetic underpinnings of this potential incipient speciation.

Within the eastern Pacific lineage of the ENA clade, freshwater stickleback have undergone rapid local adaptation (e.g. Barrett et al. 2011; Deagle et al. 2012; Rennison et al. 2016), whereas Withler and McPhail’s (1985) results left the impression that marine stickleback have not experienced divergence, let alone local adaptation. These results were interpreted for marine stickleback across the northern hemisphere – what seemed true of Pacific North America was surely true as well for both sides of the Atlantic. As such, most stickleback studies have focussed on freshwater populations and examined a single marine population only as a point of comparison. However, at least 34 studies, including the clade studies mentioned above, have sequenced loci in multiple marine populations from any clade (Table 1-2). Of these, only 17 reported FST for pairwise comparisons of marine stickleback populations (two additional studies used other measures of divergence). Often only averages were reported, with an overall FST of 0.031. These 34 studies demonstrated a greater interest in freshwater stickleback. Removing three studies that did not report the number of marine stickleback populations used, 37% of all sampling locations, and 42% of all sequenced stickleback, were marine.

The average marine FST of 0.031 is potentially biologically relevant, but interpretations were limited by the limited resolution achieved across the genome. Larger regions of the genome could be covered through Next-Generation Sequencing, the first study of which was published by Hohenlohe et al. (2010). Two marine populations from Alaska, separated by > 1000 km of coastline, were sequenced at > 45 000 SNPs, resulting in FST of 0.0076. Catchen et al. (2013a) replicated Hohenlohe et al.’s (2010) protocol using marine and coastal and insular freshwater populations from Oregon. The three marine populations had pairwise FSTs of 0.01, 0.009, and 0.003, although Structure analysis could nevertheless partition individuals into distinct groups (Catchen et al. 2013a). Deagle et al. (2013) included 11 marine populations in their study of Haida Gwaii. Neighbour-joining and PCA both grouped most marine stickleback together, but one or two populations were genetically distinct. Beyond their representation in a figure, these results are not discussed in the paper, nor were FST estimates provided. Roesti et al.

(2014) claimed a median FST of 0 for SNPs sequenced in two Vancouver Island

populations, but the mean was 0.023. Collectively these results suggest low but possibly important variation along the Pacific coast. Studies have recently examined global parallel selection on the genome in freshwater stickleback and included geographically paired marine and freshwater populations. The initial expectation was that freshwater ecotypes evolved independently rather than from a single freshwater ancestor (McKinnon et al. 2004). Once this was established, genomic regions would be identified that, irrespective of geography, consistently differentiated marine from freshwater stickleback (DeFaveri et al. 2011; Jones et al. 2012ab). Less than 0.5% of the genome showed signs of parallel evolution (Jones et al. 2012a). One of these studies (Jones et al. 2012b) explicitly included marine stickleback with “typically marine” phenotypes (see their Figure 1 caption). Jones et al. (2012a) identified a split between northern Pacific freshwater populations on the one hand, and Pacific marine and southern Pacific freshwater populations on the other, but apart from a single mention in the results, this surprising finding is not discussed further. Despite all of this highly suggestive work, marine stickleback have not been sampled in appreciable numbers along the Pacific coast of North America to determine population structure.

In contrast, numerous studies from Europe have reported significant FST between marine populations in the Baltic Sea or elsewhere, ranging from 0.005 to 0.102 (Table 1-

2). Cano et al. (2008) found that males had lower FST (0.001) than females (0.010), suggesting that males contributed disproportionately to gene flow. Lind and Grahn (2011) were the first to report evidence for directional selection in the marine environment, with mill effluent in the Baltic Sea contributing to pairwise FST values as high as 0.24. To date, the most thorough analyses of marine stickleback have come from a series of papers published by the Merilä lab. DeFaveri and Merilä (2013) sampled 1288 individuals from 38 populations across the Baltic Sea and assessed variation in plate morph, at the Eda allele, and at 20 neutral microsatellite loci for 14 of these populations. Note here the difficulty of defining a marine stickleback. All populations occur in the Baltic Sea, where salinity varies from full strength seawater to fresh water. Indeed, typically freshwater fish species such as tench and northern pike can be found in the Baltic Sea (HELCOM 2006).

Global FST for the neutral markers was 0.006, but at Eda it was 0.102. This PST – FSTQ –

FST comparative approach provided the only study to date demonstrating selection in marine stickleback, even when migration is high and drift is low (but see Leinonen et al.

2006 for an early PST-FST comparison using body morphology). Further work using the same 38 Baltic marine populations found higher FST in genic (0.026) than nongenic (0.008) microsatellites, with a strong correlation between allele frequency and salinity. The 38 populations clustered into five distinct groups based on the genic loci, but a single cluster based on the nongenic loci (DeFaveri et al. 2013a). Patterns of differentiation were inconsistent with passive dispersal via oceanographic currents, but were instead likely caused by selection. DeFaveri et al. (2013b) subsequently assessed 240 marine stickleback from ten sites within the Atlantic lineage. Using 138 microsatellite loci, FST was relatively high for loci under balancing selection (0.011), was higher yet for neutral loci (0.033) and was the highest for loci under directional selection (0.097). Of these latter, the authors found significant associations between some functionally relevant microsatellites and environmental conditions, including salinity and temperature. FST estimates appeared to be temporally stable (DeFaveri and Merilä 2015). Collectively, these European results indicate that even marine populations can become structured when barriers to gene flow are limited, and that both drift and selection can play a role. This contrasts with the conclusions for Pacific North America that marine stickleback are genetically homogeneous; however, sampling in this region has not been sufficiently extensive to test this adequately.

1.4.4 Population size

Marine threespine stickleback are widely assumed to be ubiquitous in the northern hemisphere. Large effective population sizes is often cited as one reason for high connectivity along the coastline (e.g. Bell 1976). Indeed, estimates of stickleback biomass in bays can be extraordinarily high – for instance fishermen in the White Sea routinely report 1 ton hauls of stickleback per 40 m beach seine (Yershov and Sukhotin 2015). In the Baltic Sea stickleback numbers have been rising, with estimates in the billions (Jurvelius et al. 1996) and known catch per unit effort within the cooling intake

of nuclear power plants approach 180 000 stickleback per day (Bergström et al. 2015). This abundance coupled with assumptions of limited phenotypic and genetic variation between populations has resulted in little conservation attention paid to marine stickleback. This may be justified in most regions, but the White Sea stickleback population collapsed to near extinction during the early 1960s, and began to show recovery only in the late 1990s (Yershov and Sukhotin 2015). Multiple genetic methods to calculate effective population size of Baltic Sea stickleback generated estimates of 100 to 900 stickleback, which is well below the 10% Ne/Nc ratio commonly observed in fishes (DeFaveri and Merilä 2015). This raises two possibilities: either there is a systematic bias in all effective population size estimation methods that make them particularly useless for large populations; or the census sizes are smaller than appreciated, due perhaps to biologically relevant population structure caused by natal homing. If the latter, it undercuts the assumption of panmixia and suggest that marine stickleback might warrant conservation attention.

1.4.5 Evolution in marine stickleback

Contemporary marine populations are used as proxies of the ancestral condition in part because it is assumed that they have undergone little recent evolution, as illustrated by the following quotations. “They [marine stickleback] are also essentially ‘living fossils’ that have changed little morphologically over the last 10 million years…[The ocean ancestor is] evolutionarily static…” (Foster et al. 2003, p. 11). “We assume that the present oceanic stickleback represents the ancestral phenotype which gave rise to the recent post-glacial freshwater populations, and we draw conclusions about the direction and extent of evolution in fresh water based on this assumption” (Baker et al. 2008, p. 580-581). “The general marine environment faced by oceanic stickleback likely has not changed markedly since the last glacial period because oceanic stickleback have the option to shift their range in the face of changing environments due to glacial cycles” (Baker et al. 2008, p. 581). The evolutionary stasis of marine stickleback was explicitly tested based on twelve morphological traits in two fossil marine stickleback (a 10-million year old

specimen from Lompoc, California (Bell 1977) and a 13-million year old specimen from Palo Verdes, California), as well as fish from three contemporary marine/anadromous populations from Poland, Quebec, and Alaska, and from three Alaskan freshwater populations with moderate to high armature (Bell et al. 2009). Based on Principal Component Analysis, the fossil forms differed little from modern-day marine stickleback. They were located at either end of PC1 and towards one extreme of PC2, all within the cluster of marine fish. I replicated this analysis using Bell et al.’s (2009) methods, but only marine stickleback that I had collected from Alaska to southcentral California (see Chapter Two) (Table 1-3). Palo Verdes was at the extreme end of the distribution of contemporary fish for PC1 and was outside the distribution for PC2 (Figure 1-7), while Lompoc occurred within the fully-plated marine group. A portion of PC1 was not occupied by either fossil, but was extensively populated by contemporary fish. This simple analysis indicates that the fossil specimens are relatively similar morphologically to some of the contemporary fish, but they are unlike others, suggesting that marine stickleback have not been as morphologically static as previously assumed. Whether the variation in marine forms is due to evolution or plasticity remains to be ascertained. However, inferring stasis from two fossil specimens seems questionable at best.

1.5 The way forward

Shared SGV plays an important role in adaptive divergence, particularly in parallel evolution during ecological speciation. However, inferences about the role of SGV versus alternatives such as plasticity and de novo mutation requires that the ancestral condition be properly characterized. Threespine stickleback are icons of parallel evolution, and Ectodysplasin is the example of adaptation from SGV. However, the assumptions made about marine stickleback that make them so useful for ancestral- derived comparisons – namely, that phenotypic and genetic variation is consistent across their range with no evidence of local adaptation or contemporary evolution – are based on insufficient data from northern Pacific studies, and are contradicted by studies from Europe and Japan. Here I briefly discuss how this thesis contributes to understanding of variation in marine threespine stickleback and its role in adaptive divergence.

Table 1-3 - Mean phenotypic measurements (mm) for threespine stickleback from seven populations, and measurement of two fossil specimens (LO, Lompoc, and PV, Palo Verdes). Measurements were made on the right side of the fish for pelvic spine, ectocoracoid length, and plate count as described by Bell et al. (2009). No. corresponds to the labels on Figure 1-7.

No. Trait LO PV CA01 CA02 CA03 OR01 OR02 BC01 AK01 1 Head length 14.1 17.8 11.9 12.2 11.2 11.2 13.3 15.2 15.3 2 Body depth 8.1 13.6 9.9 10.7 9.3 9.6 11.8 11.4 11.5 3 Caudal depth 2.3 4.4 1.7 2 2.2 1.5 1.9 2.3 1.7 4 Head to anal fin 33.6 55.2 29.6 30.8 27.4 28.5 34 35.5 41.8 5 1st dorsal spine length 4.9 4.7 4 3.9 4.3 4.3 4.7 5.7 5 6 2nd dorsal spine length 5.6 8 4.6 4.3 4.7 4.8 5.3 6 5.4 7 Pelvic spine length 7.5 14.1 5.9 6.8 6.2 6.3 6.6 9 7.1 8 Ectocoracoid length 12.1 13.9 6.9 7.7 7.1 7 7.6 9.3 11.7 9 Plate count 25 29 10.3 13.5 9.3 32.2 16.1 33.6 29.9 10 Standard length 51 83.4 42.3 43.1 38.6 41.3 47.2 52.1 58.3

Figure 1-7 - PCA of ten phenotypic traits in extant marine stickleback from seven localities (dots) and two fossil specimens from California (triangles). LO = Lompoc, PV = Palo Verdes. See Table 1-3 for phenotypic traits and Chapter Two for the sampling localities. Arrows show the loadings of each trait, with numbers 1-10 defined in Table 1- 3. PC1 explains 65% of total variance, PC2 explains 11%. Variables were centered and scaled to unit variance.

To avoid repetition, Chapter Two outlines the marine populations studied in Chapters Three through Five, and associated collection methods. Note that this is the largest collection of its kind of Pacific North American threespine stickleback since the genomics era, and previous sampling tended to capture stickleback from rivers. Chapter Three is an initial exploration of phenotypic variation that has traditionally been used to assess ecogeographic rules and local adaptation. Specifically, I explore Jordan’s Rule (increasing vertebral number with increasing latitude) and two related rules: Bergmann’s Rule (increasing body size with increasing latitude) and pleomerism (vertebral number and body size vary positively). This is the first study of its kind for Pacific stickleback, and encompasses a substantially greater range than a comparable study on the western Atlantic coast. Collectively, this study demonstrates potentially adaptive phenotypic variation among marine stickleback populations In Chapter Four, I combine phenotypic data (plate morph and 3D morphometrics) with the sequencing of over 300 000 loci (yielding over 5000 SNPs) to determine population genetic structure, effective population size, and selection in marine threespine stickleback populations ranging from southcentral California to Alaska. Using non-genic loci as a conservative estimate for neutral FST distribution, a PST-FSTQ-FST analysis was conducted for plate count and body shape. To test the significance of genetic variation among marine populations for inferences regarding the pace and form of evolution in freshwater environments, a single freshwater population from Vancouver Island was sequenced and outliers of selection examined for each marine-freshwater pairwise combination. In short, this chapter provides evidence for population genetic structuring in marine stickleback, and suggests that marine stickleback have continued to evolve in a manner that affects inferences of freshwater adaptation. Mechanisms for maintaining large levels of SGV in nature are poorly resolved, but one prediction is that SGV is maintained by a form of heterozygote advantage – namely, that individuals with greater levels of genome-wide heterozygosity are better buffered against developmental noise than individuals with low levels of heterozygosity. Alternatively, developmental noise may be enhanced in single-locus heterozygotes for mutations of large effect. Chapter Five tests these competing hypotheses using the

presence/absence of plates on each myomere as a measure of asymmetry. This study complicates interpretations regarding the maintenance of SGV at Eda in the marine environment. Characterizing SGV among contemporary ancestors allows for more robust justifications for one’s choice of marine stickleback population when comparing adaptive divergence between marine and freshwater ecotypes. Barrett et al. (2011) found that cold tolerance had evolved in parallel in freshwater stickleback from British Columbia. Morris et al. (2014) demonstrated that this was associated with the evolution of gene expression, particularly for mitochondrial-related genes. Chapter Six follows up on this work by comparing a single marine population to a single freshwater population, both from Vancouver Island, British Columbia. Since Vancouver Island marine stickleback are part of the genetic cluster that extends from Washington to Alaska, using other marine populations from the region could be constituted as pseudoreplication – something that could not be demonstrated prior to this thesis. Marine and freshwater stickleback were raised at different temperature treatments and mitochondrial biogenesis measured from cardiac and pectoral muscle imaged using transmission electron microscopy. Mitochondrial biogenesis has been measured in freshwater stickleback but not marine stickleback, and is an adaptive phenotype likely associated with cold adaptation. This chapter provides a reflection on the role of pre-existing plasticity, rather than pre-existing genetic variation, on the adaptation process. Collectively, this thesis will demonstrate that phenotypic and genetic variation does occur in the marine environment, and this has implications both for local adaptation in marine stickleback, and for our understanding of the role of SGV during adaptive divergence between ancestral and derived forms.

Stickleback collection

Chapters Three through Five use stickleback collected from the same locations along the Pacific coast of North America. Rather than repeat the same sampling information in each chapter, it is described once here. Any variation from these procedures will be outlined in the respective chapter.

2.1 Sampling effort

Danielle Morris and I sampled marine stickleback from Vancouver Island to Monterey Bay (summer 2013) and northern British Columbia (BC) (summer 2015). The Vancouver Island specimens were collected in conjunction with the marine fishes course at the Bamfield Marine Sciences Centre. We sampled an additional freshwater location, Brannen Lake, Vancouver Island, BC, during the summer of 2011 (during my MSc). Sampling was directed to locations with published records of marine stickleback. A total of 24 marine locations were sampled using minnow traps (most localities), seines (Bamfield Inlet, BC; Little Clam Bay, Washington), kick seines (South Slough, Oregon), or dip nets (Little Clam Bay, BC). Traps were set for  10 h and collected at low tide. A total of 2008 trap hours, plus seining and dip netting, yielded 821 juvenile and adult marine threespine stickleback from seven locations: Elkhorn Slough, Monterey Bay, California (CA01), Doran Park, Bodega Bay, California (CA02), Arcata Marsh, Arcata Bay, California (CA03), South Slough, Coos Bay, Oregon (OR01), Tillamook Bay, Oregon (OR02), Little Clam Bay, Puget Sound, Washington (WA01), and Bamfield Inlet, Vancouver Island, British Columbia (BC01) (Table 2-1). Additional marine stickleback were collected by Ella Bowles (summer of 2012) from Swikshak Lagoon, Alaska (AK01) (Figure 2-1, Table 2-1). No stickleback were recovered from the vicinity of Prince Rupert, BC, despite sampling at several locations. In total, stickleback were sampled along a 21.8 latitudinal range. A total of 368 threespine stickleback were retained (Table 2-1), including 283 adults (defined as > 30 mm length, at completion of plate development; Schluter et al. 2010). Juveniles included all 50 individuals retained from WA01 and 32 of the 51 individuals from OR01. Therefore, any measure of morphology excluded WA01 and over 50

Table 2-1 - Collection sites and sampling characteristics.

Pop Water Region State/ Date Latitude Longitude N N > Collection Euthanasia Permit ID body Province collected 30 method mm CA01 Elkhorn Monterey California July 26- 36°49’45N 121°44’07W 35 35 Minnow MS-222 SC- Slough Bay 27, 2013 traps 12743 CA02 Doran Bodega California July 28, 38°18’52N 123°01’55W 50 48 Minnow MS-222 SC- Park Bay 2013 traps 12743 CA03 Arcata Arcata California July 22, 40°51’23N 124°05’24W 50 46 Minnow MS-222 SC- Marsh Bay 29, 2013 traps 12743 OR01 South Coos Bay Oregon July 17- 43°17’35N 124°19’26W 51 19 Kick MS-222 18088 Slough 18, 30, seine 2013 OR02 Tillamook Tillamok Oregon July 14, 45°28’52N 123°53’49W 50 50 Minnow MS-222 18088 tidal gate Bay 2013 traps WA01 Little Puget Washington July 9- 47°34’32N 122°32’43W 50 0 Dip net MS-222 13- Clam Bay Sound 10, 2013 226a BC01 Bamfield Vancouver British Late 48°49’55N 125°08’17W 51 51 Beach MS-222 XR Inlet Island Columbia June, seine 108 2013 2013 AK01 Swikshak Kodiak Alaska 2012 58°37’14N 153°44’44W 31 31

Lagoon Island BCFW Brannen Vancouver British 2011 49°12’54N 124°03’16W 50 NA Minnow Eugenol Lake Island Columbia traps

Figure 2-1 - Locations from which threespine stickleback were collected, including eight marine sites (triangles) and one freshwater site (circle).

half of OR01. All stickleback were treated according to Canadian Council on Animal Care (CCAC) guidelines (AUP AC13-0040) and state/provincial/national collection and import permits (see Table 2-1). Traps were initially baited with hot dogs, but the unwanted capture of crabs made this practice impractical, and so traps were set unbaited. Stickleback were euthanized using buffered tricaine methanesulfonate (MS-222) or Eugenol (clove oil) and preserved in 70% ethanol. Fin clips were preserved in 95% ethanol for later sex determination and genetic sequencing.

2.2 The marine environment

Stickleback were collected in coastal habitat. All localities except WA01 had individuals in reproductive condition (gravid females, red-throated males). Salinity at the time of sampling was > 30 ppt at all locations except OR02 (~12 ppt). Fish were collected at BC01 and AK01 by other individuals and will not be described. Elkhorn Slough (CA01) is an 11-km long tidal estuary with full-strength seawater along most of its reach. It contains a patchy distribution of Zostera eelgrass, which is dominated by adult threespine stickleback and shiner surf-perch during spring and summer. Stickleback leave the estuary during the winter, but their overwintering habitat is unknown (Grant 2009). In addition to stickleback, samples included the obligate marine longjaw mudsucker (Gallichthys mirabilis) and a smelt (tentatively identified as Atherinops affinis) (Figure 2-2). No stickleback were acquired from Bodega Bay (CA02) directly, but a population was collected in a culvert in a saltmarsh off Doran Park (Figure 2-3). This region of the bay drains at low tide, resulting in a large tidal mud flat. The culvert retains marine water at very high salinities (> 38 ppt at low tide). This water flows through a saltmarsh before abruptly terminating at a freshwater marsh that has been extensively choked by plants. A slight trickle of fresh water flowed from this marsh into the slough. The transition from saltwater to fresh water in the slough was abrupt, with salinity still 20 ppt a few metres from the start of the choked marshland (Figure 2-3). Stickleback were collected along the entire reach from the culvert to near the freshwater marsh. No other species were collected.

Figure 2-2 - Elkhorn Slough, Monterey Bay, California, U.S.A. (Top left) Danielle Morris setting minnow traps on saltwater plants. (Top right) Adult male marine threespine stickleback. (Bottom left) Longjaw mudsucker. (Bottom right) Unidentified species of smelt.

Figure 2-3 - Doran Park, Bodega Bay, California, U.S.A. (Top left) Doran Park marshland showing culvert. (Top right) The first minnow trap haul with marine stickleback. (Bottom left) A close-up of an adult marine threespine stickleback. (Bottom right) The saltmarsh terminated at a freshwater marsh.

Arcata Marsh (CA03) is a muddy-bottomed saltmarsh and parkland at the south end of Arcata Bay, between the cities of Eureka and Arcata. Adult and juvenile stickleback were collected in a saltwater sidechannel of the tidal marsh. Traps were partially exposed at low tide, but covered by over a metre of water at high tide. Stickleback were collected along with Pacific staghorn sculpin (Leptocottus armatus) and many juvenile Dungeness crab (Metacarcinus magister) (Figure 2-4). The South Slough (OR01) collecting site was a narrow, shallow muddy-bottomed saltwater channel at the south end of Coos Bay. Juvenile and adult stickleback were collected along with juvenile sculpin (Figure 2-5). The slough is vast – sampling from regions specified in the literature failed to collect any marine stickleback. South Slough National Estuary Research Reserve staff were contacted and searched for stickleback while I moved on to California – they located adults and juveniles in an area they referred to as Hidden Creek, which we sampled together on my return trip. Tillamook Bay (OR02) stickleback were collected at a tidal gate on the south end of the bay, off the south end of Delta Island at the intersection of Tillamook and Wilson rivers. Stickleback inhabit the entire bay during the winter, but move to the saltwater marshes at the south of the bay during the breeding season (Bottom and Forsberg 1978; Ellis 2002). Only adults were collected, but recently-hatched juveniles schooled extensively in the area. Traps were nearly exposed at low tide, but covered to several metres at high tide. Stickleback were collected along with Pacific staghorn sculpin, prickly sculpin (tentatively Cottus asper), bay pipefish (Sygnathus leptorhynchus), and saddleback gunnel (Pholis ornate) (Figure 2-6). Salinity at this location varies from full- strength seawater to < 1 ppt, depending on river flow (Ellis 2002). At the time of sampling salinity at low tide was 12 ppt. WA01 was the only rocky-bottom coastal marine habitat sampled. Water depth varied too much between low and high tide to use minnow traps. Large schools of juveniles were observed along shore at high tide – stickleback were sampled using aquarium dip nets, with 100 juvenile stickleback readily captured in < 30 min. Seining failed to recover any stickleback – but a single crescent gunnel (Pholis laeta) was captured (Figure 2-7).

Figure 2-4 - Arcata Marsh, Arcata Bay, California, U.S.A. (Top left) Low tide. (Top right) Adult marine threespine stickleback. (Bottom left) Juvenile Dungeness crab. (Bottom right) Zostera, a typical nursery plant for marine stickleback.

Figure 2-5 - South Slough, Coos Bay, Oregon, U.S.A. (Top left) South Slough National Estuarine Research Reserve staff help me use a kick-seine at low tide to find marine stickleback. (Top right) Juvenile marine threespine stickleback. (Bottom left) Preserving fin clips for sequencing. (Bottom right) Fin clips preserved in 95% ethanol.

Figure 2-6 - Tillamook Bay, Oregon, U.S.A. (Top left) East side of tidal gate at low tide. (Top right) Adult marine threespine stickleback. (Bottom left) Saddleback gunnel. (Bottom right) Sculpin spp. Bay pipefish were also collected but no pictures were taken.

Exploring Jordan’s Rule in Pacific threespine stickleback

3.1 Introduction

Ecogeographic rules, concerning patterns of phenotypic variation along some geographic continuum, were originally formulated as compelling examples of adaptation. Desert animals with larger ears or polar animals with thicker fur than temperate conspecifics provided early evidence that selection modified phenotypes to best suit difficult environments (Allen 1877). Jordan (1891) stated that fish from colder waters (higher latitudes, open oceans, deep waters) tended to have more vertebrae than related species from warmer waters (lower latitudes, coastal seas, shallow waters). He suspected that temperature shaped the biotic and abiotic interactions that formed the selective environment of the species. Through a process Jordan described as “ichthyization”, warm-water species evolved reduced vertebral number as a consequence of skeletal specialization driven by enhanced competition (Jordan 1891). He later reversed the order of events when fossil evidence revealed increased vertebral number to be the derived phenotype - relaxed selection under colder temperatures reduced skeletal specialization (Jordan 1905). Although Jordan’s explanation of his rule is largely dismissed today (but see McDowall 2008), Jordan’s Rule continues to be a research focus. In many investigated species vertebral number increases linearly with increasing latitude (Billerbeck et al. 1997; McDowall 2003a; McBride and Horodysky 2004; Yamahira et al. 2006; Barriga et al. 2013; although see Resh et al. 1976; Shikano and Merilä 2011). In some cases, vertebral number has been explicitly associated with temperature (Tåning 1952; Lindsey 1954; Seymour 1959; Lindsey and Harrington 1972; Baumann et al. 2012; Reimchen and Cox 2016; Ackerly and Ward 2016). The function of vertebral number variation is partially understood. Vertebral number affects flexibility (Long and Nipper 1996; McDowall 2003a), maximum body curvature for escape swimming (Brainerd and Patek 1998), body elongation associated with ambush predation (Maxwell and Wilson 2013), burst swimming (Swain 1992a), and C-start velocity (Ackerly and Ward 2016). Although the reasons are poorly understood, vertebral number is an important differentiator of fish stocks (Swain et al. 2001), ecomorphs (Aguirre et al. 2014), and life-history strategies (McDowall 2003b), impacts

female reproductive investment in northern pike (Tibblin et al. 2016), and evolves under predation in multiple fish species (Swain 1988; Ackerly and Ward 2016; Tibblin et al. 2016). Geographic patterns consistent with Jordan’s Rule may reflect benefits for larvae swimming in colder or more viscous waters (Fuiman and Batty 1997; Hunt von Herbing 2002). The causes of Jordan’s Rule are similarly poorly understood. Despite much evidence consistent with Jordan’s Rule, whether it reflects temperature-induced plasticity, local adaptation, or both is largely unknown (but see Yamahira and Nishida 2009; Baumann et al. 2012). Furthermore, it has been argued that Jordan’s Rule is superfluous, and that Bergmann’s Rule, whereby body size increases with decreasing temperature, and pleomerism, whereby larger individuals have more vertebrae, are together sufficient to explain Jordan’s Rule (Lindsey 1975). Marine threespine stickleback is an ideal system in which to study Jordan’s Rule. Stickleback vertebral number is moderately heritable (h2 = 0.37 – 0.55, Hermida et al. 2002; Alho et al. 2011) with known QTLs (Berner et al. 2014; Miller et al. 2014). Several studies have described functional consequences of variation in vertebral number in stickleback (Swain and Lindsey 1984; Swain 1992b; Walker and Bell 2000; Spoljaric and Reimchen 2007; Aguirre et al. 2014), including the ratio of caudal to abdominal vertebrae (Swain 1992a; Aguirre et al. 2014). Juvenile stickleback with more vertebrae prefer cooler waters, potentially to optimize swimming efficiency under a given temperature/viscosity (Reimchen and Cox 2015). Given that juveniles may migrate > 100 km offshore to escape predation (Cowen et al. 1991), greater flexibility and predator- escape abilities would be more important in colder, more viscous waters. Vertebral number therefore has a genetic component and has adaptive significance – all of the prerequisites for Jordan’s Rule to evolve. Temperature-induced plasticity also influences vertebral number. In threespine stickleback, vertebral number is established prior to hatching (Lindsey 1962) and is plastically influenced by temperature during this brief developmental period. The fewest vertebrae develop in fish subject to “intermediate” temperatures (18-24°C) during this period in several stickleback genotypes (Lindsey 1962). If developmental temperature

varies latitudinally, plasticity alone could account for Jordan’s Rule. The role of plasticity can be assessed by observing how other temperature-induced meristic traits vary with latitude. Lindsey (1962) found that numbers of dorsal- and anal-fin rays and basal supports (hereafter “basals”) increased with decreasing temperature, whereas the number of pectoral-fin rays exhibited an inverted-V reaction norm. Fin-ray and vertebral numbers have different heritability (Hermida et al. 2002), with different QTLs (Miller et al. 2014), and in related fourspine stickleback are not correlated in the wild (Krueger 1961). Therefore, analysis of latitudinal patterns of variation in other meristic traits may help disentangle the relations between Jordan’s Rule, local adaptation, and plasticity. Here I test Jordan’s Rule by investigating variation in vertebral number for marine threespine stickleback along a 21.8 latitudinal range, testing the following predictions: (1) The number of vertebrae should increase with increasing latitude; (2) Jordan’s Rule should persist after accounting for Bergmann’s Rule; (3) Jordan’s Rule may differ between the sexes, reflecting different functional significances of Jordan’s Rule and pointing to possible genetic mechanisms to explain these patterns; (4) Other meristic traits will also be correlated with latitude in a manner suggesting a role for plasticity. Collectively, these predictions seek to test the applicability of Jordan’s Rule to threespine stickleback and provide a first step in determining its causes and consequences, while also determining whether phenotypic variation exists within a putative “panmictic” population.

3.2 Materials and Methods

3.2.1 Collection and sex determination

Stickleback were collected as described in Chapter Two, across a latitudinal range with a known temperature gradient (Figure 3-1). DNA was extracted from fin clips using the phenol-chloroform method (Chomczynski and Sacchi 1987). Sex was determined

Figure 3-1 - Variation in average sea surface temperature (C), 2010-2014, across the study region, based on data from the National Oceanic and Atmospheric Administration.

using primers developed by Peichel et al. (2004) that amplify sex-specific alleles at the idh locus. Alleles were visualized in a 2% agarose gel.

3.2.2 Sample preparation and phenotypic measurements

Radiographs were taken of individual fish positioned right laterally using a cabinet X-ray apparatus (Kubtec XPERT 80-L), at 35 kV and 1000 μA for 15 seconds with lateral projection. Total, caudal, and abdominal vertebrae were counted from the digital image using ImageJ, beginning with the basioccipital and ending with the vertebra preceding the urostyle (Figure 3-2). Caudal vertebrae were assigned as per Aguirre et al. (2014), with the first caudal vertebra usually readily distinguished by the presence of a haemal spine and haemal arch, a lack of ribs, and a close association with the first anal pterygiophore. Transitional vertebrae that had haemal arches but no haemal spines, or had relatively short haemal arches that were not associated with the first anal pterygiophore, were scored as abdominal vertebrae, following Aguirre et al. (2014). Standard length (mm) was measured in ImageJ. Fish were stained in Alizarin red for visualization of fin rays and basals. Dorsal- fin rays were counted beginning with the first soft ray posterior to the third dorsal spine. Anal-fin ray counts included the first spiny fin ray. Pectoral-fin ray counts were calculated as the average for both pectoral fins, to account for asymmetry. Branched rays were counted as single fin rays. Dorsal- and anal-fin basals were also counted, but as each fin ray always had a basal associated with it, only counts of basals posterior to the last associated fin ray were included. These could be thought of as “extra” basals that extend beyond the fin. Three BC01 fish were sent to another facility before these additional phenotypes could be measured, so for all fin ray and basal counts the sample size was reduced to 48 for this population.

3.2.3 Statistical analysis

Goodness-of-fit Chi-square tests were used to test for equality in sex ratio for each site. Logistic regression was used to determine the association between sex ratio and

Figure 3-2 - Radiograph of a marine threespine stickleback. The urostyle was not included in total vertebral counts.

latitude. Because ecogeographic rules usually assume linear relations with latitude, linear regression was used as a first approximation of the influence of latitude on vertebral number (total, abdominal, and caudal) and standard length. Analysis of Variance (ANOVA) was conducted for each phenotype using site as a factor, with post-hoc Tukey HSD tests used to contrast all sampling sites. It became apparent that fish at sites south of British Columbia showed little variation, so vertebral number and standard length were again tested using a linear regression for only these “southern” sites. Jordan’s Rule and Bergmann’s Rule make different but potentially closely related predictions, so a linear regression was run for Jordan’s Rule with standard length as a covariate. The interaction term and main effects were systematically dropped, and the model with the lowest Akaike Information Criterion (AIC) selected for further analysis. Type II SS were calculated for the final model. Welch’s t-test was used to assess differences between the sexes for vertebral number (total, abdominal, caudal) and standard length. Linear regressions for Jordan’s Rule and Bergmann’s Rule were conducted separately for each sex. Spearman correlations for standard length vs. vertebral number were also analyzed separately for each sex. Spearman correlations were estimated for comparisons between vertebral number (total, abdominal, caudal), standard length, and average pectoral-fin ray, dorsal-fin ray, anal-fin ray, dorsal-fin basal, and anal-fin basal number, with α corrected to 0.05/36. Linear regressions were conducted for all fin rays and basals in association with latitude. Pleomerism was tested using a Pearson correlation between the average vertebral number for a population and the maximum standard length sampled from the population (Lindsey 1975). All statistical analyses were conducted with R (v3.2.1, R Core Team 2016).

3.3 Results

3.3.1 Sex

The proportion of males and females qualitatively differed among sites (Table 3- 1). Three sites (CA02, CA03, and OR01) had male:female ratios that did not differ from an expectation of 1:1. Of the rest, CA01 had a M:F ratio of 3:1, OR02 and AK01 of 1:3,

and BC01 of 12:1 (Chi-square test, χ2 > 6.63 for all four tests, 1 d.f., p < 0.01). These site-specific differences did not vary systematically with latitude (logistic regression, z = -1.5, 279 d.f., p > 0.1).

3.3.2 Vertebral number

Vertebral number tended to increase with increasing latitude. Total vertebral number varied from 30 to 34 (Table 3-2). Vertebral number increased 0.04 vertebrae per 2 increase in degree north latitude (linear regression, F1,278 = 35.2, p < 0.001, R = 0.11, regression coefficient b + SE = 0.04 + 0.006). However, a linear regression using only

Californian and Oregonian populations was not significant (linear regression, F1,196 = 3.4, p = 0.06, R2 = 0.02, b + SE = -0.03 + 0.01). Based on ANOVA and post-hoc Tukey HSD tests, total vertebral counts did not vary from CA01 to OR01 (mean + SE = 31.51 + 0.05 vertebrae), or between BC01 and AK01; however, fish from BC01 and AK01 had significantly more vertebrae (32.27 + 0.08 vertebrae) than those from the southern sites (Figure 3-3, Table 3-1). Abdominal vertebral number varied from 14 to 19 (Table 3-2). On average, AK01 fish had significantly more abdominal vertebrae than those from BC01 (ANOVA with Tukey HSD p < 0.001) (Table 3-1). Caudal vertebral number varied from 14 to 18 (Table 3-2) and was significantly higher, on average, for BC01 fish than for those from all other sites, including AK01 (Tukey HSD p < 0.001) (Table 3-1). Only caudal vertebral number varied consistently positively with latitude (caudal: linear regression, F1,278 = 18.7, p < 0.001, R2 = 0.06, b + SE = 0.03 + 0.007) (Figure 3-3). The ratio of caudal to abdominal 2 vertebrae slightly increased with latitude (linear regression, F1,278 = 3.4, p = 0.07, R =

Figure 3-3 - Mean ( SE) of variation among sampling sites, plotted against latitude, for A) total vertebral number, B) number of abdominal vertebrae, C) number of caudal vertebrae, and D) standard length (mm). Phenotypes are separated by sex: circles (female), triangles (male). Only four females were captured at 48.8°N.

Table 3-1 - Average meristic phenotypes and standard length (+ SD) of marine threespine stickleback collected along the Pacific coast from central California to Alaska. Different letters denote significantly different ANOVA post-hoc Tukey HSD contrasts at α < 0.0023, within each column only. 1 N = 48 for dorsal-, pectoral-, and anal-fin rays.

Site Lat N N Sex Standard Max Total Abdominal Caudal Dorsal-fin Anal-fin Pectoral (M/F) length (mm) length vertebral vertebral vertebral ray number ray fin-ray (mm) number number number number number CA01 36°49 35 26/9* 42.3+3.6AB 50.22 31.6+0.8A 15.8+0.7AC 15.8+ 0.7A 11.3+ 9.2+0.7A 10.0+ 0.8AB 0.2B CA02 38°18 48 20/28 43.2+ 3.35B 50.90 31.6+0.6A 15.8+0.6AC 15.8+ 0.8A 11.3+ 8.8+0.9A 10.3+ 0.9AB 0.4A CA03 40°51 46 28/18 39.2+ 5.8A 54.16 31.5+0.6A 15.5+ 0.5A 16.0 + 11.3+ 9.1+0.6A 10.0+ 0.6A 0.7)AB 0.2B OR01 43°17 19 7/12 41.9+7.0AB 52.15 31.3+0.6A 15.8+0.5AC 15.5+ .7A 11.3+ 9.2+0.6A 10.0+ 0.7AB 0.4B OR02 45°28 50 13/37* 47.3+ 3.6D 54.72 31.4+0.7A 15.5+0.7A 15.9+ 0.6A 10.7+ 0.8A 9.0+0.6A 9.9+ 0.2B BC01 48°49 511 47/4* 52.0+ 2.4E 57.23 32.3+0.7B 15.5+0.5A 16.8+ 0.7B 12.3+0.7C 10.5+ 10.0+ 0.6B 0.1B AK01 58°37 31 8/23* 58.8+ 8.1F 75.05 32.3+0.7B 16.2+0.4BC 16.1+ 0.5A 11.6+ 0.9B 9.3+0.7A 10.0+ 0.2B

Table 3-2 - Distributions of total, abdominal, and caudal vertebral numbers for threespine stickleback sampled from each of seven populations. The first row under the header lists the vertebral phenotype (30-34 total vertebrae, 14-19 abdominal vertebrae, and 14-18 caudal vertebrae). No individual had 18 abdominal vertebrae, and so that phenotype is not included in this table.

Total Abdominal Caudal Site Latitude 30 31 32 33 34 14 15 16 17 19 14 15 16 17 18 CA01 36°49’N 2 14 15 4 0 2 7 21 5 0 1 11 18 5 0 CA02 38°18’N 1 17 28 2 0 1 11 30 6 0 0 22 15 10 1 CA03 40°51’N 2 21 23 0 0 0 23 23 0 0 0 10 28 8 0 OR01 43°17’N 1 11 7 0 0 0 5 13 1 0 1 8 9 1 0 OR02 45°28’N 3 24 21 2 0 0 28 21 0 1 1 9 32 8 0 BC01 48°49’N 1 3 29 17 1 0 26 25 0 0 0 2 13 30 6 AK01 58°37’N 0 3 18 9 1 0 1 24 6 0 0 2 24 5 0

0.01, b + SE = 0.001 + 0.0007). Caudal and abdominal vertebrae varied negatively (Table 3-3, Figure 3-4).

3.3.3 Standard length

Standard length varied significantly among sites (Table 3-1). The northernmost site (AK01) had the longest fish (mean 58.8 mm) and standard length varied positively with latitude (Figure 3-3), increasing 0.9 mm per degree north latitude, on average (linear 2 regression, F1,278 = 310.2, p < 0.001, R = 0.53, b + SE = 0.86 + 0.05). Unlike vertebral number, the regression remained significant after removing the two northernmost 2 populations (linear regression, F1,196 = 19.8, p < 0.001, R = 0.09, b + SE = 0.50 + 0.11). Post-hoc Tukey HSD showed more variation in standard length among neighbouring sites than was found for vertebral number (Table 3-1). Standard length varied positively with total vertebral number (Table 3-3, Figure 3-5) and caudal vertebral number, but not with abdominal vertebral number (Table 3-3). Based on multiple regression, total vertebral number varied positively with latitude (partial regression coefficient, b + SE = 0.02 + 0.009, p = 0.02) after accounting for significant variation in standard length (b + SE = 0.02 + 0.008, p < 0.01: overall F2,277 = 21, R2 = 0.13, p < 0.001) with a non-significant interaction term. Among the seven populations, average vertebral number tended to vary positively with maximum standard length, although this relation was weakly statistically significant (Pearson correlation, r = 0.71, d.f. = 5, t = 2.3, p = 0.07).

3.3.4 Sex-specific differences

The sexes differed for several traits (Figure 3-3 to Figure 3-6). On average, males had significantly more vertebrae than females (31.87 vs. 31.57, Welch’s t-test, t = -3.4, d.f. = 271, p < 0.001), despite having smaller standard lengths (45.4 mm vs. 47.5 mm, Welch’s t-test, t = 2.3, d.f. = 262, p < 0.05). However, vertebral number varied positively 2 with latitude for both sexes (linear regression, males: F1,147 = 24.0, p < 0.0001, R = 0.14, 2 b + SE = 0.05 + 0.009; females: F1,129 = 18.3, p < 0.0001, R = 0.12, b + SE = 0.04 +

Table 3-3 - Spearman correlations for pairs of meristic traits and standard length. Bold indicates statistical significance at p < 0.001. The number of individuals varied depending on ability to calculate the phenotype – and so degrees of freedom are 275 for any comparison that included fin rays or basal plates, and 278 for the rest.

Anal Dorsal Vertebral Standard Caudal Abdominal Pectoral- Dorsal fin- basal number length vertebrae vertebrae fin rays fin-rays rays plates Standard 0.33 length Caudal 0.65 0.29 vertebrae Abdominal 0.33 0.03 -0.46 vertebrae Pectoral fin- 0.01 -0.16 -0.03 0.08 rays Dorsal fin- 0.42 0.27 0.42 -0.02 -0.06 rays Anal fin-rays 0.41 0.32 0.48 -0.13 -0.14 0.62 Dorsal basal -0.06 0.06 -0.06 0.00 0.00 -0.43 -0.24 plates Anal basal 0.01 0.09 -0.04 0.05 -0.01 -0.15 -0.36 0.44 plates

Figure 3-4 - Relation between the numbers of abdominal and caudal vertebrae ( SE) among populations and sexes. Shaded or darkly hatched symbols are female, white or lightly crosshatched symbols are male. Diamonds = CA01, triangles = CA02, squares without hatching = CA03, circles = OR01, X = OR02, rectangle = BC01, hatched square = AK01. Only four females were captured at BC01.

Figure 3-5 - Relation between standard length (mm) and total vertebral number ( SE) among populations and sexes. Shaded or darkly hatched symbols are female, white or lightly crosshatched symbols are male. Diamonds = CA01, triangles = CA02, squares without hatching = CA03, circles = OR01, X = OR02, rectangle = BC01, hatched square = AK01. Only four females were captured at BC01.

0.009). Females had more abdominal vertebrae than males (15.9 vs. 15.5, Welch’s t-test, t = 4.3, d.f. = 254, p < 0.0001). Males had more caudal vertebrae than females (16.3 vs. 15.7, Welch’s t-test, t = -7.2, d.f. = 278, p < 0.0001). However, the ratio of caudal to abdominal vertebrae varied significantly with latitude for males only (linear regression, 2 males: F1,147 = 7.9, p < 0.01, R = 0.05, b + SE = 0.003 + 0.001; females: F1,129 = 1.0, p = 0.3, b + SE = 0.001 + 0.0008). Standard length increased with latitude for both sexes 2 (linear regression, males: F1,147 = 174.6, p < 0.0001, R = 0.54, b + SE = 0.91 + 0.07; 2 females: F1,129 = 133.9, p < 0.0001, R = 0.51, b + SE = 0.81 + 0.07). Similarly, the standard length—vertebral number relation held for both sexes (Spearman correlation, males: d.f. = 147, rS = 0.43, p < 0.0001; females: d.f. = 129, rS = 0.27, p < 0.01).

3.3.5 Fin rays and basals

Meristic phenotypes were linearly associated with latitude (Figure 3-6). Pectoral- fin ray counts decreased by 0.01 rays per increase in degree latitude (linear regression, 2 F1,275 = 13.6, p < 0.001, R = 0.05, b + SE = -0.01 + 0.003). Other traits increased with 2 latitude: dorsal-fin ray counts (linear regression, F1,275 = 12.1, p < 0.001, R = 0.04, b + 2 SE = 0.03 + 0.009), dorsal-fin basal counts (linear regression, F1,275 = 5.0, p = 0.03, R =

0.02, b + SE = 0.02 + 0.007), anal-fin ray counts (linear regression, F1,275 = 18.7, p < 0.0001, R2 = 0.06, b + SE = 0.03 + 0.008), and anal-fin basal counts (linear regression, 2 F1,275 = 12.2, p < 0.001, R = 0.04, b + SE = 0.02 + 0.007). Dorsal- and anal-fin ray counts correlated positively with each other and with standard length, total vertebral number, and caudal vertebral number (Table 3-3). Anal- and dorsal-fin basals also correlated positively with each other, and negatively with their respective fin rays. Pectoral-fin rays showed no correlations with any trait (Table 3-3).

3.4 Discussion

3.4.1 Jordan’s Rule

In this study of threespine stickleback, vertebral number increased with increasing latitude, although not strictly as predicted by Jordan’s rule and assuming that latitude and

relevant temperatures are correlated. In particular, stickleback north of Washington had more vertebrae than their Oregonian and Californian counterparts. Miller and Hubbs (1969) discovered a similar pattern in this region for freshwater stickleback. Although they did not present averages for the total population, the range of averages for different plate morphs suggested that stickleback from Sitkalidak Island, AK (31.87 to 32.00 vertebrae) had more vertebrae than those from Carmel River, CA (31.24 to 31.33 vertebrae). My results contrast those of a study of marine and freshwater stickleback on the Atlantic coast of North America, which reported no evidence for Jordan’s Rule (Garside and Hamor 1973). However, the latitudinal range from that study was much smaller (< 6 latitude). Latitude explained limited variation in vertebral number, as it did not vary significantly among populations from southern California to northern Oregon. Only fish from BC01 and AK01 had significantly more vertebrae. Jordan (1891) made no claims about the linearity of vertebral number variation with temperature or latitude; presuming that the northern stickleback persist in cooler waters than southern stickleback, Jordan’s original formulation of his rule has been demonstrated in threespine stickleback. More sites will have to be sampled in the future to determine where this transition begins and whether it applies throughout central and northern British Columbia and western Alaska.

3.4.2 Bergmann’s Rule and pleomerism

Standard length varied strongly with latitude, as predicted by Bergmann’s Rule (for complications and comparisons with other fish species, see Blackburn et al. 1999; Belk and Houston 2002; Rypel 2014). Furthermore, fish with longer standard length tended to have more vertebrae, and there was weak evidence, despite low power, for Lindsey’s pleomerism. It has been suggested that Bergmann’s Rule combined with a positive body length—vertebral number relation renders Jordan’s Rule redundant (McDowall 2008). That is, Jordan’s Rule is simply a function of larger size being selected in colder waters. This does not fully explain my results, as latitude significantly affected vertebral number even when accounting for standard length. Furthermore, it has been proposed that Bergmann’s Rule reflects increased longevity caused by slow growth

and reduced competition (Angilletta and Dunham 2003; Rypel 2014). If so, this cannot explain Jordan’s Rule in threespine stickleback, as vertebral number is determined early in development (Lindsey 1962). Ecogeographic rules have long been provided as evidence of adaptation in nature – it is possible standard length and vertebral number co- vary with latitude to adapt stickleback for moving optimally through temperature-specific water viscosities (Hunt von Herbing 2002; McDowall 2003a; Aguirre et al. 2014).

3.4.3 Sexual dimorphism

Male stickleback had more caudal vertebrae than females, whereas females had more abdominal vertebrae than males. There was a tendency for individuals with more caudal vertebrae to have fewer abdominal vertebrae and vice versa, patterns which have also been reported for Alaskan freshwater and anadromous stickleback (Aguirre et al. 2014). However, the greater abdominal vertebrae in females did not fully compensate for the greater caudal vertebrae in males, as males had more total vertebrae than females. This is consistent with observations on freshwater threespine stickleback from Haida Gwaii, BC (Reimchen and Nelson 1987). Jordan’s Rule is apparent for both sexes, but the underlying patterns differed. Males and females both showed increased vertebral number with latitude, but only males showed an increase in the ratio of caudal to abdominal vertebrae. In other words, Jordan’s Rule is evident in increased caudal vertebral number with increasing latitude in males, but increased total vertebral number in females. The functional significance of these altered ratios is not completely understood, but a larger ratio is important for predator avoidance (Swain 1992a), whereas a smaller ratio may allow expansion of the abdominal cavity during egg production (Aguirre et al. 2014). These results suggest a potential role for selection in generating Jordan’s Rule in this species.

3.4.4 Plasticity and meristic traits

Lindsey (1962) explored temperature-induced plasticity in a variety of meristic traits in a freshwater population of threespine stickleback. He found that total vertebral number was plastically influenced by developmental temperature according to a V-

shaped reaction norm, with the trough of the “V” located between 18 and 24°C, depending on the genotype (in this case, fish with 2 to 5 plates). Decreasing the hatching temperature below this range resulted in development of more vertebrae, although temperatures below 16°C were not examined. Dorsal- and anal-fin rays and, to a lesser extent, dorsal- and anal-fin basals, also increased with decreasing temperature, whereas pectoral-fin rays showed a peaked relation with temperature (Lindsey 1962). Since the breeding grounds of threespine stickleback in central California can experience temperatures as high as the temperatures that induced the fewest vertebrae in Lindsey’s experiment (Figure 3-1, NOAA National Estuarine Research Reserve System 2015), we can anticipate similar patterns to Lindsey’s observations if developmental temperature is driving Jordan’s Rule. These predictions did in fact hold true, although the amount of variation explained by latitude for each meristic trait was always less than 10%.

3.4.5 Causes of Jordan’s Rule

At least three definitions of Jordan’s Rule have been used over the years. Jordan’s initial formulation of his rule was general and lacked a firm mechanism. It was applied only to related species or genera, with selection on heritable variation as the ascribed cause (e.g. Günther 1862; Jordan 1891). Hubbs (1922) focused on within-population variation of a single species, and proposed that Jordan’s Rule was a direct consequence of temperature affecting the rate of early development (McDowall 2008). Between these extremes, the current definition addresses interpopulational variation within a single species spread over a broad latitudinal range (e.g. McDowall 2003ab; Baumann et al. 2012; Barriga et al. 2013). Latitude is considered a proxy for temperature, with an assumed linear relationship. Under this definition there is no a priori reason (selection or development) for temperature to shape Jordan’s Rule. Abiotic factors proposed to influence Jordan’s Rule include light intensity (MacCrimmon and Kwain 1969), oxygen concentration (Garside 1966), salinity (Fahy and O’Hara 1977), and temperature. Overall, temperature has the most empirical support (Jordan 1891; Hubbs 1922; Baumann et al. 2012; Ackerly and Ward 2016). An examination of Bergmann’s rule in multiple fish species reported similar results

irrespective of whether body size was regressed on latitude or temperature (Rypel 2014). Vertebral number, unlike body size, is set early in development, so the extent to which these rules are comparable is unclear. Jordan’s Rule could be caused by plasticity, selection, or both. The best experimental evidence for the causes of Jordan’s Rule comes from silversides (Baumann et al. 2012). In Atlantic silversides vertebral number varies more strongly with latitude than for topsmelt from the Pacific, presumably associated with the stronger latitudinal temperature gradient in the Atlantic. Plasticity only partially explained this variation in a common garden environment, suggesting that Jordan’s Rule was caused in part by natural selection on genetically-heritable variation. Marine stickleback are generally considered to be genetically homogeneous along the Pacific coast (Withler and McPhail 1985; Hohenlohe et al. 2010), suggesting plasticity as the only explanation for Jordan’s Rule. However, some population structure may exist in marine environments (Catchen et al. 2013a; DeFaveri et al. 2013ab), so genetic causes for Jordan’s Rule cannot be ruled out. My results are consistent with a dual role of selection and plasticity – variation in meristic traits suggests that populations with more vertebrae developed under cooler temperatures, whereas sex-specific differences in Jordan’s Rule suggest that selection has shaped the responses each sex has to temperature. Further investigation is clearly required to disentangle these complex relations. Linear regression of vertebral number to latitude, although significant, did not account for much of the total variation in vertebral number. This is in part because populations south of BC01 had similar low vertebral numbers, whereas BC01 and AK01 had similar high numbers. Plasticity and/or selection could explain these trends. Although average sea surface temperatures vary linearly with latitude (Baumann et al. 2012), marine threespine stickleback may seek similar preferred temperatures along the coast in which to build their nests (Mori 1994) to optimize the amount of oxygen available to their eggs (Reebs et al. 1984; Hopkins et al. 2011); plastic responses to developmental temperature would not be evidenced as a linear relation with latitude. Furthermore, waters at a range of warmer temperatures vary less in viscosity than do waters at a similar

range of cold temperatures (Hunt von Herbing 2002), such that predator escape velocity would not be impacted across a range of warm temperatures. Selection would therefore also not be evidenced as a linear relation with latitude.

3.5 Future directions

Ecogeographic rules were once interpreted as compelling evidence for the role of natural selection, but today the causes of these rules are recognized as being harder to disentangle. Although selection is certainly responsible for some patterns or some instances of patterns (e.g. Allen’s Rule), plasticity alone may explain other patterns. Studies of the ecological and evolutionary processes underlying ecogeographic patterns will increase understanding of the processes generating systematic variation in diversity that led to these rules in nature. The threespine stickleback is an important model within both evolutionary and developmental biology. This study suggests that Jordan’s Rule and Bergmann’s Rule hold in marine stickleback, but a number of questions have been highlighted that require further exploration. What is the functional significance of ecogeographic variation in stickleback? What is the influence of temperature and what temperatures do stickleback experience during their lives? Does population genetic structure conform to the phenotypic patterns observed here? This latter question is the topic of the next chapter.

Contemporary ancestor? Adaptive divergence from standing genetic variation in Pacific marine threespine stickleback

4.1 Introduction

The ecological theory of adaptive divergence predicts that populations diverge phenotypically and genetically if they reside in distinct environments (Bernatchez 2004; Rundle and Nosil 2005), potentially resulting in speciation. Some of the most striking examples of adaptive divergence come from species in which contemporary populations persist in environments putatively occupied by ancestral populations (e.g. Berry et al. 1978; Losos et al. 2000; Hendry 2001; Des Roches et al. 2011; Domingues et al. 2012; Cahill et al. 2013; Yavno and Fox 2013; Morris et al. 2014; Dennenmoser et al. 2017). To the extent that such population have not undergone evolution, contemporary descendants of populations in ancestral environments can be used as a proxy for the ancestor, with phenotypic and genetic differences between these “contemporary ancestors” and “derived” populations being used to infer the direction, source, and pace of adaptation to derived environments. However, recent changes in the ancestral environment may stimulate evolutionary responses in the contemporary populations that occupy it, complicating their utility as a proxy. Standing genetic variation (SGV), defined as the variety of alleles segregating in population (Przeworski et al. 2005; Barrett and Schluter 2008), is expected to play an important role in parallel evolution. In particular, SGV permits rapid adaptation compared to de novo mutation, and increases the likelihood that the same beneficial allele will be present in different derived populations (Elmer and Meyer 2011; Hendry 2013; Peichel and Marques 2017). The role of SGV in adaptive divergence is readily measurable: if an allele fixed in the derived population is present in the contemporary ancestor at low frequencies, it likely contributed to adaptation (Barrett and Schluter 2008). However, the inference that an allele present in the contemporary ancestral population resulted in adaptation via SGV requires three assumptions. (1) The subset of individuals that originally colonised the derived environment must have contained the rare adaptive allele at some frequency; otherwise it arose from de novo mutation or

subsequent gene flow. (2) The contemporary ancestor has undergone little evolution, including gene flow from the derived population (e.g., Schluter and Conte 2009). (3) The ancestral population has to have been properly characterized (i.e., population structure and allele frequencies associated with SGV). If there are multiple potential ancestral populations, each with a different pool of SGV, inferences about the source and pace of evolution in the derived population will vary depending on which putative ancestral population is investigated (e.g., Przeworski et. al 2005). These assumptions must be verified to characterize accurately the role of SGV during population divergence. Threespine stickleback (Gasterosteus aculeatus) provide perhaps the best documented examples of adaptation from SGV. Marine threespine stickleback occur widely in the northern hemisphere, including along the Pacific coast of North America from Alaska south to southcentral California. Across the north Pacific coast, much freshwater habitat formed recently (~10 000 – 20 000 years ago) in association with isostatic rebound following glacial retreat. Subsequent colonisation of this habitat by stickleback allows tests of the significance of de novo mutation and SGV for adaptation (e.g. Colosimo et al. 2005; Schluter and Conte 2009; Chan et al. 2010). For instance, marine stickleback bodies are often covered by >29 bony lateral plates, but fewer plates (0-10) have evolved in parallel in freshwater populations through selection on a rare marine allele (Colosimo et al. 2005). Despite numerous studies indicating the role of SGV at either a single locus for platedness (Eda), or for multiple loci with unknown phenotypic effects (Hohenlohe et al. 2010; Jones et al. 2012ab), assumptions about the appropriateness of considering extant marine sticklebacks as representative of the ancestors of freshwater populations remains untested. Despite evident genetic variation in threespine stickleback among geographic clades (Ortí et al. 1994; Mäkinen and Merilä 2008; Lescak et al. 2015), marine stickleback on the eastern Pacific coast are largely assumed to constitute a single population. This assumption it justified by the absence of barriers to gene flow in the marine environment, the migratory capacity of marine stickleback (Williams and Delbeek 1989), and the low marine population structure reported from several local studies (Withler and McPhail 1985; Hohenlohe et al. 2010, but see Catchen et al. 2013a). Nevertheless, substantial evidence indicates local

adaptation even in highly migratory marine fishes (Hemmer-Hansen et al. 2014; Carvalho et al. 2016; Machado et al. 2017), and indeed in Baltic Sea threespine stickleback (DeFaveri et al. 2013ab; DeFaveri and Merilä 2013). If Pacific threespine stickleback constitute a single population, their large effective population size should limit the effects of genetic drift and high gene flow should offset local adaptation (Bernatchez 2016). Given these conditions, marine stickleback should not exhibit local differentiation that would generate regional differences in the initial colonists of freshwater lakes and streams. SGV should be the same along the Pacific coast - and all freshwater populations could have evolved from the same initial pool of marine SGV facilitating parallel adaptation. These assumptions require formal testing to elucidate the role of SGV during adaptive divergence. In this study, I consider phenotypic and genotypic variation of > 200 marine threespine stickleback from eight locations from California to Alaska to test hypotheses about the genetic structure of marine stickleback and its evolutionary consequences. Based on variation in plate phenotypes and genotypes associated with SGV at Eda, three- dimensional body morphology from micro-computed tomography (μCT) scans quantified using geometric morphometrics, and Genotype-by-Sequencing (Elshire et al. 2011), I assess whether marine stickleback along the Pacific coast constitute a single population. By doing so, I test assumptions about the distribution of SGV in “contemporary ancestors” of freshwater stickleback. Additionally, I test predictions regarding the influence of SGV on the source and pace of adaptation based on genomic sequences for a freshwater population from British Columbia. I specifically test the following predictions. (1) Marine stickleback harbour SGV. (2) As with other marine fish, marine stickleback exhibit population genetic structure, which may influence the SGV available for selection. (3) Marine populations vary in effective population size, with larger populations harbouring more SGV. (4) Marine populations exhibit phenotypic divergence for body shape, but not for platedness. (5) Differences in SGV among populations show evidence of having been shaped, at least in part, by natural selection. (6) Geographic proximity to a freshwater population determines the extent of genetic divergence between marine and freshwater stickleback. (7) Differences in SGV among marine populations

determine the loci that contribute to adaptation in freshwater stickleback populations.

4.2 Materials and Methods

4.2.1 Sex identification

Stickleback were collected as outlined in Chapter Two, including stickleback from eight marine locations and one freshwater location (Brannen Lake, BC). Sex was identified using primers developed by Peichel et al. (2004) that amplify sex-specific alleles at the idh locus. Alleles were visualized in a 2% agarose gel for 367 individuals. A total of 205 males and 162 females were sampled (see Chapter Three). All three BC01 females were included in the sequencing run (see 4.2.2). Otherwise, individuals were chosen randomly for sequencing (Table B-2).

4.2.2 Library preparation and analysis

In order to assess population structure, a reduced representation method was used to yield DNA for sequencing. 200 ng total genomic DNA was extracted per fish in January 2016 using Qiagen DNeasy Blood and Tissue kits (n = 265) and digested with EcoRI and MseI restriction enzymes. 30-35 individuals were included from each marine location, and 15 from Brannen Lake, BC (Table 4-1). After digestion-ligation (Appendix D), fish were pooled into groups of nine. Cleanup and size selection were performed simultaneously using SPRI beads (Beckman Coulter), at a bead ratio of 0.8x and 0.61x for left and right-side cleanup, respectively. This left a fragment range between 250 and 600 bp. Pooled samples were divided into three replicates to ameliorate stochastic differences during PCR and were amplified. Replicates were pooled and left-side cleaned using SPRI beads. Pooled samples were quantified using a 2200 TapeStation (Agilent) and Qubit (Thermofisher) dsDNA high sensitivity assay. Equal volumes of each 2 nM pooled sample were pooled to make the final library. Library preparation followed the Illumina protocols for the Illumina NextSeq 500 Mid-Output kits with version 2 chemistry. Two sequencing runs were completed on the Illumina Next Seq 500 using 150

Table 4-1 - Numbers of stickleback used for plate counts, 3-D morphometrics and included in the sequencing run after process_radtags filtering.

Site Plate counts 3-D morphometrics Sequenced CA01 35 35 29 CA02 48 48 28 CA03 46 44 28 OR01 19 20 29 OR02 50 48 30 WA01 0 0 25 BC01 51 48 31 AK01 31 29 24 BCFW NA NA 15

cycles and different final library concentrations - the first at 1.8 pM final concentration, the second at 1.1 pM final concentration. A 20% PhiX spike-in was used for both runs to compensate for the low diversity nature of the library. Results from the two sequencing runs were merged. Initial population statistics were performed on cleaned sequenced reads using Stacks (Catchen et al. 2011, 2013b). Single Nucleotide Polymorphisms (SNPs) were called using a bounded SNP model with an upper error rate of 0.1 and stringent filtering criteria: i) log likelihood threshold > -60; ii) sequenced in more than 75% of individuals; iii) in 6 of 8 populations; iv) 4x coverage; v) minor allele frequency of 2%, and vi) FIS > - 0.3. Individuals were included if they retained > 10 000 RAD-loci after cleaning (n = 239, Table 4-1). GSnap (Wu and Watanabe 2005) was used to align reads to the stickleback genome (Ensembl release 72, Flicek et al. 2014), allowing for five mismatches with soft masking disabled. All population statistics except for FST were calculated using the populations module of Stacks.

4.2.3 Population genetic structure

Pairwise global and per-locus FST were calculated using the Weir and Cockerham (1984) adaptation implemented in hierfstat v0.04-22 (Goudet 2005) in R (R Core Team 2016). Discriminant Analysis of Principal Components (DAPC) (Jombart et al. 2010) was used to assess population structure using Adegenet v2.0.1 (Jombart 2008) (Appendix A). The optimal number of Principal Components (PCs) to retain was calculated using both xvalDAPC and a-scores, which gave similar answers (Appendix A). The optimal number of clusters was assigned based on the lowest Bayesian Information Criteria (BIC) score of k-means clustering. As several possible k clusters had similarly low BIC scores, analyses were run and compared using 3 to 8 clusters. An analysis of molecular variance (AMOVA), implemented in poppr v2.3.0 (Kamvar et al. 2014, 2015) using the Ade4 package (Dray and Dufour 2007), was used to determine the proportions of genetic variance among versus within sampling sites or Adegenet-recognized clusters. Missing values were replaced with the average frequency for a locus; ignoring missing values did not alter overall patterns. To explore the 89

possibility of cryptic population structure, each sampling site was further analyzed individually using Stacks and Adegenet. The distance between each sampling site was measured as distance along the coast using Google Maps. Distances were measured to or from the mouth of each bay. Neighbouring localities were separated by 242-479 km, except for BC01-AK01, which were separated by approximately 2500 km of coastline. The location of WA01 in Puget Sound resulted in all locations south of Washington being closer to BC01 than they were to WA01. Genetic distance was calculated using the pairwise global Weir and

Cockerham estimates of FST from hierfstat. Geographic and genetic distance matrices were compared using a Mantel test from the Adegenet package with 999 replications to determine Isolation-by-Distance (IBD). Population statistics were also estimated for the full dataset in Stacks using the optimal Adegenet-identified clusters, rather than marine sites. A phylogenetic network was calculated using SNPhylo (Lee et al. 2014) and visualized using FigTree v.1.4.3 (Rambaut 2007). Individuals were colour-coded according to their recognized genetic cluster.

4.2.4 Effective population size

Effective population size (Ne) was estimated using NeEstimator v2.01 (Do et al.

2014), which implements the LDNe algorithm (Waples and Do 2008). Minor alleles were filtered at 5% frequency. Because SNPs on a single RAD-locus are likely to be linked, only the first SNP present on each RAD-locus was used. To assess the association between Ne and SGV among populations, I estimated Pearson correlations between average Ne and measures of genetic diversity: the numbers of SNPs and private alleles, observed heterozygosity, or nucleotide diversity.

4.2.5 Platedness

Variation among populations was assessed using plate number and Eda genotype. Adult stickleback (i.e., fish > 30 mm standard length) (n = 280, Table 4-1) were stained

in Alizarin red. Plate number, including keel, was counted on both sides of the body and summed. Low-plated individuals without keel (LPNK) were defined as individuals with < 20 anterior plates. Partially-plated keeled (PPK) fish had 21 – 59 plates, including at least one plate at the caudal keel. Fully-plated keeled (FPK) fish had  60 plates. Additionally, some low-plated fish had a keel (LPK) and were defined as having < 20 anterior plates plus additional plates at the caudal keel. Partially plated stickleback that lacked a keel (PPNK) had > 20 anterior plates but no plates at the caudal keel. Individuals were also genotyped at the Stn382 locus (Colosimo et al. 2005) as this microsatellite is linked to an indel in intron 1 of the Eda gene, yielding a 218 bp “fully- plated” allele (C) or a 158 bp “low-plated” allele (L) (Lucek et al. 2012). Genotyping followed the protocol of Peichel et al. (2004). Individuals were genotyped as LL (homozygous for the low-plated allele), CL (heterozygous), or CC (homozygous for the fully-plated allele). This approach allowed juveniles (< 30 mm SL) to be included in the analysis (total n = 361), and provided genetic information at a locus with known adaptive significance that was not recovered from sequencing. Hardy-Weinberg equilibrium (HWE) was assessed for Stn382 for each marine site using a goodness-of-fit Chi-square test.

4.2.6 Morphometrics

Phenotypic variation among populations was further assessed using morphometric analysis. Stickleback > 30 mm SL that had been preserved with relatively little bending (n = 272) were straightened, and spines and fins held flat against the body using plastic wrap. μCT scanning at a resolution of 20 μm was conducted in a standardized fashion for all individuals using a Scanco μCT35 instrument (Scanco AG). Three-dimensional images were generated from the anterior point of the premaxilla to the posterior tip of the pelvic spine, using standardized isosurface thresholds in Amira 5.4 (FEI Visualization Sciences Group). 55 landmarks were plotted on the left side of each fish using Amira 5.4 (Figure 4-1, Table B-1) and raw landmark scores were exported to MorphoJ v1.06a (Klingenberg 2011) for further analyses. A prior study had removed the operculum on the left side of all AK01 stickleback, so landmarks were plotted on their right sides. Data 91

Figure 4-1 - Positions of 55 landmarks used for the morphometric analysis. See Table B- 1 for identity of landmarks.

were first transformed to remove differences associated with isometric scaling, rotation and translation using Procrustes superimposition. Residuals from a within-marine site multivariate regression on centroid size were estimated and used in all subsequent analyses. Principal Components Analysis (PCA) determined the major axes of phenotypic variation. Canonical Variate Analysis (CVA) was used to determine Procrustes distances using sex and marine site as categorical variables, although BC01 had only one female and OR01 included one individual of unknown sex. The significance of Procrustes distances among pairwise comparisons of marine site-sex combinations was determined based on 10 000 permutations and a corrected α of 0.0005. Discriminant Function Analysis (DFA) was used to determine the likelihood that individuals could be reassigned to their site of origin, given their phenotypes. For this analysis the effect of sex on reassignment success was not assessed.

4.2.7 Selection on phenotypic variation

Patterns of phenotypic variation among populations may be the result of genetic drift, natural selection, or phenotypic plasticity. One way to rule out neutral evolutionary processes is to compare estimates of phenotypic divergence (PST) with neutral expectations based in part on observed neutral genetic differentiation. I estimated observed phenotypic divergence as: 2 2 2 PST = σ B/(σ B + 2σ W), 2 2 where σ B and σ W were the between- and within-population components of variance, respectively, for plate count and the first four PCs from the morphometric analysis (as per Kaeuffer et al. 2012; Ravinet et al. 2013b). Variance components were estimated for all marine sampling sites together (global PST) and pairwise using lme4 (Bates et al. 2015), with population as a random effect. Genetic divergence at the Stn382 locus for Eda (FSTQ) was estimated using the Weir and Cockerham method in Genepop V4 (Rousset 2008).

Neutral genetic divergence (FST) was estimated in hierfstat using non-genic SNPs identified from my dataset using Biomart (Smedley et al. 2015). Non-genic SNPs may still be linked to loci under selection, so this approach provides a conservative estimate of neutrality. 93

Selection was inferred based on two methods. The first assessed the association between PST-FST and FSTQ-FST using Mantel tests. This measure is based on the expectation that phenotypic or QTL divergence will be uncorrelated with neutral genetic divergence – by extension implicating selection to explain such patterns. The second test involved Whitlock and Guillaume’s (2009) method using the R-code from Lind et al. (2011). In brief, the expected between-population variance component for a neutral phenotype was estimated by using observed non-genic FST and the observed within- population variance component for the phenotype: 2 2 σ B = 2FST σ W/(1-FST) 2 2 As per Lind et al. (2011), the distribution of neutral σ B was estimated by generating a χ distribution with six degrees of freedom (one less the number of sampling sites excluding 2 Washington), and multiplying a randomly drawn value from this distribution by σ B. 2 From this new distribution expected neutral σ B were drawn 10 000 times and used to create a distribution of neutral PST-FST. The observed PST-FST was then compared to this distribution and the quantile of the neutral distribution that lay beyond the observed value was used as the probability of the observed outcome in the absence of selection, p. Under the expectation of no selection, p is > 0; selection is evidenced if p = 0. FSTQ-FST was also compared to the neutral PST-FST, as per DeFaveri and Merilä (2013). To quantify the relation between phenotype and Stn382 genotype, I fit a generalized linear model (GLM) in R, using the glm routine, as per DeFaveri and Merilä (2013), with plate number as the dependent variable, genotype as a fixed effect, and using a log-link function with a quasi-Poisson error distribution. Furthermore, a Mantel test was used to estimate the correlation between pairwise FSTQ and PST measures.

4.2.8 Marine-marine genetic divergence

To identify loci with unusually high between-population variation in allele frequencies, per-locus FST estimates were calculated in hierfstat using information from all Adegenet-identified clusters. Loci were included only if they were sequenced in all five clusters and in > 75% of individuals; otherwise the Stacks parameters were as described in 4.2.2. Per-locus FST was estimated for all marine clusters considered 94

collectively, as was global pairwise FST, using the Weir and Cockerham method. Manhattan plots were generated using qqman (Turner 2014). Outlier loci were defined as

the largest 5% of FST values (Narum and Hess 2011). Genic outliers were identified using BioMart (Smedley et al. 2015).

4.2.9 Causes of population structure

To determine the contributions of outliers to population structure, outliers and

non-outliers were assessed separately for pairwise FST, clustering assignment in Adegenet, AMOVA, and IBD, as described in 4.2.2, 4.2.3, and 4.2.8. For this analysis, populations were identified based on sampling location.

4.2.10 Marine-freshwater genetic divergence

To assess the extent to which the choice of putative “contemporary ancestor”

affected inference about adaptation to fresh water I estimated per-locus FST for each of

eight marine-freshwater comparisons, and the minimum and maximum FST per locus was

retained for analysis. The expectation was that minimum and maximum FST should approximately follow a 1-to-1 line if choice of “contemporary ancestor” did not influence the outcome. Outliers were identified for each marine-freshwater pair as per 4.2.8, with the expectation that the same outliers should be recovered from each marine-freshwater

comparison. In order to estimate FST, analyses were conducted as described in 4.2.2 and 4.2.8, with the exception that the number of SNPs was reduced by applying more stringent criteria. For loci to be retained they needed to have been sequenced in all eight marine sites and the freshwater population, and in > 75% of individuals.

4.3 Results

4.3.1 Sequencing results

Over 192 million reads passed initial filters (Table B-3). 282 RAD-loci were excluded due to excess heterozygosity. Filtering minor alleles at a threshold of 2%

substantially reduced the number of retained loci by 32%. After filtering, between 230 010 (AK01) and 426 018 loci (OR01) were retained for each site, generating between 1877 (AK01) and 5204 (OR01) SNPs (Table 4-2).

4.3.2 Standing genetic variation

All marine samples exhibited SGV, ranging from an average of 0.82% (AK01) to 1.22% (OR01) of the sequenced genome; however, the pool of SGV varied from California to Alaska (Table 4-2). Samples from each marine site included multiple private alleles (alleles found only at that location) (Figure B-1) and were polymorphic for a portion of the variant loci (loci that were polymorphic in at least one marine site). Polymorphism among variant loci varied from 57% (AK01) to 80% (OR01). For variant loci, the average frequency of the major allele (present in > 50% of all sequenced stickleback) ranged from 88% (OR02) to 93% (AK01) (Figure B-2), suggesting that the frequencies of SGV also differed among locations. OR01 and OR02 had qualitatively distinct major allele frequency distributions relative to the other marine sites, with fewer high-frequency alleles (Figure B-2). Heterozygosity ranged from 0.10 (AK01) to 0.17

(OR02) for variant loci. Population average FIS over all variant loci varied from 0.032

(CA01) to 0.064 (OR01) (Table 4-2). Although FIS was close to 0 for most loci, it approached 1 for a few loci (Figure B-3).

4.3.3 Population genetic structure

All pairwise comparisons of global FST significantly exceeded 0, and ranged from

0.020 to 0.181 (Table 4-3). Pairwise FST between the northern marine groups (WA01, BC01, and AK01) were all small (< 0.05), although other comparisons showed moderate (0.05 - 0.15), and three showed great differentiation (0.15-0.25). Significant population genetic structure was detected. The best-supported number of clusters from the eight marine groups sampled was five (BIC = 1379, Figure 4-2, Table B-4). The five clusters were, from south to north, CA01, CA02, CA03 & OR01, OR02, and WA01 & BC01 & AK01. The CA03 & OR01 cluster also contained seven

Table 4-2 - Population genetic statistics for the filtered dataset of marine stickleback for (top) all variant loci and (bottom) all sequenced loci. N = average number of individuals used; Loci = average number of sequenced loci; Variant = number of sequenced loci polymorphic in at least one marine site; SNP = number of Single Nucleotide Polymorphisms; % Poly = proportion of variant loci (top) or sequenced loci (bottom) that were polymorphic. See the List of Symbols, Abbreviations and Nomenclature for definitions of other abbreviations.

% Pop Private N P HetO HO HetE HE π FIS Poly CA01 131 25.46 56.32 0.905 0.131 0.869 0.138 0.862 0.141 0.032 CA02 103 25.02 69.39 0.901 0.141 0.859 0.149 0.851 0.153 0.041 CA03 8 23.57 68.84 0.906 0.130 0.870 0.141 0.859 0.144 0.050 OR01 17 25.57 80.42 0.897 0.145 0.856 0.158 0.842 0.162 0.064 OR02 67 26.56 79.52 0.877 0.166 0.834 0.178 0.822 0.181 0.055 WA01 4 21.74 65.57 0.915 0.119 0.881 0.130 0.871 0.133 0.049 BC01 25 27.43 62.22 0.916 0.115 0.886 0.125 0.875 0.128 0.048 AK01 1 20.00 57.09 0.927 0.103 0.897 0.111 0.889 0.114 0.048

Pop Loci Variant SNP % Poly N P HetO HO HetE HE π FIS CA01 407925 6124 3449 0.846 25.9 0.9986 0.0020 0.9980 0.0021 0.9979 0.0021 0.0005 CA02 422173 6406 4445 1.053 25.4 0.9985 0.0021 0.9979 0.0023 0.9977 0.0023 0.0006

CA03 398661 5812 4001 1.004 23.9 0.9986 0.0019 0.9981 0.0021 0.9979 0.0021 0.0007 OR01 426018 6471 5204 1.222 26.0 0.9984 0.0022 0.9978 0.0024 0.9976 0.0025 0.0010 OR02 410111 6216 4943 1.205 26.9 0.9981 0.0025 0.9975 0.0027 0.9973 0.0027 0.0008 WA01 395684 5963 3910 0.988 21.9 0.9987 0.0018 0.9982 0.0020 0.9980 0.0020 0.0007 BC01 400271 5972 3716 0.928 27.7 0.9988 0.0017 0.9983 0.0019 0.9981 0.0019 0.0007 AK01 230010 3288 1877 0.816 20.2 0.9990 0.0015 0.9985 0.0016 0.9984 0.0016 0.0007

Table 4-3 - Pairwise geographic distances (km: above diagonal) and global pairwise

Weir and Cockerham FST (below diagonal). Note that all FSTs are significantly greater than 0.

CA01 CA02 CA03 OR01 OR02 WA01 BC01 AK01 CA01 242 584 885 1137 1616 1532 4037 CA02 0.096 342 643 895 1374 1291 3795 CA03 0.110 0.059 301 553 1032 948 3453 OR01 0.121 0.063 0.044 252 731 648 3152 OR02 0.181 0.134 0.100 0.082 478 395 2900 WA01 0.146 0.082 0.079 0.045 0.142 207 2711 BC01 0.157 0.095 0.094 0.056 0.159 0.027 2505 AK01 0.096 0.052 0.058 0.032 0.092 0.020 0.022

Figure 4-2 - Adegenet-identified clusters for k = 5. Inset shows hypothetical range of each cluster. Note that the cluster identified as CA03, OR01 contains one AK01 and seven OR02 individuals.

100

individuals from OR02 and a single individual from AK01; otherwise individuals clustered with others from their sampling locality. The possibility of a single genetic cluster was as well-supported as ten genetic clusters (BIC = 1392). Three to six clusters had BIC values that differed little (difference BIC < 2) from the best-supported model. Altering the number of putative clusters revealed different population structures (Table B-4), with WA01 & AK01 continuing to cluster until k = 13. Cryptic population structure was not evident for any sampling locality (Table B-5). Even when WA01, BC01, and AK01 were included in a single analysis, k = 1 was the best supported cluster (BIC = 436.9). However, k = 2 (BIC = 438.2) and k = 3 (BIC = 440.2) still separated individuals by locality. The genetic variance partitioned between clusters by AMOVA was low (19%) compared to within clusters (81%). When clusters were not considered, 33% of variation occurred between marine sites. A Mantel test of pairwise geographic distances and pairwise Weir and Cockerham FST was non-significant when all sites were included (r = - 0.2, p = 0.8). This was largely driven by the extreme distance between the WA01, BC01, and AK01 sites. If AK01 is excluded from the analysis the association between geographic and genetic distance was weakly significant (r = 0.5, p = 0.02) (Figure B-4). A total of 4299 variant loci were sequenced in all Adegenet clusters, with 2441 (CA01) to 3869 (CA03 & OR01) SNPs per cluster. The CA01 sample included the most private alleles (n = 80), but the cluster of WA01 & BC01 & AK01, which individually had few private alleles (1 to 25), now had 47 (Figure B-5, Table B-7). The proportion of variant loci that were polymorphic within a cluster varied from 57% (CA01) to 90% (CA03 & OR01). Observed heterozygosity for variant loci varied from 13% (CA01) to

17% (OR02). FIS was lowest in southern California and highest in CA03 & OR01 (Table B-7). The phylogenetic network based on SNPs that passed Linkage Disequilibrium filtering largely agreed with Adegenet assignment (Figure 4-3). The network revealed greater intermixing of groups than did Adegenet, but WA01, BC01, and AK01 still largely clustered together and comprised a separate lineage from most southern stickleback. CA01 and CA02 constituted distinct lineages. Most OR02 individuals

101

Figure 4-3 - Result of the phylogenetic analysis using SNPhylo. Individuals are colour- coded according to their five Adegenet-recognized clusters (Figure 4-2).

102

appeared to be derived from the CA03 & OR01 clade, with 75% bootstrapping confidence.

4.3.4 Effective population size

Effective population sizes varied among marine sites (Table 4-4). CA02 had the smallest Ne of 19.7. The largest populations occurred in the north, with Ne > 1000 for each of WA01, BC01, and AK01. The estimates for AK01 were bounded by infinity. Ne for the remaining sites varied from 350 (OR02) to 615 (OR01). Effective population size did not correlate with the number of SNPs or private alleles (p > 0.05), but Ne varied weakly negatively with observed heterozygosity (Pearson’s correlation: t = -2.7, d.f. = 6, r = -0.74, p = 0.03) and nucleotide diversity (Pearson’s correlation: t = -2.4, d.f. = 6, r = - 0.71, p = 0.05).

4.3.5 Platedness

Fish sampled from each site differed in the frequencies of plate morphs (Table B- 2). FPK morphs comprised 100% of samples from BC01 and OR01. Four LPNK individuals were sampled from AK01, with the rest being FPK. All other sites were at least trimorphic for LPNK, PPK, and FPK. California in particular had high frequencies of LPNK stickleback, comprising 77% of samples. Five individuals from OR02 and CA01 exhibited the rare LPK morph, and a single individual from OR02 was a PPNK morph. Juvenile and adult morphs could be estimated using Stn382 genotypes (Figure 4- 4, Table B-2). Only 2 of 50 WA01 individuals were heterozygous CL; the remainder were CC. Among juvenile OR01 there was a single LL, 11 CL, and 17 CC individuals. Furthermore, although all OR01 adults were FPK, six of these were CL heterozygotes (see Chapter Five). All polymorphic groups were in Hardy-Weiberg equilibrium for Stn382, except for AK01 (observed 24 CC, 0 CL, 3 LL, expected 21.3 CC, 5.3 CL, 0.3 LL; Chi-Square test: 1 d.f., p = 0) (Table B-2).

103

Table 4-4 - Effective population sizes of marine threespine stickleback at sampled sites. The harmonic sample size, the observed and expected correlations between SNPs, and the parametric confidence intervals (CI) are included.

2 2 Site Harmonic N Overall r Expected r Ne CI CA01 22.2 0.05131 0.05064 458.4 370.0 - 600.9 CA02 22.1 0.06523 0.05096 19.7 19.5 - 20.0 CA03 20.3 0.05764 0.05620 211.8 188.2 - 242.0 OR01 22.7 0.04980 0.04930 613.6 496.3 - 802.1 OR02 24.1 0.04718 0.04629 347.3 308.1 - 397.7 WA01 19.4 0.05932 0.05912 1526.1 778.5 - 32 859.3 BC01 24.4 0.04590 0.04562 1121.3 712.4 - 2608.5 AK01 17.2 0.06813 0.06793 1560.6 473.3 - ∞ BCFW 12.8 0.09767 0.09763 7298.7 3342.9 - ∞

104

Figure 4-4 - Frequency distributions of Eda genotypes using the Stn382 marker for each sampling site. “The North” refers to samples from Washington, British Columbia, and Alaska. CC = homozygous for the fully-plated allele. LL = homozygous for the low- plated allele. CL = heterozygous.

105

4.3.6 Morphometrics

Phenotypes varied extensively among sites. The first eight Principal Components explained 71% of all phenotypic variance (Figure B-6). The first two Canonical Variates (CVs) explained 57% of the variation among combinations of site and sex, and the first four CVs explained 77% (Table B-8). CV1, after accounting for differences in centroid size, revealed that BC01 fish had narrow, streamlined bodies with dorsal and ventral landmarks both shifting inward relative to the consensus fish (Figure 4-5, Figure 4-6). Californian fish were grouped close together on CV1 and had squatter, less streamlined bodies with dorsal and ventral landmarks shifted away from one another relative to the consensus. AK01, OR01 and OR02 had intermediate phenotypes between BC01 and California. CV2 showed a gradual transition from CA01 to BC01, but AK01 was clearly distinct from all other sites along this axis. AK01 showed substantial dorsolateral and anterior-posterior constriction of the body relative to all other sites (Figure 4-5, Figure 4- 6). The sexes differed morphologically at all sites except AK01 (Table B-9), but the sexes still largely grouped together according to sampling location. CA02 and OR02 were exceptions, with males from both sites clustering with OR01 males. Similarly, CA01 and CA02 females had morphologies that were not significantly distinct. The DFA revealed that most fish could be classified according to marine site (based on Procrustes distance, p < 0.001 for all pairwise comparisons), with a single fish misclassified (Table B-10). Cross-validation misclassified an average of 2.8 fish per pairwise comparison (n = 59 total misclassifications), but this varied from 0 to 7 (CA02 – OR02), 8 (CA03 – OR02), 9 (CA01 – CA02), and 10 (CA02 – CA03). Only three fish were misclassified when comparing sites from within an Adegenet-recognized cluster. Thus, most misclassifications occurred among, rather than within, genetic clusters.

106

Figure 4-5 – Among-group variation from a 3D morphometric analysis, for (a) Canonical Variate (CV) 1 vs. CV2 and (b) CV3 vs. CV4 for body shape.

107

Figure 4-6 - Wireframes of stickleback oriented (left) left laterally, showing the head to the anterior tip of the pelvic spine, and (right) superiorly. (A) CV1 for a BC01-type body shape; (B) CV1 for a CA01-type body shape; (C) CV2 for an AK01-type body shape; (D) CV2 for a BC01-type body shape. Light blue wireframe shows the consensus morphology, whereas dark blue shows the conformational change.

108

4.3.7 PST-FST and FSTQ-FST comparisons

PST was estimated as 0.46 for platedness, 0.60 for PC1, 0.29 for PC2, 0.23 for

PC3, and 0.09 for PC4. FSTQ was 0.60. Plate PST and FSTQ greatly exceeded the range of the neutral PST-FST distribution (p = 0 for both), as did PST for PC1 (p = 0). PST for PC2 was marginally significant, but within the tail of the neutral distribution (p = 0.002), whereas PST for PC3 (p = 0.02) and PC4 (p = 0.7) were well within the neutral distributions (Figure 4-7).

Of all of the Mantel tests between PST, FST, FSTQ, and geography, only one was significant and two were marginally significant, with corrected α = 0.05/18 = 0.0027:

PC1 PST correlated positively with geographic distance, plate count PST correlated positively with FSTQ, and, surprisingly, PC1 PST correlated positively with FSTQ (Figure

B-7, Figure B-8, Table 4-5). The relation between plate PST and FSTQ was substantiated with a generalized linear model which showed a decrease in plate number with number of L alleles (null deviance = 4813 with 277 d.f., residual deviance = 349 with 275 d.f., p < 0.0001, Hosmer and Lemeshow goodness of fit text: χ2 = -5x10-26, d.f. = 8, p = 1).

4.3.8 Outlier analysis for marine-marine comparisons

Based on the five Adegent-recognized clusters, 4299 loci were identified in marine stickleback for which global FST could be estimated. The top 5% (n = 215) were flagged as outliers (Figure 4-8). The average FST for outlier loci was 0.47 (range 0.37 - 0.65). 198 outliers were from known linkage groups (LG), the remainder were from scaffolds. The number of outliers varied strongly and positively with the number of ln- transformed SNPs identified on a LG or scaffold (glm with Poisson distribution, null deviance = 785 with 72 d.f., residual deviance = 220 with 71 d.f., p < 0.0001, Figure B- 9). Even so, > 12% of SNPs from linkage groups IV, VII, and XIX were outliers, whereas VI, XIV, XV, XVII, and XVIII had no outliers despite > 100 SNPs each. 47 outliers occurred within introns or exons of 29 known genes (Table 4-6). The Gene Ontologies of these genes included 22 distinct biological processes, 5 distinct cellular components, and 21 distinct molecular functions. Most GO terms were

109

Table 4-5 - Observed correlations and probabilities expected from sampling error for the null hypothesis (p) for Mantel tests between geographic distance, neutral genetic distance

(FST), phenotypic distance (PST – for plates or the first four Principal Components (PCs) of body shape), or genetic distance at Eda (FSTQ). Bold indicates significant correlations ( = 0.0027).

Comparison Observed p correlation

Geography – neutral FST -0.29 0.7

Geography – plate number PST 0.005 0.5

Geography – FSTQ 0.26 0.1

Geography – PC1 PST 0.86 0.003

Geography – PC2 PST 0.003 0.21

Geography – PC3 PST -0.03 0.5

Geography – PC4 PST -0.31 0.95

Neutral FST – plate number PST 0.14 0.3

Neutral FST – FSTQ 0.14 0.2

Neutral FST – PC1 PST -0.17 0.7

Neutral FST – PC2 PST 0.02 0.4

Neutral FST – PC3 PST -0.3 0.9

Neutral FST – PC4 PST -0.04 0.5

FSTQ – Plate count PST 0.9 0.002

FSTQ – PC1 PST 0.5 0.025

FSTQ – PC2 PST 0.3 0.09

FSTQ – PC3 PST 0.02 0.4

FSTQ – PC4 PST -0.01 0.5

110

Table 4-6 - Genic loci flagged as FST outliers in the marine environment. LG = Linkage

Group. FST provides an average if N SNPs > 1.

Ensembl Gene ID LG Gene name N SNPs FST (ENSGACG000000) 10697 I ncam1b 3 0.488 15004 II mapk6 1 0.493 15375 II 1 0.493

15469 II 1 0.493

15576 III KCNN1 (1 of many) 1 0.491 17510 IV enox2 (1 of many) 1 0.493 17592 IV si:ch211-114c17.1 1 0.492 18231 IV abcb7 2 0.493 18337 IV unc5a 5 0.494 19242 IV cog5 1 0.473 19472 IV nav3 1 0.468 20061 VII S100P 2 0.467 20158 VII SIM2 2 0.470 20236 VII TSPOAP1 3 0.461 20379 VII rb1 1 0.469 20525 VII smyd4 1 0.472 7733 VIII CSNK1G2 (1 of many) 1 0.468 5895 X gabbr2 1 0.474 5403 XI dus1l 3 0.478 3535 XII OVGP1 (1 of many) 3 0.458 10752 XII cd63 1 0.440 11705 XVI 1 0.430

5787 XIX CCDC148 1 0.437 8182 XX cpvl (1 of many) 1 0.427 898 scaffold_27 GRM2 (1 of many) 2 0.433 1222 scaffold_27 2 0.444 1303 scaffold_27 mst1 1 0.443

111

Figure 4-7 - Expected neutral distributions of PST-FST and observed PST-FST for (a) platedness, including FSTQ-FST for the Eda allele, and principal components for body shape, including (b) PC1, (c) PC2, (d) PC3, and (e) PC4.

112

Figure 4-8 - Manhattan plot showing 5% FST outliers (above horizontal line) for linkage groups I through XXI, as well as several scaffolds, for eight marine populations.

113

represented by one or two genes, but proteolysis (n = 6), integral component of membranes (n= 8), and protein binding (n = 11) were most highly represented in each of the respective GO domains. Four outlier non-genic SNPs were located within the LGIV linkage block that differentiates marine from freshwater stickleback (e.g. Jones et al. 2012b).

4.3.9 The effect of outliers on population structure

3695 non-outlier and 211 outlier loci were recovered from the original marine site-specific dataset. As outlier loci were identified based on their extreme global FST values, it is not surprising that pairwise FST values were higher for outlier loci (average 0.34) than non-outlier loci (average 0.07) (Table B-11). The Adegenet-recognized cluster

CA03 & OR01 had moderate differentiation based on outlier loci (FST = 0.115) but low based on non-outlier loci (FST = 0.041). The “northern” cluster of WA01, BC01, and

AK01 had much lower FST values for both outlier (0.00, 0.016 and 0.041) and non-outlier (0.022, 0.024, 0.027) loci than all other comparisons. These differences were also reflected in population structure (Figure B-10). Using non-outlier loci, Adegenet assigned fish to the same three to five clusters as for the overall analysis (Figure B-10). However, for only outlier loci 8-15 clusters had similarly low BICs (BIC = 606-610), and k = 5 was comparatively less supported (BIC = 619). However, eight clusters did not resolve into the eight sampling sites (Figure B-10). AMOVA using outlier loci estimated 34% of variance between the five genetic clusters, and 43% between marine sites. For non-outliers 12% of the variance occurred between the five genetic clusters and 25% between marine sites. Isolation by distance was evident for both outliers and non-outliers (Figure B-10), but only if AK01 was removed from the analysis (with AK01 outliers r = -0.1, p = 0.6, non-outliers r = -0.2, p = 0.7; without AK01: outliers r = 0.6, p = 0.005, non-outliers r = 0.5, p = 0.003).

114

4.3.10 Outlier analysis for marine-freshwater comparisons

132 415 loci were sequenced in all eight marine and one freshwater population. As expected, marine-freshwater divergence was high and in all but one instance was >

0.25 (“very great” differentiation). The lowest pairwise FST values involved the Oregon sites (FST(BCFW, OR01) = 0.27, FST(BCFW,OR02) = 0.18) (Table B-12). Consistent with the extent of SGV, the BCFW sample contained only 12 private alleles, so that most of its alleles also occurred in at least one marine population. Consistent with adaptation from a subset of variants or strong directional selection in fresh water, BCFW had fewer polymorphic loci: only 46% of its variant loci were polymorphic as opposed to 56%-79% in marine populations. 1912 loci were polymorphic in at least one sampled locality (Table B-12). 1852 of these produced FST values for more than one marine-freshwater comparison (Table B-13). If all marine samples were approximately genetically equivalent, BCFW would be differentiated similarly from all other marine locations at individual loci. In contrast, the maximum difference in pairwise FSTs for individual loci was 0.991, whereas the average difference was 0.19 (Figure 4-9). 314 (17%) SNPs had minimum FST estimates of little genetic differentiation (< 0.05) in at least one marine-freshwater contrast, but great genetic differentiation (> 0.25) in another. Each marine-freshwater comparison had a distinct set of outlier loci (Figure B-11 through Figure B-18, Table 4-7). 7% (n = 11) of outlier loci were outliers in all pairwise comparisons, whereas 31% (n = 50) were outliers in only single marine-freshwater comparisons (Table 4-7). The above differences are partly explicable by the dissimilar divergence among marine-freshwater comparisons. Some marine populations were highly diverged from the freshwater population, and so the threshold FST above which a SNP was considered an outlier was high; in others this threshold was qualitatively lower. For instance, CA01,

WA01, BC01, and AK01 had all outliers above FST > 0.8. In contrast, OR02 was less genetically diverged from BCFW, leading to a minimum FST for an outlier of only 0.464. This leads to the interesting situation whereby a locus would be identified as an outlier in the OR02-BCFW comparison only, even though it was associated with higher FST in 115

Table 4-7 - Genic loci flagged as FST outliers in at least one marine-freshwater comparison. Min and Max refer to the smallest and largest FST values for a locus over all eight marine-freshwater comparisons; FST could not be estimated for six comparison involving the cog5 locus because of a lack of polymorphism between or within those marine-freshwater pairs. Outlier in…indicates the marine-freshwater comparison for which the locus was an outlier.

LG Gene name Min Max Mean Outlier in… I si:ch73-125k17.2 0.053 0.809 0.671 CA01, CA02, CA03 II gpc5a 0.431 0.902 0.820 All but OR02 II 0.199 1.000 0.879 All but OR02 II TUB 0.571 1.000 0.919 All IV trmt112 0.128 0.791 0.347 CA02 IV tenm1 0.253 1.000 0.623 WA01, AK01 IV cog5 0.003 0.504 0.253 OR02 XIX 0.488 0.725 0.648 OR02 XIX dkk2 0.536 0.578 0.550 OR02 VII 0.673 0.719 0.704 OR01, OR02 VII S100P 0.180 0.952 0.567 OR01, BC01 CSNK1G2 (1 to VIII many) 0.209 1.000 0.604 WA01, BC01, AK01 XII 0.050 0.976 0.660 CA02, WA01, BC01, AK01 XII lzic 0.541 0.947 0.846 All XVI 0.642 0.975 0.849 All but OR01, BC01, AK01 XVI 0.696 0.804 0.766 CA02, OR01, OR02 XVII I reps1 0.474 0.531 0.507 OR02 CA03, OR01, WA01, BC01, XX 0.366 0.969 0.792 AK01 XX cpvl (1 to many) 0.130 0.858 0.707 CA01, CA02, CA03, WA01

116

XXI RAMP3 0.716 0.854 0.795 CA01, CA03, OR01, OR02 XXI mtmr6 0.376 0.837 0.749 CA02, CA03, OR01 scaffold 27 plch2b 0.400 1.000 0.911 All but OR02

117

Figure 4-9- Per-locus minimum and maximum FST for eight marine-freshwater pairwise comparisons. The 1-to-1 line indicates the expectation if all marine populations are genetically equivalent.

118

other comparisons. Indeed, of the 42 outliers unique to OR02-BCFW, 32 had higher FST estimates in other marine-freshwater comparisons. Outliers were distributed along sixteen of the twenty-one linkage groups and a number of scaffolds. LGIV had the most absolute and fourth most proportional number of outliers of any linkage group (n = 26 outliers, or 37% of LGIV SNPs) (Figure B-11 through Figure B-18). 157 different genes were identified from the 1912 polymorphic loci, of which 22 contained outlier loci (Table 4-7). Only two of these genes were outliers in all eight comparisons (TUB on LGII, lzic on LGXII). The maximum pairwise marine-

freshwater FST for genic loci varied from 0.504 to 1.0 (average 0.851). The minimum varied from 0.003 to 0.716 (average 0.369). The functions of these genes are poorly known, but at least one (dkk2) is a context-dependent agonist/antagonist of the WNT- signalling pathway, which is implicated in plate production (O’Brown et al. 2015).

4.4 Discussion

4.4.1 Marine stickleback exhibit between-population genetic variation

The significance of SGV for parallel evolution depends on its occurrence in the ancestral populations. Marine stickleback harbour SGV, and it differs among populations. Differences in gene expression among two marine BC populations suggested this possibility (Morris et al. 2014), but here it has been quantified across an extensive latitudinal range. The extent of SGV, 0.8% – 1.2% of all sequenced loci, is intermediate to that reported from other studies (Catchen et al. 2013a; Feulner et al. 2013). Nucleotide diversity varied from 0.0016 to 0.0027, consistent with results from Alaska (0.0022 and 0.0025, Hohenlohe et al. 2010) but lower than that reported from Oregon (0.003-0.0036, Catchen et al. 2013a). All marine locations harboured some private alleles, even after ignoring minor alleles at < 2% frequency, suggesting that not just frequencies of SGV but also content of SGV can differ from site to site. Furthermore, the best-known example of SGV, Eda, was present at varying frequencies among populations and is likely under selection in the marine environment. Such variation in the content and frequency of SGV, in turn, led to compelling evidence for population genetic structuring.

119

Marine threespine stickleback showed substantial population structure along the

Pacific coast of North America. Although FST values (average FST = 0.088) were generally lower than those reported for marine-freshwater divergence (e.g. Mäkinen et al. 2006; Hohenlohe et al. 2010; Catchen et al. 2013a), they were higher than those reported for other marine stickleback populations along the North American Pacific coast (two

Alaskan populations: FST = 0.0076, Hohenlohe et al. 2010; three Oregonian populations:

FST = 0.007; Catchen et al. 2013a). However, they align with studies from Europe (Lind and Grahn 2011; DeFaveri and Merilä 2013; DeFaveri et al. 2013b). Five genetic clusters were identified for the eight sampled localities, although structuring was hierarchical, with some clusters more genetically diverged than others. The most widespread cluster occupied > 2700 km of coastline from Washington to Alaska. Marine stickleback from this genetic cluster have been well characterized, with genetic divergence reported to be low between populations separated by up to 1000 km (Withler and McPhail 1985; Taylor and McPhail 1999; Hohenlohe et al. 2010). This led to statements about the relative lack of genetic diversity among marine stickleback globally (e.g. Foster et al. 2003; Raeymaekers et al. 2005). In contrast, results from a broad range suggest that such generalizations should be restricted to this one genetic cluster – and even it contains morphological and genetic differentiation that could be adaptively significant. The southern genetic clusters were sequentially separated by a few hundred kilometres, well within the migratory ability of marine stickleback (Jones and John 1978; Quinn and Light 1989; Cowen et al. 1991). IBD was evident only after removing AK01 from the dataset, suggesting that limited migration could explain patterns of divergence among the southern genetic clusters. However, IBD needs to be interpreted with caution, as geographic distance was correlated with latitude, and latitudinal variation can be associated with environmental clines (e.g. Chapter Three). Whatever the causes that shape genetic variation among stickleback populations, it affects inference about the source and pace of selection in the freshwater environment, and complicates attempts to uncover loci that are under selection in derived populations.

120

4.4.2 Marine stickleback vary in their effective population sizes

The northern cluster had higher Ne than southern clusters. The latter low Ne estimates contradict the assumed large census sizes of marine stickleback (e.g., Bell

1976). DeFaveri and Merilä (2015) similarly reported low Ne for Baltic Sea stickleback populations, (52 - 450 individuals), compared to estimated census sizes in the billions

(Jurvelius et al. 1996). This ratio of Ne/N, which is orders of magnitude lower than the 10% normally observed in wild populations (Waples 2016), could reflect population structuring not reflected in fish counts – in the same manner, for instance, that salmon population counts taken in the open ocean would not reflect actual effective population sizes since salmon return to different rivers to breed. Threespine stickleback are generally not objects of conservation (Foster et al. 2003), but low Ne could affect both the likelihood of local adaptation and local extinction in threespine stickleback (Palstra and Ruzzante 2008; Hare et al. 2011). Curiously, among these Pacific populations, those with larger Ne did not seem to harbour more SGV. Instead, heterozygosity and nucleotide diversity decreased slightly with increasing Ne. The limited variation in the north could reflect the effects of a post-glacial range expansion from the south. However, there are complications with interpreting Ne from genetic point estimates, particularly in marine fishes (Waples 2016), and so these numbers should be interpreted with caution.

4.4.3 Marine stickleback exhibit between-population phenotypic variation

Marine stickleback are generally considered to be fully-plated (Bell 1976), yet the Eda genotype for low-platedness has an ancient marine origin (Colosimo et al. 2005). The low-plated allele is largely believed to exist in the marine environment as SGV only when transported from the freshwater environment (Schluter and Conte 2009) or when masked by marine modifying alleles (Colosimo et al. 2005). If it exists at low frequencies, behaviours that facilitate the movement of low-plated marine stickleback into fresh water are required for the consistent colonisation of rare variants into lakes and streams (Barrett et al. 2009b). I found substantial variation in the frequency of the low- plated allele, to the point that it was the major allele in some Californian and Oregonian

121

populations, but was absent from BC01. This finding is consistent with other records of low-plated marine stickleback in California (Bell 1979; but see Snyder and Dingle 1989; Colosimo et al. 2005) and high frequencies of low-platedness in European marine stickleback (Münzing 1963; Klepaker 1996; Raeymaekers et al. 2005; Lucek et al. 2012; Ferchaud and Hansen 2016). However, it contrasts with the general impression left from other Pacific and Atlantic North American studies that focussed on northern sites (Hagen and Gilbertson 1972; Hagen and Moodie 1981; Cresko et al. 2004; Karve et al. 2008; Kitano et al. 2008; Bell et al. 2010). The observed SGV at Eda could affect the rate at which adaptation to lakes occurred in the past. Indeed, it may explain why reduced plate size has evolved in some freshwater populations, rather than reduced plate number, as an alternative strategy that may have been required in the absence of SGV at Eda (Leinonen et al. 2012; Wiig et al. 2016). Such variation at Eda is a particularly striking reminder that the function of full- platedness in marine stickleback remains unknown. Marine stickleback body shape also varied extensively along the coast. Californian populations tended to have squatter body shapes that appeared to be less streamlined than their northern counterparts. The functional significance of these morphological differences requires testing – but it is interesting that the streamlined- looking fish were from a single genetic cluster. This potentially indicates extensive migration along the northern coast, while the less-streamlined populations were separated by relatively small geographic distances. Jamniczky et al. (2015) reported considerable morphological divergence between neighbouring sampling sites in British Columbia – groups presumably with little to no genetic divergence, implicating plasticity as a driver of morphological variation. However, two related analyses suggested that selection may also play a role in shaping phenotypic divergence. Based on DeFaveri and Merilä’s (2013) method, Pacific coast stickleback exhibited selection for platedness similar to that observed for Baltic Sea stickleback. There was also suggestive evidence for selection on PC1 of body shape, which largely corresponded to CV1 – more streamlined northern fish, more squat southern fish. PC1 was marginally associated with Eda genotype. To my knowledge, this is the first study to

122

suggest pleiotropic or linked effects of Eda on body shape in the marine environment (for marine-freshwater or freshwater-only evidence for pleiotropy, see Albert et al. 2008; for other phenotypes associated with Eda see Barrett et al. 2009ab; Rogers et al. 2012; Mills et al. 2014; Greenwood et al. 2016; Robertson et al. 2017), which could result in low- platedness being an indirect target of selection in some marine habitats. Although Eda is the best-characterized example of SGV, the outlier analysis revealed potential candidate genes for selection in the marine environment. Extensive differentiation was evidenced at some loci. Several of these outlier loci occur in protein- coding genes located within the Eda linkage group that differentiates marine from freshwater stickleback (Jones et al. 2012ab). Smaller bodies and distinct body shapes tend to evolve in freshwater stickleback populations (McGuigan et al. 2011; Bowles et al. 2016), often with significant correlations between morphology and freshwater biotic and abiotic factors (Magalhaes et al. 2016). Given the morphological variation among marine stickleback, elucidating whether freshwater morphology is the result of plasticity, SGV or de novo mutation requires informed decisions about what constitutes the marine ancestor.

4.4.4 Causes of a single northern genetic cluster – speculation

My data suggest an interpretation of the causes of the observed population structure in marine stickleback along the Pacific coast of North America. Local adaptation despite gene flow has been found in other marine fish populations (Knutsen et al. 2003, 2011; Hemmer-Hansen et al. 2014; Saha et al. 2017), including European marine stickleback (Lind and Grahn 2011); but the extent to which stickleback south of Washington exhibited population structure was unanticipated. Outlier and non-outlier loci provided different interpretations of population structure, with the lowest BICs for eight and five genetic clusters, respectively. In particular the outlier loci distinguished CA03 from OR01, which were considered a single cluster using non-outlier loci. Although the non-outlier loci are not necessarily fully neutral, this suggests that drift and selection both play a role in explaining underlying population structure.

123

Drift and selection do not appear to be substantially shaping between-population variation north of Oregon. The northern population, extending from Washington to Alaska, reproduces in coastal habitat that was not available during the last glaciation (Jacobs et al. 2004). Thus, the lack of contemporary population structure in this region may reflect rapid northward post-glacial expansion. Indeed, the occurrence of a Japanese clade in Alaskan and British Columbian waters could also reflect colonisation as suitable coastal environments were formed (Lescak et al. 2015). On the other hand, the morphological data suggest that southern populations are less streamlined in their morphology, and this may cause energetic constraints on migration. The northern population may have limited population structure precisely because they migrate more. These explanations are not necessarily mutually exclusive, as range expansion can also select for better migrators (Shine 2012).

4.4.5 Contemporary ancestors?

The choice of marine stickleback affected inference concerning the source and pace of adaptation in one freshwater population from Vancouver Island. Many studies involve comparisons between geographically proximate marine and freshwater stickleback pairs (e.g. Jones et al. 2012ab), presumably to account for the possibility of population structure in the marine environment. Yet FST was the lowest when the freshwater population was paired with a geographically distant population from northern Oregon, a finding that is difficult to reconcile with the assumption that the nearest marine population is the most suitable ancestral type. The occurrence of Japanese mtDNA haplotypes in Haida Gwaii lake populations that are not present in Haida Gwaii marine populations (Deagle et al. 1996; Johnson and Taylor 2004) suggests that this may not be an isolated incident. Similarly, several studies have noted, but not explained, the fact that northern marine stickleback are genetically more similar to southern than northern freshwater stickleback (Schluter and Conte 2009; Jones et al. 2012ab). Nearly one third of outlier loci were considered candidate genes of interest for only a single marine-freshwater comparison. Thus, one’s choice of marine population could produce spurious inferences about genes of interest, or miss true candidate genes. 124

For instance, had BC01 been the marine population of choice, I would be discussing the possible relevance of gpc5a, TUB, S100P, CSNK1G2, and lzic for adaptation to the freshwater environment. Had I only selected OR02, I would be discussing TUB and lzic as well as cog5, dkk2, and reps1. Of these dkk2 could even be implicated in plate development. Clearly more information is needed going forward about the relationship between marine and freshwater stickleback. Inference concerning the role of SGV during adaptive divergence in ancestral- derived comparisons requires three conditions to be met, for which I have provided varying evidence. The first, that colonists had to contain the same variants as present in the ancestral population, could not be tested here. Second, the contemporary ancestor has to have been relatively evolutionarily static. Both phenotypic and genetic evidence demonstrates that this condition does not hold for marine stickleback. Contemporary populations have diverged phenotypically in a manner beyond neutral evolutionary expectations, and genetically in a way explicable by drift and selection. This means that contemporary marine threespine stickleback populations are genetically and phenotypically distinct from their own ancestors – and it was these ancestors that also originally colonized lakes and streams along the coast. Thus the term “contemporary ancestor” is a misnomer, as contemporary marine threespine stickleback populations do not reflect the ancestral condition. Interpretations of freshwater stickleback evolution need to be tempered by marine stickleback evolutionary history. Third, the ancestral population has to have been properly characterized. Eastern Pacific stickleback exhibit some population genetic structure, although consistent with other reports (Withler and McPhail 1985; Hohenlohe et al. 2010) stickleback north of Oregon constitute a single genetic population. This means that along much of the coast freshwater environments were likely colonized by distinct marine stickleback populations, which differed in SGV frequency and content. Furthermore, it is likely that marine stickleback have exhibited range expansions along the southern and northern coasts throughout their evolutionary history, most recently in the north after the last glacial retreat. This means that there is no a priori reason to expect that a marine population currently proximate to freshwater populations are descendants of the ancestors

125

of those freshwater populations. Population structuring and evolutionary history thus changes our understanding of the ancestral condition of marine stickleback, and requires that we carefully consider the use of contemporary marine populations when addressing evolutionary questions. Looking ahead, it is clear that ecological information about marine stickleback is sorely needed. Freshwater stickleback have long been of interest due to their extensive genetic and phenotypic variation. But marine populations clearly harbour significant genetic and phenotypic variation – the ecological and evolutionary reasons for these differences await further investigation. “Contemporary ancestors” are used in a number of systems (e.g. Berry et al. 1978; Losos et al. 2000; Des Roches et al. 2011; Domingues et al. 2012; Cahill et al. 2013; Dennenmoser et al. 2017) for addressing evolutionary questions. They are particularly useful for determining the role of SGV during evolution, and for identifying the alleles involved in adaptation to new environments. Clearly, care must be exercised in characterizing these proxies of the ancestral form, as unaccounted population structure and current evolution can lead to spurious interpretations. The amount of SGV present in marine stickleback begs the question – why does it exist in the first place? It is possible that some SGV is maintained due to selection favouring polymorphisms. This is the topic of the next chapter.

126

Heterozygosity and asymmetry: Ectodysplasin as a form of genetic stress in marine threespine stickleback

5.1 Introduction

The ability of populations to adapt rapidly to new environmental conditions depends in part on the amount of genetic variation already present in the population. Such standing genetic variation (SGV) may persist under balancing selection if multiple alleles are adaptive under different contexts. For instance, alternative alleles may provide fitness advantages within a heterogeneous landscape (e.g. Delph and Kelly 2014), under different developmental contexts as organisms mature (e.g. Barrett et al. 2008), or under different genetic backgrounds (e.g. Weinig et al. 2003). An example of the latter is the hypothesis that genome-wide levels of heterozygosity may promote developmental stability by buffering against environmental disturbances during development (Lerner 1954; Clarke 1993). If true SGV could be maintained genome-wide if developmental stability enhances fitness (Roldan et al. 1998; Brambilla et al. 2014). Developmental stability is predicted to be favoured by selection through canalization (Waddington 1942). If phenotypic outcomes are easily disrupted by mutations or environmental stressors (“developmental noise”), mechanisms that reduce such noise may be selected if they enhance an individual’s fitness. Canalization has been subject to considerable theoretical and empirical investigation (e.g. Masel and Siegal 2009). For example, the actions of certain phenotypic capacitors, such as heat shock proteins, buffer development such that, if they fail to work, previously hidden phenotypic variation is expressed (Queitsch et al. 2002; Debat and Peronnet 2013). Because canalized phenotypes are common in nature (Félix and Barkoulas 2015), developmental noise is largely presumed to be maladaptive under normal conditions. However, increases in phenotypic variance caused by such noise may be adaptive under certain conditions (Rohner et al. 2013; Richard and Yvert 2014).

127

One possible phenotypic expression of developmental noise is fluctuating asymmetry (FA, Van Valen 1962), which has widely been used as a measure of developmental stability in wild populations (Beasley et al. 2013). In bilaterally symmetrical organisms, FA is typically defined as stochastic right-left asymmetry around a mean of zero. FA can increase under stress (Leung et al. 2000; Lens et al. 2002a; Weller and Ganzhorn 2004; Allenbach 2011) and directional selection (De Coster et al. 2013). FA has been found in diverse taxa, but whether it persists because selection is currently acting to reduce FA, favours FA, or does not act on FA remains largely unknown. Negative correlations between FA and fitness support Waddington’s theory (e.g. Nosil and Reimchen 2001; Hendrickx et al. 2003), but counterexamples in which FA is neutral with respect to fitness abound (Leung and Forbes 1996; Lens et al. 2002b; but see Aparicio and Bonal 2002), although the mechanisms remain unclear. Defensive, colouration, and locomotory phenotypes provide some of the most compelling evidences for asymmetry-fitness correlations, perhaps because such asymmetries reduce the capabilities of individuals to escape from predators (Reeve 1960; Swaddle 1997; Forsman and Herrström 2004; Tocts et al. 2016; but see Cuthill et al. 2006; Stevens et al. 2009). Lerner (1954) proposed that heterozygosity buffers an organism against asymmetry. Although there is debate over what Lerner originally meant (Møller and Swaddle 1997), many researchers have interpreted this to mean that individuals with higher genome-wide heterozygosity can compensate for perturbations during development, producing increasingly symmetrical phenotypes (Clarke 1993). This could be because genome-wide heterozygosity enhances the diversity of protein structures and regulatory elements, which then provide a greater range of responses in the face of developmental perturbation. Negative correlations between heterozygosity and asymmetry have been found in some (e.g. Leary et al. 1983, 1984; Shaner et al. 2013) but not all studies (e.g. Wooten and Smith 1986; Patterson and Patton 1990; Kark et al. 2001; Trokovic et al. 2012), with meta-analyses producing equivocal and perhaps taxon- specific results (Vøllestad et al. 1999). Alternatively, heterozygosity at particular major- effect loci could impair protein function or activity, reducing buffering capacity or

128

leading to random variation around a haplosufficient threshold. These contrasting possibilities require formal testing. Threespine stickleback Gasterosteus aculeatus is a useful subject for studying developmental instability. They produce a series of bony plates laterally along their bodies. Approximately 80% of phenotypic variation in plate number is attributable to the presence of a regulatory mutation at the Ectodysplasin (Eda) gene (Colosimo et al. 2005). Each plate has a position determined by the myomere in which it develops, with each myomere acting as in independent trait that can be scored for the presence/absence of plates and compared between body sides (Reimchen 1983). Plates on myomeres four through seven are “structural plates”, as they abut the dorsal basal plates (DBP) for the first two dorsal spines and the ascending process (AP) of the pelvic armature. Removal of structural plates impairs the deflection of dorsal and pelvic spines, reducing defensive capabilities (Reimchen 1983), consistent with the observation that predation by insects selects for low-plated stickleback that have lost non-structural plates (Marchinko 2009). Consistent with Waddington’s predictions, structural plates exhibit less asymmetry than non-structural plates (Bergstrom and Reimchen 2000), suggesting that selection has canalized the development of structural plates. The fitness consequences of FA in non- structural plates are partially understood (Reimchen and Nosil 2001; Bergstrom and Reimchen 2003; but see Loehr et al. 2013). A negative correlation between the number of myomeres exhibiting FA and genome-wide heterozygosity has been predicted (e.g. Reimchen and Nosil 2001; Bergstrom and Reimchen 2000, 2005; Reimchen et al. 2008; Reimchen and Bergstrom 2009; Loehr et al. 2012, 2013; but see Van Dongen et al. 2009), but not tested in threespine stickleback. Marine threespine stickleback have been proposed to be contemporary representatives of the ancestors of freshwater stickleback (Schluter and Conte 2009; Jones et al. 2012b). SGV in marine stickleback is an important source of adaptive freshwater alleles (Hohenlohe et al. 2010; McGuigan et al. 2011; Jones et al. 2012ab), particularly at the Eda locus (Colosimo et al. 2005). Reasons for the maintenance of SGV in the marine environment, at Eda and at other loci, are unknown. Negative correlation of FA with genome-wide heterozygosity could provide a potentially adaptive explanation

129

for at least some forms of SGV if FA is correlated with fitness. Alternatively, because stickleback platedness is largely controlled by the Eda locus, heterozygosity at Eda may be a form of genetic stress resulting in enhanced FA. If so, why does variation at Eda exist in the marine environment? In this study, I test the hypothesis that SGV is maintained in marine stickleback through selection on symmetric, more heterozygous individuals. I specifically assessed (1) whether marine populations vary in plate number and the association between plate morph and Eda genotype; (2) whether plate presence/absence asymmetry based on myomere position was fluctuating or directional; (3) whether plates four through seven of marine stickleback serve structural roles, as reported from freshwater individuals, and whether other plates fulfill similar functions; (4) whether structural plates exhibit less asymmetry in plate position than non-structural plates; and (5) whether asymmetry varies with genome-wide heterozygosity or Eda genotype.

5.2 Materials and Methods

5.2.1 Sequencing

Fin clips of fish from CA01 through to AK01 were preserved in 95% ethanol for DNA extraction. Briefly, sex- and plate morph-specific microsatellites (idh and Eda- linked Stn382, respectively) were amplified, and scored on agarose gel as per Chapters Three and Four. Stn382 is linked to the regulatory mutation in Eda that determines overall patterns of platedness (Peichel et al. 2004). The “L” allele is typical of low-plated fish, whereas the “C” allele is typical of fully-plated fish. Heterozygous “CL” individuals typically exhibit a partially-plated phenotype, but may have modifiers that restore the fully-plated form (Colosimo et al. 2005). The total number of male and female stickleback collected from each location were reported in Chapter Three and the number of Eda genotypes in Chapter Four and Table 5-1. DNA was extracted from fin clips and sequenced as described in 4.2.2.

130

Table 5-1 - The numbers of males and females for which plate counts, Eda genotypes, and standardized multi-locus heterozygosity measures were assessed.

Pop ID Females Males

Plate count Eda sMLH Plate count Eda sMLH

CA01 9 9 6 26 25 23

CA02 30 30 20 20 20 8

CA03 18 18 11 31 31 17

OR01 12 12 11 8 8 8

OR02 37 37 23 13 13 7

BC01 1 1 1 47 47 28

AK01 23 21 18 8 6 6

131

5.2.2 Plate counts and asymmetry

Stickleback were preserved in 70% ethanol for transport to the University of Calgary. To best visualize plates, fish were sequentially transferred to 50% ethanol, 20% ethanol, distilled water, 10% formalin, and distilled water, spending 24 h in each fluid. Fish were then stained in KOH-buffered Alizarin red for 48 h and rinsed in distilled water for 24 h. They were finally preserved in 40% isopropyl alcohol. Plate morph was defined as per Chapter Four. Briefly, Fully-Plated with Keel (FPK) fish have an average of > 30 plates, Low-Plated with No Keel (LPNK) fish had an average anterior plate count of < 10 plates, Partially-Plated with Keel (PPK) fish had a keel and > 10 but < 30 plates. Additionally, two rare variants were also found: Low- Plated with Keel (LPK) fish had < 10 anterior plates but one or more keel plates, and Partially-Plated with No Keel (PPNK) fish had > 10 anterior plates but lacked a keel. These plate-morph categories generally follow those of Lucek et al. (2012). Eda genotype was used to determine genotype-phenotype mismatches, defined as occurring when LL, CL, or CC fish did not have total plate counts within the LPNK, PPK, and FPK categories, respectively, regardless of the presence or absence of the keel. As numerous mismatches were noted for FPK fish, I used Welch’s two-sample t-test to compare average plate counts for CL and CC fish belonging to this FPK category. To identify “structural” and “non-structural” plates, plates three through nine were scored for their contact or lack thereof with the first or second dorsal basal plate (DBP1, DBP2), and plates seven and eight were scored for their associations with the ascending process (AP). Asymmetries in these associations were noted but could not be meaningfully signed for FA tests, and are included in the supplementary materials (Appendix C). Traditionally, plates 4-7 are considered “structural”, and 1-3 and 8-30 “non-structural” (Reimchen 1983) (Figure 5-1).

All plates were counted on both the left (LT) and right sides (RT) of each fish by a single investigator. Plate asymmetry was recorded in two ways after Bergstrom and

Reimchen (2000): total plate number asymmetry, PNUMA = RT - LT; and plate position asymmetry (PPOSA), for which myomeres were scored as plated (1) or plateless (0) from myomere 1 (P1) to myomere 30 (P30). The keel (K) plates were counted and summed. 132

Figure 5-1 - Bony elements of the heads of two marine threespine stickleback from (top) Bamfield Inlet, British Columbia and (bottom) Elkhorn Slough, California, showing the positions of the first nine plates, the first and second dorsal basal plate (DBP1, DBP2) and the ascending process (AP). Plates 4 - 7 are structural plates, the remainder are non- structural plates. Note that the Californian stickleback lacks the first, eighth, and ninth plates.

133

PPOSA was therefore measured as [(RP1 - LP1), (RP2 - LP2), …, (RP30 - LP30), (RKT - LKT)] with KT indicating total keel plate count. Signed and unsigned (absolute values) of PNUMA and PPOSA were calculated. For signed PPOSA, each R-L difference was summed across plates; for unsigned PPOSA the absolute value of each R-L difference was summed. Note that a fish could be scored as symmetric for PNUMA but asymmetric for PPOSA (e.g. RP1 - LP1 = 1, RP2 - LP2 = -1, PPOSA = 2 but PNUMA = 0). Plate position based on myomere number has been shown to be replicable among stickleback (Reimchen 1983). Rarely, a single myomere contained more than one plate; these additional plates were not included in the counts. Also rarely, a barely discernible thin spine-like structure oriented in an anterior-posterior direction would be seen on a plateless myomere. Given the delicate nature of these plate-like structures (they were prone to breakage with handling), and difficulty of visualizing them, they were not included in final counts. For most statistical analysis the keel was not included in signed or unsigned PPOSA. To assess measurement error, counts were conducted twice for all individuals, with replication conducted several months apart. The effect of measurement error on the ability to detect asymmetry was estimated using a two-factor mixed effects Analysis of Variance (ANOVA) using the lme4 package (Bates et al. 2015) in R (R Core Team 2016), with plate number as the dependent variable, individual fish as a random factor, and side as a fixed factor. Models with and without the side x individual interaction term were compared and a test of significance between the two models assessed using ANOVA. A significant side-by-individual interaction term would indicate detectable asymmetry despite measurement error (Palmer and Strobeck 1986; Bergstrom and Reimchen 2003). To reduce measurement error, all fish for which discrepancies in plate position were noted were re-scored and corrected, with the corrected data used in all subsequent analyses. Mean signed PPOSA was tested for the effects of population and sex, using a two-factor ANOVA and Type II SS using the car package in R. Mean signed PPOSA and PNUMA was tested against a null mean of 0 using a two-tailed t-test with a corrected α of 0.05/17 = 0.003, for all samples and for each population, Eda genotype, and sex, and

134

for structural and non-structural (both anterior P1-P3 and posterior P8-P30) plates. The Kolmogrov-Smirnov test was used to test for normality for PPOSA in each population. Kurtosis for PPOSA was measured using the moments package in R, using an unbiased (kurtosis2) estimator for all samples, for each population, Eda genotype, and sex. Antisymmetry was visually assessed by looking for bimodality within each histogram. If structural plate symmetry promotes fitness, structural plates should be less asymmetric than non-structural plates. The null hypothesis that the incidence of PPOSA did not vary with plate position was tested using χ2 tests for all plates, structural plates, non-structural (anterior, posterior) plates, and within each Eda genotype. Welch’s t-test was used to determine whether average PPOSA differed between structural and non- structural plates, and between CL and CC genotypes within the FPK morph. Pearson correlations tested for an association between plate number and the sum of unsigned PPOSA for all samples and within each Eda genotype. The symmetrical absence of plates was also noted.

5.2.3 Asymmetry and heterozygosity

Heterozygosity per individual fish was calculated using the Genotype-by- Sequencing dataset in Chapter Four (6655 variant loci) with the InbreedR package (Stoffel et al. 2016) in R. I estimated both multilocus heterozygosity (MLH), the number of heterozygous loci divided by the total number of loci sequenced for an individual, and standardized multilocus heterozygosity (sMLH), the number of heterozygous loci in an individual, divided by the sum of the population average heterozygosity for those loci that were sequenced in the individual of interest (Stoffel et al. 2016). Standardization of heterozygosity by allele frequencies within each population corrected for differences in allele frequencies between populations. The relations of MLH and sMLH with latitude was assessed using Pearson correlations. The relation of sMLH to Eda genotype was determined using two-way ANOVA with Eda genotype and population as factors. In order to determine whether asymmetry was related to heterozygosity, Pearson correlations were determined between MLH or sMLH and the sum of unsigned PPOSA. To determine if the number of non-asymmetric individuals swamped out the signal for an 135

asymmetry-heterozygosity relation, a Pearson correlation was determined using sMLH and only individuals with unsigned positive PPOSA values. To test whether symmetric and asymmetric individuals varied for sMLH, ANOVA was run on sMLH with the symmetrical state of an individual (symmetric or asymmetric) and population as factors. The relation between heterozygosity for Eda and asymmetry was assessed with ANOVA using PPOSA as the dependent variable and genotype, population, and their interactions as the fixed effect. Post-hoc Tukey HSD tests were used to determine which Eda contrasts were significant, with adjusted α of 0.017.

5.3 Results

5.3.1 Genotype, plate morph, and plate counts

278 stickleback were phenotyped and genotyped at Eda at the Stn382 locus; an additional five individuals were phenotyped only (Table 5-1). Most individuals had LL genotypes (48%), which did not completely match the proportion of genotyped LPNK morphs (46%). Populations varied extensively with respect to the distributions of Eda genotypes, and therefore also varied for plate number (Figure 5-2, Table 5-2). Plate number varied from an average of 7.0 (range: 4.0-19.5) for the LL genotype, 25.7 (range: 6.5-34) for the CL genotype, and 33.3 (range: 23.0-35.5) for the CC genotype (Figure 5- 2, Table 5-3). This variation illustrates some non-random genotype-phenotype mismatch. When mismatches occurred, individuals tended to have fewer plates than genotype- phenotype “matched” individuals. At ten plates or fewer, 128 of 129 individuals were LL homozygotes, with the sole exception being a CL heterozygote with 6.5 plates. Of the 15 fish with 12-19.5 plates, which represents the lower end of the PPK category; five (33%) were LL homozygotes and the remainder were CL heterozygotes. Fish with 20-29.5 plates included a single CC individual (23 plates) with the remainder being CL heterozygotes. In the traditional FPK morph class (> 30 plates), 24 were CL heterozygotes and 90 were CC homozygotes. Within the FPK class CL fish had significantly fewer plates (mean: 31.7 plates) than CC fish (mean: 33.4 plates: Welch two-sample t-test: t = 6.5, d.f. = 30, p < 0.001). Each increase in plate number was

136

Table 5-2 - Average numbers of plates, and the number of fish with each possible genotype-phenotype combination per population. See the List of Symbols, Abbreviations and Nomenclature for meaning of column headings. Population Mean LL CL CC plate count FPK PPK LPNK LPK PPNK FPK PPK LPNK LPK PPNK FPK PPK LPNK LPK PPNK CA01 10.2 0 3 25 1 0 1 2 0 2 0 0 0 0 0 0 CA02 13.3 0 0 34 0 0 7 9 0 0 0 0 0 0 0 0 CA03 9.2 0 0 42 0 0 1 5 0 0 0 1 0 0 0 0 OR01 32.2 0 0 0 0 0 6 0 0 0 0 14 0 0 0 0 OR02 16.5 0 0 23 1 1 9 10 1 1 0 3 1 0 0 0 BC01 33.7 0 0 0 0 0 0 0 0 0 0 48 0 0 0 0 AK01 30.0 0 0 3 0 0 0 0 0 0 0 24 0 0 0 0

137

Table 5-3 - Characteristics and sampling effort per Eda genotype, including average numbers of plates and asymmetric myomeres per individual (PPOSA), average standardized multi-locus heterozygosity (sMLH), total number of genotyped individuals, number of genotyped individuals expressing PPOSA, and the total number of plate positions expressing asymmetry among all individuals. LL = homozygous for the low- plated allele, CL = heterozygous, CC = homozygous for the fully-plated allele.

Eda Mean Mean Mean N N PPOSA Sum genotype plate unsigned sMLH genotyped PPOSA count PPOSA LL 7.0 0.56 0.987 133 59 75 CL 25.7 2.59 0.973 54 44 140 CC 33.3 0.30 1.000 91 21 27

138

Figure 5-2 - Distributions of average plate number (A) per population and (B) per Eda genotype. CC = homozygous for the “fully- plated” allele; LL = homozygous for the “low-plated” allele; CL = heterozygous.

139

associated with relatively fewer CL individuals, decreasing from an occurrence of 83% (30-30.5 plates) to 63% (31-31.5), 56% (32-32.5), 9% (33-33.5), 3% (34-34.5), and 0% (35-35.5) of FPK individuals (Figure 5-3).

5.3.2 Structural and non-structural plates

P4 through P7 are generally considered to be structural plates. For the 566 possible measures (283 stickleback x 2 sides) at least one dorsal basal plate (DBP) contacted plates in these positions (P4, 98.6% contacted DBP1; P5, 98.8% contacted DBP1; P6, 77.9% contacted DBP1, DBP2, both, or a rare third DBP; P7, 99.6% contacted DBP2). 96.1% of P7 abutted the ascending process (AP). Other plates appeared to be structural less often. 17% of P3 abutted DBP1; over half of these cases were from BC01. 61.8% of P8 and 8.5% of P9 abutted DBP2. 9.9% of P8 and 0% of P9 abutted the AP. Overall, this substantiates the identification of P4-P7 as structural plates, although P8 may also serve this role in some individuals.

5.3.3 Plate position asymmetry

Asymmetry was evident despite measurement error (model without individual x side interaction: d.f. = 4, AIC = 4244, log-likelihood = -2118; model with interaction term: d.f. = 5, AIC = 3136, log-likelihood = -1563; ANOVA comparing two models, d.f. = 1, p < 0.001) (Table 5-4). Signed PPOSA showed no variation among sexes or populations (Figure 5-4, Table 5-5). Mean signed PPOSA or PNUMA did not differ from zero under any comparison (Table 5-6), indicating no overall directional asymmetry. The Kolmogorov-Smirnov test rejected the hypothesis that PPOSA was distributed normally in all populations (Table 5-6), although this rejection was weak for OR01 and OR02. All distributions had positive kurtosis and showed no evidence of antisymmetry (Table 5-4). 41% of sampled stickleback had PNUMA. Fifty-four of these were left-biased, 61 were right-biased. An additional twelve fish exhibited PPOSA but not PNUMA. The unsigned sum of PNUMA was 170, but 245 plate positions exhibited PPOSA, further indicating that ignoring plate position obscures asymmetry. 196 of the PPOSA locations

140

Table 5-4 - Average number of asymmetric myomeres per individual (PPOSA), loci sequenced per individual, and average standardized multi-locus heterozygosity (sMLH) per population. Also shown is the kurtosis of the distribution of signed positional plate asymmetry (PPOSA). See the List of Symbols, Abbreviations and Nomenclature for meaning of column headings.

Population Mean unsigned Kurtosis Mean N Mean PPOSA sequenced loci sMLH CA01 0.943 5.4 5376 0.9991 CA02 1.040 7.0 5725 0.9926 CA03 0.755 4.1 4892 0.9848 OR01 0.500 3.0 5564 0.9788 OR02 1.960 3.0 5502 0.9824 BC01 0.146 6.6 5332 0.9856 AK01 0.258 3.6 5344 1.0000

141

Table 5-5 - Results of two-way ANOVA.

Factor Type II SS d.f. F p Signed PPOSA, excluding keel, for sex and population Sex 0.4 1 0.4 0.5 Population 2.9 6 0.5 0.8 Sex x Pop 6.7 6 1.2 0.3 Residuals 253 269 Unsigned PPOSA, excluding keel, for Eda and population Eda 81.7 2 48.0 < 0.001 Population 41.0 6 8.0 < 0.001 Eda x Pop 11.7 6 2.3 0.04 Residuals 223.9 263 sMLH, for symmetry (symmetrical or asymmetrical) and population Symmetry 0.001 1 0.04 0.8 Population 0.01 6 0.1 1.0 Symm x Pop 0.2 6 2.0 0.07 Residuals 2.9 173 sMLH, for Eda and population Eda 0.1 2 2.8 0.06 Population 0.1 6 0.8 0.5 Eda x Pop 0.1 5 0.9 0.5 Residuals 2.9 172

142

Table 5-6 - Results of Welch’s t-tests for significant deviations from means of 0 for signed positional plate asymmetry (PPOSA) for all samples, individual populations, Eda genotypes, sexes, and plate type (anterior P1-P3 or posterior P8-P30 non-structural plates, or structural plates P4-P7), or signed total plate count asymmetry (PNUMA) for all samples. Significance was set at α = 0.003. Also shown are the results of the Kolmogorov-Smirnov test for Normality (D) and its associated p-value, for each population.

Comparison Mean d.f. t p D p All samples 0.07 282 1.1 0.3 CA01 0.2 34 0.9 0.4 0.3 < 0.001 CA02 0.3 49 1.8 0.07 0.3 < 0.001 CA03 0.1 48 0.7 0.5 0.3 < 0.001 OR01 0 19 0 1 0.3 0.05 OR02 -0.12 49 -0.6 0.6 0.2 0.06 BC01 -0.02 47 -0.4 0.7 0.4 < 0.001 AK01 0 30 0 1 0.4 < 0.001 LL -0.09 137 -1.2 0.2 CL 0.59 53 2.3 0.03 CC 0.01 90 0.2 0.8 Male 0.04 152 0.5 0.6 Female 0.12 1.1 129 0.3 P1-P3 0.018 282 0.7 0.5 P4-P7 0.003 282 1 0.4 P8-P30 0.035 282 0.7 0.5 PNUMA 0.08 282 1.2 0.2 All samples

143

Figure 5-3 - Distributions of plate number for each Eda genotype of fully-plated stickleback. CC = homozygous for the “fully-plated” allele; CL = heterozygous.

144

Figure 5-4 - Distributions of signed positional plate asymmetry (PPOSA) (A) per population and (B) per Eda genotype.

145

occurred outside of the keel. Given 8773 possible asymmetric positions (30 myomere pairs * 283 stickleback + 283 keel pairs), PPOSA occurred in 2.8% of all positions. Most fish were asymmetric at only a single plate position (n = 75), and numbers declined as the number of asymmetric plate positions increased to a maximum of eight asymmetric plates on a single fish from OR02. PPOSA varied among populations (ANOVA: F6,276 = 10.9, p < 0.001). From OR01 south, 45% to 49% of collected individuals exhibited PPOSA compared to 78% of individuals from OR02 and only 15% from BC01. Excluding the keel, plate positions differed in their PPOSA (χ2 = 147, d.f. = 29, p < 0.001) (Figure 5-5). Structural plates had significantly lower average PPOSA than non- structural plates (mean structural PPOSA: 0.0009, mean non-structural PPOSA: 0.027, t = 12.4, d.f. = 8317, p < 0.0001). Structural plates P4, P5, and P7 were symmetrically present in all individuals, whereas a single OR02 individual (0.35% of all stickleback) showed asymmetry at P6. Non-structural P11, P28, and P30 were asymmetric in 0.7% of all sampled fish, whereas P8 had the maximum incidence of asymmetry at 8.8%. The frequency of PPOSA (excluding the keel) therefore varied among non-structural plates (χ2 = 104.2, d.f. = 25, p < 0.001), even when considering only anterior plates (χ2 = 17.4, d.f. = 2, p < 0.001) or only posterior plates (χ2 = 50.4, d.f. = 22, p < 0.001). 16% of individuals expressed some keel asymmetry. Patterns were particularly noticeable with respect to Stn382 genotype (Table 5-3). Of 2730 plate positions in CC individuals, only 9 exhibited asymmetry, and no variation in asymmetry was evident among the first 30 plates (χ2 = 28, d.f. = 29, p = 0.5). 18 CC individuals exhibited additional asymmetry at the keel. Although LL individuals tend to be low-plated, only the structural plates and plates 18-27 and 30 exhibited no asymmetry. Asymmetry differed among plates (χ2 = 281, d.f. = 29, p < 0.001), from 2.2% at P15 to 12.8% at P8. In CL heterozygotes, five non-structural plates had equal or lower asymmetries than structural P6. The remaining 21 plates were asymmetric in 3.7% (P8, P10, and P30) to 18.5% (P1) of individuals, with asymmetry varying significantly among plates (χ2 = 75, d.f. = 29, p < 0.001). 46% of CL individuals were asymmetric in the keel. Within the FPK plate morph class, CL heterozygotes (mean: 1.2) had greater asymmetry

146

Figure 5-5 - Frequencies of stickleback that exhibited symmetrical presence (green), symmetrical absence (red) or asymmetrical absence (blue) of plates per plate position.

147

than CC homozygotes (mean: 0.2) (Welch two sample t-test: t = -3.8, d.f. = 24, p = 0.0019). Plate count and PPOSA were not correlated (Pearson correlation: r = 0.018, t = 0.3, d.f. = 281, p = 0.8). For CC fish PPOSA increased with declining plate number (Pearson correlation: r = -0.74, t = -10.4, d.f. = 89, p < 0.001). For LL fish PPOSA varied positively with plate number (Pearson correlation: r = 0.44, t = 5.7, d.f. = 131, p < 0.001), whereas the relation was negative for CL fish (Pearson correlation: r = -0.43, t = -3.4, d.f. = 52, p = 0.001). Thus, more “FPK-like” CL fish exhibited less FA. Although usually not reported, the symmetrical absence of plates also showed interesting patterns (Figure 5-5). Only 6 plate positions exhibited symmetrical loss of a plate in CC individuals. Although LL individuals largely exhibit symmetrical absence of plates beyond P10, plates were symmetrically absent at several positions that have the potential to be plated in LPNK fish, particularly P2 (18%), P8 (19.5%), P9 (83%), P1 (86%), and P10 (93%). Symmetrical absence was most interesting in CL individuals: 18% of individuals symmetrically lacked P1, no individuals symmetrically lacked P2-P8, and symmetrical absence increased gradually from nearly 2% at P9 to 46% at P22, and then declined to nearly 2% for P28-P30 and 7% for the keel.

5.3.4 Asymmetry and heterozygosity

Heterozygosity was estimated from sequences of 187 individuals, of which 80 expressed PPOSA (Table 5-4). Observed heterozygosity varied positively with latitude (Pearson correlation: r = 0.43, t = -6.4, d.f. = 185, p < 0.001), with OR02 having the highest average. This relation was not evident when heterozygosity was standardized within each population (Pearson correlation: r = 0.006, t = 0.08, d.f. = 185, p = 0.9) (Figure 5-6). Similarly, observed heterozygosity varied positively with PPOSA (Pearson correlation: r = 0.38, t = 5.6, d.f. = 185, p < 0.001), but sMLH did not (Pearson correlation: r = - 0.005, t = -0.07, d.f. = 185, p = 0.9) (Figure 5-7), even when looking only at asymmetric individuals (Pearson correlation: r = 0.03, t = 0.3, d.f. = 78, p = 0.8). Mean sMLH did not differ for asymmetric and symmetric individuals (Table 5-5, mean 148

Figure 5-6 - Mean (+ SE) for (A) observed heterozygosity and (B) standardized multi- locus heterozygosity (sMLH), per population.

149

Figure 5-7 - Relations of standardized multilocus heterozygosity (sMLH) to the number of myomeres with asymmetrical loss of a plate per fish, per population.

150

asymmetric sMLH = 0.9857, mean symmetric = 0.9919). sMLH did not vary significantly among Eda genotypes (Table 5-3, Table 5-5). PPOSA varied significantly among Eda genotypes, even when accounting for population differences (Table 5-5). 81% of CL genotypes were asymmetric, compared to 23% of CC genotypes and 44% of LL genotypes (Table 5-3). 83% of all PPOSA was found in CL heterozygotes, even though this genotype comprised only 19% of all collected fish. The CL genotype had significantly higher asymmetry than either the CC (mean CL-CC = 2.3, p = 0) or LL (mean CL-LL = 2.0, p = 0) genotypes (mean LL - CC = 0.3, p = 0.2) (Figure 5-4).

5.4 Discussion

5.4.1 Summary

Lerner (1954) predicted that genome-wide heterozygosity should limit developmental noise, as individuals with more heterozygous loci should have greater plasticity to buffer perturbations. If true, this could explain genome-wide patterns of heterozygosity in nature. In this study I examined marine threespine stickleback, the contemporary forms of the ancestors of freshwater stickleback, from Alaska to California. Here I report that (1) there was extensive variation for those fish at Eda, with phenotypes not always matching genotypes but such mismatches having phenotypic consequences. This raises important questions about the significance of SGV at Eda in the marine environment. (2) Plate position asymmetry differed between structural and non-structural plates. In contrast to freshwater populations, only a single individual exhibited structural asymmetry among all marine individuals. (3) Finally, standardized genome-wide heterozygosity did not vary significantly with asymmetry, but was clearly associated with Eda genotype, even in genotype-phenotype mismatched individuals. Collectively, these results suggest that structural plate canalization is an ancestral condition that is reduced in freshwater populations, and that SGV at Eda constitutes a genetic stressor that elevates asymmetry in threespine stickleback.

151

5.4.2 Genotype-phenotype mismatch

Plate number did not match Eda genotype for over 10% of stickleback (also see Barrett et al. 2009a; Le Rouzic et al. 2011; Lucek et al. 2012). Eda is one of several genes that impacts plate development (Colosimo et al. 2005). Colosimo et al. (2005) suggested that genotype-phenotype mismatch in Eda represents cryptic genetic variation (Gibson and Dworkin 2004), wherein the low-plated allele persists at low frequencies in the marine environment through the influence of modifier alleles that restore the marine- adaptive fully-plated phenotype. This genetic variation could then be available for selection during colonisation of freshwater. My results suggest that Eda heterozygosity in fully-plated stickleback is not entirely cryptic, reducing the number of plates and increasing the number of asymmetric plate positions relative to homozygous fully-plated individuals. The fitness consequences of these phenotypes are unknown and warrant further exploration, but could explain why the low-plated allele is maintained in heterozygotes in some marine contexts. For instance, in freshwater populations individuals with fewer plates have greater fast-start performance (Bergstrom 2002).

5.4.3 Plate count, asymmetry and selection

Fully-plated stickleback produce lateral plates systematically throughout early development, beginning with P5 and P6 (Bell 1981). Plates anterior to and posterior to P6 develop in turn (P7, P4, P8, P3, etc.), with additional plates developing in the caudal region. The gap between the structural plates and the keel fills progressively from both directions (Bell 1981; Bańbura 1989; Bergstrom and Reimchen 2000). The structural plates are hypothesized to develop early due to their defensive role, particularly in spine deflection (Reimchen 1983). Consistent with this expectation, symmetric and asymmetric loss of structural plates is less common than loss of non-structural plates in Haida Gwaii and Scandinavian stickleback (Reimchen 1983; Bergstrom and Reimchen 2000; Loehr et al. 2013), presumably due to selection. My results decisively confirmed this association. Although marine stickleback varied genetically and morphologically from Alaska to California (Chapters Three and Four), and displayed substantial variation in plate count

152

and Eda genotype, only one individual was asymmetric for plate number for a structural plate. That is an occurrence of 0.35%, as opposed to freshwater reports varying from 1.1% to 4.8%. This contrast suggests that canalization of structural plates may be under relaxed selection in the freshwater environment (Trokovic et al. 2012), rather than strong directional selection (Bergstrom and Reimchen 2000).

5.4.4 Asymmetry and heterozygosity

Marine stickleback harbour more SGV than their freshwater counterparts (e.g. Withler and McPhail 1985; Raeymaekers et al. 2005), maintaining variation on which selection can act (Colosimo et al. 2005; Hohenlohe et al. 2010; McGuigan et al. 2011; Jones et al. 2012ab; O’Brown et al. 2015). Such variation is largely attributed to large effective populations, absence of recent genetic bottlenecks, and connectivity among marine stickleback (Hohenlohe et al. 2010). However, Chapter Four demonstrated small effective populations and limited connectivity among marine populations from California to Oregon. In such circumstances SGV could be maintained if genomic heterozygosity improves fitness. Correlations between adaptive traits and genomic heterozygosity would provide evidence for such a mechanism, but I found no such association between heterozygosity and asymmetry. This contrasts with many studies that reported lower FA in relatively heterozygous individuals (Quattro and Vrijenhoek 1989; Leary et al. 1983; Hartl et al. 1995; Roldan et al. 1998, Vøllestad et al. 1999; Shaner et al. 2013), but is consistent with findings for European stickleback (Van Dongen et al. 2009; Trokovic et al. 2012; see also Patterson and Patton 1990; Weatherhead et al. 1999; Carchini et al. 2001; Taylor 2001; White and Searle 2008). These patterns may have several explanations. (1) FA may be adaptively neutral, and therefore is not buffered by increased heterozygosity (Britten 1996). However, fitness consequences of non-structural plate asymmetry have been observed in freshwater populations (Reimchen and Nosil 2001; Bergstrom and Reimchen 2003; but see Bergstrom and Reimchen 2005; Reimchen and Bergstrom 2009), and they are not required for the heterozygosity hypothesis to be true. (2) FA for non-structural plates may not be under genetic control. In contrast, Loehr et al. (2012) reported significant heritability for fluctuating asymmetry in non-structural 153

plates (h2 = 0.21). (3) Heterozygosity may exceed a threshold needed to trigger asymmetry. In contrast, Trokovic et al. (2012) found no evidence for heterozygosity-FA associations in marine and relatively inbred freshwater populations of ninespine stickleback. (4) Asymmetry may not be buffered by genomic heterozygosity, and the expectation that it should be ignores considerable evidence of symmetry in asexual or haploid organisms (Clarke 1993). The significant heterozygosity-asymmetry associations reported previously may reflect the use of fewer microsatellite markers than could give good genome-wide estimates of heterozygosity – and may have detected associations between marker-specific heterozygosity rather than genomic heterozygosity. In contrast to the genomic heterozygosity hypothesis, variation at a single locus,

Eda, explained 38% (SSEda/SSTOTAL) of variance in PPOSA. Heterozygous individuals were significantly more asymmetric than homozygous fish. This suggests that heterozygosity at Eda constitutes a form of genetic stress. O’Brown et al. (2015) identified the major causative Single Nucleotide Polymorphism responsible for low- and fully-plated morphs: a TG mutation in a regulatory enhancer of the Eda gene. F1 heterozygotes of a marine x freshwater cross showed reduced expression of the “freshwater” G allele in anterior and posterior plate positions relative to the “marine” T allele. Furthermore, mutagenic production of the G allele in a marine stickleback substantially reduced Eda expression posterior to P7. Although O’Brown et al. (2015) did not study keel production, their results suggest why PPOSA is highest in CL heterozygotes. CC individuals produce decreasing Eda protein with myomere position beyond P6. Consequently, central myomeres produce the least Eda protein, but concentrations are usually still sufficient to produce a plate. In contrast, in CL heterozygotes the distance from the structural plates or keel results in myomeres that have Eda concentrations near the threshold for plate production. Noise in protein production would randomly cause some myomeres to exceed this threshold and others to fall below it. Because such variation is random with respect to body side, asymmetry would tend to occur. Low-plated stickleback would symmetrically fail to meet these thresholds relatively close to P6. Although this hypothesis does not explain asymmetry in P1-P3 or keel production, it does explain the peak in symmetrical loss of plates in CL

154

heterozygotes by P22 and the high incidence of CL asymmetry. It also explains why structural plates are symmetric: protein levels are always so high in these myomeres that noise cannot disrupt plate formation. Why Eda protein concentrations would vary with distance from structural plates is not known, nor is it known why structural plate asymmetry would increase in fresh water fish. How this model can be reconciled with Bell’s (1981) model of paedomorphosis requires further investigation.

5.4.5 Conclusions

FA did not vary with genomic heterozygosity, and so the hypothesis that SGV is maintained in the marine environment through selection against developmental instability was not supported. Rather, heterozygosity at a major-effect locus was implicated in asymmetry. This contrasts with other studies that have found evidence that heterozygosity in particular proteins buffers against asymmetry (Leary et al. 1983, 1984). Because the Eda mutation underlies a regulatory region as opposed to a functional protein domain, this necessarily leads to questions about the types of major-effect loci and their impacts on asymmetry. By impacting active protein abundance, loss-of-function heterozygosity or heterozygosity at regulatory regions may affect asymmetry more than that for heterozygosity in protein-coding genes that results in proteins with slightly distinct functions. To date such comparisons have not been done. This study highlights the need to focus on more mechanistic-based hypotheses for addressing the origin and maintenance of asymmetry in nature. Despite considerable evidence of the significance of SGV in Eda for parallel evolution in freshwater populations of threespine stickleback, little is known about the maintenance of Eda in marine stickleback. The transporter hypothesis (Schluter and Conte 2009) and the cryptic genetic variation hypothesis (Colosimo et al. 2005) have been suggested, but in Chapter Four I found evidence that such variation is under selection in the marine environment. To date no explanation fits all instances of the low- plated allele in marine stickleback, and this study questions the “cryptic” nature of fully- plated heterozygotes. If heterozygosity at Eda increases asymmetry, yet polymorphisms at Eda persist in the marine environment, does asymmetry have any fitness effects? Is 155

asymmetry maintained despite its costs, due to other fitness consequences of Eda or its linked genes (e.g. Albert et al. 2008; Barrett et al. 2009b; Mills et al. 2014; Greenwood et al. 2016; Robertson et al. 2017; Chapter Four)? These issues are central to the causes and consequences of SGV, and thus the capacity of populations to canalize against genetic and environmental variation. This study revealed more canalization of structural plates in marine populations than has been reported for freshwater populations, which is in accordance with increased asymmetry during directional selection. This reinforces the need to know the ancestral condition of stickleback when studying adaptive divergence. The next and final data chapter focuses on another phenotype which has only been studied in freshwater stickleback – and now that I know the genetic structuring of marine stickleback, I can choose a relatively appropriate contemporary population to act as the ancestral state.

156

The environment changes, but plasticity remains the same: mitochondrial biogenesis and its role in adaptive divergence

6.1 Introduction

Phenotypic plasticity, the capacity of an organism to develop different phenotypes under different contexts, is believed to play an important role during adaptation to new environments (Baldwin 1896; Bradshaw 1965; Agrawal 2001; Pigliucci 2001; West- Eberhard 2003; DeWitt and Scheiner 2004; Ghalambor 2007; Levis and Pfennig 2016; Schneider and Meyer 2017). Baldwin (1902) envisioned a two-step process of adaptation via plasticity: adaptive phenotypic change in new environments occurs within a generation due to pre-existing plasticity, permitting initial colonisation success (‘plasticity-mediated persistence’, Morris 2014); and plasticity subsequently evolves to further enhance fitness. Depending on the expression and utility of plasticity in the new environment, such evolution could increase, decrease, shift, or lose plasticity, or maintain plasticity until other phenotypes evolve (Morris 2014). Although evidence suggests that plasticity is important during colonisation (Thibert-Plante and Hendry 2011; Morris 2014), its role in ecological speciation is less clear. This is because plasticity can simultaneously result in populations exhibiting adaptive phenotypic divergence in divergent environments and permit gene flow such that adaptive genetic divergence is inhibited (Thibert-Plante and Hendry 2011). Identification of adaptively significant plastic phenotypes that have diverged in genetically-distinct ecotypes is a necessary first step in linking plasticity to population persistence and adaptive evolution. Temperature is one of the most important drivers of adaptation and macroevolutionary change (Benton 2009; Erwin 2009; Figueirido et al. 2011; Merilä and Hendry 2014; Davis et al. 2016). Because temperatures fluctuate daily and seasonally, organisms exhibit an array of highly plastic physiological traits for acclimating to and compensating for challenging temperatures (Podrabsky and Somero 2004; Somero 2010; O’Brien 2011; Tattersall et al. 2012; Yampolsky et al. 2014). Ectotherms in particular may experience internal temperature shifts of 20°C or more. Cold temperatures can

157

increase blood and membrane viscosity, reduce oxygen and metabolite diffusion, and lower ATP production below the levels necessary for basic biological function. Physiological plasticity provides a means of compensating for each of these effects, at least partially (Johnston and Dunn 1987; Ramløv 2000; O’Brien 2011; Dos Santos et al.

2013; Cooper et al. 2014). The critical thermal minimum (CTmin) for bodily function (Cowles and Bogert 1944; Currie et al. 1998) is a complex trait that is plastic (Martínez et al. 2016) and can evolve (Doughty 1994; Barrett et al. 2011; Wallace et al. 2014).

However, the underlying phenotypes responsible for adaptive divergence in CTmin and the loci responsible remain unknown (e.g. Coppe et al. 2013). Adaptation has occurred rapidly in freshwater populations of threespine stickleback (Gasterosteus aculeatus) along the Pacific coast from Alaska to Washington, with marine ancestors colonising and adapting within approximately 10 000 generations. These freshwater populations have evolved in parallel for a number of readily observable morphological traits, including reduced lateral bony plates and pelvic armature (Colosimo et al. 2005; Chan et al. 2010). However, fresh water also poses various physiological challenges, from coping with reduced salinity (McCairns et al. 2010) to physiologically responding to novel pathogens (Hohenlohe et al. 2010). Temperatures differ predictably between marine and freshwater environments, with lakes being colder during the winter than the ocean, on average (Lee and Bell 1999; Barrett et al. 2011). Whereas marine stickleback can migrate to their preferred temperatures as winter progresses, freshwater stickleback are limited by the temperature profiles of their lakes.

In response to these stressors, lower CTmin has evolved in freshwater populations relative to marine conspecifics (Barrett et al. 2011; Gibbons et al. 2016). Furthermore, marine stickleback placed in experimental freshwater ponds adapted rapidly to freshwater temperatures, presumably through selection acting on standing genetic variation (SGV) present in marine ancestors (Barrett and Schluter 2008; Barrett et al. 2011; Morris et al. 2014) or epigenetic modification (Shama et al. 2014, 2016; Gibbons et al. 2016). The underlying causes of increased cold tolerance in freshwater populations, however, remains unknown.

158

Jones et al. (2012b) reported a number of genomic regions in stickleback that have repeatedly undergone selection in the freshwater environment. One of these multi- genic regions included the gene peroxisome proliferator-activated receptor alpha a (PPARAa), although the precise locus under selection could not be determined. PPARAa regulates various genes involved in mitochondrial biogenesis and fatty acid metabolism (Li et al. 2005; Ventura-Clapier et al. 2008; McMullen et al. 2014). Morris et al. (2014) found that a suite of genes involved in mitochondrial activity, including PPARAa, were up-regulated in white muscle tissue under cold temperatures in freshwater but not in marine stickleback. Mitochondrial biogenesis is induced by cold temperatures in the pectoral muscle of freshwater stickleback (Orczewska et al. 2010). Therefore mitochondrial biogenesis may evolve in freshwater stickleback as a means of coping with the cold. However, mitochondrial biogenesis in marine stickleback has not been reported. In ectotherms, mitochondria could influence cold tolerance in several ways (Guderley and Johnston 1996; Moyes 2003; Guderley 2004; Lannig et al. 2005; Lucassen et al. 2006; Pörtner er al. 2007; O’Brien 2011; Bremer et al. 2012; Colinet et al. 2017). Most significantly, mitochondria synthesize ATP and aid in oxygen diffusion. Therefore mitochondrial biogenesis could provide more molecules of aerobic metabolic enzymes to compensate for slower ATP production, and/or greater lipid membrane concentration for improved oxygen diffusion. This could permit ATP-hungry tissues to meet their metabolic demands under increasingly cold conditions. I hypothesize that lower CTmin evolved in freshwater stickleback in association with greater plasticity in cold-induced mitochondrial biogenesis. I therefore predict the following: (1) Given the importance of mitochondrial biogenesis for coping with cold temperatures, marine and freshwater stickleback should both exhibit plasticity in mitochondrial volume density. That is, marine stickleback should be preadapted to fresh water by having an adaptively plastic phenotype. To test this, marine stickleback were challenged in salinities just above fresh water, to demonstrate that reduced salinity produces an adaptive response. (2) Because

CTmin is a measure of loss of equilibrium, and not death itself, ATP-hungry pectoral muscle will demonstrate this plastic compensatory response more than cardiac tissue, which is still able to function even after ecological death. (3) Cold-induced mitochondrial

159

biogenesis will be greater in the pectoral muscle of freshwater than marine stickleback. That is, plasticity will have evolved such that it has been enhanced, rather than lost in freshwater fish. This is because the upper thermal limit of marine and freshwater stickleback is the same, necessitating similar warm-acclimation responses in mitochondrial volume density (Barrett et al. 2011). If these predictions are supported, this study would fill a missing piece in the chain from genotype (Jones et al. 2012b) to gene expression (Morris et al. 2014) to complex physiological phenotypes (Barrett et al. 2011) for a plastic trait that has undergone adaptive divergence.

6.2 Materials and Methods

6.2.1 Sampling and experimental design

Live stickleback were acquired during fall 2014. Marine threespine stickleback were captured from Roquefeuil Bay and freshwater stickleback from Frederick Lake, both located on Vancouver Island, British Columbia, Canada (Table 6-1). Fish were initially maintained at the Bamfield Marine Sciences Centre in 113 L tanks. From November 5-13, 2014, salinity for the marine stickleback was reduced 2 ppt per day from 32 ppt to 14 ppt. On November 17, 2014, stickleback were transported to the University of Calgary’s Life and Environmental Sciences Animal Resource Centre (LESARC), where salinity was further reduced upon acclimation to 6 ppt by December 8, 2014. Fish were fed thawed chironomid larvae (bloodworms) ad libitum once daily. On January 19, 2015, stickleback began acclimation to one of three temperature treatments with two replicate tanks per treatment: cold (6.2°C + 0.7 SD), maintained by JBJ Arctica 1/10 hp titanium aquarium chillers (one per tank), moderate (due to limited freshwater fish, this treatment was done for marine stickleback only) (14.5°C + 0.4 SD), and warm temperature (20.6°C + 0.7 SD) maintained by Fluval E50 electronic heaters (1 per tank). Temperatures were measured daily. Four marine and four freshwater stickleback were housed in each tank, or eight marine stickleback in each room temperature tank. At the beginning of the temperature trial, stickleback ranged from 42-55 mm standard length. Stickleback were allowed to acclimate to their respective temperatures until March 31,

160

Population Latitude/ Ecotype Temp Cardiac Pectoral Avg. Avg. Longitude N N Cardiac Pectoral Vv Vv (+ SE) (+ SE) Frederick 48°51'16.41", FW 21°C 6 6 0.234 0.164 Lake, BC -125°1'30.46" (0.023) (0.040) 6°C 8 7 0.229 0.196 (0.013) (0.013) Roquefeuil 48°51'24.59", Marine 21°C 6 6 0.223 0.123 Bay, BC -125°6'41.20" (0.015) (0.015) 14.5°C 8 7 0.246 0.152 (0.013) (0.016) 6°C 8 8 0.262 0.195 (0.012) (0.009)

161

2015, when eighteen stickleback, randomly selected from all temperature treatments and populations, were euthanized in eugenol and dissected. On April 6, 2015, the remaining 20 fish were euthanized and dissected. The duration of the experiment exceeded the time required to induce mitochondrial biogenesis in freshwater stickleback from Alaska (Orczewska et al. 2010). One marine and one freshwater stickleback, both from the warm-acclimated treatment, died during the experiment, reducing the sample of each to seven. All procedures were approved by the Animal Care Committee (AUP AC13-0040).

6.2.2 Tissue fixation and imaging – cardiac muscle

In preparation of mitochondrial imaging, whole hearts were removed from each stickleback. Cardiac muscle was rinsed in a solution of chilled sodium cacodylate buffer and sliced once to expose the inner heart tissue. After rinsing, cardiac muscle was placed in 100 ml of fixative containing 2% paraformaldehyde and 2% glutaraldehyde in 0.1M sodium cacodylate (modified Karnovsky’s fixative). Suspended tissues were rotated in a rotator overnight and then shipped to the University of Alberta’s Transmission Electron Microscope Facility for further processing under the supervision of Dr. Nasser Tahbaz. Upon arrival, cardiac muscle was minced into < 1 mm cubes and washed in modified Karnovsky’s fixative. After 1 h in fixative, samples were rinsed with sodium cacodylate buffer, followed by 1 h of post-fixation with 1% osmium tetroxide in 0.1M sodium cacodylate buffer, on ice in the dark. Samples were rinsed three times in water and then stained in 2% uranyl acetate. Pellets were dehydrated using a series of rinses with increasing ethanol concentration, from 70-100%, followed by infiltration with propylene oxide and resin. Ultrathin sections were sliced with an Ultracut E (Reichert-Jung) and collected on 300-mesh copper grids. Images were taken using a Philips 410 transmission electron microscope (TEM). 10-15 micrographs at 11 500x final magnification were taken per individual using the aligned-systematics-quadrats-subsampling method (Cruz- Orive and Weibel 1981). A warm-acclimated marine and a warm-acclimated freshwater fish did not yield usable data for cardiac tissue, leaving six samples (Figure 6-1, Table 6- 1).

162

Figure 6-1 - Sample micrographs: (A) cold-acclimated pectoral muscle (2000x), (B) warm-acclimated pectoral muscle (2000x), (C) cold-acclimated cardiac muscle, (11 500x), and (D) warm-acclimated cardiac muscle (11 500x) from freshwater threespine stickleback.

163

6.2.3 Tissue fixation and imaging – pectoral muscle

Pectoral muscle was treated as above, with a few differences. During dissection and initial fixation, pectoral muscle remained connected to cartilage and was diced into 2 mm cubes. After fixation, samples were processed at the University of Calgary’s Microscopy and Imaging Facility by Dr. Wei-Xiang Dong. Pectoral tissue was collected on 200-mesh copper grids. I photographed 10-15 micrographs per individual on a Philips 410 TEM using the aligned-systematics-quadrats-subsampling method (Cruz-Orive and Weibel 1981), which were viewed at 2000x magnification (Figure 6-1, Figure 6-2). Three stickleback (one each of warm-acclimated marine, warm-acclimated freshwater, and room-temperature marine) could not be processed by Dr. Dong for unknown reasons. However, the two fish that did not yield usable cardiac tissue did yield usable pectoral tissue (Table 6-1).

6.2.4 Statistics

Temperature-induced mitochondrial biogenesis was determined by estimating mitochondrial volume density for each fish. Between six and eight stickleback were assessed per temperature treatment for mitochondrial volume density. A total of 523 cardiac tissue and 380 pectoral tissue images were analysed. Counting mitochondria and measuring their surface area or volume from micrographs is complicated because they are three-dimensional and were sliced at random within three-dimensional muscle. To overcome these difficulties, a point-counting method was used on 10-15 micrographs per fish from one to two slices through muscle tissue (Weibel 1979; Cruz-Orive and Weibel 1981). Images were viewed in Stepanizer v.1.8 (Tschanz et al. 2011). For cardiac muscle, a square lattice containing 81 intersection points was placed over each image. Adjacent grid intersections were separated by 0.77 μm. For pectoral muscle, a 90-point square lattice was used, with adjacent intersections separated by 2.18 μm. Stepanizer reduced image quality substantially. To counter this, a screenshot was taken of an open Stepanizer image with the lattice overlaying the muscle tissue. This screenshot was viewed in PowerPoint and an identical lattice created. Images could then be viewed in PowerPoint,

164

Figure 6-2 - Micrographs of threespine stickleback pectoral tissue at different magnifications. (A) Cross-section of muscle fiber showing crystalline arrangements, 12 000x. (B) Longitudinal section showing banding patterns, 10 000x. (C) A mitochondrion surrounded by muscle fiber. “%”-shaped dots are unidentified parasitic cysts, ~10 000x. (D) Copper mesh showing a region of torn pectoral tissue, 200x. Each micrograph was taken from the same upper left corner of different hexagonal grids.

165

size-adjusted by 121% so that the scale bars were of a consistent size, and point counts conducted on these higher-quality images. Mitochondrial volume density (Vv) was calculated by counting the lattice intersection points that overlaid a mitochondrion (Weibel 1979). The number of points with mitochondria was summed across all micrographs for an individual fish and divided by the total number of points for that fish. Total point number could be less than the total number on the image if a section of image did not contain tissue, or if tissue folds caused dark bands that could not be viewed clearly. Vv is often size-corrected (Weibel 1979). Because pectoral muscle tissue could not be weighed, owing to its connection with other tissues, Vv was size corrected by dividing by the standard length of the fish. Both raw and size-corrected measures were used for subsequent statistics. Analysis of Variance (ANOVA) of cold and warm-acclimated fish was used to investigate ecotype-by-temperature interaction. To further investigate the nature of plasticity in marine stickleback, a linear regression was performed using all three temperature treatments. Finally, a Pearson correlation was used to estimate the association between cardiac and pectoral mitochondrial volume density.

6.3 Results

For cardiac muscle, Vv varied from 0.144 (warm-acclimated freshwater fish) to 0.323 (cold-acclimated marine fish), with an overall mean of 0.24 (based on 42 191 test points). Vv did not vary significantly between populations, temperatures, or with their interactions using raw data (Table 6-2). For size-corrected data Vv differed between populations (Table 6-2), with marine stickleback having higher mitochondrial volume densities relative to their size than freshwater stickleback. Including room temperature, marine stickleback did not exhibit significant plasticity, although there was a trend of

increasing Vv with decreasing temperature (raw data: F1,20 = 4.2, p = 0.053; size-

corrected data F1,20 = 2.9, p = 0.11). Vv for pectoral muscle ranged from 0.037 (cold-acclimated freshwater fish) to 0.298 (warm-acclimated freshwater fish) with a mean of 0.168 (based on 33 542 test points). The lowest-scoring fish was unusual, having half as many mitochondria as the 166

Table 6-2 - Results of Analysis of Variance comparing mitochondrial volume density in cardiac and pectoral tissue between ecotypes (marine or freshwater) and temperatures (6°C or 21°C). Results with the interaction term are shown. * indicates significant effect.

Factor Type III SS d.f. F p Cardiac raw data Temperature 0.00199 1 1.2 0.29 Ecotype 0.00149 1 0.5 0.36 Temp x Ecotype 0.00329 1 1.9 0.18 Residuals 0.04 24 Cardiac size-corrected data Temperature 0.0000005 1 0.7 0.40 Ecotype 0.0000041 1 5.8 0.02 * Temp x Ecotype 0.0000009 1 1.2 0.28 Residuals 0.000017 24 Pectoral raw data, outlier removed Temperature 0.01817 1 6.3 0.02 * Ecotype 0.00237 1 0.8 0.37 Temp x Ecotype 0.00274 1 0.9 0.34 Residuals 0.07 23 Pectoral size-corrected data, outlier removed Temperature 0.00000647 1 6.3 0.02 * Ecotype 0.00000004 1 0.04 0.84 Temp x Ecotype 0.00000128 1 1.2 0.28 Residuals 0.000023 23

167

next-lowest-scored fish. It was the only fish for which mitochondria could not be found in some micrographs. This fish was treated as an outlier and the data was compared with and without this fish. Cold treatment elevated the average Vv in both marine and freshwater fish (Figure 6-3, Table 6-1). Including room temperature, plasticity was

particularly strong in marine stickleback (raw data: F1,19 = 16.5, p = < 0.001; size-

corrected data F1,19 = 16.4, p = < 0.001). Population and its interaction with temperature did not significantly affect Vv (Table 6-2). Cardiac Vv and pectoral Vv were not significantly correlated (r = 0.31, t = 1.8, d.f. =30, p = 0.08) (Figure 6-4).

6.4 Discussion

Linking genotype to phenotype is an essential pursuit in evolutionary biology, and this is particularly challenging for physiological traits that have diverged during ecological speciation (Rogers and Bernatchez 2007; Pavey et al. 2010). Rapid adaptation to temperature has occurred in freshwater threespine stickleback, and sequencing, gene expression, and functional knowledge of mitochondria suggest that mitochondria are involved. Mitochondrial biogenesis and its associated effects have been studied in freshwater stickleback (Orczewska et al. 2010; O’Brien 2011; Teigen et al. 2015; Keenan et al. 2017) and the activity of aerobic metabolic enzymes in the cold, but not biogenesis, has been studied in anadromous stickleback (Guderley and Leroy 2001; Guderley et al. 2001; Guderley 2004). The results of this study complement these studies by demonstrating that both marine and freshwater stickleback (1) exhibit plasticity in mitochondrial volume density (2) in pectoral tissue but not in cardiac muscle, but (3) with no difference in the degree of plasticity expressed by each population.

6.4.1 Mitochondrial biogenesis as an adaptation to the cold

Mitochondrial biogenesis is well-established as an adaptively plastic response to cold temperatures in fishes. In particular, it is linked with increased ATP production and oxygen diffusion to compensate for the effects of the cold. Although I have demonstrated

168

Figure 6-3 - Boxplots of mitochondrial volume density for marine and freshwater threespine stickleback, acclimated to 6°C (green), 14.5°C (purple, marine only) or 21°C (peach), for (A) cardiac and (B) pectoral muscle tissue.

169

Figure 6-4 - The relation between pectoral and cardiac mitochondrial volume density. Ecotype (marine or freshwater (FW)) and temperature acclimation treatment are shown.

170

this plastic response in marine and freshwater populations of stickleback, the underlying consequences remain unknown because biogenesis can take one of two forms. Mitochondria can proliferate their membranes but not metabolic enzymes to compensate for oxygen diffusion alone (O’Brien 2011). Similarly, mitochondria can increase metabolic enzyme concentration without biogenesis occurring (Orczweska et al. 2010), which this study could not detect. However, given that pectoral mitochondrial biogenesis in the cold in freshwater stickleback is coupled with increased citrate synthase and cytochrome c oxidase activity and the up-regulation of aerobic metabolic and biogenesis- regulating genes (Orczewska et al. 2010; Morris et al. 2014), it is likely that the plasticity observed here involved compensation for metabolic depression and oxygen diffusion. My results are consistent with those reported in other studies. For freshwater stickleback from Alaska, 3-4 individuals were acclimated to each of 8°C or 20°C temperatures for nine weeks. Mitochondrial volume density in pectoral muscle differed from 12% under warm temperatures to 24% under cold temperatures (Orczewska et al. 2010) – somewhat more dramatic than but consistent with the 16% to 20% in freshwater and 12% to 20% in marine stickleback reported here. In contrast, liver tissue showed low mitochondrial volume densities of ~4-5% under both temperature treatments. This contrast demonstrates that not all tissue are capable of altering mitochondrial volume density with temperature. Indeed, I report no plasticity but high mitochondrial volume densities in cardiac muscle, which is consistent with a lack of change in the mitochondrial lipid composition of sea bream cardiac mitochondria (Trigari et al. 1992). Morris et al. (2014) speculated that mitochondrial biogenesis would be an important determiner of CTmin, based in part on the fact that PPARAa, an important gene for mitochondrial biogenesis that resides in a region of marine-freshwater genomic divergence (Jones et al. 2012b), was up-regulated in the cold in multiple freshwater but not marine stickleback populations. However, I report no difference in plasticity for biogenesis in the pectoral muscle of marine and freshwater stickleback. A closer examination of the Morris et al. (2014) dataset for genes implicated in biogenesis by O’Brien (2011) show that some genes involved in biogenesis were up-regulated in both marine and freshwater stickleback, notably mitochondrial transcription factor B1

171

(TFB1M) and mitochondrial DNA-directed RNA polymerase (POLRMT), while PPARAa, nuclear respiratory factor-1 (NRF-1), and mitochondrial transcription factor B2 (TFB2M) were up-regulated in freshwater stickleback only. Since the publication of Morris et al. (2014) MitoCarta 2.0 was released (Calvo et al. 2016) with an updated inventory of genes involved in mammalian mitochondrial function. Although similarities between the pathways of mammalian and fish mitochondrial function are unknown, 1044 mammalian genes had orthologues on the stickleback microarray from Morris et al. (2014). Of these 1082 mammalian orthologues and stickleback mitochondrial genes, 422 (39%) were upregulated in the cold in both marine and freshwater stickleback, 173 (16%) were up-regulated in the cold in freshwater stickleback only, and 52 (5%) were up- regulated in the cold in marine stickleback only. An additional 50 (5%) were up-regulated in the cold in marine but down-regulated in the cold in freshwater stickleback (dataset not provided for thesis). Given the large degree of similarities between marine and freshwater stickleback for mitochondrial function, it perhaps is not surprising that they would demonstrate similar plasticity in mitochondrial volume density. It is possible that freshwater CTmin has evolved through other physiological pathways (Kammer et al. 2011; Shama et al. 2014, 2016; Teigen et al. 2015; Keenan et al. 2017) – perhaps in genes uniquely up-regulated in freshwater stickleback (Morris et al. 2014). It is also possible that PPARAa has differentiated marine from freshwater stickleback for reasons other than mitochondrial biogenesis, such as fatty acid oxidation.

6.4.2 Phenotypic plasticity and adaptive divergence

Ecological speciation requires that populations colonise and survive in new environments. In some cases, evolutionary rescue through de novo mutations may have to occur to enable persistence (Barrett and Hendry 2012; Gonzalez et al. 2013); in others selection on SGV may permit a subset of individuals to survive (Barrett and Schluter 2008). Baldwin (1902) recognized a third possibility – that plasticity mediates persistence during colonisation by meeting the phenotypic demands imposed by the novel environment. Plasticity-mediated persistence can be demonstrated in several ways (Morris 2014). “Hard PMP” involves experimental evidence to show that without 172

plasticity, no persistence would be possible. “Soft PMP” involves indirect demonstrations that plasticity was likely involved in persistence. Here I have demonstrated that an important phenotype for cold acclimation, mitochondrial biogenesis, is plastically regulated in stickleback pectoral tissue, depending on temperature. Furthermore, this plasticity is present in both a marine and a freshwater population, under derived (fresh water) conditions. The use of a single marine population as typical of the ancestral state may be questionable given the preceding chapters, but there is one line of evidence to justify its inclusion: marine stickleback north of Oregon comprise a single genetic cluster (Chapter Four). It is more likely that this population is representative of the ancestral condition than if a comparable study had been done in Oregon or California, although caution must be exercised given the possibility that these marine stickleback have subsequently evolved. Under the assumption of limited marine stickleback evolution, the first colonisers of freshwater lakes appear to be preadapted to meeting at least some of the physiological challenges of temperature in a low salinity environment. Evidence for hard PMP is lacking (Morris 2014); mitochondrial biogenesis in stickleback is a prime candidate for further research into this important component of adaptive speciation. Baldwin (1902) also recognized that plasticity may not be sufficient for long-term persistence. Subsequent evolution may have to occur to enable persistence of populations in new environments (Chevin et al. 2013; Morris 2014). Lower CTmin has evolved in freshwater stickleback (Barrett et al. 2011), but the results of this study suggest that this has likely not occurred through the evolution of biogenesis in pectoral muscle, as plasticity has not been altered. This contrasts with the “flexible stem model” for stickleback morphology, wherein plasticity in trophic morphology evident in marine populations has become canalized in freshwater populations at either end of the plasticity spectrum, “locking in” benthic and limnetic stickleback adaptive phenotypes. However, once locked-in these trophic morphologies have been further enhanced by, presumably, subsequent mutation that increases their fit with the environment (Wund et al. 2008, 2012). Mitochondrial biogenesis may permit survival of marine colonisers in fresh water such that subsequent variation could be selected to improve performance during cold winter months, what Baldwin (1902) referred to as correlated variants. The nature of this

173

variation remains to be determined, but it could include increases in metabolic enzyme activity that are decoupled from biogenesis (O’Brien 2011), or increased abundance of epithelial calcium channels (Gibbons et al. 2016) that may partially compensate for salinity-temperature effects. Finding the link between genotype and physiological adaptations during ecological speciation remains challenging (Schneider and Meyer 2017), particularly given the high plasticity of physiological phenotypes. Lower CTmin in freshwater stickleback likely reflects the evolution of mitochondrial performance, as suggested by both genome scans and gene expression studies. I have ruled out changes in mitochondrial biogenesis in pectoral muscle as a contributor to divergence, but maintain that it was likely important during initial colonisation. Cold adaptation can also involve modifications to mitochondrial performance, so future work should compare aerobic metabolic capacity between marine and freshwater stickleback acclimated to the cold.

174

Conclusion

I began this thesis by highlighting three major assumptions about standing genetic variation that affect our interpretations of adaptive divergence. First, SGV is meaningful if only it was present in individuals that encountered the new environment. Ancestral and derived populations could both have the same allele but not through descent, if the same de novo mutation occurred in the derived population. I could not test this assumption in this thesis. One approach would be to have high coverage across the genome for different freshwater populations, and assess the existence of hard versus soft sweeps around loci with SGV in the marine environment. The shotgun genotyping approach that I used could not produce such coverage in the one freshwater population I sequenced. The second assumption was that “contemporary ancestors” have not undergone evolution since the separation of freshwater and marine lineages. If marine fish comprise a single population, and it has evolved, little can be learned about the direction of adaptation in freshwater populations. Characterization of the ancestral state is even more challenging if multiple marine populations exhibit local adaptation. This thesis has made several contributions in this context. First, I demonstrated that marine stickleback populations exhibit adaptive phenotypic variation. Body size and vertebral number both varied in accord with ecogeographic rules. Whether caused by genetic variation, plasticity, or both, ecogeographic rules have long been held as evidence for adaptation. Ideally, this finding would stimulate analysis using a protocol similar to Baumann et al. (2012), combining observations of wild fish with those of F1 individuals raised under different temperatures in a common garden environment. This would more reliably disentangle genetic from environmental effects that, for various reasons beyond my control, I was unable to do myself. If vertebral number continued to distinguish northern stickleback from southern stickleback, it would be interesting to conduct a QTL analysis on F2 hybrids, and/or sample more extensively to conduct a Genome-Wide Association Study identifying alleles underlying vertebral number variation. Regardless, freshwater stickleback vary in body size and vertebral number, and so determining the direction of evolution will require informed decisions about what constitutes the ancestral phenotype.

175

The second way in which it was apparent that marine stickleback have recently evolved involved variation in plate number. Eda is the single best example of parallel evolution from standing genetic variation, and yet this was a frankly surprising discovery, one that had me hitting the literature for other instances of low-plated marine stickleback (see the results of this in Chapter One) to justify to myself that this was not an outrageous finding. This variation is probably not clinal, as there are fully-plated marine populations in California. Whether this variation is related to different types of marine habitat

(estuaries, bays, open coastline, etc.) remains to be assessed. Not only did Eda vary, FSTQ at Stn382 and PST for plate number exceeded values expected from genetic drift. Eda is multifaceted and complicated – it should be no surprise that its existence in the marine environment is as complicated as in fresh water. Even without this evidence, the iconic image of the fully-plated marine stickleback evolving into numerous freshwater populations seems questionable, as the low-plated phenotype came from marine SGV. Why marine populations evolved in such interesting ways is another issue entirely – one complicated by four observations from this thesis: (1) Fully-plated fish that were heterozygous for Eda produced fewer plates than those that were homozygous for the fully-plated allele, reducing the “cryptic” nature of these heterozygotes and leading to questions about the fitness consequences of such heterozygosity. (2) Eda heterozygotes exhibited greater asymmetry at non-structural plates than either homozygote genotype, leading to questions about the fitness consequences of such asymmetry. (3) Eda was associated with body shape, either pleiotropically or due to its strong linkage with other genes. (4) All populations except Alaska were in Hardy-Weinberg equilibrium at Stn382 as adults, suggesting that heterozygotes are neither favoured nor selected against. Further research is required, but this highlights that the reasons for the commonness of fully- plated stickleback in the marine environment is poorly understood. Sequencing also suggested that marine stickleback have adapted to their local environments, with candidate loci for selection being identified. I will say nothing more about that aspect of the thesis, largely because lists of loci provide little information when the functional and ecological significances of those loci are unknown. Interestingly,

176

genes in Linkage Group IV was very prominent among these candidates, and this is the chromosome that contains Eda. It was also apparent that some traits of marine stickleback have likely not evolved. In particular, stickleback varied little in asymmetry of structural plates. This pattern is quite distinct from that observed for non-structural plates. Furthermore, this variation was lower than has been observed in several freshwater stickleback populations – perhaps because life in fresh water is more stressful, or selection in fresh water is stressful, to the point of reducing canalization in freshwater stickleback. I cannot definitively conclude that differences in canalization at structural plates have not evolved in marine stickleback, as buffering mechanisms against perturbations that reduced canalization may have evolved in some populations. This cryptic evolution (genetic or plastic compensation – Grether 2005; Morris and Rogers 2013), if it occurred, would not be detectable from phenotypes in the wild. Mitochondrial biogenesis also seems to be evolutionarily conservative. This plastic phenotype was, barring small samples, relatively consistent in pectoral muscle between marine and freshwater stickleback, suggesting that the ancestral phenotype has been retained in the derived population. It would be extremely interesting to follow up on this work in two ways: (1) by determining whether mitochondrial respiration differs between marine and freshwater populations, thereby accounting for differences in critical thermal minima (an experiment that can be subtly handled using the Oroboros device); and (2) by determining whether mitochondrial biogenesis is identical along a latitudinal gradient, or exhibits adaptive differences in marine populations and/or freshwater populations along the coast. Whether mitochondrial biogenesis is identical under saline and freshwater conditions in the marine population also warrants attention, as alteration of biogenesis by salinity would imply that biogenesis was a cryptic phenotype in the marine population. The third assumption I raised proposed that contemporary ancestors should not have structured populations, because structure would lead to the problem of having to determine which contemporary form represents the ancestors of the freshwater population of interest. Contrary to this assumption SGV differed between marine sites – with eight

177

sampling locations resolving into five genetic clusters. Furthermore, the northern cluster was consistent with post-glacial range expansion, whereas the southern populations have been known to shift their ranges with past climate change. Thus, freshwater populations may have diverged from marine populations that no longer reside in the vicinity. For instance, freshwater populations occur in Baja California, Mexico, further south than any current marine population. Consequently, sampling marine populations as representative of the ancestors of geographically proximate freshwater populations may often not be justified. I also demonstrated that using the “wrong” marine population has significant implications for the detection of outliers of selection. These findings indicate that inference about ancestral states based on characteristics of contemporary samples must be tempered to the extent that correspondence between ancestral and contemporary phenotypes cannot be verified. Freshwater stickleback have evolved independently in numerous streams and lakes along the Pacific coast. Prior to this thesis, it was believed that the initial colonists of freshwater environments all contained the same subset of marine SGV, at least in principle. This thesis has called this assumption into question by demonstrating that multiple loosely connected marine populations exist. The consequences of such variation for the evolution of freshwater stickleback are numerous and require further attention. For instance, is it possible that some lakes do not currently contain threespine stickleback because initial colonists lacked the SGV necessary for adaptation, and failed to be rescued by de novo mutation? Was parallel evolution constrained by variation in the presence of SGV among the colonists of different lakes? To what extent does variation in polymorphic loci among freshwater lakes reflect differences in the initial frequency of alleles, and therefore the probability of fixation? How likely are rare alleles in the marine environment to be present in freshwater colonists? Might SGV that is adaptive in fresh water be transported among marine populations through marine transport? The results of this thesis therefore signals that careful attention is needed in the choice of comparator marine stickleback, as it affects inferences about the pace and form of evolution in the freshwater environment.

178

All of the above illustrate that marine stickleback are interesting in their own right and are not homogeneous phenotypically or genetically. They are an important food source for a host of commercially important fishes. If marine stickleback are not a single large panmictic population along the coast, but rather harbour local adaptations and possibly unique phenotypic variants, the question about their continued persistence becomes inevitable. Even seemingly large populations can collapse, as occurred in the White Sea. The proper management of marine stickleback requires knowing their genetic structuring to make informed decisions for the future. Looking ahead, it is clear that ecological information about marine stickleback is sorely needed. We know very little about environmental differences among the coastal habitats utilized by marine stickleback. How many marine populations remain in bays year round, and how many migrate? When they migrate, where do they go? Do they return to their natal site? Are there different marine stickleback ecotypes? How does coastal habitat relate to genetic and phenotypic variation? Freshwater stickleback have long been of interest due to their incredible array of genetic and phenotypic variation, but the marine environment clearly harbours an important amount of genetic and phenotypic variation that remains to be explored. My hope is that this thesis has contributed to the recognition of marine stickleback as interesting creatures in their own right. For threespine stickleback to truly be a “supermodel” (Gibson 2005) for testing evolutionary questions in nature, their evolutionary history needs to be properly characterized. Simplifying assumptions about marine threespine stickleback have permitted a great deal of useful work to be accomplished, but have also masked evolutionary processes occurring in the ocean that impact interpretations of evolution in fresh water. Marine stickleback are not some Platonic form – there is no ideal type that represents the ancestral condition. This thesis has unequivocally demonstrated that marine stickleback are genetically and phenotypically heterogeneous, both within and between populations, and that they are continuing to evolve. The significance of threespine stickleback for testing evolutionary questions will depend on our ability to incorporate this heterogeneity into our evolutionary models. I predict that contemporary ancestors will come under increasing scrutiny over the next several years. This thesis is

179

the first in what will hopefully be a series of voices calling attention to the importance of better understanding “contemporary ancestors” in order to better understand the significance of SGV during adaptation.

180

References

Ackerly KE, Ward AB (2016) How temperature-induced variation in musculoskeletal anatomy affects escape performance and survival of zebrafish (Danio rerio). Journal of Experimental Zoology Part A: Ecological Genetics and Physiology. 325: 25-40.

Agrawal AA (2001) Phenotypic plasticity in the interactions and evolution of species. Science. 294: 321-326.

Aguirre WE (2009) Microgeographical diversification of threespine stickleback: body shape-habitat correlations in a small, ecologically diverse Alaskan drainage. Biological Journal of the Linnean Society. 98: 139-151.

Aguirre WE, Bell MA (2012) Twenty years of body shape evolution in a threespine stickleback population adapting to a lake environment. Biological Journal of the Linnean Society. 105: 817-831.

Aguirre WE, Walker K, Gideon S (2014) Tinkering with the axial skeleton: vertebral number variation in ecologically divergent threespine stickleback populations. Biological Journal of the Linnean Society. 113: 204-219.

Albert AY, Sawaya S, Vines TH, Knecht AK, Miller CT, Summers BR, Balabhadra S, Kingsley DM, Schluter D (2008) The genetics of adaptive shape shift in stickleback: pleiotropy and effect size. Evolution. 62: 76-85.

Alho JS, Leinonen T, Merilä J (2011) Inheritance of vertebral number in the three-spined stickleback (Gasoterosteus aculeatus). PLOS One. 6:e19579.

Allen JA (1877) The influence of physical conditions in the genesis of species. Radical Review. 1: 108-140.

181

Allenbach DM (2011) Fluctuating asymmetry and exogenous stress in fishes: a review. Reviews in Fish Biology and Fisheries. 21: 355-376.

Anderson E, Stebbins Jr. GL (1954) Hybridization as an evolutionary stimulus. Evolution. 8: 378-388.

Angilletta Jr. MJ, Dunham AE (2003) The temperature-size rule in ectotherms: simple evolutionary explanations may not be general. American Naturalist. 162: 332-342.

Aparicio JM, Bonal R (2002) Why do some traits show higher fluctuating asymmetry than others? A test of hypotheses with tail feathers of birds. Heredity. 89: 139-144.

Baker JA, Foster SA, Bell MA (1995) Armor morphology and reproductive output in threespine stickleback, Gasterosteus aculeatus. In: Ecomorphology of fishes. Eds. JJ Luczkovich, PJ Motta, SF Norton, KF Liem. Springer: Netherlands. pp. 225-233.

Baker JA, Heins DC, Foster SA, King RW (2008) An overview of life-history variation in female threespine stickleback. Behaviour. 145: 579-602.

Baldwin JM (1896) A new factor in evolution. American Naturalist. 30: 441-451.

Baldwin JM (1902) Development and evolution. The MacMillan Company, New York.

Bańbura J (1994) Lateral plate morph differentiation of freshwater and marine populations of the three-spined stickleback, Gasterosteus aculeatus, in Poland. Journal of Fish Biology. 44: 773-783.

Barrett RDH, Hendry AP (2012) Evolutionary rescue under environmental change? In Behavioural responses to a changing world. Eds. U Candolin, BBM Wong. Oxford University Press: New York. pp. 216-233.

182

Barrett RDH, Schluter D (2008) Adaptation from standing genetic variation. Trends in Ecology and Evolution. 23: 38-44.

Barrett RDH, Rogers SM, Schluter D (2008) Natural selection on a major armor gene in threespine stickleback. Science. 322: 255-257.

Barrett RDH, Rogers SM, Schluter D (2009a) Environment specific pleiotropy facilitates divergence at the Ectodysplasin locus in threespine stickleback. Evolution. 63-11: 2831- 2837.

Barrett RDH, Vines TH, Bystriansky JS, Schulte PM (2009b) Should I stay or should I go? The Ectodysplasin locus is associated with behavioural differences in threespine stickleback. Biology Letters. 5: 788-791.

Barrett RDH, Paccard A, Healy TM, Bergek S, Schulte PM, Schluter D, Rogers SM (2011) Rapid evolution of cold tolerance in stickleback. Proceedings of the Royal Society B. 278: 233-238.

Barriga JP, Milano D, Cussac VE (2013) Variation in vertebral number and its morphological implication in Galaxias platei. Journal of Fish Biology. 83: 1321-1333.

Barry T (n.d.) Investigating the relationship between phenotypic covariation structure, population genetic structure and ecology in Gasterosteus aculeatus in Pender Harbour, BC. MSc Thesis Proposal, University of Calgary.

Bates D, Maechler M, Bolker B, Walker S (2015) Fitting linear mixed-effect models using lme4. Journal of Statistical Software. 67: 1-48.

183

Baumann H, Casián JAR, Conover DO (2012) Contrasting latitudinal variations in vertebral number and sex determination in Pacific versus Atlantic silverside fishes. Copeia. 2012: 341-350.

Baumgartner JV, Bell MA (1984) Lateral plate morph variation in California populations of the threespine stickleback, Gasterosteus aculeatus. Evolution. 38: 665-674.

Beasley DAE, Bonisoli-Alquati A, Mousseau TA (2013) The use of fluctuating asymmetry as a measure of environmentally induced developmental instability: A meta- analysis. Ecological Indicators. 30: 218-226.

Belk MC, Houston DD (2002) Bergmann’s rule in ectotherms: a test using freshwater fishes. American Naturalist. 160: 803-808.

Bell MA (1976) Evolution of phenotypic diversity in Gasterosteus aculeatus Superspecies on the Pacific coast of North America. Systematic Zoology. 25: 211-227.

Bell MA (1977) A late Miocene marine threespine stickleback, Gasterosteus aculeatus aculeatus, and its zoogeographic and evolutionary significance. Copeia. 2: 277-282.

Bell MA (1979) Low-plate morph of the threespine stickleback breeding in salt water. Copeia. 1979: 529-533.

Bell MA (1981) Lateral plate polymorphism and ontogeny of the complete plate morph of threespine sticklebacks (Gasterosteus aculeatus). 35: 67-74.

Bell MA, Foster SA (1994) Introduction to the evolutionary biology of the threespine stickleback. In: The evolutionary biology of the threespine stickleback. Eds. MA Bell, SA Foster. Oxford University Press: NY. pp. 1-27.

184

Bell MA, Aguirre WE, Buck NJ (2004) Twelve years of contemporary armor evolution in a threespine stickleback population. Evolution. 58: 814-824.

Bell MA, Stewart JD, Park PJ (2009) The world’s oldest fossil threespine stickleback fish. Copeia. 2009: 256-265.

Bell MA, Gangavalli AK, Bewick A, Aguirre WE (2010) Frequency of Ectodysplasin alleles and limited introgression between sympatric threespine stickleback populations. Environmental Biology of Fishes. 89: 189-198.

Benton MJ (2009) The red queen and the court jester: Species diversity and the role of biotic and abiotic factors through time. Science. 323: 728-732.

Bergstrom CA (2002) Fast-start swimming performance and reduction in lateral plate number in threespine stickleback. Canadian Journal of Zoology. 80: 207-213.

Bergstrom CA, Reimchen TE (2000) Functional implications of fluctuating asymmetry among endemic populations of Gasterosteus aculeatus. Behaviour. 137: 1097-1112.

Bergstrom CA, Reimchen TE (2003) Asymmetry in structural defenses: insights into selective predation in the wild. Evolution. 57: 2128-2138.

Bergstrom CA, Reimchen TE (2005) Habitat dependent associations between parasitism and fluctuating asymmetry among endemic stickleback populations. Journal of Evolutionary Biology. 18: 939-948.

Bergström U, Olsson J, Casini M, Eriksson BK, Fredriksson R, Wennhage H, Appelberg M (2015) Stickleback increase in the Baltic Sea - a thorny issue for coastal predatory fish. Estuarine, Coastal and Shelf Science. 163: 134-142.

185

Bernatchez L (2004) Ecological theory of adaptive radiation: an empirical assessment from coregonine fishes (Salmoniformes). In: Evolution illuminated: salmon and their relatives. Eds. AP Hendry, SC Stearns. Oxford University Press: NY. pp. 175-207

Bernatchez L (2016) On the maintenance of genetic variation and adaptation to environmental change: considerations from population genomics in fish. Journal of Fish Biology. 89: 2519-2556.

Berner D, Moser D, Roesti M, Buescher H, Salzburger W (2014) Genetic architecture of skeletal evolution in European lake and stream stickleback. Evolution. 68: 1792-1805.

Berry RJ, Jakobson ME, Peters J (1978) The house mice of the Faroe Islands: a study in microdifferentiation. Journal of Zoology. 185: 573-592.

Bertin L (1925) Recherches bionomiques, biométriques et systématiques sur les épinoches. Annals de l’Institut Océanographique Monaco. 2: 1-208.

Billerbeck JM, Ortí G, Conover DO (1997) Latitudinal variation in vertebral number has a genetic basis in the Atlantic silverside, Menidia menidia. Canadian Journal of Fisheries and Aquatic Sciences. 54: 1796-1801.

Black EA (1977) Population differentiation of the three-spined stickleback (Gasterosteus aculeatus). MSc thesis, University of British Columbia,

Blackburn TM, Gaston KJ, Loder N (1999) Geographic gradients in body size: a clarification of Bergmann’s rule. Diversity and Distributions. 5: 165-174.

Blouw DM (1996) Evolution of offspring desertion in a stickleback fish. Écoscience. 1: 18-24.

186

Blouw DM, Hagen DW (1990) Breeding ecology and evidence of reproductive isolation of a widespread stickleback fish (Gasterosteidae) in Nova Scotia, Canada. Biological Journal of the Linnean Society. 39: 195-217.

Borg B (1985) Field studies on three-spined sticklebacks in the Baltic. Behaviour. 93: 153-157.

Bottom D, Forsberg B (1978) The fishes of Tillamook Bay. Tillamook Estuary Study, Federal Aid Progress Reports Fisheries. F-100-R.

Bowles E, Johnston RA, Vanderzwan SL, Rogers SM (2016) Genetic basis for body size variation between an anadromous and two derived lacustrine populations of threespine stickleback Gasterosteus aculeatus in southwest Alaska. Current Zoology. 62: 71-78.

Bradic M, Teotónio H, Borowsky RL (2013) The population genomics of repeated evolution in the blind cavefish Astyanax mexicanus. Molecular Biology and Evolution. 30: 2383-2400.

Bradshaw AD (1965) Evolutionary significance of phenotypic plasticity in plants. Advances in Genetics. 13: 115-155.

Brainerd EL, Patek SN (1998) Vertebral column morphology, C-start curvature, and the evolution of mechanical defenses in Tetraodontiform fishes. Copeia. 1998: 971-984.

Brambilla A, Biebatch I, Bassano B, Bogliani G, von Hardenberg A (2014) Direct and indirect causal effects of heterozygosity on fitness-related traits in Alpine ibex. Proceedings of the Royal Society B. 282: 20141873.

Brawand D, Wagner CE, Li YI, Malinsky M, Keller I, Fan S, Simakov O, Ng AY, Lim ZW, Bezault E, Turner-Maier J, Johnson J, Alcazar R, Noh HJ, Russell P, Aken B,

187

Alföldi J, Amemiya C, Azzouzi N, Baroiller J-F, Barloy-Huber F, Berlin A, Bloomquist R, Carleton KL, Conte MA, D’Cotta H, Eshel O, Gaffney L, Galibert F, Gante HF, Gnerre S, Greuter L, Guyon R, Haddad NS, Haerty W, Harris RM, Hofmann HA, Hourlier T, Hulata G, Jaffe DB, Lara M, Lee AP, MacCallum I, Mwaiko S, Nikaido M, Nishihara H, Ozouf-Costaz C, Penman DJ, Przybylski D, Rakotomanga M, Renn SCP, Ribeiro FJ, Ron M, Salzburger W, Sanchez-Pulido L, Santos ME, Searle S, Sharpe T, Swofford R, Tan FJ, Williams L, Young S, Yin S, Okada N, Kocher TD, Miska EA, Lander ES, Venkatesh B, Fernald RD, Meyer A, Ponting CP, Streelman JT, Lindblad- Toh K, Seehausen O, Di Palma F (2014) The genomic substrate for adaptive radiation in African cichlid fish. Nature. 513: 375-381.

Bremer K, Monk CT, Gurd BJ, Moyes CD (2012) Transcriptional regulation of temperature-induced remodeling of muscle bioenergetics in goldfish. American Journal of Physiology - Regulatory, Integrative and Comparative Anatomy. 303: R150-R158.

Britten HB (1996) Meta-analyses of the association between multilocus heterozygosity and fitness. Evolution. 50: 2158-2164.

Brown WW, Cheng C (1946) Investigations into the food of the cod (Gadus callarius L.) off Bear Island, and of the cod and haddock (G. aeglefinus L.) off Iceland and the Murman Coast. Hull Bulletins of Marine Ecology. 3: 35-71.

Burke MK (2012) How does adaptation sweep through the genome? Insights from long- term selection experiments. Proceedings of the Royal Society B. 279: 5029-5038.

Buth DG, Haglund TR (1994) Allozyme variation in the Gasterosteus aculeatus complex. In: The evolutionary biology of the threespine stickleback. Eds. MA Bell, SA Foster. Oxford University Press: NY. pp. 61-84.

188

Cahill JA, Green RE, Fulton TL, Stiller M, Jay F, Ovsyanikov N, Salamzade R, St. John J, Stirling I, Slatkin M, Shapiro B (2013) Genomic evidence for island population conversion resolves conflicting theories of polar bear evolution. PLOS Genetics. 9: e1003345.

Calvo SE, Clauser KR, Mootha VK (2016) MitoCarta2.0: an updated inventory of mammalian mitochondrial proteins. Nucleic Acids Research. 44: D1251-D1257.

Canadian Endangered Species Conservation Council (CESCC) (2001) Wild Species 2000: The General Status of Species in Canada. Ottawa: Minister of Public Works and Government Services Canada.

Canadian Endangered Species Conservation Council (CESCC) (2002) Wild Species 2000: The General Status of Species in Canada. Updated database generated in 2002 for butterflies and freshwater fishes. Ottawa: Minister of Public Works and Government Services Canada.

Canadian Endangered Species Conservation Council (CESCC) (2006) Wild Species 2005: The General Status of Species in Canada. Ottawa: Minister of Public Works and Government Services Canada.

Cano JM, Mäkinen HS, Merilä J (2008) Genetic evidence for male-biased dispersal in the three-spined stickleback (Gasterosteus aculeatus). Molecular Ecology. 17: 3234-3242.

Carchini G, Chiarotti F, Di Domenico M, Mattoccia M, Paganotti G (2001) Fluctuating asymmetry, mating success, body size and heterozygosity in Coenagrion scitulum (Rambur) (Odonata: Coenagrionidae). Animal Behaviour. 61: 661-669.

189

Carvalho GR, Hauser L, Martinsohn J, Naish K (2016) Fish, genes and genomes: contributions to ecology, evolution and management. Journal of Fish Biology. 89: 2471- 2478.

Catchen J, Amores A, Hohenlohe P, Cresko W, Postlethwait JH (2011) Stacks: building and genotyping loci de novo from short-read sequences. G3: Genes, Genomes, Genetics. 1: 171-182.

Catchen J, Bassham S, Wilson T, Currey M, O’Brien C, Yeates Q, Cresko WA (2013a) The population structure and recent colonization history of Oregon threespine stickleback determined using restriction-site associated DNA-sequencing. Molecular Ecology. 22: 2864-2883.

Catchen J, Hohenlohe PA, Bassham S, Amores A, Cresko WA (2013b) Stacks: an analysis tool set for population genomics. Molecular Ecology. 22: 3124-3140.

Chan YF, Marks ME, Jones FC, Villarreal Jr. G, Shapiro MD, Brady SD, Southwick AM, Absher DM, Grimwood J, Schmutz J, Myers RM, Petrov D, Jónsson B, Schluter D, Bell MA, Kingsley DM (2010) Adaptive evolution of pelvic reduction in sticklebacks by recurrent deletion of a Pitx1 enhancer. Science. 327: 302-305.

Chevin L-M, Gallet R, Gomulkiewicz R, Holt RD, Fellous S (2013) Phenotypic plasticity in evolutionary rescue experiments. Philosophical Transactions of the Royal Society B. 386: 20120089.

Chomczynski P, Sacchi N (1987) Single-step method of RNA isolation by acid guanidinium thiocyanate-phenol-chloroform extraction. Analytical Biochemistry. 162: 156-159.

190

Clarke GM (1993) The genetic basis of developmental stability. I. Relationships between stability, heterozygosity and genomic coadaptation. Genetica. 89: 15-23.

Colinet H, Renault D, Roussel D (2017) Cold acclimation allows Drosophila flies to maintain mitochondrial functioning under cold stress. Insect Biochemistry and Molecular Biology. 80: 52-60.

Colosimo PF, Hosemann KE, Balabhadra S, Villarreal Jr. G, Dickson M, Grimwood J, Schmutz J, Myers RM, Schluter D, Kingsley DM (2005) Widespread parallel evolution in sticklebacks by repeated fixation of ectodysplasin alleles. Science. 307: 1928-1933.

Cooper BS, Hammad LA, Montooth KL (2014) Thermal adaptation of cellular membranes in natural populations of Drosophila melanogaster. Functional Ecology. 28: 886-894.

Coppe A, Agostini C, Marino IA, Zane L, Bagelloni L, Bortoluzzi S, Patarnello T (2013) Genome evolution in the cold: Antarctic icefish muscle transcriptome reveals selective duplications increasing mitochondrial function. Genome Biology and Evolution. 5: 45- 60.

Cowen RK, Chiarella LA, Gomez CJ, Bell MA (1991) Offshore distribution, size, age, and lateral plate variation of late larval/early juvenile sticklebacks (Gasterosteus) off the Atlantic coast of New Jersey and New York. Canadian Journal of Fisheries and Aquatic Sciences. 48: 1679-1684.

Cowles RB, Bogert CM (1944) A preliminary study of the thermal requirements of desert reptiles. Bulletin of the American Museum of Natural History. 83: 265-296.

Cresko WA, Amores A, Wilson C, Murphy J, Currey M, Phillips P, Bell MA, Kimmel CB, Postlethwait JH (2004) Parallel genetic basis for repeated evolution of armor loss in

191

Alaskan threespine stickleback populations. Proceedings of the National Academy of Sciences USA. 101: 6050-6055.

Crivelli AJ, Britton RH (1987) Life history adaptations of Gasterosteus aculeatus in a Mediterranean wetland. Environmental Biology of Fishes. 18: 109-125.

Cruz-Orive L-M, Weibel ER (1981) Sampling designs for stereology. Journal of Microscopy. 122: 235-257.

Currie RJ, Bennett WA, Beitinger TL (1998) Critical thermal minima and maxima of three freshwater game-fish species acclimated to constant temperatures. Environmental Biology of Fishes. 51: 187-200.

Cuthill IC, Stevens M, Windsor AMM, Walker HJ (2006) The effects of pattern symmetry on detection of disruptive and background-matching coloration. Behavioral Ecology. 17: 828-832.

Dartnall HJG, Walkey M (1979) Parasites of marine stickleback. Journal of Fish Biology. 14: 471-474.

Davis KE, Hill J, Astrop TI, Wills MA (2016) Global cooling as a driver of diversification in a major marine clade. Nature Communications. 7: 13003.

Deagle BE, Reimchen TE, Levin DB (1996) Origins of endemic stickleback from the Queen Charlotte Islands: mitochondrial and morphological evidence. Canadian Journal of Zoology. 74: 1045-1056.

Deagle BE, Jones FC, Chan YF, Absher DM, Kingsley DM, Reimchen TE (2012) Population genomics of parallel phenotypic evolution in stickleback across stream - lake ecological transitions. Proceedings of the Royal Society B. 279: 1277-1286.

192

Deagle BE, Jones FC, Absher DM, Kingsley DM, Reimchen TJ (2013) Phylogeography and adaptation genetics of stickleback from the Haida Gwaii archipelago revealed using genome-wide single nucleotide polymorphism genotyping. Molecular Ecology. 22: 1917- 1932.

Debat V, Peronnet F (2013) Asymmetric flies: the control of developmental noise in Drosophila. Fly. 7: 70-77.

De Coster G, Van Dongen S, Malaki P, Muchane M, Alcántara-Exposito A, Matheve H, Lens L (2013) Fluctuating asymmetry and environmental stress: understanding the role of trait history. PLOS One. 8: e57966.

DeFaveri J, Merilä J (2013) Evidence for adaptive phenotypic differentiation in Baltic Sea sticklebacks. Journal of Evolutionary Biology. 26: 1700-1715.

DeFaveri J, Merilä J (2015) Temporal stability of genetic variability and differentiation in the three-spined stickleback (Gasterosteus aculeatus). PLOS One. 10: e0123891.

DeFaveri J, Shikano T, Shimada Y, Goto A, Merilä J (2011) Global analysis of genes involved in freshwater adaptation in threespine sticklebacks (Gasterosteus aculeatus). Evolution. 65-6: 1800-1807.

DeFaveri J, Jonsson PR, Merilä J (2013a) Heterogeneous genomic differentiation in marine threespine sticklebacks: adaptation along an environmental gradient. Evolution. 67-9: 2530-2546.

DeFaveri J, Shikano T, Shimada Y, Merilä J (2013b) High degree of genetic differentiation in marine three-spined sticklebacks (Gasterosteus aculeatus). Molecular Ecology. 22: 4811-4828.

193

Delph LF, Kelly JK (2014) On the importance of balancing selection in plants. New Phytologist. 201: 45-56.

Demchuk A, Ivanov M, Ivanova T, Polyakova N, Mas-Martí E, Lajus D (2015) Feeding patterns in seagrass beds of three-spined stickleback Gasterosteus aculeatus juveniles at different growth stages. Journal of the Marine Biological Association of the United Kingdom. 95: 1635-1643.

Dennenmoser S, Vamosi SM, Nolte AW, Rogers SM (2017) Pool-seq reveals adaptive genomic divergence under high gene flow between freshwater and brackish-water ecotypes of prickly sculpin (Cottus asper). Molecular Ecology. 26: 25-42.

Des Roches S, Robertson JM, Harmon LJ, Rosenblum EB (2011) Ecological release in White Sands lizards. Ecology and Evolution. 1: 571-578.

DeWitt TJ, Sheiner SM (2004) Phenotypic plasticity: functional and conceptual approaches. Oxford University Press: USA.

Dittmar EL, Oakley CG, Conner JK, Gould BA, Schemske DW (2016) Factors influencing the effect size distribution of adaptive substitutions. Proceedings of the Royal Society B. 283: 20153065.

Do C, Waples RS, Peel D, Macbeth GM, Tillett BJ, Ovenden JR (2014) NeEstimator v2: a re-implementation of software for the estimation of contemporary effective population size (Ne) from genetic data. Molecular Ecology Resources. 14: 209-214.

Domingues VS, Poh Y-P, Peterson BK, Pennings PS, Jensen JD, Hoekstra HE (2012) Evidence of adapation from ancestral variation in young populations of beach mice. Evolution. 66-10: 3209-3223.

194

Dos Santos RS, Galina A, Da-Silva WS (2013) Cold acclimation increases mitochondrial oxidative capacity without inducing mitochondrial uncoupling in goldfish white skeletal muscle. Biology Open. 2: 82-87.

Doughty P (1994) Critical thermal minima of garter snakes (Thamnophis) depend on species and body size. Copeia. 1994: 537-540.

Dray S, Dufour AB (2007) The ade-4 package: implementing the duality diagram for ecologists. Journal of Statistical Software. 22: 1-20.

Drevecky CJ, Falco R, Aguirre WE (2013) Genetic divergence of a sympatric lake- resident-anadromous three-spined stickleback Gasterosteus aculeatus species pair. Journal of Fish Biology. 83: 111-132.

Ellis RH (2002) Fish use of Tillamook Bay. Synthesis report for monitoring conducted 1999 through 2001. Tillamook County Estuary Partnership. 111 pp.

Elmer KR, Meyer A (2011) Adaptation in the age of ecological genomics: insights from parallelism and convergence. Trends in Ecology and Evolution. 26: 298-306.

Elshire RJ, Glaubitz JC, Sun Q, Poland JA, Kawamoto K, Buckler ES, Mitchell SE (2011) A robust, simple Genotyping-by-Sequencing (GBS) approach for high diversity species. PLOS One. 6: e19379.

Erwin DH (2009) Climate as a driver of evolutionary change. Current Biology. 19: R575- R583.

Fahy WE, O’Hara RK (1977) Does salinity influence the number of vertebrae developing in fishes? Conseil Permanent International pour l’Exploration de la Mer. 37: 156-161.

195

Félix M-A, Barkoulas M (2015) Pervasive robustness in biological systems. Nature Reviews Genetics. 16: 483-496.

Ferchaud A-L, Hansen MM (2016) The impact of selection, gene flow and demographic history on heterogeneous genomic divergence: three-spine sticklebacks in divergent environments. Molecular Ecology. 25: 238-259.

Feulner PGD, Chain FJJ, Panchal M, Eizaguirre C, Kalbe M, Lenz TL, Mundry M, Samonte IE, Stoll M, Milinski M, Reusch TBH, Bornberg-Bauer E (2013) Genome-wide patterns of standing genetic variation in a marine population of three-spined sticklebacks. Molecular Ecology. 22: 635-649.

Figueirido B, Janis CM, Pérez-Claros JA, De Renzi M, Palmqvist P (2011) Cenozoic climate change influences mammalian evolutionary dynamics. Proceedings of the National Academy of Sciences USA. 109: 722-727.

Flicek P, Amode MR, Barrell D, Beal K, Billis K, Brent S, Carvalho-Silva D, Clapham P, Coates G, Fitzgerald S (2014) Ensembl 2014. Nucleic Acids Research. 42: D749-D755.

Forsman A, Herrström J (2004) Asymmetry in size, shape, and color impairs the protective value of conspicuous color patterns. Behavioral Ecology. 15: 141-147.

Foster SA, Baker JA, Bell MA (2003) The case for conserving threespine stickleback populations. Fisheries. 28: 10-18.

Freeman VJ (1951) Studies on the virulence of the bacteriophage-infected strains of Corynebacterium diphtheriae. Journal of Bacteriology. 61: 675-688.

196

Fuiman LA, Batty RS (1997) What a drag it is getting cold: partitioning the physical and physiological effects of temperature on fish swimming. Journal of Experimental Biology. 200: 1745-1755.

Gach MH, Reimchen TE (1989) Mitochondrial DNA patterns among endemic stickleback from the Queen Charlotte Islands: a preliminary survey. Canadian Journal of Zoology. 67: 1324-1328.

Garside ET (1966) Developmental rate and vertebral number in salmonids. Journal of the Fisheries Research Board of Canada. 23: 1537-1551.

Garside ET, Hamor T (1973) Meristic variation in the threespine stickleback, Gasterosteus aculeatus L., in eastern Canadian waters. Canadian Journal of Zoology. 51: 547-551.

Ghalambor CK, McKay JK, Carroll SP, Reznick DN (2007) Adaptive versus non- adaptive phenotypic plasticity and the potential for contemporary adaptation in new environments. Functional Ecology. 21: 394-407.

Gibbons TC, Rudman SM, Schulte PM (2016) Responses to simulated winter conditions differ between threespine stickleback ecotypes. Molecular Ecology. 25: 764-775.

Gibson G (2005) The synthesis and evolution of a supermodel. Science. 307: 1891.

Gibson G, Dworkin I (2004) Uncovering cryptic genetic variation. Nature Reviews Genetics. 5: 681-690.

Glazer AM, Cleves PA, Erickson PA, Lam AY, Miller CT (2014) Parallel developmental genetic features underlie stickleback gill raker evolution. EvoDevo. 5: 19.

197

Gompel N, Prud’homme B (2009) The causes of repeated genetic evolution. Developmental Biology. 332: 36-47.

Gonzalez A, Ronce O, Ferrière R, Hochberg ME (2013) Evolutionary rescue: an emerging focus at the intersection between ecology and evolution. Philosophical Transactions of the Royal Society B. 368: 20120079.

Goudet J (2005) Hierfstat, a package for R to compute and test hierarchical F-statistics. Molecular Ecology Resources. 5: 184-186.

Grant NE (2009) Changes in cover and use of Zostera marina habitat in Elkhorn Slough, California. MSc Thesis, California State University Monterey Bay.

Greenwood AK, Mills MG, Wark AR, Archambeault SL, Peichel CL (2016) Evolution of schooling behavior in threespine sticklebacks is shaped by the Eda gene. Genetics. 203: 677-681.

Grether GF (2005) Environmental change, phenotypic plasticity, and genetic compensation. American Naturalist. 166: E115-E123.

Guderley H (2004) Metabolic responses to low temperature in fish muscle. Biology Reviews. 79: 409-427.

Guderley H, Johnston IA (1996) Plasticity of fish muscle mitochondria with thermal acclimation. Journal of Experimental Biology. 199: 1311-1317.

Guderley H, Leroy PH (2001) Family origin and the response of threespine stickleback, Gasterosteus aculeatus, to thermal acclimation. Journal of Comparative Physiology B. 171: 91-101.

198

Günther A (1862) Catalogue of the fishes in the British Museum, Volume Fourth. London: British Museum (Natural History).

Hagen DW (1967) Isolating mechanisms in threespine stickleback (Gasterosteus). Journal of the Fisheries Research Board of Canada. 24: 1637-1692.

Hagen DW, Gilbertson LG (1972) Geographic variation and environmental selection in Gasterosteus aculeatus L. in the Pacific Northwest, America. Evolution. 26: 32-51.

Hagen DW, Moodie GEE (1981) Polymorphism for plate morphs in Gasterosteus aculeatus on the east coast of Canada and an hypothesis for their global distribution. Canadian Journal of Zoology. 60: 1032-1042.

Haglund T, Buth DG, Blouw DM (1990) Allozyme variation and the recognition of the “white stickleback”. Biochemical Systematics and Evolution. 18: 559-563.

Haglund TR, Buth DG, Lawson R (1992) Allozyme variation and phylogenetic relationships of Asian, North American, and European populations of the threespine stickleback, Gasterosteus aculeatus. Copeia. 1992: 432-443.

Hare MP, Nunney L, Schwartz MK, Ruzzante DE, Burford M, Waples RS, Ruegg K, Palstra F (2011) Understanding and estimating effective population size for practical application in marine species management. Conservation Biology. 25: 438-449.

Hartl GB, Suchentrunk F, Willing R, Petznek R (1995) Allozyme heterozygosity and fluctuating asymmetry in the brown hare (Lepus europaeus): a test of the developmental homeostasis hypothesis. Philosophical Transactions of the Royal Society B. 350: 313- 323.

199

Hartley CJ, Newcomb RD, Russell RJ, Yong CG, Stevens JR, Yeates DK, La Salle J, Oakeshott JG (2006) Amplification of DNA from preserved specimens shows blowflies were preadapted for the rapid evolution of insecticide resistance. Proceedings of the National Academy of Sciences USA. 103: 8757-8762.

Harvey CJ, Ruggerone GT, Rogers DE (1997) Migrations of the three-spined stickleback, nine-spined stickleback and pond smelt in the Chignik catchment, Alaska. Journal of Fish Biology. 50: 1133-1137.

Hedrick PW (2013) Adaptive introgression in animals: examples and comparison to new mutation and standing variation as sources of adaptive variation. Molecular Ecology. 22: 4606-4618.

HELCOM (2006) Assessment of coastal fish in the Baltic Sea. Baltic Sea Environment Proceedings. No. 103 A.

Hemmer-Hansen J, Therkildsen NO, Pujolar JM (2014) Population genomics of marine fishes: Next-Generation prospects and challenges. The Biological Bulletin. 227: 117-132.

Hendrickx F, Maelfait J-P, Lens L (2003) Relationship between fluctuating asymmetry and fitness within and between stressed and unstressed populations of the wolf spider Pirata piraticus. Journal of Evolutionary Biology. 16: 1270-1279.

Hendry AP (2001) Adaptive divergence and the evolution of reproductive isolation in the wild: an empirical demonstration using introduced sockeye salmon. Genetica. 1: 515-534.

Hendry AP (2009) Ecological speciation! Or the lack thereof? Canadian Journal of Fisheries and Aquatic Sciences. 66: 1383-1398.

200

Hendry AP (2013) Key questions in the genetics and genomics of eco-evolutionary dynamics. Heredity. 111: 456-466.

Hermida M, Fernández C, Amaro R, San Miguel E (2002) Heritability and “evolvability” of meristic characters in a natural population of Gasterosteus aculeatus. Canadian Journal of Zoology. 80: 532-541.

Hermisson J, Pennings PS (2005) Soft sweeps: molecular population genetics of adaptation from standing genetic variation. Genetics. 169: 2335-2352.

Higham TE, Rogers SM, Langerhans RB, Jamniczky HA, Lauder GV, Stewart WJ, Martin CH, Reznick DN (2016) Speciation through the lens of biomechanics: locomotion, prey capture and reproductive isolation. Proceedings of the Royal Society B. 283: 20161294.

Higuchi M, Goto A (1996) Genetic evidence supporting the existence of two distinct species in the genus Gasterosteus around Japan. Environmental Biology of Fishes. 47: 1- 16.

Higuchi M, Sakai H, Goto A (2014) A new threespine stickleback, Gasterosteus nipponicus sp nov. (Teleostei: Gasterosteidae), from the Japan Sea region. Ichthylogical Research. 61: 341-351.

Hohenlohe PA, Bassham S, Etter PD, Stiffler N, Johnson EA, Cresko WA (2010) Population genomics of parallel adaptation in threespine stickleback using sequenced RAD tags. PLOS Genetics. 6: e1000862.

Hopkins K, Moss BR, Gill AB (2011) Increased ambient temperature alters the parental care behaviour and reproductive success of the three-spined stickleback (Gasterosteus aculeatus). Environmental Biology of Fishes. 90: 121-129.

201

Hubbs CL (1922) Variations in the number of vertebrae and other meristic characters of fishes correlated with temperature of water during development. American Naturalist. 56: 360-372.

Hunt von Herbing I (2002) Effects of temperature on larval fish swimming performance: the importance of physics to physiology. Journal of Fish Biology. 61: 865-876.

Ishikawa M, Mori M, Nagata Y (2006) Intraspecific differences in patterns of courtship behaviours between the Pacific Ocean and Japan Sea forms of the three-spined stickleback Gasterosteus aculeatus. Journal of Fish Biology. 69: 938-944.

Ishikawa A, Kusakabe M, Kume M, Kitano J (2016) Comparison of freshwater tolerance during spawning migration between two sympatric Japanese marine threespine stickleback species. Evolutionary Ecology Research. 17: 525-534.

Jacobs DK, Haney TA, Louie KD (2004) Genes, diversity, and geologic process on the Pacific coast. Annual Review of Earth Planet Science. 32: 601-652.

Jamieson IG, Blouw DM, Colgan PW (1992a) Parental care as a constraint on male mating success in fishes: a comparative study of threespine and white sticklebacks. Canadian Journal of Zoology. 70: 956-962.

Jamieson IG, Blouw DM, Colgan PW (1992b) Field observations on the reproductive biology of a newly discovered stickleback (Gasterosteus). Canadian Journal of Zoology. 70: 1057-1063.

Jamniczky HA, Barry TN, Rogers SM (2015) Eco-evo-devo in the study of adaptive divergence: examples from threespine stickleback (Gasterosteus aculeatus). Integrative and Comparative Biology. 55: 166-178.

202

Jarosz DF, Lindquist S (2010) Hsp90 and environmental stress transform the adaptive value of natural genetic variation. Science. 330: 1820-1824.

Johnson LS, Taylor EB (2004) The distribution of divergent mitochondrial DNA lineages of threespine stickleback (Gasterosteus aculeatus) in the northeastern Pacific basin: post- glacial dispersal and lake accessibility. Journal of Biogeography. 31: 1073-1083.

Johnston IA, Dunn J (1987) Temperature acclimation and metabolism in ectotherms with particular reference to teleost fish. Symposia of the Society for Experimental Biology. 41: 67-93.

Jombart T (2008) adegenet: a R package for the multivariate analysis of genetic markers. Bioinformatics. 24: 1403-1405.

Jombart T, Devillard S, Balloux F (2010) Discriminant analysis of principal components: a new method for the analysis of genetically structured populations. BMC Genetics. 11:94.

Jones DH, John AWG (1978) A three-spined stickleback, Gasterosteus aculeatus L. from the North Atlantic. Journal of Fish Biology. 13: 231-236.

Jones FC, Brown C, Pemberton JM, Braithwaite VA (2006) Reproductive isolation in a threespine stickleback hybrid zone. Journal of Evolutionary Biology. 19: 1531-1544.

Jones FC, Chan YF, Schmutz J, Grimwood J, Brady SD, Southwick AM, Absher DM, Myers RM, Reimchen TE, Deagle BE, Schluter D, Kingsley DM (2012a) A genome- wide SNP genotyping array reveals patterns of global and repeated species-pair divergence in sticklebacks. Current Biology. 22: 83-90.

203

Jones FC, Grabherr MG, Chan YF, Russell P, Mauceli E, Johnson J, Swofford R, Pirun M, Zody MC, White S, Birney E, Searle S, Schmutz J, Grimwood J, Dickson MC, Myers RM, Miller CT, Summers BR, Knecht AK, Brady SD, Zhang H, Pollen AA, Howes T, Amemiya C, Broad Institute Genome Sequencing Platform and Whole Genome Assembly Team, Lander ES, Di Palma F, Lindblad-Toh K, Kingsley DM (2012b) The genomic basis of adaptive evolution in threespine stickleback. Nature. 484: 55-61.

Jordan DS (1891) Relations of temperature to vertebrae among fishes. Science. 18: 104- 107.

Jordan DS (1905) A guide to the study of fishes, Volume I. Henry Holt and Company: New York.

Jurvelius J, Leinikki J, Mamylov V, Pushkin S (1996) Stock assessment of pelagic three- spined stickleback (Gasterosteus aculeatus): a simultaneous up- and down-looking echo- sounding study. Fisheries Research. 27: 227-241.

Kaeuffer R, Peichel CL, Bolnick DI, Hendry AP (2012) Parallel and nonparallel aspects of ecological, phenotypic, and genetic divergence across replicate population pairs of lake and stream stickleback. Evolution. 66: 402-418.

Kammer AR, Orczewska JI, O’Brien KM (2011) Oxidative stress is transient and tissue specific during cold acclimation of threespine stickleback. Journal of Experimental Biology. 214: 1248-1256.

Kamvar ZN, Tabima JF, Grünwald NJ (2014) Poppr: an R package for genetic analysis of populations with clonal, partially clonal, and/or sexual reproduction. PeerJ. 2: e281.

Kamvar ZN, Brooks JC, Grünwald NJ (2015) Novel R tools for analysis of genome-wide population genetic data with emphasis on clonality. Frontiers in Genetics. 6: 208.

204

Kark S, Safriel UN, Tabarroni C, Randi E (2001) Relationship between heterozygosity and asymmetry: a test across the distribution range. Heredity. 86: 119-127.

Karve AD, von Hippel FA, Bell MA (2008) Isolation between sympatric anadromous and resident threespine stickleback species in Mud Lake, Alaska. Environmental Biology of Fishes. 81: 287-296.

Kassen R, Schluter D, McPhail JD (1995) Evolutionary history of threespine sticklebacks (Gasterosteus spp.) in British Columbia: insights from a physiological clock. Canadian Journal of Zoology. 73: 2154-2158.

Kaufman R (2014) Standing genetic variation and latitudinal clines in threespine stickleback. ECOL 528 paper, University of Calgary.

Keenan K, Hoffman M, Dullen K, O’Brien KM (2017) Molecular drivers of mitochondrial membrane proliferation in response to cold acclimation in threespine stickleback. Comparative Biochemistry and Physiology, Part A. 203: 109-114.

Kimura M, Ohta T (1971) Theoretical aspects of population genetics. Princeton University Press: USA.

Kitano J, Mori S, Peichel CL (2007) Phenotypic divergence and reproductive isolation between sympatric forms of Japanese threespine sticklebacks. Biological Journal of the Linnean Society. 91: 671-685.

Kitano J, Bolnick DI, Beauchamp DA, Mazur MM, Mori S, Nakano T, Peichel CL (2008) Reverse evolution of armor plates in the threespine stickleback. Current Biology. 18: 769-774.

205

Kitano J, Ishikawa A, Kume M, Mori S (2012) Physiological and genetic basis for variation in migratory behavior in the three-spined stickleback, Gasterosteus aculeatus. Ichthyological Research. 59: 293-303.

Klepaker T (1996) Lateral plate polymorphism in marine and estuarine populations of the threespine stickleback (Gasterosteus aculeatus) along the coast of Norway. Copeia. 1996: 832-838.

Klingenberg CP (2011) MorphoJ: an integrated software package for geometric morphometrics. Molecular Ecology Resources. 11: 353-357.

Knecht AK, Hosemann KE, Kingsley DM (2007) Constraints on utilization of the EDA- signaling pathway in threespine stickleback evolution. Evolution and Development. 9: 141-154.

Knutsen H, Jorde PE, André C, Stenseth NC (2003) Fine-scaled geographical population structuring in a highly mobile marine species: the Atlantic cod. Molecular Ecology. 12: 385-394.

Knutsen H, Olsen EM, Jorde PE, Espeland SH, André C, Stenseth NC (2011) Are low but statistically significant levels of genetic differentiation in marine fishes ‘biologically meaningful’? A case study of coastal Atlantic cod. Molecular Ecology. 20: 768-783.

Krueger WH (1961) Meristic variation in the fourspine stickleback, Apeltes quadracus. Copeia. 1961: 442-450.

Lannig G, Storch D, Pörtner H-O (2005) Aerobic mitochondrial capacities in Antarctic and temperate eelpout (Zoarcidae) subjected to warm versus cold acclimation. Polar Biology. 28: 575-584.

206

Leary RF, Allendorf FW, Knudson RL (1983) Developmental stability and enzyme heterozygosity in rainbow trout. Nature. 301: 71-72.

Leary RF, Allendorf FW, Knudson RL (1984) Superior developmental stability of heterozygotes of enzyme loci in salmonid fishes. American Naturalist. 124: 540-551.

Lee CE, Bell MA (1999) Causes and consequences of recent freshwater invasions by saltwater animals. Trends in Ecology and Evolution. 14: 284-288.

Lee T-H, Guo H, Wang X, Kim C, Paterson AH (2014) SNPhylo: a pipeline to construct a phylogenetic tree from huge SNP data. BMC Genomics. 15: 162.

Leinonen T, Cano JM, Mäkinen H, Merilä J (2006) Contrasting patterns of body shape and neutral genetic divergence in marine and lake populations of threespine sticklebacks. Evolution. 19: 1803-1812.

Leinonen T, McCairns RJS, Herczeg G, Merilä J (2012) Multiple evolutionary pathways to decreased lateral plate coverage in freshwater threespine sticklebacks. Evolution. 66- 12: 3866-3875.

Lens L, Van Dongen S, Kark S, Matthysen E (2002b) Fluctuating asymmetry as an indicator of fitness: can we bridge the gap between studies? Biological Reviews. 77: 27- 38.

Lens L, Van Dongen S, Matthysen E (2002a) Fluctuating asymmetry as an early warning system in the critically endangered taita thrush. Conservation Biology. 16: 479-487.

Lepais O, Weir JT (2014) SimRAD: an R package for simulation-based prediction of the number of loci expected in RADseq and similar genotyping by sequencing approaches. Molecular Ecology Resources. 14: 1314-1321.

207

Lerner MI (1954) Genetic homeostasis. John Wiley & Sons, Inc.: New York.

Le Rouzic A, Østbye K, Klepaker TO, Hansen TF, Bernatchez L, Schluter D, Vøllestad LA (2011) Strong and consistent natural selection associated with armour reduction in sticklebacks. Molecular Ecology. 20: 2483-2493.

Lescak EA, Marcotte RW, Kenney LA, von Hippel FA, Cresko WA, Sherbick ML, Colgren JJ, López JA (2015) Admixture of ancient mitochondrial lineages in three-spined stickleback populations from the North Pacific. Journal of Biogeography. 42: 532-539.

Leung B, Forbes MR (1996) Fluctuating asymmetry in relation to stress and fitness: effects of trait type as revealed by meta-analysis. Ecoscience. 3: 400-413.

Leung B, Forbes MR, Houle D (2000) Fluctuating asymmetry as a bioindicator of stress: comparing efficacy of analyses involving multiple traits. American Naturalist. 155: 101- 115.

Levis NA, Pfennig DW (2016) Evaluating ‘plasticity-first’ evolution in nature: key criteria and empirical approaches. Trends in Ecology and Evolution. 31: 563-574.

Li P, Zhu Z, Lu Y, Granneman JG (2005) Metabolic and cellular plasticity in white adipose tissue II: role of peroxisome proliferator-activated receptor-α. American Journal of Physiology - Endocrinology and Metabolism. 289: E617-E626.

Lind EE, Grahn M (2011) Directional genetic selection by pulp mill effluent on multiple natural populations of three-spined stickleback (Gasterosteus aculeatus). Ecotoxicology. 20: 503-512.

208

Lind MI, Ingvarsson PK, Johansson H, Hall D, Johansson F (2011) Gene flow and selection on phenotypic plasticity in an island system of Rana temporaria. Evolution. 65: 684-697.

Lindsey CC (1954) Temperature-controlled meristic variation in the paradise fish Macropodus opercularis (L.). Canadian Journal of Zoology. 32: 87-98.

Lindsey CC (1962) Experimental study of meristic variation in a population of threespine sticklebacks, Gasteroteus aculeatus. Canadian Journal of Zoology. 40: 271-312.

Lindsey CC (1975) Pleomerism, the widespread tendency among related fish species for vertebral number to be correlated with maximum body length. Journal of the Fisheries Research Board of Canada. 32: 2453-2469.

Lindsey CC, Harrington Jr. RW (1972) Extreme vertebral variation induced by temperature in a homozygous clone of the self-fertilizing cyprinodontid fish Rivulus marmoratus. Canadian Journal of Zoology. 50: 733-744.

Liu S, Hansen MM, Jacobson MW (2016) Region-wide and ecotype-specific differences in demographic histories of threespine stickleback populations, estimated from whole genome sequencing. Molecular Ecology. 25: 5187-5202.

Loehr J, Leinonen T, Herczeg G, O’Hara RB, Merilä J (2012) Heritabiliy of asymmetry and lateral plate number in the threespine stickleback. PLOS One. 7: e39843.

Loehr J, Herczeg G, Leinonen T, Gonda A, Van Dongen S, Merilä J (2013) Asymmetry in threespine stickleback lateral plates. Journal of Zoology. 289: 279-284.

Long Jr. JH, Nipper KS (1996) The importance of body stiffness in undulatory propulsion. American Zoologist. 36: 678-694.

209

Losos JB, Creer DA, Glossip D, Goellner R, Hampton A, Roberts G, Haskell N, Taylor P, Etling J (2000) Evolutionary implications of phenotypic plasticity in the hindlimb of the lizard Anolis sagrei. Evolution. 54: 301-305.

Lucassen M, Koschnik N, Eckerle LG, Pörtner HO (2006) Mitochondrial mechanisms of cold adaptation in cod (Gadus morhua L.) populations from different climatic zones. Journal of Experimental Biology. 209: 2462-2471.

Lucek K, Haesler MP, Sivasundar A (2012) When phenotypes do not match genotypes – unexpected phenotypic diversity and potential environmental constraints in Icelandic stickleback. Journal of Heredity. 103: 579-584.

MacCrimmon HR, Kwain W-H (1969) Influence of light on early development and meristic characters in the rainbow trout, Salmo gairdneri Richardson. Canadian Journal of Zoology. 47: 631-637.

Machado LF, Damasceno JD, Bertoncini AA, Tosta VC, Farro APC, Hostim-Silva M, Oliveira C (2017) Population genetic structure and demographic history of the spadefish, Chaetodipterus faber (Ephippidae) from Southwestern Atlantic. Journal of Experimental Marine Biology and Ecology. 487: 45-52.

Magalhaes IS, D’Agostino D, Hohenlohe PA, MacColl ADC (2016) The ecology of an adaptive radiation of three-spined stickleback from North Uist, Scotland. Molecular Ecology. 25: 4319-4336.

Mäkinen HS, Merilä J (2008) Mitochondrial DNA phylogeography of the three-spined stickleback (Gasterosteus aculeatus) in Europe - evidence for multiple glacial refugia. Molecular Phylogenetics and Evolution. 46: 167-182.

210

Mäkinen HS, Cano JM, Merilä J (2006) Genetic relationships among marine and freshwater populations of the European three-spined stickleback (Gasterosteus aculeatus) revealed by microsatellites. Molecular Ecology. 15: 1519-1534.

Mäkinen HS, Cano JM, Merilä J (2008) Identifying footprints of directional and balancing selection in marine and freshwater three-spined stickleback (Gasterosteus aculeatus) populations. Molecular Ecology. 17: 3565-3582.

Marchinko K (2009) Predation’s role in repeated phenotypic and genetic divergence of armor in threespine stickleback. Evolution. 63: 127-138.

Marchinko KB, Schluter D (2007) Parallel evolution by correlated response: lateral plate reduction in threespine stickleback. Evolution. 61: 1084-1090.

Marchinko K, Matthews B, Arnegard M, Rogers SM, Schluter D (2014) Maintenance of a genetic polymorphism with disruptive natural selection in stickleback. Current Biology. 24: 1289-1292.

Marques DA, Taylor JS, Jones FC, Di Palma F, Kingsley DM, Reimchen TE (2017) Convergent evolution of SWS2 opsin facilitates adaptive radiation of threespine stickleback into different light environments. PLOS Biology. 15: e2001627.

Martínez JD, Cadena CD, Torres M (2016) Critical thermal limits of Poecilia caucana (Steindachner, 1880) (Cyprinodontiformes: Poeciliidae). Neotropical Ichthyology. 14: 2016.

Masel J, Siegal ML (2009) Robustness: mechanisms and consequences. Trends in Genetics. 25: 395-403.

211

Matuszewski S, Hermisson J, Kopp M (2015) Catch me if you can: adaptation from standing genetic variation to a moving phenotypic optimum. Genetics. 200: 1255-1274.

Maxwell EE, Wilson LAB (2013) Regionalization of the axial skeleton in the ‘ambush predator’ guild – are there developmental rules underlying body shape evolution in ray- finned fishes? BMC Evolutionary Biology. 13: 265.

Mazzarella AB, Boessenkool S, Østbye K, Vøllestad LA, Trucchi E (2016) Genomic signatures of the plateless phenotype in the threespine stickleback. Ecology and Evolution. 6: 3161-3173.

McBride RS, Horodysky AZ (2004) Mechanisms maintaining sympatric distributions of two ladyfish (Elopidae: Elops) morphs in the Gulf of Mexico and Western North Atlantic Ocean. Limnology and Oceanography. 49: 1173-1181.

McCairns RJS, Bernatchez L (2010) Adaptive divergence between freshwater and marine sticklebacks: insights into the role of phenotypic plasticity from an integrated analysis of candidate gene expression. Evolution. 64-4: 1029-1047.

McDowall RM (2003a) Variation in vertebral number in galaxiid fishes, how fishes swim and a possible reason for pleomerism. Reviews in Fish Biology and Fisheries. 13: 247- 263.

McDowall RM (2003b) Variation in vertebral number in galaxiid fishes (Teleostei: Galaxiidae): a legacy of life history, latitude and length. Environmental Biology of Fishes. 66: 361-381.

McDowall RM (2008) Jordan’s and other ecogeographical rules, and the vertebral number in fishes. Journal of Biogeography. 35: 501-508.

212

McGuigan K, Nishimura N, Currey M, Hurwit D, Cresko WA (2011) Cryptic genetic variation and body size evolution in threespine stickleback. Evolution. 65: 1203-1211.

McKinnon JS, Rundle HD (2002) Speciation in nature: the threespine stickleback model systems. Trends in Ecology and Evolution. 17: 480-488.

McKinnon JS, Mori S, Blackman BK, David L, Kingsley DM, Jamieson L, Chou J, Schluter D (2004) Evidence for ecology’s role in speciation. Nature. 429: 294-298.

McManus KF, Taravella AM, Henn BM, Bustamante CD, Sikora M, Cornejo OE (2017) Population genetic analysis of the DARC locus (Duffy) reveals adaptation from standing variation associated with malaria resistance in humans. PLOS Genetics. 13: e1006560.

McMullen PD, Bhattacharya S, Woods CG, Sun B, Yarborough K, Ross SM, Miller ME, McBride MT, LeCluyse EL, Clewell RA, Andersen ME (2014) A map of the PPARα transcription regulatory network for primary human hepatocytes. Chemo-Biological Interactions. 209: 14-24.

McPhail JD (1984) Ecology and evolution of sympatric sticklebacks (Gasterosteus): morphological and genetic evidence for a species pair in Enos Lake, British Columbia. Canadian Journal of Zoology. 62: 1402-1408.

Merilä J, Hendry AP (2014) Climate change, adaptation, and phenotypic plasticity: the problem and the evidence. Evolutionary Applications. 7: 1-14.

Messer PW, Petrov DA (2013) Population genomics of rapid adaptation by soft selective sweeps. Trends in Ecology and Evolution. 28: 1-11.

213

Miller CT, Beleza S, Pollen AA, Schluter D, Kittles RA, Shriver MD, Kingsley DM (2007) cis-Regulatory changes in Kit ligand expression and parallel evolution of pigmentation in sticklebacks and humans. Cell. 131: 1179-1189.

Miller CT, Glazer AM, Summers BR, Blackman BK, Norman AR, Shapiro MD, Cole BL, Peichel CL, Schluter D, Kingsley DM (2014) Modular skeletal evolution in sticklebacks is controlled by additive and clustered Quantitative Trait Loci. Genetics. 197: 405-420.

Miller RR, Hubbs CL (1969) Systematics of Gasterosteus aculeatus, with particular reference to intergradation and introgression along the Pacific coast of North America: a commentary on a recent contribution. Copeia. 1969: 52-69.

Mills MG, Greenwood AK, Peichel CL (2014) Pleiotropic effects of a single gene on skeletal development and sensory system patterning in sticklebacks. EvoDevo. 5:5.

Møller AP, Swaddle JP (1997) Asymmetry, developmental stability, and evolution. Oxford Series in Ecology and Evolution. Oxford University Press: New York.

Mori S (1994) Nest site choice by the three-spined stickleback, Gasterosteus aculeatus (form leiurus), in spring-fed waters. Journal of Fish Biology. 45: 279-289.

Morris MRJ (2014) Plasticity-mediated persistence in new and changing environments. International Journal of Evolutionary Biology. 2014: 416497.

Morris MRJ, Rogers SM (2013) Overcoming maladaptive plasticity through plastic compensation. Current Zoology. 59: 526-536.

214

Morris MRJ, Richard R, Leder EH, Barrett RDH, Aubin-Horth N, Rogers SM (2014) Gene expression plasticity evolves in response to colonization of freshwater lakes in threespine stickleback. Molecular Ecology. 23: 3226-3240.

Moyes CD (2003) Controlling muscle mitochondrial content. Journal of Experimental Biology. 206: 4385-4391.

Münzing J (1963) The evolution of variation and distributional patterns in European populations of the three-spined stickleback, Gasterosteus aculeatus. Evolution. 17: 320.

Muramoto J-I, Igarashi K, Itoh M, Makino S (1969) A study of the chromosomes and enzymatic patterns of sticklebacks in Japan. Proceedings of the Japan Academy. 45: 803- 807.

Myhre F, Klepaker T (2009) Body armour and lateral-plate reduction in freshwater three- spined stickleback Gasterosteus aculeatus: adaptations to a different buoyancy regime? Journal of Fish Biology. 75: 2062-2074.

Narum SR, Hesse JE (2011) Comparison of F(ST) outlier tests for SNP loci under selection. Molecular Ecology Resources. 11 Suppl 1: 184-194.

Nosil P, Reimchen TE (2001) Tarsal asymmetry, nutritional condition, and survival in water boatmen (Callicorixa vulnerata). Evolution. 55: 712-720.

Nosil P, Harmon LJ, Seehausen O (2009) Ecological explanations for (incomplete) speciation. Trends in Ecology and Evolution. 24: 145-156.

O’Brien KM (2011) Mitochondrial biogenesis in cold-bodied fishes. Journal of Experimental Biology. 214: 275-285.

215

O’Brown NM, Summers BR, Jones FC, Brady SD, Kingsley DM (2015) A recurrent regulatory change underlying altered expression and Wnt response of the stickleback armor plates gene EDA. eLife. 4: e05290.

Oravec TJ, Reimchen TE (2012) Divergent reproductive life histories in Haida Gwaii stickleback (Gasterosteus spp.) Canadian Journal of Zoology. 91: 17-24.

Orczewska JI, Hartleben G, O’Brien KM (2010) The molecular basis of aerobic metabolic remodeling differs between oxidative muscle and liver of threespine sticklebacks in response to cold acclimation. American Journal of Physiology - Regulatory, Integrative and Comparative Anatomy. 299: R352-R364.

O’Reilly P, Reimchen TE, Beech R, Strobeck C (1993) Mitochondrial DNA in Gasterosteus and Pleistocene glacial refugium on the Queen Charlotte Islands, British Columbia. Evolution. 47: 678-684.

Orr HA (2005) The genetic theory of adaptation: a brief history. Nature Reviews Genetics. 6: 119-127.

Orr HA, Betancourt AJ (2001) Haldane’s sieve and adaptation from standing genetic variation. Genetics. 157: 875-884.

Orr HA, Unckless RL (2014) The population genetics of evolutionary rescue. PLOS Genetics. 10: e1004551.

Ortí G, Bell MA, Reimchen TE, Meyer A (1994) Global survey of mitochondrial DNA sequences in the threespine stickleback: evidence for recent migrations. Evolution. 48: 608-622.

216

Paaby AB, Rockman MV (2014) Cryptic genetic variation: evolution’s hidden substrate. Nature Reviews Genetics. 15: 247-258.

Palmer AR, Strobeck C (1986) Fluctuating asymmetry: measurement, analysis, patterns. Annual Review of Ecology and Systematics. 17: 391-421.

Palstra FP, Ruzzante DE (2008) Genetic estimates of contemporary effective population size: what can they tell us about the importance of genetic stochasticity for wild population persistence? Molecular Ecology. 17: 3428-3447.

Patterson BD, Patton JL (1990) Fluctuating asymmetry and allozymic heterozygosity among natural populations of pocket gophers (Thomomys bottae). Biological Journal of the Linnean Society. 40: 21-36.

Pavey SA, Collin H, Nosil P, Rogers SM (2010) The role of gene expression in ecological speciation. Annals of the New York Academy of Sciences. 1206: 110-129.

Peichel CL, Marques DA (2017) The genetic and molecular architecture of phenotypic diversity in sticklebacks. Philosophical Transactions of the Royal Society B. 372: 20150486.

Peichel CL, Ross JA, Matson CK, Dickson M, Grimwood J, Schmutz J, Myers RM, Mori S, Schluter D, Kingsley DM (2004) The master sex-determination locus in threespine sticklebacks is on a nascent Y chromosome. Current Biology. 14: 1416-1424.

Pelz H-J, Rost S, Hünerberg M, Fregin A, Heiberg A-C, Baert K, MacNicoll AD, Prescott CV, Walker A-S, Oldenburg J, Müller CR (2005) The genetic basis of resistance to anticoagulants in rats. Genetics. 170: 1839-1847.

217

Pennings PS (2012) Standing genetic variation and the evolution of drug resistance in HIV. PLOS Computational Biology. 8: e1002527.

Peter BM, Huerta-Sanchez E, Nielsen R (2012) Distinguishing between selective sweeps from standing variation and from a de novo mutation. PLOS Genetics. 8: e1003011.

Petrovitch E (2013) Assessing Jordan’s Rule in coastal Pacific threespine stickleback. ECOL 507 paper, University of Calgary.

Pigliucci M (2001) Phenotypic plasticity: beyond nature and nurture. JHU Press: USA.

Podrabsky JE, Somero GN (2004) Changes in gene expression associated with acclimation to constant temperatures and fluctuating daily temperatures in an annual killifish Austrofundulus limnaeus. Journal of Experimental Biology. 207: 2237-2254.

Pörtner HO, Peck L, Somero G (2007) Thermal limits and adaptation in marine Antarctic ectotherms: an integrative view. Philosophical Transactions of the Royal Society of London B. 362: 2233-2258.

Poulin R, FitzGerald GJ (1989) Early life histories of three sympatric sticklebacks in a salt-marsh. Journal of Fish Biology. 34: 207-221.

Przeworski M, Coop G, Wall JD (2005) The signature of positive selection on standing genetic variation. Evolution. 59: 2312-2323.

Quattro JM, Vrijenhoek RC (1989) Fitness differences among remnant populations of the endangered sonoran topminnnow. Science. 245: 976-978.

Queitsch C, Sangster TA, Lindquist S (2002) Hsp90 as a capacitor of phenotypic variation. Nature. 417: 618-624.

218

Quinn TP, Brodeur RD (1991) Intra-specific variations in the movement patterns of marine animals. American Zoologist. 31: 231-241.

Quinn TP, Light JT (1989) Occurrence of threespine sticklebacks (Gasterosteus aculeatus) in the open North Pacific Ocean: migration or drift? Canadian Journal of Zoology. 67: 2850-2852.

R Core Team (2016) R: A language and environment for statistical computing. R Foundation for Statistical Computing. Vienna, Austria. https://www.R-projects.org

Radwan J, Babik W (2012) The genomics of adaptation. Proceedings of the Royal Society B. 279: 5024-5028.

Raeymaekers JAM, Maes GE, Audenaert E, Volckaert FAM (2005) Detecting Holocene divergence in the anadromous-freshwater three-spined stickleback (Gasterosteus aculeatus) system. Molecular Ecology. 14: 1001-1014.

Raeymaekers JAM, Konijnendik N, Larmuseau MHD, Hellemans B, de Meester L, Volckaert FAM (2014) A gene with major phenotypic effects as a target for selection vs. homogenizing gene flow. Molecular Ecology. 23: 162-181.

Rafiński J, Bańgbura J, Przybylski M (1989) Genetic differentiation of freshwater and marine sticklebacks (Gasterosteus aculeatus) of Eastern Europe. Journal of Zoological Systematics and Evolutionary Research. 27: 33-43.

Rajon E, Plotkin JB (2013) The evolution of genetic architectures underlying quantitative traits. Proceedings of the Royal Society B. 280: 20131552.

Rambaut A (2007) FigTree. http://tree.bio.ed.ac.uk/software/figtree/

219

Ramløv H (2000) Aspects of natural cold tolerance in ectothermic animals. Human Reproduction. 15: 26-46.

Raunich L, Callegarini C, Cucchi C (1972) Ecological aspects of hemoglobin polymorphism in Gasterosteus aculeatus (Teleostei). In: Fifth European marine biology symposium. Ed. B Battaglia. pp. 153-162.

Ravinet M, Prodöhl PA, Harrod C (2013a) On Irish stickleback: morphological diversification in a secondary contact zone. Evolutionary Ecology Research. 15: 271-294.

Ravinet M, Prodöhl PA, Harrod C (2013b) Parallel and nonparallel ecological, morphological and genetic divergence in lake-stream stickleback from a single catchment. Journal of Evolutionary Biology. 26: 186-204.

Ravinet M, Takeuchi N, Kume M, Mori S, Kitano J (2014) Comparative analysis of Japanese three-spined stickleback clades reveals the Pacific Ocean lineage has adapted to freshwater environments while the Japan Sea has not. PLOS One. 9: e112404.

Reebs SG, Whoriskey FG, FitzGerald GJ (1984) Diel patterns of fanning activity, egg respiration, and the nocturnal behavior of male three-spined sticklebacks, Gasterosteus aculeatus L. (f. trachurus). Canadian Journal of Zoology. 62: 329-334.

Reeve ECR (1960) Some genetic tests on asymmetry of sternopleural chaeta number in Drosophila. Genetic Research. 1: 151-172.

Reimchen TE (1983) Structural relationships between spines and lateral plates in threespine stickleback (Gasterosteus aculeatus). Evolution. 37: 931-946.

220

Reimchen TE (2000) Predator handling failures of lateral plate morphs in Gasterosteus aculeatus: functional implications for the ancestral plate condition. Behaviour. 137: 1081-1096.

Reimchen TE, Bergstrom CA (2009) The ecology of asymmetry in stickleback defense structures. Evolution. 63: 115-126.

Reimchen TE, Cox KD (2016) Differential temperature preferences of vertebral phenotypes in Gasterosteus. Canadian Journal of Zoology. 94:1-5.

Reimchen TE, Nelson JS (1987) Habitat and morphological correlates to vertebral number as shown in a Teleost, Gasterosteus aculeatus. Copeia. 1987: 868-874.

Reimchen TE, Nosil P (2001) Lateral plate asymmetry, diet, and parasitism in threespine stickleback. Journal of Evolutionary Biology. 14: 632-645.

Reimchen TE, Ingram T, Hansen SC (2008) Assessing niche differences of sex, armour and asymmetry phenotypes using stable isotope analyses in Haida Gwaii sticklebacks. Behaviour. 145: 561-577.

Rennison DJ, Heilbron K, Barrett RDH, Schluter D (2015) Discriminating selection on lateral plate phenotype and its underlying gene, Ectodysplasin, in threespine stickleback. American Naturalist. 185: 150-156.

Rennison DJ, Owens GL, Heckman N, Schluter D, Veen T (2016) Rapid adaptive evolution of colour vision in the threespine stickleback radiation. Proceedings of the Royal Society B. 283: 20160242.

Resh VH, White DS, Elbert SA, Jennings DE, Krumholz LA (1976) Vertebral variation in the emerald shiner Notropis atherinoides from the Ohio River: an apparent

221

contradiction to “Jordan’s Rule”. Bulletin of the Southern California Academy of Sciences. 75: 76-84.

Richard M, Yvert G (2014) How does evolution tune biological noise? Frontiers in Genetics. 5: 374.

Robertson S, Bradley JE, MacColl AD (2017) Eda haplotypes in three-spined stickleback are associated with variation in immune gene expression. Scientific Reports. 7: 42677.

Rockman MV (2012) The QTN program and the alleles that matter for evolution: all that’s gold does not glitter. Evolution. 66-1: 1-17.

Roesti M, Gavrilets S, Hendry AP, Salzburger W, Berner D (2014) The genomic signature of parallel adaptation from shared genetic variation. Molecular Ecology. 23: 3944-3956.

Rogers SM, Bernatchez L (2007) The genetic architecture of ecological speciation and the association with signatures of selection in natural lake whitefish (Coregonus sp. Salmonidae) species pairs. Molecular Biology and Evolution. 24: 1423-1438.

Rogers SM, Tamkee P, Summers B, Balabahadra B, Marks M, Kingsley DM, Schluter D (2012) Genetic signature of adaptive peak shift in threespine stickleback. Evolution. 66- 8: 2439–2450.

Rohner N, Jarosz DF, Kowalko JE, Yoshizawa M, Jeffery WR, Borowsky RL, Lindquist S, Tabin CJ (2013) Cryptic variation in morphological evolution: Hsp90 as a capacitor for loss of eyes in cavefish. Science. 342: 1372-1375.

222

Roldan ERS, Cassinello J, Abaigar T, Gomendio M (1998) Inbreeding, fluctuating asymmetry, and ejaculate quality in an endangered ungulate. Proceedings of the Royal Society of London B. 265: 243-248.

Rousset F (2008) Genepop ‘007: a complete reimplementation of the Genepop software for Windows and Linux. Molecular Ecology Resources. 8: 103-106.

Rowland WJ (1989) The ethological basis of mate choice in male threespine sticklebacks, Gasterosteus aculeatus. Animal Behaviour. 38: 112-120.

Rundle HD, Nosil P (2005) Ecological speciation. Ecology Letters. 8: 336-352.

Rundle HD, Schluter D (2004) Natural selection and ecological speciation in sticklebacks. In: Adaptive speciation. Eds. U Dieckmann, M Doebeli, JAJ Metz, D Tautz. Cambridge University Press: NY. pp. 192-209.

Rybkina EV, Ivanova TS, Ivanov MV, Kucheryavyy AV, Lajus DL (2016) Habitat preference of three-spined stickleback juveniles in experimental conditions and in wild eelgrass. Journal of the Marine Biological Association of the United Kingdom. 96: 1-9.

Rypel AL (2014) The cold-water connection: Bergmann’s Rule in North American freshwater fishes. American Naturalist. 183: 147-156.

Saha A, Johansen T, Hedeholm R, Nielsen EE, Westgaard JI, Hauser L, Planque B, Cadrin SX, Boje J (2017) Geographic extent of introgression in Sebastes mentella and its effect on genetic population structure. Evolutionary Applications. 10: 77-90.

Saimoto RK (1993a) Life history of marine threespine stickleback in Oyster Lagoon, British Columbia. MSc thesis, University of British Columbia.

223

Saimoto RS (1993b) Reproductive and natal homing of marine threespine sticklebacks (Gasterosteus aculeatus). MSc thesis, University of British Columbia.

Samuk K, Iritani D, Schluter D (2014) Reversed brain size sexual dimorphism accompanies loss of parental care in white sticklebacks. Ecology and Evolution. 4: 3236- 3243.

Scarcelli N, Kover PX (2009) Standing genetic variation in FRIGIDA mediates experimental evolution of flowering time in Arabidopsis. Molecular Ecology. 18: 2039- 2049.

Schluter D, Conte GL (2009) Genetics and ecological speciation. Proceedings of the National Academy of Sciences USA. 106: 9955-9962.

Schluter D, Clifford EA, Nemethy M, McKinnon JS (2004) Parallel evolution and inheritance of quantitative traits. American Naturalist. 163: 809-822.

Schluter D, Marchinko KB, Barrett RDH, Rogers SM (2010) Natural selection and the genetics of adaptation in threespine stickleback. Philosophical Transactions of the Royal Society of London B. 365: 2479-2486.

Schneider RF, Meyer A (2017) How plasticity, genetic assimilation and cryptic genetic variation may contribute to adaptive radiations. Molecular Ecology. 26: 330-350.

Seymour A (1959) Effects of temperature upon the formation of vertebrae and fin rays in young Chinook salmon. Transactions of the American Fisheries Society. 88: 58-69.

Shama LNS, Strobel A, Mark FC, Wegner KM (2014) Transgenerational plasticity in marine sticklebacks: maternal effects mediate impacts of a warming ocean. Functional Ecology. 28: 1482-1493.

224

Shama LNS, Mark FC, Strobel A, Lokmer A, John U, Wegner KM (2016) Transgenerational effects persist down the maternal line in marine sticklebacks: gene expression matches physiology in a warming ocean. Evolutionary Applications. 9: 1096- 1111.

Shaner P-JL, Chen Y-R, Lin J-W, Kolbe JJ, Lin S-M (2013) Sex-specific correlations of individual heterozygosity, parasite load, and scalation asymmetry in a sexually dichromatic lizard. PLOS One. 8: e56720.

Shikano T, Merilä J (2011) Body size and the number of vertebrae in the nine-spined stickleback (Pungitius pungitius). Biological Journal of the Linnean Society. 104: 378- 385.

Shimada Y, Shikano T, Merilä J (2011) A high incidence of selection on physiologically important genes in the three-spined stickleback, Gasterosteus aculeatus. Molecular Biology and Evolution. 28: 181-193.

Shine R (2012) Invasive species as drivers of evolutionary change: cane toads in tropical Australia. Evolutionary Applications. 5: 107-116.

Simenstad CA, Hood WG, Thom RM, Levy DA, Bottom DL (2002) Landscape structure and scale constraints on restoring estuarine wetlands for Pacific coast juvenile fishes. In: Concepts and controversies in tidal marsh ecology. Eds. MP Weinstein, DA Kreeger. Kluwer Academic Publishers: New York. pp. 597-630.

Smedley D, Haider S, Durinck S, Pandini L, Provero P, Allen J, Arnaiz O, Awedh MH, Baldock R, Barbiera G, Bardou P, Beck T, Blake A, Bonierbale M, Brookes AJ, Bucci G, Buetti I, Burge S, Cabau C, Carlson JW, Chelela C, Chrysostomou C, Cittaro D, Collin O, Cordova R, Cutts RJ, Dassi E, Di Genova A, Djari A, Esposito A, Estrella H, Eyras E,

225

Fernandez-Banet J, Forbes S, Free RC, Fujisawa T, Gadaleta E, Garcia-Manteiga JM, Goodstein D, Gray K, Guerra-Assunção JA, Haggarty B, Han DJ, Han BW, Harris T, Harshbarger J, Hastings RK, Hayes RD, Hoede C, Hu S, Hu ZL, Hutchins L, Kan Z, Kawaji H, Keliet A, Kerhornou A, Kim S, Kinsella R, Klopp C, Kong L, Lawson D, Lazarevic D, Lee JH, Letellier T, Li CY, Lio P, Liu CJ, Luo J, Maass A, Mariette J, Maurel T, Merella S, Mohamed AM, Moreews F, Nabihoudine I, Ndegwa N, Noirot C, Perez-Llamas C, Primig M, Quattrone A, Quesneville H, Rambaldi D, Reecy J, Riba M, Rosanoff S, Saddiq AA, Salas E, Sallou O, Shepherd R, Simon R, Sperling L, Spooner W, Staines DM, Steinbach D, Stone K, Stupka E, Teague JW, Dayem Ullah AZ, Wang J, Ware D, Wong-Erasmus M, Youens-Clark K, Zadissa A, Zhang SJ, Kasprzyk A (2015) The BioMart community portal: an innovative alternative to large, centralized data repositories. Nucleic Acids Research. 43: W589-W598.

Smith C, Spence R, Barber I, Przybylski M, Wootton RJ (2014) The role of calcium and predation on plate morph evolution in the three-spined stickleback (Gasterosteus aculeatus). Ecology and Evolution. 4: 3550-3554.

Snyder RJ, Dingle H (1989) Adaptive, genetically based differences in life history between estuary and freshwater threespine stickleback (Gasterosteus aculeatus L.) Canadian Journal of Zoology. 67: 2448-2454.

Snyder RJ, Dingle H (1990) Effects of freshwater and marine overwintering environments on life histories of threespine sticklebacks: evidence for adaptive variation between anadromous and resident freshwater populations. Oecologia. 84: 386-390.

Somero GN (2010) The physiology of climate change: how potentials for acclimatization and genetic adaptation will determine ‘winners’ and ‘losers’. Journal of Experimental Biology. 213: 912-920.

226

Spoljaric MA, Reimchen TE (2007) 10 000 years later: evolution of body shape in Haida Gwaii three-spined stickleback. Journal of Fish Biology. 70: 1484-1503.

Stapley J, Reger J, Feulner PGD, Smadja C, Galindo J, Ekblom R, Bennison C, Ball AD, Beckerman AP, Slate J (2010) Adaptation genomics: the next generation. Trends in Ecology and Evolution. 25: 705-712.

Stern DL (2013) The genetic causes of convergent evolution. Nature Reviews Genetics. 14: 751-764.

Stevens M, Castor-Perry SA, Price JRF (2009) The protective value of conspicuous signals is not impaired by shape, size, or position asymmetry. Behavioral Ecology. 20: 96-102.

Stoffel MA, Esser M, Kardos M, Humble E, Nichols H, David P, Hoffman JI (2016) inbreedR: an R package for the analysis of inbreeding based on genetic markers. Methods in Ecology and Evolution. 7: 1331-1339.

Swaddle JP (1997) Developmental stability and predation success in an insect predator- prey system. Behavioral Ecology. 8: 433-436.

Swain DP (1988) Evidence of selection for vertebral number of fry in peamouth, Mylocheilus caurinus. Canadian Journal of Fisheries and Aquatic Sciences. 45: 1279- 1290.

Swain DP (1992a) The functional basis of natural selection for vertebral traits of larvae in the stickleback Gasterosteus aculeatus. Evolution. 46: 987-997.

Swain DP (1992b) Selective predation for vertebral phenotype in Gasterosteus aculeatus: reversal in the direction of selection at different larval sizes. Evolution. 46: 998-1013.

227

Swain DP, Lindsey CC (1984) Selective predation for vertebral number of young sticklebacks, Gasterosteus aculeatus. Canadian Journal of Fisheries and Aquatic Sciences. 41: 1231-1233.

Swain DP, Frank KT, Maillet G (2001) Delineating stocks of Atlantic cod (Gadus morhua) in the Gulf of St Lawrence and Cabot Strait areas using vertebral number. ICES Journal of Marine Sciences. 58: 253-269.

Taniguchi N, Honma Y, Kawamata K (1990) Genetic differentiation of freshwater and anadromous threespine sticklebacks (Gasterosteus aculeatus) from northern Japan. Japanese Journal of Ichthyology. 37: 230-238.

Tåning ÅV (1952) Experimental study of meristic characters in fishes. Biological Reviews. 27: 169-193.

Tattersall GJ, Sinclair BJ, Withers PC, Fields PA, Seebacher F, Cooper CE, Maloney SK (2012) Coping with thermal challenges: physiological adaptations to environmental temperatures. Comparative Physiology. 2: 2151-2202.

Taugbøl A, Junge C, Quinn TP, Herland A, Vøllestad LA (2014) Genetic and morphometric divergence in threespine stickleback in the Chignik catchment, Alaska. Ecology and Evolution. 4: 144-156.

Taylor DS (2001) Physical variability and fluctuating asymmetry in heterozygous and homozygous populations of Rivulus marmoratus. Canadian Journal of Zoology. 79: 766- 778.

228

Taylor EB, McPhail JD (1986) Prolonged and burst swimming in anadromous and freshwater threespine stickleback, Gasterosteus aculeatus. Canadian Journal of Zoology. 64: 416-420.

Taylor EB, McPhail JD (1999) Evolutionary history of an adaptive radiation in species pairs of threespine sticklebacks (Gasterosteus): insights from mitochondrial DNA. Biological Journal of the Linnean Society. 66: 271-291.

Teigen LE, Orczewska JI, McLaughlin J, O’Brien KM (2015) Cold acclimation increases levels of some heat shock protein and sirtuin isoforms in threespine stickleback. Comparative Biochemistry and Physiology, Part A. 188: 139-147.

Thibert-Plante X, Hendry AP (2011) The consequences of phenotypic plasticity for ecological speciation. Journal of Evolutionary Biology. 24: 326-342.

Thompson CE, Taylor EB, McPhail JD (1997) Parallel evolution of lake-stream pairs of threespine stickleback (Gasterosteus) inferred from mitochondrial DNA variation. Evolution. 51: 1955-1965.

Tibblin P, Berggren H, Nordahl O, Larsson P, Forsman A (2016) Causes and consequences of intra-specific variation in vertebral number. Scientific Reports. 6: 26372.

Tocts AMS, Johnson DW, Carter AJR (2016) Strong nonlinear selection against fluctuating asymmetry in wild populations of a marine fish. Evolution. 70: 2899-2908.

Trigari G, Pirini M, Ventrella V, Pagliarani A, Trombetti F, Borgatti AR (1992) Lipid composition and mitochondrial respiration in warm- and cold-adapted sea bass. Lipids. 27: 371-377.

229

Trokovic N, Herczeg G, Ab Ghani NI, Shikano T, Merilä J (2012) High levels of fluctuating asymmetry in isolated stickleback populations. BMC Evolutionary Biology. 12: 115.

Tschanz SA, Burri PH, Weibel ER (2011) A simple tool for stereological assessment of digital images: the STEPanizer. Journal of Microscopy. 243: 47-59.

Turchin MC, Chiang CWK, Palmer CD, Sankararaman S, Reich D, Genetic Investigation of Anthropometric Traits (GIANT) Consortium, Hirschhorn JN (2012) Evidence of widespread selection on standing variation in Europe at height-associated SNPs. Nature Genetics. 44: 1015-1019.

Turner SD (2014) qqman: an R package for visualizing GWAS results using Q-Q and manhattan plots. biorXiv.

USFWS (2009) Unarmored threespine stickleback (Gasterosteus aculeatus williamsoni) 5-year review: summary and evaluation. United States Fish and Wildlife Service. 36 pp.

Vamosi SM (2006) Contemporary evolution of armour and body size in a recently introduced population of threespine stickleback Gasterosteus aculeatus. Current Zoology. 52: 483-490.

Van Dongen S, Lens L, Pape E, Volckaert FAM, Raeymaekers JAM (2009) Evolutionary history shapes the association between developmental instability and population-level genetic variation in three-spined sticklebacks. Journal of Evolutionary Biology. 22: 1695- 1707.

Van Valen L (1962) A study of fluctuating asymmmetry. Evolution. 16: 125-142.

230

Ventura-Clapier R, Garnier A, Veksler V (2008) Spotlight issue on signaling in cardiac metabolism: transcriptional control of mitochondrial biogenesis. The central role of PGC1-α. Cardiovascular Research. Oxford University Press. CVR-2007-1636R2.

Vøllestad LA, Hindar K, Møller AP (1999) A meta-analysis of fluctuating asymmetry in relation to heterozygosity. Heredity. 83: 206-218.

Waddington CH (1942) Canalization of development and the inheritance of acquired characters. Nature. 150: 563-565.

Walker JA, Bell MA (2000) Net evolutionary trajectories of body shape evolution within a microgeographic radiation of threespine stickleback (Gasterosteus aculeatus). Journal of Zoology. 252: 293-302.

Wallace GT, Kim TL, Neufeld CJ (2014) Interpopulational variation in the cold tolerance of a broadly distributed marine copepod. Conservation Physiology. 2: cou041.

Waples RS (2016) Tiny estimates of the Ne/N ratio in marine fishes: Are they real? Journal of Fish Biology. 89: 2479-2504.

Waples RS, Do C (2008) LDNe: a program for estimating effective population size from data on linkage disequilibrium. Molecular Ecology Resources. 8: 753-756.

Weatherhead PJ, Dufour KW, Lougheed SC, Eckert CG (1999) A test of the good-genes- as-heterozygosity hypothesis using red-winged blackbirds. Behavioral Ecology. 10: 619- 625.

Weibel E (1979) Stereological methods volume 1: practical methods for biological morphometry. Academic Press: NY.

231

Weinig C, Dorn LA, Kane NC, German ZM, Halldorsdottir SS, Ungerer MC, Toyonaga Y, Mackay TF, Purugganan MD, Schmitt J (2003) Heterogeneous selection at specific loci in natural environments in Arabidopsis thaliana. Genetics. 165: 321-329.

Weir BC, Cockerham CC (1984) Estimating F-statistics for the analysis of population- structure. Evolution. 38: 1358-1370.

Weller B, Ganzhorn JU (2004) Carabid beetle community composition, body size, and fluctuating asymmetry along an urban-rural gradient. Basic and Applied Ecology. 5: 193- 201.

West-Eberhard MJ (2003) Developmental plasticity and evolution. Oxford University Press: USA.

White TA, Searle JB (2008) Mandible asymmetry and genetic diversity in island populations of the common shrew, Sorex araneus. Journal of Evolutionary Biology. 21: 636-641.

Whitlock MC, Guillaume F (2009) Testing for spatially divergent selection: comparing

QST to FST. Genetics. 183: 1055-1063.

Whoriskey FG, FitzGerald GJ (1994) Ecology of the threespine stickleback on the breeding grounds. In: The evolutionary biology of the threespine stickleback. Eds. MA Bell, SA Foster. Oxford University Press: NY. pp. 188-206.

Wiig E, Reseland JE, Østbye K, Haugen HJ, Vøllestad LA (2016) Variation in lateral plate quality in threespine stickleback from fresh, brackish and marine water: a micro- computed tomography study. PLOS One. 11: e0164578.

232

Williams DD, Delbeek JC (1989) Biology of the threespine stickleback, Gasterosteus aculeatus, and the blackspotted stickleback, G. wheatlandi, during their marine pelagic phase in the Bay of Fundy, Canada. Environmental Biology of Fishes. 24: 33-41.

Withler RE, McPhail JD (1985) Genetic variability in freshwater and anadromous sticklebacks (Gasterosteus aculeatus) of southern British Columbia. Canadian Journal of Zoology. 63: 528-533.

Wooten MC, Smith MH (1986) Fluctuating asymmetry and genetic variability in a natural population of Mus musculus. Journal of Mammalogy. 67: 725-732.

Wootton RJ (1976) The biology of the sticklebacks. Academic Press: USA.

Wootton RJ (1984) A functional biology of sticklebacks. University of California Press: USA.

Worgan JP, FitzGerald GJ (1981) Habitat segregation in a salt marsh among adult sticklebacks (Gasterosteidae). Environmental Biology of Fishes. 6: 105-109.

Wu TD, Watanabe CK (2005) GMAP: a genomic mapping and alignment program for mRNA and EST sequences. Bioinformatics. 21: 1859-1875.

Wund MA, Baker JA, Clancy B, Golub JL, Foster SA (2008) A test of the “flexible stem” model of evolution: ancestral plasticity, genetic accommodation, and morphological divergence in the threespine stickleback radiation. American Naturalist. 172: 449-462.

Wund MA, Valena S, Wood S, Baker JA (2012) Ancestral plasticity and allometry in threespine stickleback reveal phenotypes associated with derived, freshwater ecotypes. Biological Journal of the Linnean Society. 105: 573-583.

233

Yamada M, Higuchi M, Goto A (2001) Extensive introgression of mitochondrial DNA found between two genetically divergent forms of threespine stickleback, Gasterosteus aculeatus, around Japan. Environmental Biology of Fishes. 61: 269-284.

Yamada M, Higuchi M, Goto A (2007) Long-term occurrence of hybrids between Japan Sea and Pacific Ocean forms of threespine stickleback, Gasterosteus aculeatus, in Hokkaido Island, Japan. Environmental Biology of Fishes. 80: 435-443.

Yamahira K, Nishida T (2009) Latitudinal variation in axial patterning of the medaka (Actinopterygii: Adrianichthyidae): Jordan’s rule is substantiated by genetic variation in abdominal vertebral number. Biological Journal of the Linnean Society. 96: 856-866.

Yamahira K, Lankford Jr. TE, Conover DO (2006) Intra- and interspecific latitudinal variation in vertebral number of Menidia spp. (Teleostei: Atherinopsidae). Copeia. 2006: 431-436.

Yampolsky LY, Zeng E, Lopez J, Williams PJ, Dick KB, Colbourne JK, Pfrender ME (2014) Functional genomics of acclimation and adaptation in response to thermal stress in Daphnia. BMC Genomics. 15: 859.

Yavno S, Fox MG (2013) Morphological change and phenotypic plasticity in native and non-native pumpkinseed sunfish in response to sustained water velocities. Journal of Evolutionary Biology. 26: 2383–2395.

Yershov P, Sukhotin A (2015) Age and growth of marine three-spined stickleback in the White Sea 50 years after a population collapse. Polar Biology. 38: 1813-1823.

Yoshida K, Miyagi R, Mori S, Takahashi A, Makino T, Toyoda A, Fujiyama A, Kitano J (2016) Whole-genome sequencing reveals small genomic regions of introgression in an

234

introduced crater lake population of threespine stickleback. Ecology and Evolution. 6: 2190-2204.

Zietara MS (1989) Electrophoretic and malate dehydrogenase polymorphism in populations of the threespine (Gasterosteus aculeatus) and the ninespine (Pungitius pungitius) sticklebacks from Vistula river and Gdansk Gulf. Biochemical Systematics and Ecology. 17: 253-259.

235

APPENDIX A: EXPLORING ADEGENET

A.1. Introduction Structure is the program commonly used to assess the genetic structuring of populations, and has been used extensively in the Rogers lab. My data violated the assumptions of Structure in several essential ways: (1) SNPs violate assumptions of independence, largely because if multiple SNPs are contained on a single 150 bp sequenced read, those SNP frequencies may be correlated due to linkage. It is possible to overcome this by only using the first, or some randomly-called, SNP from each RAD- fragment. This however excludes a large amount of potentially important data. (2) Structure is ideal for exploring data under an island model but performs poorly under a stepping stone model. A stepping stone model intuitively seemed like a better fit for latitudinally-sampled fish populations. The Discriminant Analysis of Principal Components implemented by Adegenet largely overcomes these issues by conducting a multi-step procedure. First, the multidimensional genetic data is collapsed into a few dimensions through a Principal Components Analysis (PCA). This eliminates correlations among SNPs (Jombart et al. 2010). Second, a Discriminant Function Analysis (DFA) is done on this simplified dataset. This reduces the between-individual variation while maximizing the between- cluster variation, so that those genetic variants that distinguish populations can be readily discovered. DAPC does not assume an island or stepping stone model; although it performs better under an island model, it outperforms Structure at stepping stone models (Jombart et al. 2010). It also does not assume Hardy-Weinberg equilibrium. Despite reducing the number of dimensions, PCA still generates a large number of dimensions; > 200 were generated from my dataset. Using all of the dimensions increases the likelihood of perfectly discriminating each individual, essentially resulting in as many identified “populations” as there are individuals (if allowing Adegenet to identify clusters), or clusters perfectly matching to a priori information. Not using enough dimensions, however, removes large amounts of genetic variation from the data set, reducing the ability to resolve genetic clusters (Jombart et al. 2010). I wanted to

236

know how Adegenet would perform on my data using different numbers of PCs – particularly when individuals were randomly assigned to eight new “populations.” Similar to Structure, DAPC uses Bayesian Information Criteria (BIC) in order to estimate the likely number of genetic clusters present in the data. Lower BIC scores are more likely reflective of the true number of genetic clusters (k). Thus I could both force the program to maximize variation among genetic clusters that I assigned, or I could have the program determine the optimal number of genetic clusters represented by my samples without any a priori knowledge about the number of sampling locations. Briefly, k-means clustering works using the algorithm of Hartigan and Wong (1979), which seeks to minimize the within-cluster sum of squares. First, the user sets a range of potential clusters, which the algorithm tests by randomly assigning cluster means to the data (for k = 3, three means are assigned, for k = 4, four means are assigned, etc.). k clusters is then generated by assigning each individual to its closest mean using Euclidean distances. The centroid of each cluster is then calculated and is used as the new mean, to which individuals are reassigned. This continues until individual assignments within a cluster no longer change, at which point the optimal arrangement for that cluster has been found. This continues until all clusters have been removed from the “live” set (Hartigan and Wong 1979). Bayesian Information Criteria (BIC) are then used to determine the most likely number of clusters, with the model with the lowest BIC preferred. Note that the initial step, of randomly assigning means, means that not all possible ways of partitioning individuals into the assigned number of clusters is compared. The algorithm may end at local optima rather than more realistic assignments (Hartigan and Wong 1979). Also note that the shapes of the clusters are constrained to be approximately spherical and of similar variance and sample size (Robertson 2015). The consequences of the first of these issues could be assessed by running the same data multiple times through Adegenet. Adegenet was used in Chapter Four to determine population genetic structure in the form of genetic clustering. In order to get those results, some initial exploration of Adegenet was required, particularly around the consequences of filtering and the populations used. This is explored below.

237

A.2. Methods Multiple analyses were done in Adegenet: (1) I wanted to know the consequences of PC retention on cluster assignment. To explore this I pooled all stickleback together and randomly assigned them to eight artificial populations using a random number generator in Excel. I then utilized an option in Adegenet in which I can tell the program the origin of each individual, causing it to attempt to maximize the distance among these randomly-generated populations. I did this while retaining different numbers of PCs, from all PCs to those suggested as the optimal number to retain using the a-score values that were acquired from my actual data. I note that a-scores is no longer the preferred method for determining the number of PCs to retain – xvalDAPC is instead preferred. This method requires a priori population assignments and uses cross-validation techniques. However, it does not perform well when data is missing, and I called SNPs even if they were missing in up to 25% of individuals. For this reason I opted to use a- scores, although I note that by replacing missing values with the average frequency at that locus (see below) I was able to use xvalDAPC, and it returned 20 PCs as the optimal number to retain. This was similar to the a-scores value of 22 or 23. Retaining 20 vs. 23 PCs produced identical clustering results. (2) To determine the effects of local optima, I ran Adegenet using identical datasets ten times at a k-value of 5, and compared clustering assignments. (3) I explored the effects of removing blacklisted loci and loci with low minor allele frequencies (maf) by comparing the unfiltered and FIS/maf-filtered dataset. (4) Since missing data are a common problem with Genotype-by-Sequencing analyses, multiple solutions have been proposed. I explored the consequences of ignoring missing values, removing missing values if missing from > 25% of individuals, replacing missing values with the average allele frequency for that locus, or setting missing values to zero, using the poppr v.2.1.1 package in R (Kamvar et al. 2014, 2015). (5) Lastly, I explored the effects of clustering assignment when using only marine stickleback, or when including two freshwater populations.

A.3. Results and discussion A.3.1. Effect of retaining Principal Components for Discriminant Function Analysis

238

There were approximately 250 PCs that could be retained for further analysis. The rule of thumb is to use no more than N/3 PCs, which for 224 fish means no more than 74 PCs. The a-scores showed that the optimal PC retention was 22-23 PCs, which accounted for 30% of the total genetic variance. Using randomly-generated populations, Adegenet was able to find ways of partitioning the genetic variance when all PCs were retained such that genetic clusters could be identified. This ability was lessened when using 74 PCs and virtually eliminated by 23 PCs (Figure A-1). The optimal number of PCs using a-scores was indeed sufficient to remove false clustering, with assignment success to these random populations varying from 4% to 48% (average 32%) (Table A-1), in contrast to the analysis used in Chapter Four, where clusters found through k-means had assignment success between 96% and 100%.

A.3.2. The effect of local optima As I was exploring the data, it became evident that the BIC values sometimes changed when running the same dataset multiple times through find.clusters. To see the effects of these slight changes I ran the filtered dataset ten times (Table A-2). The assignment of individuals was identical for each test run with the exception of run 4, which grouped all of OR01 and some OR02 individuals in the WA01 & BC01 & AK01 cluster rather than with the CA03 cluster.

A.3.3. The effect of filtering A large number of minor alleles were filtered from the dataset reported in Chapter Four. To explore the consequences of removing these alleles, the unfiltered and filtered datasets were compared. There were minor differences between datasets, but the overall patterns were highly similar. Unfiltered and maf-filtered datasets had the lowest BIC for k = 5 clusters, and the assignment of individuals to each of these five clusters was nearly identical. The only difference was an additional stray OR02 individual in the CA03 & OR01 cluster in the filtered dataset (Table A-3).

A.3.4. The effect of removing missing values from the dataset

239

Table A-1 - The effect of PC retention on assignment scores using a dataset of pooled fish randomly assigned to one of eight populations. Varying PCs were retained for the DAPC. Columns numbered 1-8 show assignment scores for the eight randomly-created “populations”. Assignment scores are the proportion of individuals that were successfully re-assigned to their cluster after the analysis. Also included is the result of the find.clusters command in Adegenet, using the random data, and a dataset in which individuals were assigned to their actual locality. In all cases the filtered dataset was used.

Dataset N N 1 2 3 4 5 6 7 8 SNPs PCs Random 6797 224 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 Random 6797 150 0.81 0.90 0.90 0.97 0.85 0.90 0.73 0.96 Random 6797 100 0.81 0.60 0.86 0.74 0.81 0.84 0.73 0.75 Random 6797 74 0.77 0.47 0.72 0.77 0.65 0.68 0.59 0.67 Random 6796 23 0.32 0.33 0.48 0.45 0.35 0.29 0.04 0.29 Random, 6797 23 1.0 1.0 1.0 1.0 1.0 0.97 1.0 1.0 find.clusters, Assigned by 6655 23 1.0 0.96 1.0 1.0 0.97 1.0 1.0 0.92 locality

240

Table A-2 - Result of multiple runs in Adegenet using the filtered dataset and a k-means of five. The first ten runs were generated using the same data tested repeatedly. Members of a cluster are grouped in [ ], numbers are used if a few individuals from another locality were included in the cluster. Bold is used to highlight runs that differed from the rest. The last four runs indicate different methods for treating missing data: excluding missing loci if they are absent in > 25% of individuals (“threshold”), which was done twice to reflect different optima; mean, which replaced a missing value with the average value for that locus; and zero, which replaced a missing value with a frequency of 0. The first 10 rounds ignored missing loci.

Test run BIC for k = 5 Clustering assignment k = 5 lowest BIC? 1 378.96 Yes [CA01] [CA02] [CA03, OR01, 7 OR02, 1 AK01] [OR02] [WA01, BC01, AK01] 2 1379.59 No (k = [CA01] [CA02] [CA03, OR01, 7 OR02, 1 AK01] 4) [OR02] [WA01, BC01, AK01] 3 1378.97 No [CA01] [CA02] [CA03, OR01, 7 OR02, 1 AK01] (k=4) [OR02] [WA01, BC01, AK01] 4 1378.96 Yes [CA01] [CA02] [CA03] [OR02] [OR01, 2 OR02, WA01, BC01, AK01] 5 1378.96 Yes [CA01] [CA02] [CA03, OR01, 7 OR02, 1 AK01] [OR02] [WA01, BC01, AK01] 6 1378.96 No (k = [CA01] [CA02] [CA03, OR01, 7 OR02, 1 AK01] 4) [OR02] [WA01, BC01, AK01] 7 1378.96 No (k = [CA01] [CA02] [CA03, OR01, 7 OR02, 1 AK01] 4) [OR02] [WA01, BC01, AK01] 8 1378.96 Yes [CA01] [CA02] [CA03, OR01, 7 OR02, 1 AK01] [OR02] [WA01, BC01, AK01] 9 1378.96 Yes [CA01] [CA02] [CA03, OR01, 7 OR02, 1 AK01]

241

[OR02] [WA01, BC01, AK01] 10 1378.96 Yes [CA01] [CA02] [CA03, OR01, 7 OR02, 1 AK01] [OR02] [WA01, BC01, AK01] threshold 1261.64 Yes [CA01] [CA02] [CA03, 1 OR01] [OR02] [OR01, 1 OR02, WA01, BC01, AK01] threshold 1261.64 No (k = [CA01] [CA02] [CA03, OR01, 5 OR02, 1 AK01] round 2 4) [OR02] [WA01, BC01, AK01] mean 1378.96 No (k = [CA01] [CA02] [CA03, OR01, 7 OR02, 1 AK01] 4) [OR02] [WA01, BC01, AK01] zero 1577.51 No (k = [CA01] [CA02, OR01, OR02, WA01] [CA03] [BC01] 10) [3 CA01, 1 CA02, 2 CA03, 2 OR01, 3 OR02, 3 WA01, 1 BC01, AK01]

242

Table A-3 - Comparisons of how populations clustered at different k-values, for both the unfiltered and filtered datasets. Only the first run per k-value is shown. N PopXX indicates the number of individuals that were assigned to a cluster from that sampling area. Clusters are grouped in [ ].

Dataset k BIC Clusters Unfiltered 3 1459.2010 [CA01] [OR02] [CA02, CA03, OR01, 3 OR02, BC01, WA01, AK01] Filtered 3 1380.2968 [CA01] [OR02] [CA02, CA03, OR01, 2 OR02, BC01, WA01, AK01] Unfiltered 4 1459.9033 [CA01] [CA02, CA03, OR01, 1 AK01, 5 OR02] [OR02] [1 OR01, WA01, BC01, AK01] Filtered 4 1379.8241 [CA01] [CA03, OR01, 7 OR02, 1 AK01] [OR02] [CA02, WA01, BC01, AK01] Unfiltered 5 1458.983 [CA01] [CA02] [CA03, OR01, 6 OR02, 1 AK01] [OR02] [WA01, BC01, AK01] Filtered 5 1378.9557 [CA01] [CA02] [CA03, OR01, 7 OR02, 1 AK01] [OR02] [WA01, BC01, AK01] Unfiltered 8 1466.346 [CA01] [CA02] [4 CA02] [5 CA02] [1 CA02, CA03, OR01, 6 OR02, 1 AK01] [OR02] [BC01] [1 BC01, WA01, AK01] Filtered 8 1386.3846 [CA01] [CA02] [4 CA02] [CA03] [OR01, 7 OR02, 1 AK01] [OR02] [BC01] [1 BC01, WA01, AK01]

243

Figure A-1 - The relationship between clustering and PC retention. (a) Retaining 23 PCs and using Adegenet to identify eight clusters, individuals were largely assigned to their geographic location. (b) through (f) show the results of retaining different numbers of PCs for a dataset in which stickleback are randomly assigned to one of eight artificial populations, and Adegenet was provided with these population identifiers. No population structure should be evident, but Adegenet was able to partition clusters when the majority of PCs were retained. (b) All PCs; (c) 150 PCs; (d) 100 PCs; (e) Recommended maximum number of N/3 = 74 PCs; (f) a-scores recommended 23 PCs, same as in (a).

244

Four or five clusters were best supported whether ignoring, removing some threshold, or averaging across missing loci. At k = 5, assignments were remarkably similar between methods. The only method that produced wildly divergent results was setting missing values to 0, with the best supported cluster number being ten (Table A-2). Collectively these results supported ignoring missing values.

A.3.5. The effect of including freshwater stickleback Initially I had expected that marine stickleback would comprise a single genetic cluster. When they did not (Chapter Four), I checked the validity of the results against randomly-created populations of fish (see above); I further tested these expectations by including freshwater stickleback in the run. This provided a certain set of expectations: (1) that freshwater stickleback would be genetically distinct from the marine stickleback; (2) that this distinction would account for most of the variation in the discriminant components; and (3) that marine populations, even if still distinguishable, would be relatively tightly clustered together relative to the freshwater fish. All three expectations were valid (Figure A-2).

A.3.6. Summary This exploration of Adegenet justified my retention of the a-score recommended number of PCs; demonstrated that local optima can have small effects on clustering assignment, thereby requiring that a dataset be run multiple times before assigning clusters; and showed that decisions about filtering criteria and dealing with missing values can have small to profound results on clustering assignment.

245

Figure A-2 - Adegenet clustering results for k = 7. Freshwater: 5 = Brannen Lake, Vancouver Island, BC, 2 = Mayliewan Stormwater Drainage Pond, Edmonton, Alberta. The rest are the marine populations used in this thesis. (Left) DF1 on x-axis vs. DF2 on y-axis, (Right) DF3 on x-axis vs. DF4 on y-axis.

246

APPENDIX B: SUPPLEMENTAL TABLES AND FIGURES TO CHAPTER FOUR Table B-1 - List of landmarks used for 3D morphometrics. Adapted from Tegan Barry’s MSc proposal (n.d.). Landmark Description 1 Anterior tip of dentary 2 Anterior tip of premaxilla 3 Anterior tip of maxilla 4 Anterior corner of nasal ventrolateral process 5 Dorsal corner of nasal-lateral ethmoid suture 6 Dorsal maxima of lachrymal 7 Lachrymal-prefrontal suture on orbital 8 Anterior tip of articular 9 Ventral maxima of lachrymal 10 Dorsal tip of articular 11 Ventral-most tip of articular 12 Lachrymal-second orbital suture 13 Anterior tip of preoperculum 14 Dorsal-most tip of supraorbital 15 Ventral-most tip of sphenotic 16 Dorsal-most tip of third suborbital 17 Posterior minima of third suborbital 18 Ventral-most tip of third suborbital 19 Anterior minima of preoperculum 20 Anterior dorsal-most tip of preoperculum 21 Posterior maxima of preoperculum, first ridge 22 Ventral dorsal-most tip of preoperculum 23 Dorsal-most tip of interoperculum 24 Ventral maxima of preoperculum, second ridge 25 Ventral-most tip of interoperculum 26 Dorsal-most tip of suboperculum

247

27 Ventral maxima of suboperculum 28 Posterior tip of suboperculum 29 Dorsal-most tip of operculum 30 Anterior maxima of operculum 31 Anterior minima of operculum 32 Ventral-most tip of operculum 33 Posteriodorsal tip of operculum 34 Opercular hinge angle 35 Posterior tip of pterotic 36 Anterior tip of ectocoracoid 37 Posterior tip of ectocoracoid 38 Anterior tip of pelvic plate 39 Anterior midline of pelvic plate at suture point 40 Minima of pelvic plate at trochlear joint 41 Maxima of pelvic process 42 Posterior tip of pelvic process 43 Anterior minima of ascending process of pelvic plate 44 Anteriodorsal maxima of ascending process of pelvic plate 45 Posteriodorsal maxima of ascending process of pelvic plate 46 Posterioventral maxima of ascending process of pelvic plate at trochlear joint 47 Dorsal-most tip of pelvic spine 48 Ventral-most tip of pelvic spine 59 Posterior tip of pelvic spine 50 Midline of plate 4 at lateral pores 51 Ventral tip of plate 4 52 Midline of plate 5 at lateral pores 53 Ventral tip of plate 5 54 Midline of plate 6 at lateral pores 55 Midline of plate 7 at lateral pores

248

Table B-2 – The numbers of each sex, plate morph, and Stn382 genotype sampled from California to Alaska. The number of fish within each category that made it through the Stacks pipeline are indicated in brackets. See the List of Symbols, Abbreviations and Nomenclature for the meaning of row and column headings. * Indicates violation from HWE for Stn382. Note that not every fish could be both genotyped and phenotyped.

Population M F FPK PPK LPNK PPNK LPK CC CL LL CA01 26 9 (6) 2 (1) 5 (4) 25 0 (0) 3 0 (0) 5 (5) 29 (23) (21) (3) (24) CA02 20 (8) 30 7 (4) 9 (4) 34 0 (0) 0 0 (0) 16 (8) 34 (20) (20) (0) (20) CA03 31 18 2 (1) 5 (3) 42 0 (0) 0 1 (0) 6 (4) 42 (17) (11) (24) (0) (24) OR01 25 25 20 0 (0) 0 (0) 0 (0) 0 31 17 (9) 1 (1) (14) (15) (18) (0) (19) OR02 13 (7) 37 12 (5) 11 24 1 (1) 2 4 (1) 21 25 (23) (9) (14) (1) (14) (15) WA01 35 16 (7) NA NA NA NA NA 49 2 (1) 0 (0) (18) (24) BC01 47 4 (3) 48 0 (0) 0 (0) 0 (0) 0 51 0 (0) 0 (0) (28) (31) (0) (31) AK01* 8 (6) 23 27 0 (0) 4 (1) 0 (0) 0 24 0 (0) 3 (1) (18) (23) (0) (22)

249

Table B-3 - Number of reads dropped or retained by process_radtags. Measurement N reads Total sequenced reads 249 038 733 Total containing adapter 1 839 266 Ambiguous barcode drops 40 205 180 Low quality read drops 129 991 Ambiguous RAD-tag drops 13 977 761 Retained reads 192 886 535 % retained 77.4%

250

Table B-4 - Bayesian Information Criteria (BIC) for clusters k = 2 to 8, for the SNP dataset using all eight marine sites. Members of a cluster are grouped in [ ], numbers are used if a few individuals from another locality were included in the cluster.

Number of clusters BIC value Clusters 2 1385.45 [OR02] [CA01, CA02, CA03, OR01, WA01, BC01, AK01] 3 1380.30 [OR02] [CA01] [CA02, CA03, OR01, 2 OR02, WA01, BC01, AK01] 4 1379.82 [OR02] [CA01] [CA03, OR01, 7 OR02, 1 AK01] [CA02, WA01, BC01, AK01] 5 1378.96 [OR02] [CA01] [CA03, OR01, 7 OR02, 1 AK01] [CA02] [WA01, BC01, AK01] 6 1380.75 [OR02] [CA01] [OR01, 7 OR02, 1 AK01] [CA02] [CA03] [WA01, BC01, AK01] 7 1383.82 [OR02] [CA01] [OR01, 7 OR02, 1 AK01] [CA02] [CA03] [BC01] [WA01, 1 BC01, AK01] 8 1386.38 [OR02] [CA01] [OR01, 7 OR02, 1 AK01] [CA02] [CA03] [BC01] [WA01, 1 BC01, AK01] [4 CA02]

251

Table B-5 - Bayesian Information Criteria (BIC) values for k = 1 to 3, for each marine site treated on its own through the Stacks pipeline and Adegenet, to test for cryptic population structure. The number of Single Nucleotide Polymorphisms (SNPs) retained by Stacks is indicated. The final test included all fish from the Adegenet-recognized northern cluster. To maximize the potential differences between individuals, 22 PCs were retained despite being far higher than N/3. The lowest BIC values are shown in bold.

Marine site SNPs k = 1 k = 2 k = 3 CA01 15413 211.388 213.388 215.277 CA02 30492 204.388 205.175 206.2744 CA03 19185 204.941 206.615 208.571 OR01 21904 218.062 220.098 222.086 OR02 22377 229.990 231.841 233.863 WA01 17342 179.686 181.580 NA BC01 15416 226.996 229.160 231.285 AK01 8702 153.336 154.833 NA WA01 & BC01 & AK01 7308 516.189 517.536 519.439

252

Table B-6 - Pairwise FST values for Adegenet-recognized clusters.

CA01 CA02 CA03 & OR01 OR02 CA02 0.107 CA03 & OR01 0.114 0.053 OR02 0.215 0.169 0.105 WA01 & BC01 & AK01 0.161 0.094 0.061 0.206

253

Table B-7 - Full population genetic statistics when using the identified Adegenet clusters. (Top) For all variant positions. (Bottom) For all sites. Note that the CA03 & OR01 cluster includes a few individuals from OR02 and AK01 (see Chapter Four). Private = Private alleles. N = Number of individuals used. Loci = The average number of loci that were sequenced. Variant = The number of loci that were polymorphic in at least one marine site. SNP = The number of Single Nucleotide Polymorphisms. % Poly = The proportion of variant loci that were polymorphic in the marine site of interest (top) or the proportion of sequenced loci that were polymorphic in the marine site of interest (bottom). See the List of Symbols, Abbreviations and Nomenclature for the meaning of other abbreviations.

Cluster Private N % Poly P HetO HO HetE HE π FIS CA01 80 25.8 56.8 0.90 0.13 0.87 0.14 0.86 0.14 0.03 CA02 35 25.3 68.6 0.90 0.14 0.86 0.15 0.85 0.15 0.03 CA03 & OR01 14 54.9 90.0 0.89 0.14 0.86 0.16 0.84 0.16 0.09 OR02 10 21.9 77.6 0.88 0.17 0.83 0.18 0.82 0.18 0.04 WA01 & BC01 & AK01 47 66.7 79.3 0.92 0.12 0.88 0.13 0.87 0.13 0.07

Cluster Loci Variant loci Poly loci % Poly N P HetO HO HetE HE π FIS CA01 286058 4299 2441 0.85 26.3 0.999 0.002 0.998 0.002 0.998 0.002 0.0004 CA02 286058 4299 2947 1.03 25.72 0.999 0.002 0.998 0.002 0.999 0.002 0.0005 CA03 & ORO1 286058 4299 3869 1.35 56.0 0.998 0.002 0.998 0.002 0.998 0.002 0.0013 OR02 286058 4299 3335 1.17 22.1 0.998 0.003 0.997 0.003 0.997 0.003 0.0006 WA01 & BC01 & AK01 286058 4299 3410 1.19 67.7 0.999 0.002 0.998 0.002 0.998 0.002 0.0011

254

Table B-8 - 3D morphometric variation from Canonical Variate Analysis (CVA) among seven marine groups, scaled by the inverse of the within-group variation, for CVs 1 through 14.

CV Eigenvalues % Variance Cumulative % 1 48.34 32.04 32.04 2 37.48 24.85 56.89 3 17.30 11.47 68.36 4 12.90 8.55 76.91 5 8.74 5.79 82.70 6 5.64 3.74 86.44 7 5.28 3.50 89.94 8 4.29 2.84 92.78 9 3.19 2.11 94.89 10 2.15 1.43 96.32 11 1.69 1.12 97.44 12 1.63 1.08 98.52 13 1.34 0.89 99.41 14 0.89 0.59 100

255

Table B-9 - Procrustes distance (above diagonal) and p-values (below diagonal) for Canonical Variate Analysis (CVA) with site and sex as categorical variables. Note that BC01 had a single female and OR01 had a single individual not assigned to sex; the results of those two individuals are not shown. F = female, M = male, * = significance below 0.0005.

CA01, CA01, CA02, CA02, CA03, CA03, OR01, OR01, OR02, OR02, BC01, AK01, AK01,

F M F M F M F M F M M F M CA01, F 0.039 0.030 0.043 0.041 0.043 0.051 0.046 0.043 0.054 0.073 0.111 0.098 CA01, * 0.052 0.025 0.045 0.033 0.071 0.045 0.050 0.039 0.062 0.108 0.088 M CA02, F 0.007 * 0.047 0.034 0.043 0.035 0.052 0.043 0.063 0.075 0.103 0.095 CA02, * * * 0.039 0.025 0.060 0.036 0.040 0.033 0.052 0.103 0.084 M CA03, F * * * * 0.028 0.047 0.051 0.038 0.051 0.066 0.089 0.076 CA03, * * * * * 0.055 0.038 0.041 0.036 0.061 0.100 0.079 M OR01, F * * * * * * 0.052 0.042 0.067 0.067 0.096 0.091 OR01, * * * 0.003 * * * 0.039 0.034 0.052 0.111 0.090 M OR02, F * * * * * * * * 0.039 0.057 0.096 0.085 OR02, * * * * * * * 0.016 * 0.055 0.109 0.085 M BC01, * * * * * * * * * * 0.089 0.072

256

M AK01, * * * * * * * * * * * 0.003 F AK01, * * * * * * * * * * * 0.042 M

257

Table B-10 - The number of individuals assigned to the wrong location based on phenotype for the Discriminant Function Analysis (DFA) (above diagonal) and cross- validation (below diagonal).

CA01 CA02 CA03 OR01 OR02 BC01 AK01 CA01 0 0 0 0 0 0 CA02 9 1 0 0 0 0 CA03 4 10 0 0 0 0

OR01 0 3 2 0 0 0 OR02 1 7 8 2 0 0 BC01 1 0 2 1 5 0 AK01 2 0 0 0 1 1

258

Table B-11 - Weir and Cockerham global pairwise FST for non-outlier loci (above diagonal) and outlier loci (below diagonal). Loci were identified as outliers using the five Adegenet-recognized clusters, and then pulled from the initial SNP dataset identifying eight marine sites.

CA01 CA02 CA03 OR01 OR02 WA01 BC01 AK01 CA01 0.099 0.110 0.106 0.142 0.125 0.132 0.086 CA02 0.283 0.060 0.056 0.100 0.073 0.083 0.054 CA03 0.351 0.178 0.041 0.074 0.073 0.086 0.057 OR01 0.395 0.214 0.147 0.050 0.040 0.053 0.032 OR02 0.530 0.493 0.403 0.373 0.096 0.112 0.068 WA01 0.575 0.419 0.409 0.125 0.518 0.027 0.022 BC01 0.666 0.538 0.533 0.204 0.590 0.041 0.025 AK01 0.402 0.288 0.355 0.115 0.400 0.016 -0.001

259

Table B-12 - Pairwise FST values and outlier locus information for each of the eight marine-freshwater comparisons. Marine-BCFW identifies the marine group that was compared to the Brannen Lake, BC, freshwater population. N loci is the number of loci for which an FST value could be calculated. 5% threshold is the FST value above which a locus was flagged as an outlier. N outlier loci is the number of flagged loci above that threshold. N unique is the number of outliers that were only flagged as outliers in the comparison of interest. N excluded is the number that were flagged in all comparisons except the comparison of interest.

BCFW paired FST N 5% N outlier N N with loci threshold loci unique excluded CA01 0.365 1471 0.806 74 2 0 CA02 0.329 1540 0.760 77 3 0 CA03 0.308 1529 0.716 77 3 0 OR01 0.274 1610 0.672 81 0 0 OR02 0.184 1604 0.464 81 42 28 WA01 0.351 1453 0.845 73 0 0 BC01 0.373 1440 0.896 72 0 0 AK01 0.385 1337 0.875 69 0 3

260

Table B-13 - The number of polymorphic loci, out of 1912, for which FST values could be calculated for any given number of marine-freshwater pairwise comparisons. N loci is the number of Single Nucleotide Polymorphisms (SNPs) for which pairwise FST values could be calculated. N pairwise is the number of marine-freshwater pairwise comparisons which had FST values that could be calculated. In other words, 60 SNPs produced FST estimates for one of eight possible marine-freshwater comparisons, whereas 1029 produced FST estimates for all eight marine-freshwater comparisons.

N pairwise N loci 1 60 2 93 3 125 4 194 5 172 6 138 7 101 8 1029

261

Figure B-1 – The distribution of private allele frequencies per population. (a) CA01, (b) CA02, (c) CA03, (d) OR01, (e) OR02, (f) WA01, (g) BC01, (h) AK01.

262

Figure B-2 - The distribution of major allele frequencies for each population. (a) CA01, (b) CA02, (c) CA03, (d) OR01, (e) OR02, (f) WA01, (g) BC01, (h) AK01.

263

Figure B-3 - The distribution of per-locus FIS values for each population. (a) CA01, (b) CA02, (c) CA03, (d) OR01, (e) OR02, (f)

WA01, (g) BC01, (h) AK01. Note that FIS values less than -0.3 were filtered from the dataset.

264

Figure B-4 - Isolation-by-Distance as tested using a Mantel test of pairwise geographic distances (km) on pairwise genetic distances (Weir and Cockerham FST), at 999 replications. (a) For all populations, p = 0.8 (b) Excluding AK01, p = 0.02.

265

Figure B-5 - The distribution of private allele frequencies when assigning individuals based on Adegent clustering at k = 5. (a) CA01, (b) CA02, (c) CA03 & OR01 (& 7 OR02, 1 AK01), (d) OR02, (e) WA01 & BC01 & AK01. Compare to Figure B-1.

266

Figure B-6 – (A) % variance in stickleback morphology explained by each Principal Component. (B) The association between PC1 and PC2, with individuals identified based on population. Below PC1 are wireframes showing the change in shape (dark blue) relative to the consensus fish (light blue) along the PC1 axis. Note that this shape change is similar to that noted for CV1 in the main text.

267

Figure B-7 - Mantel tests for (a) geographic distance and neutral FST, (b) geographic distance and plate PST, (c) geographic distance and FSTQ, (d) neutral FST and plate PST, (e) neutral FST and FSTQ, (f) Plate PST and FSTQ. See Table 4-5 for associated statistics.

268

Figure B-8 - Mantel tests for PST for PC1 (a-c), PC2 (d-f), PC3 (g-i), and PC4 (j-l) compared to geographic distance (a,d,g,j), neutral FST (b,e,h,k), and FSTQ (c,f,i,l). See Table 4-5 for associated statistics.

269

Figure B-9 - The relationship between the number of Single Nucleotide Polymorphisms (SNPs) identified on a Linkage Group (LG) or scaffold, and the number of identified outlier loci using a 5% threshold. Linkage Groups with unusual numbers of outliers are identified.

270

Figure B-10 - From left to right: BIC-values from k-means clustering, Adegenet-identified clusters, Isolation-by-Distance including Alaska, Isolation-by-Distance excluding Alaska, for (a) non-outlier loci and (b) outlier loci.

271

Figure B-11 - Manhattan plot for the CA01-BCFW pairwise comparison. The blue line shows the threshold for identifying the top 5% of outliers.

272

Figure B-12 - Manhattan plot for the CA02-BCFW pairwise comparison. The blue line shows the threshold for identifying the top 5% of outliers.

273

Figure B-13 - Manhattan plot for the CA03-BCFW pairwise comparison. The blue line shows the threshold for identifying the top 5% of outliers.

274

Figure B-14 - Manhattan plot for the OR01-BCFW pairwise comparison. The blue line shows the threshold for identifying the top 5% of outliers.

275

Figure B-15 - Manhattan plot for the OR02-BCFW pairwise comparison. The blue line shows the threshold for identifying the top 5% of outliers.

276

Figure B-16 - Manhattan plot for the WA01-BCFW pairwise comparison. The blue line shows the threshold for identifying the top 5% of outliers.

277

Figure B-17 - Manhattan plot for the BC01-BCFW pairwise comparison. The blue line shows the threshold for identifying the top 5% of outliers.

278

Figure B-18 - Manhattan plot for the AK01-BCFW pairwise comparison. The blue line shows the threshold for identifying the top 5% of outliers.

279

APPENDIX C: RELATIONAL ASYMMETRY (SUPPLEMENT TO CHAPTER FIVE) C.1. Summary Plates P1 through P9 were potentially associated with other bony structures, including the cleithrum (P1 and P2), the dorsal basal plates (P3-P9) and the ascending process (P7 and P8). Associations with these structures in part distinguish structural from non-structural plates, but while investigating structural plates it became evident that some interesting asymmetries occurred when lateral plates were symmetrically present. That is, rather than display presence/absence asymmetry, plates were often asymmetric in their relationship with other structures, what I have called relational asymmetry (RELA). Unfortunately, the complexities of these relationships were such that no signed value could be meaningfully provided, and so no tests of fluctuating asymmetry could be made. For those interested, I present the RELA data here, if for no other reason than that I find it interesting and it may prove useful in the future. Below, non-structural plates refer to P1- P3 and P8-P9. Excluding symmetrically or asymmetrically absent plate positions from the analysis resulted in 195 RELA measures (Figure C-1). Plates defined by position exhibited variation in the incidence of asymmetry (χ2 = 136, d.f. = 10, p < 0.001) (Figure C-2). Non-structural plate asymmetries were evenly distributed (χ2 = 9, d.f. =5, p = 0.1), but structural plate asymmetries were not (χ2 = 142, d.f. = 4, p < 0.001). Non-structural plates had marginally higher average RELA than structural plates (Welch’s two-sample t- test: t = -2.0, d.f. = 559, p = 0.042). The proportion of asymmetries in non-structural plates varied from 3% (P3) to 10% (P9), at an average incidence of 8% (Figure C-2, Table C-1). Curiously, P3, which is not considered a structural plate, did abut DBP1 symmetrically in 37 individuals, and asymmetrically in 23. This phenotype was particularly common in BC01 (53% of all occurrences, symmetrical or asymmetrical). The proportion of asymmetries in structural plates varied from 0% (P7) to 21% (P6), at an average incidence of 6%, or 2% if ignoring P6. These measures reduce a variety of phenotypes per plate into a single measure of

280

asymmetry. For instance, P1 symmetrically abutted the cleithrum in 126 individuals, and symmetrically was detached from the cleithrum in 2 individuals; the remainder that symmetrically had plates were asymmetric for one of these combinations. P6 had the widest variety of possible phenotypic combinations, and the highest rate of asymmetry (Table C-1). This plate could develop such that it came into contact with the first or second dorsal basal plate, or both, or neither. The most common phenotype was symmetrical abutment of DBP2 (n = 176). Considering each side as a separate phenotype, 19% of P6 touched no dorsal basal plate, 5.3% touched the first dorsal basal plate, 69% touched the second dorsal basal plate, and 4.4% touched both. Reimchen (1983) found that P6 exhibited the least variability in presence/absence among all plates in freshwater populations, was among the first plate to develop, and contributed substantially to deflection of the pelvic and both dorsal spines. The high degree of variation noted here suggests that the function of P6 is determined plastically early in development; follow-up studies will need to focus on whether other plates compensate for this variation in structural role. In contrast, all but one individual had both P7 plates abutting DBP2, while no P5 plates abutted DBP2. P7 almost always abutted the AP (94% of individuals), but 13% of individuals had at least one P8 abutting the AP, marking it as a rare structural plate. This rare phenotype occurred predominantly in CA02 (29% of all occurrences), AK01 (29%), and BC01 (24%). PPOSA and RELA were not correlated (Pearson correlation: r = 0.05, t = 0.8, d.f. = 281, p = 0.4). There was no association between sMLH and non-structural RELA (Pearson correlation: r = 0.12, t = 1.7, d.f. = 185, p = 0.09) (Figure C-3). There was, however, a significant negative relationship between sMLH and structural RELA (Pearson correlation: r = -0.16, t = -2.2, d.f. = 185, p = 0.03), but this disappeared when Alaska was removed from the dataset (Pearson correlation: r = -0.02, t = -0.2, d.f. = 161, p = 0.8) (Figure C-3).

281

Table C-1 - Number of phenotypic variants for plates P1 to P9, including PPOSA and RELA. N = plate absent. 0 = plate does not abut the cleithrum (P1-P2) or any dorsal plate (P3-P9) or the ascending process (P7AP, P8AP). 1 = plate does abut the cleithrum (P1- P2), the first dorsal basal plate (P3-P9), or the ascending process (P7AP, P8AP). 2 = the plate abuts the second dorsal basal plate (P3-P9). 3 = the plate abuts the first and second dorsal basal plate (P4-P9). 4 = the plate abuts a rare intermediate dorsal basal plate (P4- P9). 5 = the plate does not abut but is oriented towards either the first or second dorsal basal plate (P4-P9). Columns represent the right side of the fish, rows represent the left side of the fish; numbers indicate the number of fish which have that left/right combination of phenotypes. Symmetrical phenotypes are labeled in bold.

P1 N 0 1 N 126 7 8 0 3 2 1 1 7 3 126 P2 N 0 1 N 25 5 4 0 7 87 8 1 1 20 126 P3 N 0 1 N 1 1 0 0 2 219 12 1 0 11 37 P4 N 0 1 N 0 4

282

1 4 275 P5 N 0 1 2 3 4 5 N 0 1 1 1 1 278 2 3 4 1 5 1 P6 N 0 1 2 3 4 5 N 1 0 31 3 13 1 4 1 4 6 3 4 1 2 8 176 2 2 3 2 2 7 4 3 5 3 1 5 1 P7 N 0 1 2 3 4 5 N 0 1 2 282 3 4 1 5 P8 N 0 1 2

283

N 26 4 3 0 6 45 11 1 2 6 13 169 P9 N 0 1 2 N 113 8 0 5 139 6 1 2 6 6 P7AP N 0 1 N 0 4 5 1 9 265 P8AP N 0 1 N 26 7 0 12 200 14 1 0 6 18

284

Figure C-1 – The number of individuals exhibiting different degrees of RELA per population. In this case, RELA was unsigned and summed per individual.

285

Figure C-2 – The proportion of fish exhibiting relational asymmetry for each plate position. In this case, asymmetry was not included if it involved presence/absence asymmetry (PPOSA). AP = Ascending Process.

286

Figure C-3 – (A) Structural RELA and sMLH; (B) Non-structural RELA and sMLH.

287

APPENDIX D: PROTOCOLS D.1. Sex-specific marker: idh PCR protocol Reagent Volume per one 10 μl reaction molecular grade water 6.55 μl NEB buffer 1 μl dNTP 0.25 μl reverse primer 0.5 μl forward primer 0.5 μl Taq polymerase 0.2 μl DNA (7 ng) 1 μl

Forward primer sequence: 5’-GGG ACG AGC AAG ATT TAT TG-3’ Reverse primer sequence: 5’-TTA TCG TTA GCC AGG AGA TGG-3’

Thermocycler protocol: Cycle number Temperature °C Duration 1 95 1 min 45 sec 56 45 sec 68 45 sec 2 through 36 94 45 sec 59 45 sec 68 45 sec final 68 5 min

288

D.2. Plate morphs: Stn382 PCR protocol for Eda Reagent Volume per one 10 μl reaction molecular grade water 6.6 μl

10X NEB buffer with 2 mM MgCl2 1 μl dNTP 0.3 μl reverse primer 0.5 μl forward primer 0.5 μl Taq polymerase 0.1 μl DNA (14 ng) 1 μl

Thermocycler protocol: Cycle number Temperature °C Duration 93 3 min 1 95 30 sec 56 30 sec 68 30 sec 2 through 6 94 30 sec 56 30 sec 68 30 sec 7 through 36 90 30 sec 56 30 sec 68 30 sec final 68 10 min

289

D.3. SimRAD This program (Lepais and Weir 2014) in R was used to perform in silico digestions of each stickleback linkage group, in order to determine the number of expected fragments I would get using EcoRI and MseI. An example output figure is shown below.

500

400

300

200

Number of loci of Number

356 loci between 200 and 270 bp

100

0 1000 2000 3000 4000

Locus size (bp)

Figure D-1 - Sample output from the in silico digestion of Linkage Group XXI, showing the number of fragments of type “AB+BA” that were within the 200-270 base pair range.

290

D.4. Digestion of DNA by restriction enzymes for GBS Reagent Volume per reaction 10X T4 DNA ligase buffer 1.15 μl 1M NaCl 0.60 μl BSA (1 mg/ml) 0.60 μl molecular grade water 0.25 μl MseI (10 000 units/ml) 0.12 μl High Fidelity EcoRI (20 000 units/ml) 0.28 μl

Incubation temperature (°C) Duration 37 8 hours 65 45 min

Begin with 200-600 ng starting DNA. Use 3 μl Master Mix and 6 μl DNA

291

D.5. Ligation for GBS Do 1-2 days after digestion. Make ~50% more than required. This step will ligate adapters to digested fragments – add directly to digested DNA. Reagent Volume per reaction MseI adapter (10 μM) 1 μl molecular grade water 0.072 μl 10X T4 DNA ligase buffer 0.1 μl 1 M NaCl 0.05 μl BSA (1 mg/ml) 0.05 μl T4 DNA ligase 0.1675 μl

1. Add 1.4 μl Master Mix to each digestion reaction 2. Add 1 μl of EcoRI adapter (1 μM) Incubation temperature (°C) Duration 16 6 hours

292

D.6. PCR amplification of ligated product for GBS Reagent Volume per reaction 2X NEB Q5 High-Fidelity Master Mix 10 μl 10 μM combined primers 5.34 μl

Combine 4 μl sample with 16 μl Master Mix (20 μl total) Thermocycler protocol Cycle number Temperature °C Duration 98 30 sec 1 through 19 98 20 sec 60 30 sec 72 30 sec 20 72 2 min

293

APPENDIX E: PERMISSIONS

Re: Jordan's Rule

Heather Ann Jamniczky Mon 3/20/2017 8:57 AM

To:Ella Bowles ;

Cc:Ekaterina Petrovitch ; Matthew Morris ; Sean Rogers ;

Hi Matthew,

This looks great ‐ fine with me, for both this and the 3D morphometrics piece. I’ll have comments back to you shortly ‐ I’ve got another one on my desk I have to read through first. Give me a day or two.

Heather

‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐ Heather A. Jamniczky, PhD [email protected]

On Mar 18, 2017, at 5:40 PM, Ella Bowles wrote:

Hi Matthew,

Yes, it is definitely fine for you to include the document.

Ella

294

On Sat, Mar 18, 2017 at 5:19 PM, Ekaterina Petrovitch wrote: Hi Matthew,

Yes of course! Good luck!

Cheers, Ekaterina (Katy) Petrovitch

On Mar 18, 2017, at 5:17 PM, Matthew Morris wrote:

Hi all,

As you may know, I am defending my PhD May 29. As coauthors on the Jordan's Rule paper, I need your permission to include it in my thesis.

To give you a sense of the overall scope of the thesis, it looks something like this:

Chapter 1 ‐ Intro to standing genetic variation, contemporary ancestors, and the significance of marine threespine stickleback

Chapter 2 ‐ Mitochondrial plasticity in marine and freshwater stickleback

Chapter 3 ‐ (Uninteresting sampling procedures, so I don't have to repeat this method in every subsequent chapter)

295

Chapter 4 ‐ Jordan's Rule in marine threespine sĕckleback

Chapter 5 ‐ Morphological and genetic variation in marine threespine stickleback, and evidence for selection in the marine environment (Heather ‐ this is where the 3D scans come into play!)

Chapter 6 ‐ Eda as genetic stress: asymmetry in platedness in threespine stickleback

It is sufficient that you send me an email saying that you agree to the incorporation of the document in my thesis. I will have to note the contributions of each author, including of course Ekaterina's X‐raying of the fish and her initial analyses of vertebral number.

With thanks,

Matthew

296

Figure E-1 – Permission for the use of Figure 1-2.

297