Associated Behavioral, Genetic, and Gene Expression Variation with Alternative Life History Tactics in Salmonid Fishes Ashley Chin-Baarstad Purdue University

Purdue University Purdue e-Pubs

Open Access Dissertations Theses and Dissertations

Fall 2014 Associated behavioral, genetic, and gene expression variation with alternative life history tactics in salmonid fishes Ashley Chin-Baarstad Purdue University

Follow this and additional works at: https://docs.lib.purdue.edu/open_access_dissertations Part of the Biology Commons, Cell Biology Commons, Genetics Commons, Molecular Biology Commons, and the Social and Behavioral Sciences Commons

Recommended Citation Chin-Baarstad, Ashley, "Associated behavioral, genetic, and gene expression variation with alternative life history tactics in salmonid fishes" (2014). Open Access Dissertations. 245. https://docs.lib.purdue.edu/open_access_dissertations/245

This document has been made available through Purdue e-Pubs, a service of the Purdue University Libraries. Please contact [email protected] for additional information.

ASSOCIATED BEHAVIORAL, GENETIC, AND GENE EXPRESSION VARIATION

WITH ALTERNATIVE LIFE HISTORY TACTICS IN SALMONID FISHES

A Dissertation

Submitted to the Faculty

Purdue University

Ashley Chin-Baarstad

In Partial Fulfillment of the

Requirements for the Degree

Doctor of Philosophy i

December 2014

Purdue University

West Lafayette, Indiana

For all the loved ones in my life for believing that I could

iii

ACKNOWLEDGEMENTS

I would like to thank Dr. Krista Nichols, not only for her support and advice throughout this process, but also for giving me the opportunity to enter the world of genetics. I would also like to thank each of my committee members, Dr. Esteban Fernandez-Juricic, Dr.

Rick Howard, and Dr. Greg Hunt, for providing feedback and advice throughout my graduate career. Their insightful comments and questions that have helped me grow as a researcher and a writer. I appreciate the support and friendship I received from both regular and seasonal staff at the Little Port Walter Marine Research Station, National

Marine Fisheries Service, on Baranof Island, Alaska. In particular, I was touched by the way in which “Grandpa” Frank Thrower welcomed and encouraged me with his humor and kind heart. I would also like to thank members of the Nichols lab past and present:

Sunnie McCalla, Ben Hecht, Shuai Chen, Julie Scardina, Garrett McKinney, and Matt

Hale, for their support and friendship over the years. My time at Purdue was truly enriched by both my “lab” family and all the amazing people I met throughout the years in Indiana. I thank all my friends near and far for bolstering my spirits when I needed it iii

and believing in me. Finally, I would like to thank my family for their love and support in all my endeavors.

TABLE OF CONTENTS

Page LIST OF TABLES ...... vii LIST OF FIGURES ...... ix ABSTRACT ...... xi CHAPTER 1. INDIVIDUAL BEHAVIORAL VARIATION IN JUVENILE RAINBOW TROUT, ONCORHYNCHUS MYKISS ...... 1 1.1 Abstract ...... 1 1.2 Introduction ...... 2 1.3 Materials and Methods ...... 6 1.3.1 Field sampling and genetic crosses ...... 6 1.3.2 Behavioral experiments ...... 8 1.3.3 Genetic sex-typing ...... 12 1.3.4 Statistical analyses ...... 12 1.4 Results ...... 13 1.4.1 Fish Size and Sex ...... 13 1.4.2 Behavioral traits ...... 14 1.4.3 Dispersal ...... 14

1.4.4 Exploration ...... 15 iv 1.4.5 Aggression ...... 15 1.4.6 Correlations ...... 16 1.5 Discussion ...... 16 1.6 Acknowledgements ...... 21

Page CHAPTER 2. DIFFERENTIAL GENE EXPRESSION UNDERLYING BEHAVIORAL VARIATION IN THE BRAINS OF JUVENILE RAINBOW TROUT, ONCORHYNCHUS MYKISS ...... 34 2.1 Abstract ...... 34 2.2 Introduction ...... 35 2.3 Materials and Methods ...... 38 2.3.1 Field sampling and genetic crosses ...... 38 2.3.2 Behavioral experiments ...... 39 2.3.3 Sample collection ...... 41 2.3.4 Statistical analyses and sample selection ...... 42 2.3.5 RNA extraction and sequencing ...... 44 2.3.6 Gene expression ...... 44 2.3.7 Contrasts ...... 46 2.3.8 Annotation ...... 47 2.3.9 Pathway enrichment analysis ...... 48 2.4 Results ...... 49 2.4.1 RNAseq ...... 49 2.4.2 RNAseq differential gene expression ...... 50 2.4.3 Contrasts ...... 50 v

2.4.4 Biofunction analyses ...... 53 2.4.5 Pathway analyses ...... 59 2.5 Discussion ...... 63 2.6 Acknowledgements ...... 34 CHAPTER 3. GENOME-WIDE ASSOCIATION STUDY (GWAS) OF ECOTYPES IN LAKE SUPERIOR BROOK TROUT, SALVELINUS FONTINALIS ...... 81 3.1 Abstract ...... 81 3.2 Introduction ...... 82 3.3 Materials and Methods ...... 86 3.3.1 Sampling and DNA extraction ...... 86

Page 3.3.2 RAD library construction ...... 87 3.3.3 Bioinformatics ...... 87 3.3.4 SNP filtering ...... 88 3.3.5 Genome-wide association analyses ...... 89 3.3.6 Alignment of RAD-tags and annotation of genes ...... 90 3.3.7 Population genetic analyses ...... 91 3.4 Results ...... 92 3.4.1 Samples and phenotypes ...... 92 3.4.2 SNP filtering ...... 92 3.4.3 GWAS ...... 92 3.4.4 Alignment and annotation of RAD-tags ...... 93 3.4.5 Analysis of population structure ...... 94 3.5 Discussion ...... 94 3.6 Acknowledgements ...... 81 APPENDICES Appendix A Supplementary Material for Chapter 1 ...... 123 Appendix B Supplementary Material for Chapter 2 ...... 125 Appendix C Supplementary Material for Chapter 3 ...... 127 VITA ...... 133 vi

PUBLICATION ...... 136

vii

LIST OF TABLES

Table ...... Page Table 1-1 Description of each behavioral trait including an abbreviated name used for the trait throughout the text, the behavioral experiment used to measure that trait, the mean, standard deviation, and the number of observations for each trait . 22

Table 1-2 The means, standard deviations, and numbers of observations for each behavioral trait by cross type (anadromous (AXA), resident (RXR), and hybrid and reciprocal hybrids among life history types (AXR and RXA)) .... 23

Table 1-4 The means, standard deviations, and numbers of observations for each behavioral trait in all trials by sex ...... 25

Table 1-6 Significant (α < 0.05) model effects for each trait are indicated with the presence of an 'x' in the effect column for the row trait...... 27

Table 2-1 Summary of the contrasts that were done to find differential gene expression associated with behavioral traits ...... 76

viii

Table ...... Page Table 2-3 Summary of the number of significant results for each contrast in the IPA analysis (Fisher’s exact p<0.05)...... 78

Table 3-1 Number of samples captured at each location that were sequenced at RAD-tag sites, passed pruning (greater than 60% genotyping success rate), and were used in the GWAS analyses (had complete data, including stable isotope) . 100

Table 3-3 Pairwise population FST value for individuals used in the GWAS analyses using their nine capture locations and 906 loci ...... 102

viii

LIST OF FIGURES

Figure ...... Page Figure 1-1 Dispersal apparatus used to measure dispersal behaviors of rainbow trout fry. The apparatus consisted of 15 equal sized compartments with flow through of creek water at ambient temperature. The water flowed in one direction, entering at one end and exiting at the other (indicated by the direction of the black arrows between compartments, left to right). The divider between each compartment had a hole offset from the middle (example shown in the lower left hand corner) and positioned to alternate from one side to the next along the sequence of dividers (indicated by the position of the black arrows between compartments). A test individual was placed in the center compartment (indicated with the black X) for the 30 min acclimation period, before being allowed to move throughout the experimental apparatus for the trial period...... 31

Figure 1-2 Family differences in upstream/downstream preference, dispersal activity, aggression, movement, and exploration across four trials. Families are named by their cross type and a unique number; families within the same cross-type share the same line format. a) average proportion (+/- SE) of chambers that were visited that were upstream of the starting position within the dispersal apparatus of the dispersal experiment b) average proportion (+/- SE) of chambers that were visited that were downstream of the starting position within the dispersal apparatus of the dispersal experiment c) average number of chambers visited (+/- SE) within the dispersal apparatus of the dispersal experiment d) average number of aggressive displays (+/- SE) in mirror stimulation experiment e) average latency (in minutes +/- SE) to exit the isolation box in the exploration experiment...... 32 ix

Figure 1-3 Repeatabilities (+/- SE) of behavioral traits measured within the same individuals (n=108) across four trials. These values represent the proportion of total variance accounted for by random individual intercepts while accounting for fixed effects. The asterisk indicates the behaviors that had a repeatability significantly different from zero...... 33

Figure ...... Page Figure 2-1 Histograms and box plots of behavioral phenotypes in all individuals tested (n=108), with individuals chosen for gene expression work highlighted in gray (n=36): a) number of aggressive displays, b) latency to exit isolation box, c) moving during mirror stimulation test, d) total chambers visited during dispersal experiment, e) cross-type slopes for upstream preference, f) aggression slopes, g) exploration slopes, h) total chambers slopes, i) family intercepts for total chambers visited in dispersal experiment, j) cross-type slopes for downstream preference. The lines on the box plot correspond to the median, first and third quantiles in the distribution. Whiskers drawn to the furthest point within 1.5 x interquartile range (IQR) from the box (or the upper or lower data point values if that is shorter), with potential outliers represented as disconnected points. The confidence diamond contains the mean and the upper and lower 95% of the mean. Red brackets define the densest region for the distribution...... 79

Figure 2-2 Number of significant differentially expressed genes, including the number up- and down-regulated, with FDR adjusted p-value <0.05 for all contrasts. All contrasts were high versus low, unless indicated in the label, meaning the blue represents genes more highly expressed in individuals with high phenotypic values for that trait and red represents genes more highly expressed in individuals with low phenotypic values for that trait...... 80

Figure 3-1 Capture locations within Nipigon Bay, Lake Superior of brook trout used in GWAS. The Unnamed creek is also known locally as Mazukma Creek. .. 103

Figure 3-2 First two principal components created by GAPIT color coded by capture location of individuals. Only the first PC was found to be significant and included in the GWAS to account for population structure...... 104 x

Figure 3-3 Manhattan plot of the markers across the genome and their significance when analyzed for an association with the phenotype of coaster or resident based on a discriminant function analysis of stable isotope data in Lake Superior brook trout. Line represents threshold for significance of p=0.001...... 105

Figure 3-4 Manhattan plot of the markers across the genome and their significance when analyzed for an association with stable isotope values of Carbon, representing trophic information, in Lake Superior brook trout. Line represents threshold for significance of p=0.001...... 106

Figure 3-5 FastStructure results for K=2 through K=5 for the 124 fish used in the GWAS grouped by their capture location (from left to right) including Lake Superior and then the streams ordered from West to East...... 107

ABSTRACT

Chin-Baarstad, Ashley Ph.D., Purdue University, December 2014. Associated Behavioral, Genetic, and Gene Expression Variation with Alternative Life History Tactics in Salmonid Fishes. Major Professor: Krista Nichols.

Individual differences in behavior can have potential fitness consequences and often reflect underlying genetic variation. My research focuses on three objectives related to individual level variation: 1) evaluating the innate behavioral variation within and between individuals, families, and progeny of different life-history types across time; 2) testing for differences in gene expression within the brain associated with this behavioral variation; and 3) using genetic polymorphisms to test for associations with ecotype, as well as population structure, in polymorphic populations. First, we evaluated the variation in a suite of ecologically relevant behaviors across time in juvenile progeny produced from crosses within and between migratory and resident rainbow trout (Oncorhynchus mykiss). By testing multiple behaviors repeatedly in the same individuals, we better understand the innate behavioral variation in a population containing multiple life history xi types. Our study shows that there were consistent differences between individuals across time, or personality, for dispersal, aggression, and exploration. The significant repeatabilities of these behaviors indicate that these traits may be heritable. Not only did we find evidence for habituation in all behaviors, but dispersal, aggression and exploration showed significant differences between individuals in the rate of that

xii habituation. The identification of this individual level variation is a step towards understanding which potentially heritable traits selection could influence.

Genetic variation for complex phenotypes, such as the behavioral traits identified, can arise as a function of variation in protein coding regions, or as a function of mutation in regulatory regions that modify the expression of genes. By examining the transcriptome within the brains of the fish used in the behavioral trials, we were able to identify genes differentially expressed among individuals with naturally variable behaviors. Moreover, we examined whether there is any overlap in the differentially expressed genes associated with individual differences in behavior and how these gene expression differences fit into biological functions and pathways. The results demonstrated that there are some key genes and pathways associated with the observed behavioral traits, and in many cases associated with multiple behaviors. Neuronal signaling (neurotransmitters such as gamma-aminobutyric acid (GABA) and catecholamines) was involved in multiple behavioral measures, while neuronal function and development were part of habituation in behavioral responses. Hormones (such as xii

androgens, glucocorticoids, and growth hormone) and their related pathways are shown to have expression patterns correlating with behavioral differences in these juvenile fish.

The overlap in genes that are differentially expressed between behaviors suggests pleiotropic effects of these genes on behavior with respect to personality traits and habituation. The differences in pathways, biological functions, and specific annotated genes underlying observed behavioral variation provide a starting place to finding possible proximate molecular and physiological mechanisms forming the connection between the heritable genome and the phenotypic differences observed.

xiii

Understanding patterns of extant genetic diversity is crucial for understanding not only the evolutionary history of populations, but also how populations should be managed and conserved. On a genomic level, we tested the hypothesis that ecotypic differences within and across polymorphic populations of brook trout (Salvelinus fontinalis) in Nipigon Bay, Ontario, Canada, were heritable and linked to allelic variation in the genome. We also evaluated population structure using the same genetic markers. In this study, we found that life history variation in Lake Superior brook trout appears to be heritable, but genome wide association analysis revealed only a few single nucleotide polymorphism (SNP) loci associated with ecotype. Moreover, the population genetic signature from 900 SNP markers distributed across the genome shows existing population structure across the tributaries of Nipigon Bay that should be considered when making stocking decisions. In particular, the Cypress River seems to be the main source of the sampled coaster ecotype captured in Lake Superior. Our study confirms that there is little to no genetic differentiation between life history types on a genome-wide level.

By and large, the preponderant genomic signature suggests that coaster brook trout are an xiii

ecotype derived from resident populations, rather than a distinct stock. Understanding the genetic variation for ecotypic divergence and existing population structure has important implications for the conservation and restoration of the declining coaster ecotype.

In summary, this work can have important consequences for understanding the diversification of alternative life histories, and will contribute to our understanding on the units for conservation and management in these fish species. In juvenile rainbow trout, innate behavioral differences are heritable and also reflected in differential gene expression at the molecular level in the brain. While the fitness consequences of these

xiv behaviors at this life stage, which experiences high mortality in nature, remain to be verified, the identification of the molecular mechanisms related to these behavioral differences still has importance. My research shows that there is overall no significant differentiation in behaviors between progeny produced by migratory or resident parents, but that there is considerable individual variation upon which selection could act, and maintenance of this variation would be important for any conservation and management programs. For example, care should be taken to avoid selection against aggression in hatchery breeding programs, which may be correlated with other behaviors (or even a life history variant) due to a shared molecular basis. For both species used in this study – rainbow trout and brook trout – in many areas the migrant life history type is declining

(particularly the coaster brook trout in Lake Superior). Importantly, it is necessary to parse out the genetic and environmental contributions to life history decisions to understand how extant populations can be used to conserve or restore populations. In the rainbow trout study, we find little behavioral differentiation between progeny produced by migrant or resident adults; in the brook trout, we find no genetic differentiation xiv

between life history types when considering loci distributed across the genome. While the rainbow trout are too young for one to know their eventual life history trajectory or to detect behavior or gene expression differences related to migration, our findings suggest that there are no differences between migrants and residents at this stage that might contribute to differential survival at the fry stage. For the brook trout it is likely that the environmental contribution to the coaster ecotype is the key to conservation efforts, and should be examined more directly. The identification of population structure for the brook trout also informs conservation efforts, emphasizing the need for local sources

xv when stocking, and further examination of how some streams, and the biotic and abiotic variables that influence those streams, in northern Lake Superior may differentially contribute to production of the migratory ecotype. Overall, the findings presented herein provide insight into the molecular mechanisms promoting behavioral diversity among individuals, as well as future directions for research aimed at conserving life history variation among salmonids. xv

CHAPTER 1. INDIVIDUAL BEHAVIORAL VARIATION IN JUVENILE RAINBOW TROUT, ONCORHYNCHUS MYKISS

1.1 Abstract

Behavioral variation at the individual level, especially the study of animal personalities, has recently been garnering much attention, including its role in influencing fitness. For larval fish, the behavioral decisions during the developmental shift from yolk- dependence to exogenous feeding are critical to survival, as the highest mortality is experienced during this transition. In this study, we evaluated the variation in a suite of risk-taking behaviors across time in juvenile progeny produced from crosses between migratory and resident rainbow trout (Oncorhynchus mykiss). By testing multiple ecologically relevant behaviors repeatedly in the same individuals, we were able to simultaneously test multiple hypotheses regarding personality, plasticity, and behavioral syndromes to better understand the innate behavioral variation in a population containing multiple life history types. The objectives of this study were to examine variation within and between individuals, families, and progeny of different life-history types for dispersal, exploration, and aggression across multiple trials. We also examined the correlations between behaviors within individuals and if variation in size and sex contributed to variation in these behaviors. Our study shows that there were consistent behavioral differences, or personality, between individuals across time, for dispersal, aggression, and exploration. The significant repeatabilities of these behaviors indicate that these traits 2 may be heritable. Not only did we find evidence for habituation in all behaviors, dispersal, aggression and exploration showed significant differences between individuals in the rate of that learning. The identification of this individual level variation is a step towards understanding which potentially heritable traits selection could influence.

1.2 Introduction

Individual differences in behavior can influence both resource use and an individual’s ability to cope with or react to changing environmental conditions

(Dingemanse et al. 2004; Dingemanse and Reale 2005; Reale et al. 2007). Behavioral variation can be observed in multiple contexts, such as exploring novel environments and objects, predator interactions, and conspecific encounters (Gosling 2001; Sih et al. 2004a).

If such behavioral differences among individuals result from underlying genetic variation

(Dingemanse et al. 2003; Fitzpatrick et al. 2005) and influence fitness (Reale and Festa-

Bianchet 2003; Dingemanse and Reale 2005; Reale et al. 2009), then selection can lead to population- and species-level differences over evolutionary time scales(Fitzpatrick et al.

2007).

Consistent differences between individuals across time or contexts, is defined as an animal’s personality (Sih et al. 2004a). Within-individual variance may also be due to phenotypic plasticity, however, and the same behavior can be both plastic and exhibit consistent individual differences (Westneat et al. 2011; Ensminger and Westneat 2012).

Consistent differences in behavior could be the raw material available for natural selection. Environmental changes can lead to variation in selective pressures, possibly maintaining genetic variation in personalities (Dingemanse et al. 2004). Theoretically, life-history trade-offs could promote the evolution of personalities and correlations 3 among behaviors (Wolf et al. 2007). Suites of correlated behaviors have been termed behavioral syndromes (Bell and Stamps 2004; Sih et al. 2004a; Sih et al. 2004b).

Behavioral syndromes might constrain plasticity as well as evolution. For example, selection on one trait may cause a response in another genetically correlated trait and explain why a behavior may not be optimal in certain contexts (Bell 2007). Thus, organismal personalities could be an important component of life-history strategies

(Reale et al. 2009).

Salmonids have extremely variable life histories including whether and when fish undertake long distance migrations (Rounsefell 1958; Quinn and Myers 2004). Fish that migrate to sea (anadromous) undergo a transition from a more or less solitary, substrate- bound river phase to a schooling, pelagic life-style in the ocean (Magnhagen 2008).

These migrants undergo a suite of morphological, behavioral, and physiological changes prior to migration (smoltification), which are regulated by both environmental factors

(e.g. day length, water temperature) and endogenous factors (e.g. body size) (Folmar and

Dickhoff 1980). Although the timing and propensity to migrate has a heritable 3 component (Clarke et al. 1994; Nichols et al. 2008), migrant and resident ecotypes can coexist in the same population and interbreed (Quinn and Myers 2004; Charles et al.

2005). Migration represents a trade-off (Thrower et al. 2004): residents become sexually mature earlier and potentially reproduce for a number of years; migrants delay sexual maturation, attain a larger size, have greater fecundity, and reproduce once. Migration can be thought of as a syndrome involving behaviors and other traits that function together.

Taking a more ‘ecological approach’ to studying behavioral syndromes enables us to ask whether correlated behaviors might explain individual variation in fitness-related traits (Bell 2007). In fish, differences in stress-coping styles (behavioral and neuroendocrine responses to adverse stimuli) often influence survival (Magnhagen 2008).

These behavioral and physiological differences may be stable and heritable (Pottinger and Carrick 2001; Overli et al. 2007; Magnhagen 2008). For example, individuals that resume feeding first after an environmental change attack intruders quicker, more often, and are more likely to win social dominance fights (Pottinger and Carrick 2001; Overli et al. 2007; Magnhagen 2008). In terms of taking risks, certain behavioral correlations appear to be context-dependent, such as boldness in some foraging contexts but not in exploratory contexts (Wilson and Stevens 2005).

The change in larval salmon and trout from being yolk-dependent, buried in gravel nests to swimming up and ingesting food from the environment is a critical period for survival (Barton and Bond 2007), as the highest mortality is experienced during this transition (Elliott 1993). In many species of salmonids, the juveniles establish feeding 4 territories in the stream that they defend aggressively (Grant et al. 1989; Vollestad and

Quinn 2003), to obtain the food needed for optimum growth and survival (Keenleyside and Yamamoto 1962). More dominant individuals, which are generally the more aggressive individuals, obtain the more energetically profitable stream positions

(Metcalfe 1986; Keeley 2001), and this leads to faster growth. Territory size usually increases with fish size (Grant et al. 1989; Keeley 2001). Positive feedback may exist, given the effect of body size on dominance and access to resources for faster growth.

Growth in streams is important because size affects both survival and the ability to

5 migrate in anadromous fish (Hoar 1976). Aggressive behavior and growth has been shown to have both negative correlations, that may be driven by energetic costs of aggression, (Vollestad and Quinn 2003) and positive correlations, due to the competitive ability of dominant individuals and a hormonal connection (Lahti et al. 2001). As can be expected, there is a decrease in growth and an increase in mortality with increasing resource competition in streams (Keeley 2001). Variation in dispersal distance, such as in chinook salmon (Oncorhynchus tshawytscha) fry (Bradford and Taylor 1997), and direction, as seen in several species (Raleigh 1971), within streams may also affect the amount of competition a newly emerged fry experiences.

The behavioral traits examined in our study were chosen for their ecological relevance during the transition from yolk-dependence, to the onset of exogenous feeding.

As fish exit nests, they disperse throughout the stream, begin to explore new environments, and aggressively establish territories. Testing for consistent differences between individuals in their activity, dispersal, and readiness to take risks in these ways will provide a clearer picture of the personalities that have potential consequences for 5 growth and survival. Looking at the correlations amongst behaviors gives us insight into potential trade-offs both ecologically (e.g. immediate response to an environmental factor) and evolutionarily (e.g. selection on genetically correlated traits). In addition, looking for differences among life-history types might provide insight into the ‘migration syndrome.’

In this study, we evaluated the variation in a suite of risk-taking behaviors across time in juvenile progeny produced from crosses within and between migratory and resident parents. The objectives of this study were to examine variation within and between individuals, families, and progeny of different life-history types for dispersal,

6 exploration, and aggression across multiple trials. We also examined the correlations between behaviors within individuals and if variation in size and sex contributed to variation in these behaviors. We hypothesized that there would be personality for dispersive, exploratory, and aggressive behavior, and thus we predicted that there would be significant between individual variation or repeatabilities for these behaviors. We hypothesized that the crosses made from all migrant parents would be more dispersive, exploratory, and aggressive, and thus we predicted that pure migrant cross-types would display these behaviors more than pure resident crosses, with the hybrids falling in between. We also predicted those individuals to be larger, faster growing, and mostly female, as that could be what we would expect from future migrants. By testing multiple ecologically relevant behaviors repeatedly in the same individuals, we better understand the innate behavioral variation in a population containing multiple life history types.

1.3 Materials and Methods

1.3.1 Field sampling and genetic crosses

Fish used for this study were generated from crosses using anadromous and 6 resident adult rainbow trout (Oncorhynchus mykiss) sampled from natural populations in southeast Alaska. Wild fish were collected from populations in Sashin Creek and Sashin

Lake, Alaska. Anadromous individuals were collected in Sashin Creek when returning from the ocean to their natal stream to spawn. Residents were collected from Sashin Lake, also at sexual maturity. The resident fish are separated from the anadromous fish in the creek by two barrier waterfalls, with all the anadromous individuals collected downstream of these waterfalls and all residents collected upstream. Extensive genetic

7 research has been conducted previously on this study population (Thrower et al. 2004;

Aykanat et al. 2011; Hecht et al. 2012; Hale et al. 2013).

Gametes were collected from twelve individuals (six of each sex) for each life history type (anadromous or resident) captured in the wild in Sashin Creek and Sashin

Lake, similar to methods described by McKinney et al. (in prep). At collection, fish were anesthetized with tricaine methanesulphonate (MS-222; Argent Chemicals, Redmond,

WA) and gametes were expressed by applying light pressure on the abdominal cavity.

Female gametes were collected in individually labeled zip-top bags and the male gametes were collected in individually labeled plastic bags. Adults were released after collection of gametes. Gametes were stored at 5-12°C for no more than 24 hours until the crosses were made. Twelve full-sibling crosses were made in May 2011 at the Little Port Walter

Marine Station (Baranof Island, Alaska), with each family being one of four possible cross types: anadromous (AxA), resident (RxR), and hybrid and reciprocal hybrids among life history types (AxR and RxA).

Embryos were reared at ambient creek temperatures (inflow received from Sashin 7

Creek) in flow-through stack incubators in the dark. Each family was reared in an individual container within the stack to maintain family identity. Once the fish reached the fry stage (utilization of all yolk resources), nine individuals from each family (total n=108) were selected and reared in randomized individual compartments within the stack incubators. The fish were fed commercial trout feed twice daily. The remaining fish from each family were reared in outdoor vertical raceways, for another study.

1.3.2 Behavioral experiments

The behavioral experiments included laboratory measures of dispersal, exploration, and aggression, as these were expected to be the most ecologically relevant and possibly related to migration or stream-residency. All behavioral experiments were done at the fry stage (within two months of yolk absorption) between August and

September 2011, at the Little Port Walter Marine Research Station. Three experiments, described below, were conducted on each fish in succession to quantify individual variation and repeatability for dispersal, exploration, and aggression. The three experiments were repeated every eight days, for a total of four trials for each individual in an effort to estimate individual repeatability. The order of experiments for an individual was randomized to minimize effects of experiment order. The order individuals were tested was randomized and the individuals were tested blindly (assigned an identifier not related to cross-type or family) to prevent particular families or cross-types from being grouped together temporally. For observation and later scoring, all behavioral experiments were video recorded. A single observer, to prevent any observer bias or 8 differences in reaction time, scored all behaviors from the video files. The individuals were scored blindly to prevent any observer bias based on individual, family or cross- type identity. Once behavioral experiments were completed, fish were anesthetized with

MS-222, and measured for weight and fork length. From the initial and final measurements, growth rate was calculated as [ln(L2) – ln(L1)]/[t2 – t1] X 100, where L1 and L2 are lengths or weights at times 1 and 2 (t1 and t2 in days). Caudal fin clips were collected and stored in individually labeled vials with 95% ethanol for later genetic sex- typing.

1.3.2.1 Dispersal Experiment

Propensity to disperse was measured using a dispersal apparatus (see Figure 1-1) similar to that employed in previous studies (Bradford and Taylor 1997). The dispersal apparatus consisted of a 3.6 m length of 40.6-cm diameter opaque PVC pipe that had been cut in half along its length and subdivided into 15 equal sized compartments using 3 mm opaque plastic dividers. Each divider had a 5.5 cm hole offset from the middle and positioned to alternate from one side to the next along the sequence of dividers so that straight-line movement through multiple compartments was not possible. The apparatus had flow through of creek water at ambient temperature, with water entering at one end of each unit and exiting at the other. A single individual was tested in a dispersal apparatus at any given time to avoid any social effects on movement. Four dispersal apparatuses were placed side by side, to enable the testing of multiple individuals simultaneously. A test individual was placed in the center compartment and prevented from exiting by blocking the divider openings to that compartment. After a 30 minute acclimation period, compartment dividers were opened and the individual was allowed to 9 exit and move throughout the experimental apparatus. All trials were run for one hour, with the position and movements of the test fish recorded using multiplexed video cameras mounted overhead. Three measures of movement were scored: (i) upstream preference (proportion of total chambers visited that were upstream chambers), (ii) downstream preference (proportion of total chambers visited that were downstream chambers), and (iii) activity (total number of compartments visited, including re-visits).

To check that the lack of a directional preference was not due to the dispersal apparatus being a closed system, we also explored initial direction and first end reached as possible

10 metrics, but neither of these showed significant effects.

1.3.2.2 Exploration Experiment

Exploration was tested using an exit test employed previously to relate risk taking to the activity exhibited by recently emerged brook trout while searching for prey

(Farwell and McLaughlin 2009). This test measures the time required for an individual to exit an isolation box into an open environment. Exit time reflects the fish’s perception of the relative risks associated with being inside and outside of the box. The test apparatus consisted of a 190 L aquarium with an opaque, free-standing isolation box (16 cm x 15.5 cm x 15.5 cm) positioned at one of the narrow ends of the aquarium. The box had a sliding door that could be opened manually by pulling upward to create a 4 cm opening.

The box opening was oriented toward the opposite (far) end of the aquarium. Each aquarium had opaque dividers on three sides, so that fish in adjacent tanks could not see each other, and a single transparent (long) side, so that the fish could be observed and video recorded. There were ten test aquaria, each with a camera. At the beginning of each trial, the test fish was placed into the box, the box closed, and the fish given five minutes 10 to adjust to its new surroundings. The box door was then opened and the experiment was run and video recorded for 30 minutes. The video was watched later and the time it took for the test fish to exit entirely from the box into the aquarium was recorded. If a fish failed to exit from the tube after 30 minutes, then 30 minutes was recorded.

1.3.2.3 Aggression Experiment

Aggression was assessed using a mirror image stimulation experiment similar to those used in previous studies (Riddell and Swain 1991; Vollestad and Quinn 2003).

Mirror image stimulation tests rely on the propensity for fish to respond to their mirror image as if it were a conspecific. Mirror stimulation tests have been criticized because they probably do not elicit the full diversity of behavioral responses and do not represent normal interactions (Vollestad and Quinn 2003). However, they offer greater control for complicating factors, such as differences in body size, condition, and previous history of interactions, than do intruder tests (Rhodes and Quinn 1998). In addition, earlier successful applications suggest they should be adequate to provide a standard index of innate aggressiveness for an individual that can be compared among families derived from different parental phenotypes (Vollestad and Quinn 2003). The aggression experiment was conducted in the same aquarium as the exploration experiment described above, with opaque dividers on three sides, so that fish in adjacent tanks could not see each other, and a single transparent (long) side, so that the fish could be observed and video recorded. A mirror was placed at one of the long ends of each aquarium and positioned so that the fish could see the mirror from any location within the tank. Each individual was video recorded for later observation and their behavioral responses to the 11

mirror were quantified for five minutes using the event recorder JWatcher (Blumstein and

Daniel 2007). The responses quantified were time spent moving around tank and the number of aggressive displays. The stereotypic aggressive display that was exhibited was a swim against mirror (SAM), which is characterized as an individual swimming with its snout against the mirror, sometimes accompanied by vigorous head shaking and jaw snapping (Keenleyside and Yamamoto 1962; Riddell and Swain 1991; Vollestad and

Quinn 2003).

1.3.3 Genetic sex-typing

DNA was extracted from the fin clips that were collected, using a phenol:choloroform procedure modified from Wasko et al. (2003). All individuals were genotyped at a genetic sex marker, OmyY1, which was used as a proxy for phenotypic sex in analyses (Brunelli et al. 2008). This sex marker within the Sashin Creek population was previously verified as having 97% accuracy (Hecht et al. 2012).

1.3.4 Statistical analyses

The complete data set included the quantified behaviors from the three experiments across all four trials and sizes and sex for each of the individuals from each family within each cross-type. All data were tested for assumptions of normality and homogeneity or error variance to choose the most appropriate statistical tests. Pearson correlation coefficients were used to assess the relationship between length, weight, and growth measurements. Only one measure of size (final weight) was used in the models because all of the measures were highly correlated. Growth, calculated as previously described, was tested separately from size as a fixed effect in the same models. The 12

models that were used are described below in more detail.

For all models, the fixed effects tested included final weight, sex, cross-type, trial order, and a cross-by-trial interaction with a type I error rate of 0.05. The random effects included family nested within cross, individual nested within family, and interactions between those terms and trial. The significance of random effects was tested with a likelihood ratio test (LRT), which compares the values of twice the log likelihood (-2LL) of nested mixed effects models and is distributed as a chi-square with degrees of freedom equal to the difference in the number of terms between the models. Observations were

13 repeated over trials within individuals. Linear mixed models were used when analyzing the total number of chambers visited during the dispersal experiment, the proportion of upstream/downstream chambers visited during the dispersal experiment, and amount of time an individual spent moving around the tank during the aggression experiment in

JMP (version 10.0.0; SAS Institute Inc., Cary, NC). Generalized linear mixed models with a Poisson distribution were used to analyze the amount of time to exit the isolation box in the exploration experiment and the number of aggressive displays in the aggression experiment using Proc Glimmix in SAS (version 9.3; SAS Institute Inc., Cary,

NC). By including trial order as a fixed effect, we were able to test for habituation

(changes in the behaviors across trials)..

Any term that was not significant was removed before running models to extract variance components to calculate repeatability. Adjusted repeatability was calculated, which is the proportion of total variance accounted for by random individual intercepts while accounting for fixed effects (Nakagawa and Schielzeth 2010). This represents the amount of observed phenotypic variance predicted by individual identity. As some of the 13

behavioral traits were not normally distributed, correlations between behaviors were assessed by Spearman’s rank nonparametric correlations. The correlations were tested both on the average behavioral values within an individual and for each of the four trials separately.

1.4 Results

1.4.1 Fish Size and Sex

Length, weight, and growth were all significantly correlated. Table A-1 summarizes these size measurements and Table A-2 contains the correlation matrix.

There were also positive correlations between the initial and final measurements of both weight and length, meaning fish that were larger in the beginning were also larger at the end. Growth rate was positively correlated with final weight and length. The growth rate of both length and weight were negatively correlated with both initial weight and initial length, suggesting that fish that were larger in the beginning grew more slowly. There were no significant differences between any of the size measurements of the different cross-types. Only one measure of size (final weight) was used in the analyses of the behaviors because all of the measures were highly correlated. Of the 103 fish that survived to the end of all behavioral trials and were genetically sex-typed, 63 (61%) were female and 40 (39%) were male.

Behavioral traits

Five behavioral traits from three experiments (Table 1-1) were quantified for 103 juvenile individuals, with partial data for five individuals that died during the study. The individuals were distributed evenly among four cross-types (Table 1-2), with three families within each cross type (Table 1-3). Descriptive statistics of behavioral 14

observations are summarized in of Table 1-4. Behaviors varied across the four trials

(Table 1-5). Significant effects, variance components, and calculated adjusted repeatabilities for each of the behavioral traits are reported in Table 1-6. Repeatabilities for each trait are also depicted in Figure 1-3. The results from each of the three experiments are discussed below in more detail.

1.4.2 Dispersal

On average, trial had a significant effect in the dispersal experiment (F1,106=55.00, p<0.0001; Table 1-5), with the total number of chambers visited decreasing across the

15 four trials, indicating habituation. Despite this, individuals differed significantly

(χ2=47.70, df=1, p<0.0001; repeatability = 0.40±0.08; Table 1-6), as did families

(χ2=13.21, df=1, p=0.0002) (Figure 1-2c). There were no family-by-trial effects, but there were individual-by-trial interactions (χ2=11.79, df=1, p=0.0005). Individuals also varied across trials in their upstream preference (F1,318.7=14.82, p=0.0001) and downstream preference (F1,318.2=10.60, p=0.0013), again indicating habituation (Table 1-5). However, there was no evidence for personalities (between-individual differences), family effects, or individual-by-trial effects in these preferences (Table 1-6). There were significant cross-type by trial effects for both upstream (F3,318.6=3.35, p=0.019) and downstream

(F3,318.1=3.14, p=0.025) preference. Overall, there was no effect of size, growth, or sex in the dispersal experiment (all p>0.05; Table 1-6).

1.4.3 Exploration

On average, there was habituation in the latency to exit the isolation box in the exploration experiment (F1,107=77.48, p<0.0001; Table 1-5). It took individuals longer to exit the isolation box as the trials progressed (Figure 1-2e). Despite this trend, there was 15 significant between-individual variation (χ2=1253.5, df=1, p<0.0001; repeatability =

0.46±0.10; Table 1-6). Individuals differed in habituation as well (individual-by-trial interaction: χ2=246.15, df=1, p<0.0001; Table 1-6). However, there were no size, growth, sex, cross-type or family effects (all p>0.05; Table 1-6).

1.4.4 Aggression

Habituation was also evident in the aggression experiment. The number of aggressive displays decreased significantly with trial number (F1,107=123.08, p<0.0001;

Table 1-5; Figure 1-2d). Fish also exhibited less time moving in later trials

(F1,318.1=9.4558, p=0.0023; Table 1-5; Figure 1-2e). There was still significant between individual variation in the number of aggressive displays (χ2=890.82, df=1, p<0.0001; repeatability = 0.22±0.09; Table 1-6) and time spent moving around the tank (χ2=5.97, df=1, p=0.0145; repeatability = 0.10±0.05; Table 1-6).. There were also differences between individuals in the degree of habituation (individual-by-trial interaction:

χ2=386.81, df=1, p<0.0001; Table 1-6). There were no size, growth, sex, cross-type, or family effects on traits quantified in the aggression experiment (all p>0.05; Table 1-6).

1.4.5 Correlations

There were several significant correlations between the behaviors measured

(Table 1-7). On average and across all trials, upstream and downstream preferences were significantly negatively correlated. Considering that these are mutually exclusive and were tested at the same time, this is to be expected. Also significantly negatively correlated on average and across all trials were exploration and dispersal activity; individuals that took longer to exit the isolation box (less exploratory) visited fewer chambers in the dispersal experiment (less dispersive). The other significant correlations 16 were not consistently found between the same behaviors across trials and on average, likely because there was an effect of trial on each behavioral trait.

1.5 Discussion

By testing behavioral variation within and across contexts, we aimed to understand how juvenile rainbow trout interact with and respond behaviorally to their environment early in their life, and to examine whether these early behaviors differed between progeny produced from migratory or resident parents. Our study shows that there were consistent differences between individuals across time, or personality, in

17 dispersal, aggression, and exploration. However, personality was not associated with cross-type. The significant repeatabilities of these behaviors indicate that these traits may be heritable. This is especially true for dispersal activity, which also varied between families. Moreover, we demonstrated differences between individual rates of habituation in dispersive activity, aggressive and exploratory behaviors. Although there was not enough evidence to support the existence of a behavioral syndrome, the correlation between exploration and dispersal validates our test for personality in this trait.

The consistent differences between individuals across trials support the hypothesis that personality differences exist for dispersal, aggression, and exploration even at this very early time point after the fish begin exogenous feeding. The repeatabilities provide an upward-bound estimate of heritabilities. The significant repeatability of the traits is also important, because it supports the reliability of the tests’ performance, which should be established before interpreting results for personality or behavioral syndromes (Carter et al. 2013). At this developmental stage in the natural environment, the fish would be swimming up out of gravel nests and dispersing in the 17

streams, and individual variation in these associated behaviors can have important implications for survival and growth in the wild. Since these differences were seen at such an early life stage and in individually housed animals, they are most likely innate differences.

Extreme variability in environmental selective pressures could lead to the high amount of variation in behavioral traits that we observed (Fuiman and Cowan 2003). In addition to differences between individuals, there were significant differences between families in the amount of movement (total number of chambers visited; Figure 1-2)

18 during the dispersal experiment, adding further evidence to suggest a heritable component to the amount of stream movement.. These data suggest that individuals from certain families would more actively disperse, moving more and/or further, in the streams after exiting the nests as fry. Some individuals were more exploratory than others, with some individuals taking less time to exit the isolation box than others. It was previously suggested that behaviors, such as hiding, are heritable in rainbow trout (Oncorhynchus mykiss), since they differ between clonal domesticated fish lines (Lucas et al. 2004).

Habituation is a type of learning that has also been termed behavioral flexibility, in which there is a reduction in behavioral response with repeated exposure to a stimulus

(Brown et al. 2006; Rankin et al. 2009; Adriaenssens and Johnsson 2010). In our study, the fish displayed habituation in all behaviors across trials, despite separating trials by eight days. It is impossible to know if the differences observed across trials were due exclusively to the repeated exposure to the experiments or due to increased age during the experiment, as these are confounded. To eliminate age as a possibility, a similar study would have to be done, but with the start date of the trials staggered in such a way that 18

some individuals are experiencing their first trial at the age that others are experiencing later trials. However, since over our entire study, individuals only aged a total of 24 days, and changes in behavior were observed in all measured traits of all experiments, it is reasonable to conclude that the observed trial effects were caused by habituation rather than age. In our dispersal experiment, a decrease in the total chambers visited over repeated trials was observed (Figure 1-2e). This trend is consistent with previous work in

Chinook salmon (Oncorhynchus tshawytscha) fry showing a sensitive period after emergence during which stream dispersal behaviors occur before declining (Bradford and

Taylor 1997). By testing not only for the existence of habituation but also for between- individual variation, we can see if there are differences in habituation rate. There were significant differences between individuals in the rate of habituation in all three of the experiments, including dispersal activity, aggression, and exploration. The significant results suggest that some individuals are more labile in these traits over time

Progeny from anadromous and resident parents were examined for their behaviors as post-emergent juveniles, but we cannot know their eventual life history trajectory a full two years before life history decisions are generally made in this population. There is a high heritability for life history (becoming a migratory smolt or resident) in this population (h2=0.71) (Thrower et al. 2004), but both resident and anadromous cross types are capable of producing either of the life histories. In our study, there are no consistent differences between cross-types to suggest that the behaviors measured at this early developmental stage are correlated with later life history decision.

Since multiple behaviors were tested for the same individuals, we tested for the presence of a behavioral syndrome by looking for correlations across behaviors within 19

individuals. We did not find significant and consistent correlations to suggest a behavioral syndrome. There was a consistently significant correlation between exploration and dispersal activity, which could be explained by both being measures of the same trait, such as exploration or general activity. It is also consistent with findings in brook trout (Salvelinus fontinalis), where more sedentary individuals are less exploratory

(Farwell and McLaughlin 2009). Fish that visited more chambers in the dispersal apparatus, also exited an isolation box more quickly into an open environment. This is also known as convergent validity, and has been suggested as a way of assessing the

20 validity of a test for a personality trait (Carter et al. 2013). Conversely, the lack of correlation between the amount of time spent moving during the aggression trial and either latency to exit the isolation box or total chambers visited, confirms that the exploration and dispersal experiments were not simply measuring activity levels

(discriminant validity). Although the correlations found in our study do not provide enough evidence to state that there is a behavioral syndrome, they can be used to validate tests used for personality traits. Alternatively, the lack of correlation between life history and behaviors could be an indication that there is not an integrative syndrome, but rather that associations between life-history and behavioral tendencies are environmentally determined, as has previously been suggested for the pace-of-life syndrome (Niemela et al. 2013). Our study was done in a single environment and at a single life stage. It will require further study to determine how environment affects the correlation between personality and life history, and therefore determine the existence of an integrative migration syndrome that includes both personality and life-history traits.

By evaluating the variation in a suite of risk-taking behaviors across time in 20

juvenile fish, we found personality differences between individuals in dispersal, aggression, and exploration in post emergent O. mykiss produced from resident and migratory parents. While all behaviors showed evidence of habituation, more interestingly, dispersal, aggression and exploration showed significant differences between individuals in the rate of that habituation. There was not enough evidence to suggest that any of the behaviors are correlated with later life history decision to migrate or maintain residency, nor evidence to suggest that there is a behavioral syndrome. This

21 will inform our understanding of juvenile trout behavioral strategies in response to the environment.

1.6 Acknowledgements

I would like to acknowledge the input and collaboration of my coauthors Frank Thrower and Krista Nichols. This work would not have been possible without the support of regular and seasonal staff at the Little Port Walter Marine Research Station, including, but not limited to Tommy Abbas, Angela Feldmann, Charlie Waters, and Kenneth

Sprague. We are grateful to additional help in the field from Heather Holzhauer. Jessica

Monjaras and Anthony Folck facilitated lab work. Amanda Ensminger provided help through discussion of statistics. I also thank my committee members, Dr. Esteban

Fernandez-Juricic, Dr. Rick Howard, and Dr. Greg Hunt for providing insightful comments on an earlier version of this chapter. This work was supported by the National

Science Foundation (NSF-DEB-0845265 to K.M.N.) and a Graduate Research

Fellowship from the National Science Foundation (DGE1333468 to A.C.-B.).

Table 1-1 Description of each behavioral trait including an abbreviated name used for the trait throughout the text, the behavioral experiment used to measure that trait, the mean, standard deviation, and the number of observations for each trait Abbreviated Experiment Description of trait Mean SD n Trait Name

number of upstream chambers visited upstream Dispersal divided by the total number of 0.57 0.29 426 prop chambers visited

number of downstream chambers downstream Dispersal visited divided by the total number of 0.35 0.28 426 prop chambers visited

total Dispersal total number of chambers visited 135.62 105.41 426 chambers

aggression Aggression number of aggressive displays observed 2.70 7.04 426

amount of time spent moving around moving Aggression 74287.84 89655.55 426 the tank

exploration Exploration time to exit isolation box 6.48 9.03 426

Trait mean SD N mean SD n mean SD N mean SD n

upstream prop 0.63 0.27 107 0.54 0.28 108 0.54 0.30 103 0.57 0.31 108 downstream prop 0.30 0.25 107 0.38 0.27 108 0.38 0.29 103 0.35 0.30 108 total chambers 154.93 101.50 107 137.06 115.54 108 151.73 111.17 103 99.70 82.87 108 aggression 2.52 6.19 107 2.26 5.75 108 3.63 7.98 103 2.44 8.00 108 moving 87656 95528 107 66970 83603 108 78071 96706 103 64754 81510 108 exploration 5.70 9.26 107 6.64 9.33 108 5.47 8.71 103 5.23 8.26 107

Table 1-3 The means, standard deviations, and numbers of observations for each behavioral trait by family (families named by cross type: anadromous (AXA), resident (RXR), and hybrid and reciprocal hybrids among life history types (AXR and RXA) and then numbered) upstream downstrea total_ Family Trait _prop m_prop chambers aggression moving exploration mean 0.59 0.36 128.00 0.77 94168.40 4.96 SD 0.29 0.29 83.14 3.40 102015.91 8.64 AxA31 n 35 35 35 35 35 35 mean 0.70 0.22 204.61 2.25 92169.03 4.19 SD 0.19 0.15 114.59 3.86 94817.65 7.36 AxA34 n 36 36 36 36 36 36 mean 0.60 0.31 131.44 4.50 76811.58 7.93 SD 0.30 0.26 86.74 9.09 91394.60 11.18 AxA36 n 36 36 36 36 36 36 mean 0.54 0.39 94.42 0.75 72945.64 7.48 SD 0.35 0.33 93.08 1.86 89335.45 8.59 AxR37 n 36 36 36 36 36 36 mean 0.55 0.37 195.31 1.83 57858.08 5.72 SD 0.26 0.24 150.59 4.17 76562.33 10.09 AxR38 n 36 36 36 36 36 36 mean 0.53 0.38 121.44 4.19 70106.06 6.71 SD 0.23 0.23 62.13 8.59 85983.57 9.44 AxR39 n 36 36 36 36 36 36 mean 0.52 0.39 105.70 2.03 53811.70 4.84 SD 0.32 0.30 66.42 6.27 88777.67 6.86 RxA46 n 33 33 33 33 33 33 mean 0.50 0.42 118.82 2.68 100848.00 8.54

SD 0.35 0.34 128.04 6.67 106366.34 11.60 24

RxA47 n 34 34 34 34 34 34 mean 0.61 0.34 225.00 6.00 78796.28 3.14 SD 0.21 0.20 89.18 9.93 91225.78 6.03 RxA48 n 36 36 36 36 36 36 mean 0.49 0.43 92.75 0.78 66508.44 7.54 SD 0.35 0.32 75.53 2.80 88328.35 10.35 RxR40 n 36 36 36 36 36 36 mean 0.60 0.28 83.28 5.42 51541.03 6.34 SD 0.29 0.28 74.04 12.57 65957.95 8.85 RxR41 n 36 36 36 36 36 36 mean 0.61 0.33 123.08 1.11 76211.33 1.70 SD 0.30 0.28 94.43 4.01 88549.65 1.51 RxR42 n 36 36 36 36 36 35

Table 1-4 The means, standard deviations, and numbers of observations for each behavioral trait in all trials by sex Male Female

Trait mean SD n mean SD n

upstream_prop 0.58 0.29 160 0.57 0.29 252

downstream_prop 0.34 0.27 160 0.35 0.27 252

total_chambers 122.83 92.78 160 145.10 113.41 252

aggression 3.71 9.26 160 2.11 5.26 252

moving 73438.74 91577.59 160 76894.15 90206.13 252

exploration 7.10 10.14 160 4.82 7.73 251

Table 1-5 The means, standard deviations, and numbers of observations for each behavioral trait by trial, as well as F-statistics for the significance level of trial as a fixed effect on each behavioral trait upstream downstream total Trait _prop _prop chambers aggression moving exploration mean 0.64 0.29 176.76 5.79 96777.94 2.39 SD 0.26 0.23 103.97 9.53 93948.74 5.06 Trial 1 n 108 108 108 108 108 107 mean 0.60 0.34 154.49 3.65 69443.31 3.42 SD 0.27 0.26 106.73 8.33 79736.39 5.95 Trial 2 n 108 108 108 108 108 108 mean 0.53 0.38 115.76 0.64 71332.57 6.69 SD 0.30 0.29 98.05 2.65 95881.38 9.29 Trial 3 n 106 106 106 106 106 106 mean 0.51 0.39 93.56 0.62 58975.70 10.72 SD 0.33 0.31 92.94 3.35 84974.18 11.45 Trial 4 n 104 104 104 104 104 104 DF 1 1 1 1 1 1 Fixed 318.7 318.2 106 107 318.1 107 Effect DFDen of F Ratio 14.82 10.60 55.00 123.08 9.46 77.48 Trial Prob > F 0.0001 0.0013 <.0001 <.0001 0.0023 <.0001

Table 1-6 Significant (α < 0.05) model effects for each trait are indicated with the presence of an 'x' in the effect column for the row trait. upstream downstream total Trait proportion proportion chambers aggression Moving exploration n 412 412 412 412 412 411

weight sex Fixed Effects cross-type trial X x X x X x X x crossXtrial

Individual (Vi) X x X x

Family (Vf) X Random IndividualXTrial Effects X x x (IxE) FamilyXTrial

Individual (Vi) 0.00 0.00 3021.12 0.77 786462732 0.54

(SE) 0.00 0.00 612.40 0.33 387417380 0.12

Family (Vf) 1767.03

(SE) 953.14

IndividualXTrial 730.43 2.70 0.64 (IxE) (SE) 243.95 0.37 0.08

Vr 0.08 0.07 4537.15 7139200000

(SE) 0.01 0.01 440.61 569307294

repeatability 0.04 0.06 0.40 0.22 0.10 0.46

(SE) 0.04 0.05 0.08 0.09 0.05 0.10

p-value 0.25 0.12 <0.0001 <0.0001 0.01 <0.0001

Table 1-7 Correlation matrices for the average behavioral value across the four trials (a) and each of the four trials separately (b-e), with significant correlations grey highlighted. Spearman's correlations were conducted, with the correlation coefficient placed in upper cell, p-value for correlation in center, and sample size for comparison in lower cell. a) AVERAGE upstream downstream total_ aggression moving exploration _prop _prop chambers upstream 1 -0.91 0.25 0.15 0.20 -0.20 _prop <.0001 0.01 0.14 0.05 0.04 104 104 104 104 104 104 downstream 1 -0.11 -0.20 -0.19 0.13 _prop 0.27 0.04 0.05 0.20 104 104 104 104 104 total_ 1 0.13 -0.05 -0.38 chambers 0.19 0.64 <.0001 108 108 108 108 aggression 1 -0.06 -0.08 0.53 0.39

108 108 108 moving 1 0.02 0.86

108 108 exploration 1

108

b) TRIAL 1 upstream downstream total_ aggression moving exploration _prop _prop chambers upstream 1 -0.84 -0.05 0.23 0.19 -0.15 _prop <.0001 0.62 0.01 0.05 0.13 108 108 108 108 108 107 downstream 1 0.27 -0.18 -0.11 0.08 _prop 0.00 0.06 0.28 0.42 108 108 108 108 107 total_ 1 0.09 0.00 -0.23 chambers 0.38 0.98 0.02 108 108 108 107 aggression 1 -0.08 -0.12 0.43 0.22

108 108 107 moving 1 0.01 0.94

108 107 exploration 1

107 c) TRIAL 2 upstrea downstream total_ m _prop aggression moving exploration chambers _prop upstream 1 -0.98 0.01 -0.02 0.10 -0.14 _prop <.0001 0.94 0.88 0.29 0.16 108 108 108 108 108 108 downstream_ 1 0.05 0.02 -0.08 0.13 29 prop 0.63 0.86 0.39 0.18 108 108 108 108 108 total_ 1 0.15 0.02 -0.22 chambers 0.12 0.82 0.02 108 108 108 108 aggression 1 0.01 -0.17 0.94 0.08

108 108 108 moving 1 -0.01 0.96

108 108 exploration 1

108

d) TRIAL 3 upstream downstream total_ aggression moving exploration _prop _prop chambers upstream 1 -0.92 0.33 0.04 -0.07 -0.23 _prop <.0001 0.00 0.71 0.45 0.02 106 106 106 106 106 106 downstream 1 -0.16 -0.05 0.09 0.13 _prop 0.09 0.65 0.37 0.17 106 106 106 106 106 total_ 1 -0.01 -0.08 -0.32 chambers 0.93 0.41 0.00 106 106 106 106 aggression 1 0.16 -0.20 0.10 0.04

106 106 106 moving 1 -0.03 0.75

106 106 exploration 1

106 e) TRIAL 4 upstream_ downstream total_ aggression moving exploration prop _prop chambers upstream 1 -0.96 0.37 0.15 0.02 -0.03 _prop <.0001 0.00 0.12 0.82 0.78 104 104 104 104 104 104

downstream 1 -0.23 -0.14 -0.03 -0.02 30

_prop 0.02 0.16 0.73 0.86 104 104 104 104 104 total_ 1 0.15 0.02 -0.38 chambers 0.13 0.85 <.0001 104 104 104 104 aggression 1 0.11 -0.26 0.26 0.01

104 104 104 moving 1 0.10 0.30

104 104 exploration 1

104

3.6$meters$

Figure 1-1 Dispersal apparatus used to measure dispersal behaviors of rainbow trout fry. The apparatus consisted of 15 equal sized compartments with flow through of creek water at ambient temperature. The water flowed in one direction, entering at one end and exiting at the other (indicated by the direction of the black arrows between compartments, left to right). The divider between each compartment had a hole offset from the middle

(example shown in the lower left hand corner) and positioned to alternate from one side 31

to the next along the sequence of dividers (indicated by the position of the black arrows between compartments). A test individual was placed in the center compartment (indicated with the black X) for the 30 min acclimation period, before being allowed to move throughout the experimental apparatus for the trial period.

0.9 0.8

0.8 0.7 AxA31" 0.7 0.6 0.6 0.5 0.5 0.4 AxA34" 0.4 0.3 0.3 chambersSE) (+/- 0.2 chambersSE) (+/- 0.2 AxA36" proportion of upstream of proportion

0.1 downstream of proportion 0.1 0 0 AxR37" a) 1 2 trial 3 4 b) 1 2 trial 3 4 350 20 18 300 AxR38" 16 250 14 200 12 AxR39" 10 150 8 100 6 RxA46" 4 50 aggressive displays (+/- SE) aggressive (+/- displays

total chambersSE) total (+/- visited 2 RxA47" 0 0 c) 1 2 trial 3 4 d) 1 2 trial 3 4 200 25 180 RxA48" 160 20 140 RxR40" 120 15 100 80 10 RxR41" 60 40 5 time (min) to exit (+/- SE) (+/- exit to time(min) time (sec) moving (+/- SE) (+/- time(sec) moving 20 RxR42" 0 0 1 2 3 4 f) 1 2 3 4 e) trial trial 25" Figure 1-2 Family differences in upstream/downstream preference, dispersal activity, 32 15"20" 10"5" aggression, movement, and exploration across four trials. Families are named by their 0" cross type and a unique number; families within the same cross-type share the same line 1"3" trial( format. a) average proportion (+/- SE) of chambers that were visited that were upstream of the starting position within the dispersal apparatus of the dispersal experiment b) average proportion (+/- SE) of chambers that were visited that were downstream of the starting position within the dispersal apparatus of the dispersal experiment c) average number of chambers visited (+/- SE) within the dispersal apparatus of the dispersal experiment d) average number of aggressive displays (+/- SE) in mirror stimulation experiment e) average time (in seconds +/- SE) spent actively moving around the tank in the aggression experiment f) average latency (in minutes +/- SE) to exit the isolation box latency(to(exit(isola-on(box(((+/2(SE)( in the exploration experiment.

CHAPTER 2. DIFFERENTIAL GENE EXPRESSION UNDERLYING BEHAVIORAL VARIATION IN THE BRAINS OF JUVENILE RAINBOW TROUT, ONCORHYNCHUS MYKISS

2.1 Abstract

Individual differences in behavior often reflect underlying genetic variation that can arise as a function of variation in protein coding regions, or as a function of mutation in regulatory regions that modify the expression of genes. The transcriptome provides the connection between the genotype and phenotype of an organism, as it can constantly be changing both in identity and magnitude of expressed genes. The goal of our study was to examine transcriptome-wide patterns of differential gene expression in the brains of juvenile fish with naturally occurring behavioral diversity. To address the question of how differences in gene expression may be associated with behavioral differences on the individual level, we used O. mykiss progeny produced from wild migratory and resident parents in a well-studied system in southeast Alaska. These progeny had undergone a series of behavioral experiments prior to sampling of brain tissue for differential gene expression. Moreover, we examine whether there is any overlap in the differentially 34

expressed genes associated with individual differences in multiple behaviors and how these gene expression differences fit into biological functions and pathways. The results demonstrate that there are some key genes and pathways associated with the observed behavioral traits, and in many cases associated with multiple behaviors. Neuronal

35 signaling (neurotransmitters such as gamma-aminobutyric acid (GABA) and catecholamines) was involved in multiple behavioral measures, while neuronal function and development were part of habituation in behavioral responses. Hormones (such as androgens, glucocorticoids, and growth hormone) and their related pathways are shown to have expression patterns correlating with behavioral differences in these juvenile fish.

The overlap between behaviors in the form of gene expression suggests pleiotropic effects of these genes on behavior with respect to personality traits and habituation or learning in behaviors. By examining differential gene expression, including the functions and pathways of these genes, we found possible proximate molecular and physiological mechanisms underlying observed behavioral variation very early in life.

2.2 Introduction

Individual differences in animal personalities (or behavioral syndromes) can influence both resource use and an individual’s ability to cope or react to changing environmental conditions (Dingemanse et al. 2004; Dingemanse and Reale 2005; Reale et al. 2007), and these differences often reflect underlying genetic variation (Dingemanse et 35

al. 2003; Fitzpatrick et al. 2005). A number of candidate genes have been identified for their association with complex behaviors in insects, birds, fish and mammals (Fitzpatrick et al. 2005). For example, allelic variation in the foraging gene (for) in Drosophila is associated with foraging behavior and movement in Drosophila (Fitzpatrick et al. 2007).

In other candidate gene studies, gene expression differences, rather than allelic variation within candidate genes, has been shown to be associated with behavioral polymorphism

(see (Fitzpatrick et al. 2005) for review); gene expression differences can be modified by heritable variation in transcriptional regulators, developmental differences, or by

36 environmental variation. Genetic variation for complex phenotypes can arise as a function of variation in protein coding regions, or as a function of mutation in regulatory regions that modify the expression of genes. Due to the fixed nature of the genome of an organism, it can offer only a partial understanding of the processes contributing to phenotypic variation. Meanwhile, the transcriptome provides the bridge between the genotype and phenotype of an organism, as it can constantly be changing both in identity and magnitude of expressed genes.

Rainbow trout (Oncorhynchus mykiss) exhibit multiple life history strategies, including anadromy, landlocked populations, and stream resident populations (Behnke and Tomelleri 2002; Quinn and Myers 2004). Anadromous rainbow trout, also known as steelhead, vary greatly in their distribution at sea, incidence of repeat spawning, and associated morphological, physiological, and behavioral traits (Quinn and Myers 2004).

With the amount of variation in the life histories and ecology of rainbow trout, it can be expected that there will be behavioral variation across populations, and even between individuals. There is evidence that behaviors, such as hiding, foraging, and aggression, 36

are heritable since they differ between clonal rainbow trout lines (Lucas et al. 2004).

Certain behaviors are context-dependent and correlated within an individual, such as boldness in varying foraging contexts, but are not related to the same behavior in a different context (e.g. exploration of a swim flume) (Wilson and Stevens 2005). Rainbow trout exhibit different coping styles that appear related to survival (Sloman et al. 2006;

Magnhagen 2008). These behavioral differences can be tied to physiological measures, are stable individual traits, and are heritable (Pottinger and Carrick 2001; Sloman et al.

2006; Overli et al. 2007; Magnhagen 2008). For example, in fish subjected to hypoxia,

37 non-surviving fish displayed strenuous avoidance behavior but surviving fish remained calm (Sloman et al. 2006). These behavioral differences were associated with differences in plasma catecholamine and cortisol levels (Sloman et al. 2006; Magnhagen 2008).

Differences in hormonal responses, in particular those involving cortisol, also influence memory mechanisms (Overli et al. 2007; Magnhagen 2008).

Though the propensity to migrate is heritable in many species (Liedvogel et al.

2011), this decision is also influenced by environmental conditions that individuals encounter during development (Dingle 1991). One of the main regulators of development into a migratory phenotype is the brain, both through interpretation of seasonal cycles via circadian rhythms (Froy et al. 2003; Davie et al. 2009; Mueller et al. 2011) and through production, reception, and signaling of hormones (Schneider et al. 1994; Kitano et al.

2010; Bjornsson et al. 2011).

In this study, we examined differential gene expression between O. mykiss progeny produced from migratory and resident parents in a well-studied system in southeast Alaska. The two populations differ in life history type but share a recent co- 37

ancestry (Thrower et al. 2004). Several studies have shown a strong genetic basis for life history variation in these two populations, including a high heritability for life history type (Thrower et al. 2004), differential gene expression in smoltification-related genes

(Aubin-Horth et al. 2005; Aykanat et al. 2011; Sutherland et al. 2014), quantitative trait loci (QTL) for smoltification-related traits (Hecht et al. 2012), and SNPs associated with life history (Hale et al. 2013; Hecht et al. 2013). Since the brain is a key site regulating behavior, we expect to find some genes differentially expressed among individuals that vary in behavior. To address the question of how differences in gene expression may be

38 associated with behavioral differences on the individual level, we used progeny produced from wild migratory and resident parents that had undergone a series of behavioral experiments prior to sampling for gene expression (see Chapter 1). Moreover, we examine whether there is any overlap in the differentially expressed genes associated with individual differences in multiple behaviors and how these gene expression differences fit into biological functions and pathways. Combining evidence from specific differentially expressed genes, changes in biological functions, and enriched pathways, will give insight into the physiological and molecular mechanisms in the brain promoting behavioral diversity.

2.3 Materials and Methods

2.3.1 Field sampling and genetic crosses

Fish used for this study were generated from crosses using anadromous and resident adult rainbow trout (Oncorhynchus mykiss) sampled from natural populations in southeast Alaska, as described in Chapter 1. Briefly, wild fish were collected from populations in Sashin Creek and Sashin Lake, Alaska. Anadromous individuals were 38

collected in Sashin Creek when returning from the ocean to their natal stream to spawn.

Residents were collected from Sashin Lake, also at sexual maturity. The resident fish are separated from the anadromous fish in the creek by two barrier waterfalls, with all the anadromous individuals collected downstream of these waterfalls and all residents collected upstream, and extensive genetic research has been conducted previously on this study population (Thrower et al. 2004; Aykanat et al. 2011; Hecht et al. 2012; Hale et al.

2013).

Gametes were collected from twelve individuals (six of each sex) for each life history type (anadromous or resident) captured in the wild in Sashin Creek and Sashin

Lake, similar to methods described by McKinney (2013). At collection, fish were anesthetized with tricaine methanesulphonate (MS-222; Argent Chemicals, Redmond,

WA) and gametes were expressed by applying light pressure on the abdominal cavity.

Marine Station (Baranof Island, Alaska), with each family being one of four possible cross-types: anadromous (AxA), resident (RxR), and hybrid and reciprocal hybrids among life history types (AxR and RxA).

Embryos were reared at ambient creek temperatures (inflow received from Sashin

Creek) in flow-through stack incubators in the dark. Each family was reared in an individual container within the stack to maintain family identity. Once the fish reached 39 the fry stage (utilization of all yolk resources), nine individuals from each family (total n=108) were selected and reared in randomized individual compartments within the stack incubators. The fish were fed commercial trout feed twice daily. The remaining fish from each family were reared in outdoor vertical raceways, for another study.

2.3.2 Behavioral experiments

The fish chosen to evaluate gene expression had previously undergone a series of behavioral experiments, as described in Chapter 1. Briefly, the behavioral experiments included laboratory measures of dispersal, exploration, and aggression, as these were

40 expected to be the most ecologically relevant and possibly related to migration or stream- residency. All behavioral experiments were performed at the fry stage (within two months of yolk-absorption) between August and September 2011, at the Little Port

Walter Marine Research Station. Three experiments, described below, were conducted on each fish in succession to quantify individual variation and repeatability for dispersal, exploration, and aggression.

The dispersal experiment consisted of a raceway divided into 15 compartments with offset holes in the dividers between compartments. During the one hour trial, three measures of movement were scored: (i) upstream preference (proportion of total chambers visited that were upstream chambers), (ii) downstream preference (proportion of total chambers visited that were downstream chambers), and (iii) activity (total number of compartments visited, including re-visits). Exploration was tested using an exit test that measured the time required for an individual to exit an isolation box into an open environment, up to thirty minutes. Aggression was assessed using a mirror image stimulation experiment similar to those used in previous studies (Riddell and Swain 1991; 40

Vollestad and Quinn 2003). The following responses were quantified for five minutes using the event recorder JWatcher (Blumstein and Daniel 2007): time spent moving around tank and the number of aggressive displays. The stereotypic aggressive display that was exhibited was a swim against mirror (SAM), which is characterized as an individual swimming with its snout again the mirror, sometimes accompanied by vigorous head shaking and jaw snapping (Keenleyside and Yamamoto 1962; Riddell and

Swain 1991; Vollestad and Quinn 2003).

The three experiments were repeated every eight days, for a total of four trials for each individual in an effort to estimate individual repeatability. The order of experiments for an individual was randomized to minimize effects of experiment order. The order individuals were tested was randomized and the individuals were tested blindly (assigned an alpha-numeric identifier not related to cross-type or family), to prevent particular families or cross-types from being grouped together temporally. For observation and later scoring, all behavioral experiments were video recorded. A single observer scored all behaviors from the video files to prevent any observer-bias or differences in reaction time.

The individuals were scored blindly to prevent any observer bias based on individual, family or cross-type identity.

2.3.3 Sample collection

Once behavioral experiments were completed, fish were euthanized with a lethal dose of MS-222 (Argent Chemicals, Redmond, WA) and measured for weight and fork length. Caudal fin clips were collected and stored in individually labeled vials with 95% ethanol for later DNA extraction and genetic sex typing. Whole brains were immediately 41

dissected and stored in vials of RNAlater (Applied Biosystems, Foster City, CA) at -

80°C until extraction. From the initial and final measurements, growth rate was calculated as [ln(L2) – ln(L1)]/[t2 – t1] X 100, where L1 and L2 are lengths or weights at times 1 and 2 (t1 and t2 in days). DNA was extracted from the fin clips that were collected, using a phenol:choloroform procedure modified from Wasko et al. (2003). All individuals were genotyped at a genetic sex marker, OmyY1, which was used as a proxy for phenotypic sex in analyses (Brunelli et al. 2008). This sex marker within the Sashin

Creek population was previously verified as having 97% accuracy (Hecht et al. 2012).

2.3.4 Statistical analyses and sample selection

The complete data set included the quantified behaviors from the three experiments across all four trials and sizes and genotypic sex for each of the nine individuals from each family within each cross-type (n=108). Complete statistical methods were previously described (see Chapter 1). Briefly, for all models, the fixed effects tested included weight, sex, cross-type, trial order, and a cross-by-trial interaction; the random effects included family nested within cross, individual nested within family, and interactions between those terms and trial. Observations were repeated over trials within individuals. Linear mixed models were used when analyzing the total number of chambers visited during the dispersal experiment, the proportion of upstream/downstream chambers visited during the dispersal experiment, and amount of time an individual spent moving around the tank during the aggression experiment in JMP (version 10.0.0; SAS

Institute Inc., Cary, NC). Generalized linear mixed models with a Poisson distribution were used to analyze the amount of time to exit the isolation box in the exploration experiment and the number of aggressive displays in the aggression experiment using 42

PROC GLIMMIX in SAS (version 9.3; SAS Institute Inc., Cary, NC). We tested whether fish of various cross-types, sexes, and sizes differed in behavior by including them as fixed effects with a type I error rate of 0.05. The significance of random effects was tested with a likelihood ratio test (LRT), which compares the values of twice the restricted log likelihood (-2LL) of nested mixed effects models and is distributed as a chi- square with degrees of freedom equal to the difference in the number of terms between the models. Any term that was not significant was removed before running models to extract variance genes or calculate estimates.

To choose samples to test hypotheses that behavioral variation is associated with differential gene expression in the brain, estimates were produced from the behavioral statistical models for each of the significant effects for use in the analyses. When personalities were found, individual estimates were produced from the models. These are like individual intercepts and represent differences between individuals in the elevation of their reaction norms, while taking into account other significant effects. Individual differences, or personalities, were significant for measures of aggressive displays, exploratory tendency, dispersal activity, and moving during the aggression experiment

(see Chapter 1). When there were significant individual differences in habituation, individual estimates of changes across trials were produced. There were only significant individual by trial effects for aggression, exploration, and dispersal. For dispersal activity, there were significant differences between families, and so estimates for each family were calculated for dispersal activity. The fixed effect of the interaction between cross-type and trial was found to be significant when analyzing the upstream and downstream preferences of the individuals, so the estimates for each of the four cross- 43

types for these traits were calculated. Out of the available samples, thirty-six individuals were selected for RNA sequencing. Individuals were chosen from across the distributions of the behavioral traits in which we were interested to obtain a representative sample for each phenotype to be tested (Figure 2-1). Sixteen individuals were chosen from each of the two pure cross-types (AxA and RxR), with the individuals spread among three families within each cross-type. At least one male and one female were chosen from each family for a total of twenty females and sixteen males. These selected individuals were

44 binned into binary traits with reference to their behavioral differences (described in more detail below) for testing in models for differential gene expression.

2.3.5 RNA extraction and sequencing

RNA was extracted using Trizol (Ambion, Foster City, CA) following the manufacturer’s protocol. Whole brains were homogenized and RNA extracted. Library preparation and sequencing were conducted at the Purdue University Genomics Core

Facility. Samples were analyzed for concentration and quality using a Nanodrop

(Thermo Scientific, Palm Beach, FL). Final quality control was conducted on a 2100

Bioanalyzer using a NRA 6000 Nano chip (Agilent Technologies, Santa Clara, CA).

Libraries were prepared using TruSeq RNA sample preparation kits (Illumina, San Diego,

CA) and sequencing was conducted on an Illumina Hi-Seq 2500 (v3 chemistry) using

100bp paired-end reads. The thirty-six samples were individually barcoded and split across two lanes, for eighteen samples per lane, and a full experimental replicate was run on an additional two lanes for a total of four lanes of sequencing. The initial titer of each library was determined using qPCR (KAPA) and a pool created based on these titers. An 44

initial lane of sequencing was undertaken for each pool of samples to more accurately estimate the titer of each library. Subsequent lanes were then sequenced with a new pool rebalanced to obtain more even read counts among individuals.

2.3.6 Gene expression

The resulting sequence data from each individual was quality filtered using

Trimmomatic v.0.17 (Lohse et al. 2012). Sequence adapters were removed, sequence ends with quality (Phred) scores less than 20 were trimmed, and sequences less than 30 nucleotides in length were removed. To quantify gene expression, sequence reads for

45 each individual were aligned against a de novo transcriptome assembly produced from brain and embryo tissue in the same population (McKinney 2013). The assembly was built from contiguous sequences (contigs) reconstructed from overlapping sequence reads.

Contigs that share portions of sequence with other contigs, as a result of alternative splicing or allelic variation, and potentially gene duplication, are identified as belonging to the same component. To obtain a conservative estimate of gene expression and prevent inadvertent quantification of allele specific or isoform specific expression, quantification of expression was done at the component level. To avoid confusion from this terminology, we will refer to components as genes, as this is what they represent.

Alignment to the transcriptome was done using RNA-Seq by Expectation Maximization

(RSEM v.1.2.0) (Li and Dewey 2011) with default settings using the transcript-to-gene- map option to obtain read counts for each gene. Read counts for each gene were imported into edgeR v.3.2.3 (Robinson et al. 2010) for statistical analysis. Raw reads counts were converted to counts per million (cpm) to take into account differences in library size between individuals. Genes were required to pass a count threshold where a 45

minimum of 18 out of 36 individuals in a contrast must have at least 1 cpm; genes that did not pass this threshold were removed. Remaining gene read counts were then normalized using trimmed mean of m-values (TMM) normalization to account for differences in library composition. TMM normalization estimates a scaling factor that minimizes differentially expressed genes between libraries under the assumption that the majority of genes are not differentially expressed (Robinson and Oshlack 2010).

2.3.7 Contrasts

Within edgeR v.3.2.3 (Robinson et al. 2010), thirteen contrasts were evaluated to test hypotheses of differential gene expression for different behavioral phenotypes. First, to evaluate whether sex or cross-type contributed to variation in gene expression, counts were analyzed by combining sex and cross-type into a two-way factor, which allowed direct comparison between different factor levels. To examine differences in progeny produced from different cross-types, contrasts were done between the different cross- types but within the same sex; for example, females from resident parents were compared to females from anadromous parents. The other nine contrasts that were done were to compare gene expression based on behavioral differences that were significant in the previous study and are summarized in Table 2-1. At the individual level, there were significant differences for aggression, exploration, dispersal, and moving during the aggression experiment. Across trials, individuals differed in their rate of habituation for aggression, exploration, and dispersal. For these seven individual level effects, the 36 fish being tested for gene expression differences were categorized as either above or below 46

the median value found across all 108 individuals that were used in the behavioral study

(See Figure 2-1). These categories were used to find the genes that were significantly up or down-regulated for each behavior. For example, by doing a contrast between the most aggressive and least aggressive individuals, we can identify the gene associated with consistent individual level differences in aggression. The other two significant behavioral differences found in the previous study were family differences for dispersal and cross-type differences in habituation in upstream and downstream preference. For dispersal, the six families that were being tested for gene expression differences were

47 categorized as above or below the median value for the twelve total families that were used in the behavioral study. When testing for gene expression differences between cross-types for habituation in upstream and downstream preferences, only one contrast was done because the pure anadromous cross-type had a lower magnitude habituation for both upstream and downstream preference compared to the pure resident cross-type. In all of the contrasts, to correct for multiple testing, a false discovery rate (FDR) threshold of q<0.05 was used to indicate significance (Benjamini and Hochberg 1995).

2.3.8 Annotation

Contigs from genes that passed count thresholds were aligned against the

ENSEMBL databases (Flicek et al. 2013) for human (assembly GRCh37, release 73) and zebrafish (assembly Zv9, release 73), as previously described by McKinney (2013). The

ENSEMBL databases included both the cDNA (containing all transcripts resulting from

ENSEMBL known, novel, and pseudo gene predictions) and protein (containing all translations resulting from ENSEMBL known or novel gene predictions) databases.

BLASTn was used for alignments to nucleotide databases while BLASTx was used for 47

alignments to protein databases. All BLAST alignments used an e-value threshold of 1E-

5 for annotation. For final annotation of genes for pathway analysis, the contig with the lowest e-value hit for human protein or nucleotide databases was chosen; in the absence of human protein or nucleotide hits, the lowest e-value zebrafish protein or nucleotide hit was used for annotation. Preference was given to human annotations as IPA is primarily based on mammalian literature.

2.3.9 Pathway enrichment analysis

Pathway analysis was conducted using Ingenuity Systems Pathways Analysis

(IPA) software (Qiagen, Redwood City, CA) to determine if there were biofunctions or pathways that were enriched with differentially expressed transcripts. Biofunctions are groupings of genes that have been demonstrated to be involved in similar biological functions; however genes within a biofunction may or may not have interactions with each other. Biofunctions are also grouped into categories based on similarity of function, for example, the cell morphology category is comprised of biofunctions such as shape change of axons, neuritogenesis, and formation of cellular protrusions, among others. In contrast, canonical pathways are a curated set of well-characterized metabolic and cell- signaling pathways where interactions between genes are known. Data including transcript annotations, log fold-change, and p values from EdgeR analysis were uploaded into IPA using the human or zebrafish annotations from ENSEMBL, as described above.

Thresholds within IPA for considering a gene differentially expressed utilized p-values and fold-changes from edgeR and were set at p<0.001. Since IPA is not testing the 48

significance of genes as single entities, but rather their significance when combined in networks with other genes, multiple testing correction was not done on data loaded into

IPA to prevent inflated type-II error which is of particular concern when sample size is low and variability is high (Cole et al. 2003). Since the downstream pathway analyses are highly dependent upon the initial list of differentially expressed genes we instead focused our analysis on pathways and biofunctions that were identified in IPA as being significantly enriched for differentially expressed genes using Fisher’s exact test with a threshold of p<0.05, following the methods of Harding et al. (2013). Pathways and

49 biofunctions that are identified as enriched contain more differentially expressed genes than expected by chance based on the number and identity of differentially expressed genes in our dataset. Biofunctions can also be identified as functionally altered based on prior knowledge of the expected causal effects of differentially expressed genes on a biofunction. IPA uses a regulation Z-score algorithm to estimate the probability that a particular biofunction will be increased or decreased based on the differentially expressed genes in the dataset (Ingenuity Systems, www.ingenuity.com). Regulation Z-scores of

>2 or <-2 are considered significant.

2.4 Results

2.4.1 RNAseq

In total 1,684,011,616 paired reads were generated through Illumina sequencing in the 36 individuals chosen in this study. After removing low quality sequence,

1,646,801,248 (98% of total) paired reads remained (Table B-1). Paired sequence reads from each sample were aligned against the transcriptome to generate read counts. For individual samples, the number of reads used for quantifying gene expression ranged 49

from 17,358,671 to 36,171,809 paired reads for individual samples, of which, between 55% and 62% aligned to between 70,349 and 147,551 genes of the assembly (Table B-2). This alignment rate is similar to the read mapping proportion observed form individuals that were used to make the transcriptome assembly (McKinney 2013). After removing genes that failed to meet minimum count thresholds (a minimum of 18 out of 36 individuals in a contrast must have at least 1 cpm), a total of 25,529 genes remained and were tested for differential gene expression.

2.4.2 RNAseq differential gene expression

Between 0 and 722 genes (0% - 2.83% of total) were differentially expressed with an FDR adjusted p-value<0.05 for each of the 13 contrasts (Table 2-2). The largest number of differentially expressed genes (n=722) occurred between AxA cross-type, that had a strong upstream preference, and the RxR cross-type, that had a strong downstream preference. In contrast, no differentially expressed genes occurred between the sexes of anadromous fish or individuals with high and low exploration. The proportion of up- regulated to down-regulated genes varied across contrasts (Figure 2-2).

2.4.3 Contrasts

For the contrasts between sexes within the resident cross-type there was only one significant gene, which was annotated as collagen and was down-regulated in females compared to males. There were no significant genes for the contrast between the sexes within the anadromous cross-type. For the contrast between cross-type within females and within males there were 217 and 67 significant genes respectively, with most (114 and 40 respectively) down-regulated in the resident cross-type compared with the 50

anadromous cross-type (Table 2-2). There were 23 shared, differentially expressed genes between those two contrasts of cross-type within the sexes.

The contrast between the fish that displayed high aggression and low aggression produced 390 significant genes (Table 2-2). The majority of the genes (85%) had lower expression in the highly aggressive fish compared with the weakly aggressive fish

(Figure 2-2). The contrast between individuals that displayed high dispersal activity and those that displayed low dispersal activity produced seven significant genes, which were expressed at higher levels in the fish that displayed high activity compared with those that

51 displayed low dispersal activity. There were no significant genes in the contrasts between the fish with high exploration and low exploration. There were two genes with significantly greater expression in individuals with high movement during the aggression experiment, compared to those with low movement. Those two genes were annotated as growth hormone 1 (GH1) and proopiomelanocortin (POMC). Overall, there were 9 significant differentially expressed genes between individuals that did or did not change the exploratory behavior greatly across time. Most of those (67%) were more highly expressed in individuals that had smaller changes in their exploratory tendencies. One annotation in this contrast was the prolactin (PRL) gene, which was actually more highly expressed in individuals with larger changes in their exploratory behavior. The other two slope or habituation traits that were examined were aggression and dispersal activity, which had 8 and 12 significant genes respectively. Aggression, like exploration, had more genes with greater expression in the individuals who had low habituation (or less change across trials) compared with those that greatly adjusted their behaviors across trials. In contrast, the opposite trend was true for dispersal activity, where most of the 51

genes were expressed at higher levels in individuals that changed their dispersal activity more dramatically across trials. There were 98 differentially expressed genes identified in the contrast between families that had high and low dispersal activity (Table 2-2). Most of the genes were up-regulated in the families with high dispersal activity compared with those that had low dispersal activity. Overall, there were 722 genes significantly differentially expressed between the cross-types that differed in their habituation of upstream/downstream preference (Table 2-2). Half of the genes are more highly expressed in the anadromous cross-type, which has lower habituation in

52 upstream/downstream preference, compared with the resident cross-type, which has higher habituation rates (Figure 2-2). Interestingly, two of the significant genes in the contrast between cross-types for their upstream/downstream preference habituation were annotated as Circadian Locomotor Output Cycles Kaput (CLOCK) and a neurotransmitter transporter for gamma-aminobutyric acid (GABA). CLOCK was more highly expressed in the RxR cross-type that had higher habituation in upstream/downstream preference, while the GABA transporter was more highly expressed in the AxA cross-type that had lower habituation in upstream/downstream preference. All of the significant genes across contrasts, along with their logFC, FDR corrected p-values, and annotations are listed in the supplementary tables.

There was some overlap in the differentially expressed genes identified in fish with divergent behaviors and from different cross-types. For example, a single gene was significant in six contrasts: aggression, cross-type within females, cross-type upstream preference, dispersal activity, family dispersal activity, and habituation in aggression. It was annotated as deoxyribonuclease I-like 3 (DNASE1L3). ATPase, Na+/K+ transporting, 52

alpha 3 (ATP1A3) was up-regulated in individuals with high dispersal activity, low habituation in aggression, high habituation in dispersal activity, and low habituation in upstream preference AxA cross-type. One gene was shared with the contrast between cross-types for their habituation in upstream/downstream preference and the contrast between families for their dispersal activity, annotated as heat shock 70kDa protein 8

(HSPA8). The two genes mentioned earlier with expression significantly associated with movement during the aggression experiment, growth hormone 1 (GH1) and proopiomelanocortin (POMC), were also significant in the contrasts between cross-types

53 within females and between individuals with high and low habituation in exploration.

They were up-regulated in the anadromous fish compared with the resident fish within females, as well as in the fish that dramatically decreased their exploratory behavior across trials compared with those that did not change or only slightly changed.

2.4.4 Biofunction analyses

Among the contrasts, there were between 9 and 84 biofunction categories, which can contain multiple biofunctions, significantly enriched for genes that were differentially expressed (p<0.05) (Table 2-3). When looking at the individual biofunctions, there were between 5 and 309 significant biofunctions. The supplementary tables contain all of the significant biofunctions and their annotations, sorted by category, for all contrasts, including their p-values. For the contrasts between the sexes, the most significant biofunctions were in the categories of cellular development, cell morphology, and cellular assembly and organization. Reproductive development was also amongst the significant categories. For the contrast between the sexes within resident fish, the inflammatory response biofunction category (including four different biofunctions) was 53

identified as functionally altered with an increased predicted activation state (Z-score>2).

The biofunctions pertain to inflammation of the lung, body cavity, body region, and organ were all predicted to be significantly increased in females compared to male resident fish, based on the differential gene expression patterns. While the number of significant categories and biofunctions was uneven between the anadromous and resident sex contrasts, the cross-type contrasts within females and males were roughly equal.

Small molecule biochemistry and cell-to-cell signaling were among the top categories for both as well. There were some behavior biofunctions that were divergent between the

54 cross-types, such as: place preference, circadian phase shifting, navigation, emotional behavior, and gait. In the contrast between the cross-types within females, based on the up-regulation of POMC, Homer protein homolog 1 (HOMER1), and monoamine oxidase

A (MAOA) and the down-regulation of opsin 4 (OPN4), Glutamate Receptor, Ionotropic,

AMPA 2 (GRIA2), and Myelin Associated Glycoprotein (MAG), there is a predicted decrease in emotional behavior in the anadromous cross-type. The biofunction of navigation is also associated with HOMER1 expression, while circadian phase shifting is predicted to be decreased in the anadromous cross-type due to the decrease in OPN4 expression. Differential expression of POMC, HOMER1, and GRIA2 also significantly associate place preference with this contrast, although the predicted effect on this biofunction is mixed. The up-regulation of POMC and down-regulation of GRIA2 might cause an increase a conditioned preference for a place (Petraschka et al. 2007; Cai et al.

2013) in the anadromous cross-type, while the increase in HOMER1 expression could decrease it (Szumlinski et al. 2004).

Among the individual level behavioral contrasts, aggression had the greatest 54

number of significant biofunctions (n=276) and the greatest number of biofunction categories overall (n=84). The top two categories of molecular and cellular functions associated with aggression were gene expression and protein synthesis. Interestingly, 67 of the 68 molecules associated with these biofunction categories were down-regulated in the more aggressive fish compared with the less aggressive fish. Which also accounts for the majority (n=16/20) of the significant predicated activation states that were decreased

(activation z-scores <-2). These altered biofunctions are mostly related to infectious disease and cell death and survival, with less aggressive fish having decreased

55 biofunctions such as viral infection, infection by RNA viruses, infection by HIV-1, cell survival, and cell viability. Another significant biofunction was related to the signaling of dopaminergic neurons, which is predicted to be decreased in more aggressive individuals based on the lower expression of KCNN3 (potassium intermediate/small conductance calcium-activated channel, subfamily N, member 3).

For exploration differences, the top molecular and cellular function categories were drug metabolism, cell signaling, molecular transport, vitamin and mineral metabolism, and amino acid metabolism. Molecular transport included the biofunctions of concentration of aldosterone and the release of neurotransmitter, while the amino acid metabolism category was included the biofunction of synthesis of GABA (all three driven by the up-regulation of parvalbumin (PVALB) in less exploratory fish). PVALB is known to be involved in the synthesis of GABA (Csillik et al. 2010), which is a neurotransmitter implicated in stress coping in humans (Overli et al. 2007). Another significant category that only contained one biofunction was organismal function, which only covered lethargy (affected by the up-regulation of Collagen, Type X, Alpha 1 (COL10A1) in less 55

exploratory fish).

Differences between individuals for movement during the aggression trials produced gene expression differences in 74 biofunction categories, with the top one for molecular and cellular functions being gene expression. Seven of the eight genes associated with this category were up-regulated in the fish that moved more compared with those that move less (GH1, MYC Associated Factor X (MAX), MIF4G Domain-

Containing Protein (MIF4GD), nuclear factor of kappa light polypeptide gene enhancer in B-cells inhibitor, alpha (NFKBIA), POMC, PRL, and v-rel avian reticuloendotheliosis

56 viral oncogene homolog A (RELA), but not luteinizing hormone/choriogonadotropin receptor (LHCGR)). Also under the category of gene expression were two biofunctions, transcription of RNA and expression of RNA, which were identified as functionally altered (due to the expression patterns of those same eight molecules). They are predicted to have an increased activation (z-scores>2) in fish that moved more compared with those that moved less during the aggression experiment. Some of these same molecules account for the significance of regulation of cholinergic neurons (increased), production of neuronal progenitor cells (increased), and function of neurons (affected) in individuals with higher movement. An interesting biofunction under the significant category of behavior was aggressive behavior, which has two genes associated with it: POMC and beta-2-microglobulin (B2M) (both up-regulated in the fish that moved more compared with those that moved less, although they have reported opposite associations with aggression).

Individual differences in dispersal activity, or the total number of chambers visited, had the fewest number of biofunction categories (n=9) and biofunctions (n=5) 56

significantly associated with it. They were all based on a single molecule, either carnosine dipeptidase 1 (CNDP1) or zinc finger protein 277 (ZNF277), which were up- regulated in individuals with high dispersal activity compared with those with low. These two molecules play a role in the molecular and cellular functions of cell cycle, post- transcriptional modification, protein degradation, and protein synthesis.

For the habituation or plasticity that was found in behavioral traits, aggression was associated with 43 biofunctions, of which the top categories were: cell-to-cell signaling and interaction, molecular transport, amino acid metabolism, cell cycle, and

57 cellular development. Cell-to-cell signaling and interaction was affected by the down regulation of Pro-Melanin-Concentrating Hormone (PMCH) in the individuals that showed more habituation or decreased their aggression more dramatically, the biofunctions include: firing of nucleus accumbens shell (predicted to be decreased), firing of neurons (predicted to be increased), turnover of norepinephrine (predicted to be decreased), plasticity of neuronal synapse (predicted to be affected), and release of acetylcholine (predicted to be decreased).

For habituation in exploration, there were 78 significant biofunction categories, with the top molecular and cellular functions being: cell-to-cell signaling and interaction, cellular growth and proliferation, cellular assembly and organization, cellular function and maintenance, and gene expression. Under the cellular function and maintenance category, the biofunction of production of neuronal progenitor cells (increased), function of neurons (affected), synaptogenesis of brain (decreased), and the regulation of cholinergic neurons (increased) are all related to individuals with higher habituation in exploratory behavior (molecules: POMC, PRL, tectorin alpha (TECTA), and 57

Thrombospondin 1 (THBS1). Under the significant behavior category, there were two biofunctions: operant performance and exercise. Operant performance was affected by the up-regulation of POMC in fish that showed more habituation in exploration, while exercise is predicted decrease because of the down-regulation of protein phosphatase 1, regulatory subunit 3A (PPP1R3A) in fish who decreased their exploratory behavior the most.

For the contrast for habituation in dispersal activity there were 74 associated biofunction categories, with the top molecular and cellular functions being: cellular

58 development, cellular growth and proliferation, cell-to-cell signaling and interaction, cellular function and maintenance, and cell cycle. Cellular development includes functions such as the expansion of neural precursor cells, development of T helper type

17 (Th17) cells, proliferation of dendritic epidermal T cells, differentiation of naïve helper T cells, and clonal expansion of T lymphocytes (all predicted to be affected by the down-regulation of Aryl Hydrocarbon Receptor (AHR)). One more significant biofunction was the synthesis of GABA, because of the higher expression of PVALB in individuals who demonstrated more habituation in dispersal activity.

On the level of family differences, the contrast looking at dispersal activity revealed 79 biofunction categories with 273 significant biofunctions. Under the significant category of behavior, there were four functions: locomotion, associative learning, grooming, and eyeblink conditioning. Four of the molecules associated with locomotion were up-regulated (Ceroid-Lipofuscinosis, Neuronal 6, Late Infantile, Variant

(CLN6), cadherin, EGF LAG seven-pass G-type receptor 1 (CELSR1), polymorphisms of heat shock protein 70 (HSPA1A/HSPA1B), and junctophilin 3 (JPH3)) and three were 58

down-regulated (Ceroid-Lipofuscinosis, Neuronal 8 (CLN8), Down Syndrome Cell

Adhesion Molecule (DSCAM), and MAG) in highly dispersive families compared with low dispersing families.

When comparing habituation in upstream and downstream preferences on the cross-type level, there were the most significant biofunctions (n=309) across 78 categories. One of the highest molecular and cellular function categories is cell-to-cell signaling and interaction, under which the higher expression of KCNN3 leads to an expected increase in afterhyperpolarization of dopaminergic neurons. The third highest

59 physiological system development and function category is behavior, under which there are seven functions: tone fear conditioning, taste aversion learning, circadian rhythm, memory, olfactory memory, chaining behavior, and fasting. The six molecules associated with circadian rhythm are split with half up-regulated (Tenascin C (TNC), Neural Cell

Adhesion Molecule 1 (NCAM1), and nuclear receptor subfamily 1, group D, member 1

(NR1D1)) and half down-regulated (AHR, CLOCK, and CHRNB2) in individuals from cross-types with more habituation in upstream/downstream preference compared to those with less habituation. The two molecules with opposite expression would both decrease olfactory memory (PLAU and CHRNB2). Plasminogen activator, urokinase (PLAU) is up-regulated, while cholinergic receptor, nicotinic, beta 2 (CHRNB2) is down-regulated, in individuals from the AxA cross-types with more habituation in upstream/downstream preference compared to those from the RxR cross-type with less habituation in upstream/downstream preference. There are six biofunctions identified as functionally altered: metastasis of lung, response of lymphocytes, gastrointestinal neoplasia, response of mononuclear leukocytes, breast or colorectal cancer, and hepatic steatosis. The first 59

five are predicted to be decreased in the AxA cross-type which lower habituation in upstream/downstream preference (z-score<-2).

2.4.5 Pathway analyses

Pathway analysis revealed up to 80 altered canonical pathways for each contrast

(Table 2-3), with the greatest number of altered pathways occurring between individuals that moved more than average and those that moved less than average in the aggression trial. The contrast between individuals with high and low dispersal activity (total chambers intercepts) and between individuals with high and low habituation in

60 aggression (aggression slopes) had no significant pathways. The supplementary tables summarize all the significant canonical pathways across the contrasts, including the negative log of the p-values. The top pathways for each contrast are presented briefly below.

For the contrast between sexes of the anadromous cross-type, the top three significant pathways relate to catecholamines (dopamine, noradrenaline, and adrenaline), which function as neurotransmitters and hormones. L-dopa degradation, the most significant pathway, involves its role as a precursor metabolite in the biosynthesis of the catecholamines. Dopamine degradation and noradrenaline and adrenaline degradation are the next two most significant pathways and are also significant because of the higher expression of leucine rich transmembrane and O-methyltransferase domain containing

(LRTOMT) in females from the anadromous cross-type. Meanwhile, the contrast between the sexes of the resident cross-type, some of the significant pathways including the third most significant pathway which was glucocorticoid receptor signaling, were related to glucocorticoids. This is because of the lower expression of PLAU, POMC, and RELA in 60

females from the resident cross-type.

The creatine-phosphate synthesis pathway was significant for both cross-type comparisons, and was the most significant pathway in the contrast between the cross- types within females. In both contrasts, the significance of this pathway is associated with the gene creatine kinase, mitochondrial 2 (CKMT2), which was down-regulated in the anadromous cross-type compared with the resident cross-type.

There were 31 significant pathways in the contrast between high and low aggressive individuals, with the most significant being eIF2 (eukaryotic initiation factor-

2) signaling (involved in protein synthesis and gene expression). This pathway appears to be more active in less aggressive fish (z-score=-3.317). A variety of stimuli, including insulin, modulate eIF2 activity, which in turn regulates mRNA translation. Another pathway that was identified as having a significant activity pattern, based on prior knowledge of the expected causal effects of differentially expressed genes, is the antioxidant action of vitamin C (z-score=2.066). The observed gene expressed fits a pattern of the pathway being activated in more aggressive fish. Interestingly, androgen signaling was also among the significant pathways for the aggression comparison.

Specifically, the genes guanine nucleotide binding protein gamma 2 (GNG2), and several genes that encode RNA polymerase subunits (POLR2I, POLR2J, POLR2K, and POLR2L) are all down-regulated in more aggressive individuals.

The contrast between highly exploratory and less exploratory individuals only resulted in 5 significant pathways, the highest one being intrinsic prothrombin activation, which has to do with fibrin formation and healing. The gene COL10A1’s higher expression in less exploratory fish is responsible for the significance of this pathway. 61

Glucocorticoid receptor signaling was also among the 80 significant pathways in the contrast between individuals that moved more than average and those that moved less than average in the aggression trial. This was due to the up-regulation of NFKBIA,

POMC, PRL, and RELA in fish that moved more. The expression of NFKBIA and RELA were also responsible for the significance of protein kinase A (PKA) signaling in this contrast.

For the comparisons between individuals with high and low habituation in exploration and dispersal activity, they both had eight significant pathways, although they

62 were different. Significantly related to habituation in exploration were glucocorticoid receptor signaling and dopamine receptor signaling. When the dopamine receptor signaling pathway is active, the expression of PPP1R3A and PRL is expected to decrease, but in fish with high habituation in exploration PPP1R3A expression went down but PRL went up compared to those with low habituation. The up-regulation of PRL and POMC expression accounts for the significance of the glucocorticoid receptor signaling pathway.

Within the contrast between individuals with high and low habituation in dispersal activity the highest significant pathway was glutamine degradation.

On the family level of analysis, for the comparison between high and low dispersal activity, GABA receptor signaling came up as a significant pathway. The significance of this pathway can be attributed to the lower expression of adaptor-related protein complex 2, mu 1 subunit (AP2M1) and solute carrier family 6, member 13

(SLC6A13) in families with a higher dispersal activity.

One of the most interesting findings was that in the comparison of habituation in upstream and downstream preferences among the cross-types, circadian rhythm signaling 62

showed up as significant. Within this pathway, CLOCK was down-regulated while glutamate receptor, ionotropic, N-methyl D-aspartate-associated protein 1 (GRINA) and

NR1D1 were up-regulated in the AxA cross-type with low habituation in upstream preference. Also within the habituation of upstream preference contrast, the LXR/RXR

(liver X receptor/retinoid C receptor) activation appears to be active in the low habituating AxA cross-type (z-score=2.035) based on the gene expression patterns of apolipoprotein H (APOH), chemokine (C-C motif) ligand 2 (CCL2), fatty acid synthase

(FASN), hydroxyacyl-CoA dehydrogenase (HADH), prostaglandin-endoperoxide

63 synthase 2 (PTGS2), and serpin peptidase inhibitor, clade A, member 1 (SERPINA).

LXR/RXR is involved in the regulation of lipid metabolism, inflammation, and cholesterol to bile acid catabolism.

There were several pathways that were significant in multiple contrasts. As mentioned above, intrinsic prothrombin activation was the most significant pathway in the contrast between highly exploratory and less exploratory individuals. This was also one of the top pathways in the comparison between individuals with high and low habituation in exploration, as well as in the comparison between individuals with high and low habituation in dispersal activity. Hepatic fibrosis/hepatic stellate cell activation was a significant pathway in the contrasts for individual exploration differences, as well as habituation in both exploration and dispersal activity. Dendritic cell maturation was a significant pathway related to individual differences in exploration and moving during the aggression experiment, in addition to habituation in exploration. Glucocorticoid receptor signaling was among the 80 significant pathways in the contrast between individuals that moved more than average and those that moved less than average in the 63

aggression trial, as well as between individuals that differed in rates of habitation for exploration (as previously mentioned).

2.5 Discussion

Here, we report the results of a transcriptome-wide study on gene expression in the brains of behaviorally diverse juvenile rainbow trout. The results demonstrate that there are some key genes and pathways associated with the observed behavioral traits.

Neuronal signaling (neurotransmitters such as gamma-aminobutyric acid (GABA) and catecholamines) was involved in behavioral measures possibly related to stress, while

64 neuronal function and development were part of habituation in behavioral responses.

Hormones (such as androgens, glucocorticoids, and growth hormone) and their related pathways with known behavioral connections, are shown here to have expression patterns correlating with behavioral differences in these juvenile fish. In addition to the genes that are known to play a role in salmonid behavioral phenotypes (such as: GH1, POMC, PRL, and CLOCK), we also identified some novel genes (such as: COL10A1, PPP1R3A, and

HOMER1), through the use of biofunction and pathway analyses. This study combines multiple lines of evidence, including specific differentially expressed genes, changes in biological functions, and enriched pathways, to identify possible proximate molecular and physiological mechanisms underlying observed behavioral variation. We examined the correlation between gene expression and behavior with respect to personality traits, habituation or learning of behaviors, and overlap between behaviors in the form of gene expression that may suggest a mechanism for behavioral syndromes, with the major findings discussed in more detail below.

As we would expect from evaluating differential gene expression in the brain, 64

there were many genes and differentially regulated pathways and biological functions that are associated with neuronal signaling. Catecholamines (dopamine, noradrenaline, and adrenaline), which function as neurotransmitters and hormones, have important physiological regulatory roles. Catecholamine related pathways are significantly associated with differences between the sexes in the anadromous cross-type. The higher expression KCNN3, which affects dopaminergic neurons, was significantly associated with the low habituation in upstream/downstream preference of the anadromous cross- type and less aggressive fish. Also related to dopamine (a catecholamine) and its receptor

65 signaling is the habituation of exploratory behavior. Dopamine is important to vital brain functions like motor control and short-term memory. Dopamine is a monoamine neurotransmitter that has been associated with calm behavior during hypoxic challenges, social stress, and the proactive/reactive behavioral spectrum in rainbow trout (Overli et al.

1999; Schjolden and Winberg 2007), rodent, and lizard brains (Overli et al. 2007). The catecholamine neurotransmitters, especially dopamine, and their related pathways play an important role in multiple behavioral traits (aggression, upstream preference, and habituation in exploration).

Another neurotransmitter, GABA (gamma-aminobutyric acid), was associated with differences in dispersal and exploratory behaviors. These two behavioral traits are highly correlated within individuals (see Chapter 1). A transporter of GABA, an inhibitory neurotransmitter, was more highly expressed in the anadromous cross-type, which displayed less habituation in upstream/downstream preference compared with the resident cross-type. GABA receptor signaling was associated with differences between families differing in dispersal activity. Synthesis of GABA was significantly associated 65

with individual differences in exploration and habituation in dispersal activity, because of higher expression of PVALB in individuals with higher habituation and lower exploration.

Expression of GABA in the hypothalamus has been associated with stress coping in humans (Overli et al. 2007). The significant association of GABA and its pathways with multiple behavioral traits, particularly the correlated traits of dispersal and exploration, suggests a shared proximate mechanism. Overlapping genetic control for traits has been suggested as the mechanism behind behavioral syndromes, or suites of correlated behaviors (Reale et al. 2007).

The nervous system plays an important mechanistic role underlying habituation

(Rankin et al. 2009), and we find several biological functions related to neuron functioning and development associated with habituation in behavioral traits.

Differentially expressed genes that play a role in dendritic cell maturation were significantly related to exploration, habituation in exploration, and movement during the aggression experiment. Higher rates of habituation in exploratory behavior and higher movement were also related to the differential expression of genes as they related to the function of neurons, increased production of neuronal progenitor cells, and the increased regulation of cholinergic neurons. Based on the observed gene expression patterns, individuals with higher habituation rates for aggressive behavior are predicted to have increased firing of neurons and differences in plasticity of neuronal response when compared to individuals that displayed low habituation. The expansion of neural precursor cells was significantly associated with habituation in dispersal activity. The higher expression of KCNN3, which is related to an increase in afterhyperpolarization

(AHP) of dopaminergic neurons, was significantly associated with the low habituation in 66

upstream/downstream preference of the anadromous cross-type and less aggressive fish.

This is notable because the AHP, the hyperpolarizing phase of a neuron’s action potential, has been implicated in learning and aging (Matthews et al. 2009) as well as being regulated by the circadian clock (Cloues and Sather 2003). Overall, based on the observed gene expression patterns, it appears that the functioning and production of neurons is strongly associated with habituation. This is not shocking, considering that throughout the life of a fish, brain development itself is plastic in response to the environment (Näslund et al. 2012).

Hormones and their pathways have long since known to influence behavior

(Kappeler and (ed.) 2010). In our study, we find a few notable hormone pathways that are differentially regulated with respect to behavior (i.e. androgens, glucocorticoids, and growth hormone), each of which we will discuss in turn. It was interesting to find androgen signaling was among the significant pathways in the contrast between high and low aggressive individuals. The environmentally sensitive hormone testosterone has been associated with several physiological and behavioral traits in juncos, including territorial aggression (Atwell et al. 2014). The expression of androgen and estrogen receptors in the brains of African cichlids has been associated with social dominance (St-Cyr and Aubin-

Horth 2009). The differentially expressed genes from the androgen signaling pathway

(GNG2, POLR2I, POLR2J, POLR2K, and POLR2L), are good candidates for regulators of socially aggressive behavior.

From our study, genes involved in glucocorticoid pathways (NFKBIA, POMC,

PLAU, PRL, and REA) appear to play a significant role in differences between behaviorally diverse fish. Glucocorticoids, a major subclass of steroid hormones, 67

regulate a large number of immune, metabolic, cardiovascular and behavioral functions

(Wada 2008). They are produced and released from the adrenal cortex under the control of the hypothalamic-pituitary-interrenal (HPI) axis. Glucocorticoids produce their effect on responsive cells by activating the glucocorticoid receptor directly or by indirectly modulating the transcription of target genes (such as decreasing expression of PRL and

POMC (Sakai et al. 1988; Drouin et al. 1993)). Glucocorticoid receptor signaling is a significant pathway associated with movement in the aggression trial, as well as between individuals that differed in rates of habituation for exploration. Glucocorticoid hormones

68 in the HPI axis, have been associated with the control of pain in subordinates (Ashley et al. 2009), calm behavior during hypoxic challenges (Schjolden and Winberg 2007), aggression (Pottinger and Carrick 2001; Overli et al. 2007), the proactive/reactive stress coping styles (Schjolden and Winberg 2007), competitive ability (Overli et al. 1999;

Pottinger and Carrick 2001), memory (Wada 2008), learning (Overli et al. 2007), food intake (Pottinger and Carrick 2001; Overli et al. 2007), and locomotor behaviors (Overli et al. 1999; Wada 2008). Our results confirm the role of the HPI axis in memory and learning (with respect to exploratory behavior), as well locomotor behavior (movement during the aggression tests), which are important even at this early developmental stage, as soon as the fish emerge from the gravel nests.

Growth is important at this developmental stage for both behavioral and life history diversity, and the principal regulator of somatic growth is growth hormone (GH1)

(Björnsson 1997). Growth hormone (GH1) has also been associated with appetite, dominance, and anti-predator behavior in salmonids and the European eel (Schjolden and

Winberg 2007; St-Cyr and Aubin-Horth 2009). We found that GH1 has significantly 68

higher expression in anadromous females (compared with resident females), individuals with high habituation in exploratory behaviors, and individuals with high movement during the aggression experiment. Individuals that moved the most in the aggression experiment had over an 18-fold higher expression of GH1 compared with those that moved the least, corresponding with findings that growth hormone increases swimming activity in juvenile salmonids (Björnsson 1997).The more than 23-fold increase in expression of GH1 in anadromous females compared to resident females is most likely due to heritable differences in growth trajectories of the different life history types

(Thrower et al. 2004). Fry from resident parents have been shown to grow faster

(Thrower et al. 2004) and thus we would expect to have higher levels of GH1 (Björnsson

1997), yet that is not what we observe. This may be due to the proposed trade-off between growth rate and behaviors that is used to explain why natural fish populations normally grow at sub-maximal rates (Björnsson 1997).

Pro-opiomelanocortin (POMC) is another differentially regulated gene with a known central role in the HPI axis and behavior. POMC was up-regulated more than 15- fold in anadromous females, individuals with high habituation in exploration, and individuals that displayed high movement in the aggression tests. The transcription of

POMC is regulated by glucocorticoids, through a negative glucocorticoid response element proximal to the POMC promoter (Drouin et al. 1993). In our study, differences in the expression of POMC associated with habituation in exploratory behavior created a significant association with the behavioral biofunctions of operant performance and exercise. Differences in the POMC locus were associated with the ability to acquire a behavior through reinforcement, or operant performance, in mice (Grahame et al. 1998). 69

A significant association between time spent moving during the aggression experiment and the biofunction of aggressive behavior was caused by differences in POMC expression, as well as the differential expression of B2M (beta-2-microglobulin). The up- regulation of POMC in the fish that swam around the tank the most in the aggression experiment could be explained by its known involvement in the response of rainbow trout to social stress (Schjolden and Winberg 2007). The fish that were moving around the tank the most were probably the ones that were experiencing the most social stress from the mirror stimulation. POMC is also a known precursor of beta endorphin and is involved in

70 stress coping (Overli et al. 2007). In rainbow trout POMC has also been associated with feeding behavior, energy balance, and stress response (Metz et al. 2006). The overlap in differentially expressed genes among the multiple contrasts made between behavioral phenotypes in our study suggest that both growth hormone and pro-opiomelanocortin, play an important role as the proximate mechanisms underlying the diversity of behaviors observed in the early life history of these fish.

Prolactin (PRL) has long been known for its involvement in osmoregulation in fish (Manzon 2002), but in our study, differential expression of this gene was also associated with behavioral differences. PRL is a hormone, regulated by the pituitary, involved in the glucocorticoid receptor signaling pathway, in that glucocorticoids regulate the transcription of PRL. There are regions upstream of the prolactin gene that have negative or positive control of transcription, depending on which sequence in the

DNA that the glucocorticoid hormone-receptor complex binds (repression when bound to negative glucocorticoid response element and enhancement when bound to glucocorticoid response element) (Sakai et al. 1988). Prolactin is known to be released in 70

response to stress, including psychological stress like being exposed to a novel environment, with acute stress increasing prolactin levels and chronic exposure to a stressor resulting in both the behavioral and prolactin response becoming habituated

(Yelvington et al. 1985). Contrary to what we would expect, we found that PRL was up- regulated in individuals that displayed more habituation in their exploratory behaviors.

Those individuals that were more plastic or learned faster had 3.4x greater expression of

PRL in their brains. PRL is also capable of regulating Na+/K+ ATPase activity in teleosts

(Folmar and Dickhoff 1980). ATPase, Na+/K+ transporting, alpha 3 (ATP1A3) was up-

71 regulated in individuals with high dispersal activity, low habituation in aggression, high habituation in dispersal activity, and the upstream preferring AxA cross-type. The alpha 3 isoform of Na+/K+-ATPase is found exclusively in neurons. Consistent with our results, it has been shown in mice that ATP1A3 is important for spatial learning and memory, as well as being associated with locomotor activity (Moseley et al. 2007).

One of the more compelling findings in our study was that genes involved in regulation of the circadian rhythm were associated with differential gene regulation between fish with a low habituation in upstream/downstream preference (progeny produced from the anadromous cross-type) vs. high habituation in upstream/downstream preference (the resident cross-type). A circadian rhythm is an approximately 24 hour periodicity in the various biochemical and physiological processes of living beings.

Within this signaling pathway, CLOCK was down-regulated in the brain while GRINA and NR1D1 were up-regulated in the AxA cross-type with low habituation for upstream/downstream preference. In this same contrast there were seven significant behavior biofunctions: tone fear conditioning, taste aversion learning, circadian rhythm, 71

memory, olfactory memory, chaining behavior, and fasting. The six molecules associated with circadian rhythm are split with half up-regulated (TNC, NCAM1, and NR1D1) and half down-regulated (AHR, CLOCK, and CHRNB2) in individuals from cross-types with lower habituation compared to those with higher habituation in upstream/downstream preference. The two molecules with opposite expression would both decrease olfactory memory (PLAU and CHRNB2). Plasminogen activator, urokinase (PLAU) is up-regulated, while cholinergic receptor, nicotinic, beta 2 (CHRNB2) is down-regulated, in individuals from the anadromous cross-type with lower habituation for upstream/downstream

72 preference compared to the resident with more habituation. Clock is a transcription factor that has been associated with locomotor rhythms, eclosion rhythms, and cocaine sensitivity in Drosophila (Sokolowski 2001). Circadian rhythm gene polymorphisms are also related to differences in migration timing in salmon (O'Malley et al. 2007; O'Malley and Banks 2008; O'Malley et al. 2010a; O'Malley et al. 2010b). Though the timing of migration in these fish would be a few years later, our study suggest very early differential regulation of the expression of these genes in development between individuals with behavioral differences and among individuals descended from anadromous or resident parents.

Though it is impossible to discuss all differentially regulated genes, pathways, and biological functions revealed by our study, there were a few additional biological functions identified as being of interest as they relate to behavioral and life history diversity. In the comparison of fish with differing levels of exploratory behavior, the organismal biofunction of lethargy showed up as significant (driven by the up-regulation of COL10A1 (collagen, type X, alpha 1) in less exploratory fish). More exploratory fish 72

expressing lower levels of COL10A1 makes sense considering mice that have this gene knocked out showed lethargy (Gress and Jacenko 2000). In the habituation in exploration contrast, exercise showed up as a significant behavior biofunction based on gene expression differences. The lower expression of PPP1R3A (protein phosphatase 1, regulatory subunit 3A) in fish that decreased their exploratory behavior the most, fits our predictions based on an observed decreased duration of exercise in knockout mice for this gene (Aschenbach et al. 2001). Without the biofunction analyses, we would not have

73 identified COL10A1 and PPP1R3A as genes potentially responsible for a decrease in exploratory behavior.

Though unexpected so early in development, not long after these fish would have emerged from the gravel for first feeding, a biofunction found to be differentially regulated included that for ‘navigation.’ The biological function, navigation, was differentially regulated between female progeny produced from anadromous vs. resident parents, demonstrating early differences in behavior and associated gene expression with fish produced from parents with different life histories. The differential regulation of this biological function is due to the up-regulation of HOMER1 in anadromous females when compared to resident females. HOMER1 isoform expression in the brains of mice has been associated with differences in working memory and sensorimotor function

(Lominac et al. 2005). A constitutively expressed isoform of HOMER1 in the prefrontal cortex facilitates stimulates glutamate release, thereby regulating excitatory neurotransmission and mediating normal cognitive, sensorimotor, learning, and emotional processing (Lominac et al. 2005). Along with the up-regulation of HOMER1, 73

the up-regulation of POMC (discussed above) was also associated with the behavioral biofunction of place preference. In mice these genes have been shown to affect place preference and locomotor activity (Szumlinski et al. 2004; Petraschka et al. 2007). Based on their functioning in mice, the up-regulation of POMC might increase a conditioned preference for a place (Petraschka et al. 2007) and down-regulation of GRIA2 could inhibit the acquisition as well as extinction of a conditioned place preference (Cai et al.

2013) in the anadromous cross-type, while the increase in HOMER1 expression would decrease it (Szumlinski et al. 2004). While resident fish may appear to have a higher

74 preference for a place, anadromous fish use imprinting on their natal stream to later return to that location. These genes may therefore play different roles in the acquisition, maintenance, and expressed behavior of preference for the natal stream.

The goal of our study was to examine transcriptome-wide patterns of differential gene expression in juvenile fish with naturally occurring behavioral phenotypic diversity.

The results demonstrate that there are some key genes and pathways associated with the observed behavioral traits, and in many cases associated with multiple behaviors.

Neuronal signaling, function, and development play a complex role in the potential regulation of behaviors and possibly playing an early role in shaping differences between life histories. Potential hormonal regulators and their pathways (particularly glucocorticoids) also have complex effects, both on other molecules as well as behaviors.

The seemingly pleiotropic effects of these genes could possible constrain adaptation or cause selection to affect a trait other than its target. By examining differential gene expression (as well as altered biofunctions and enriched pathways), we were interested in finding out on a proximate level, what are some of the mechanisms promoting differences 74

in behavior very early in life.

2.6 Acknowledgements

I would like to acknowledge the input and collaboration of my coauthors Frank

Thrower and Krista Nichols. This work would not have been possible without considerable support at the Little Port Walter Marine Research Station, National Marine

Fisheries Service, on Baranof Island, Alaska. Tommy Abbas, Angela Feldmann, Charlie

Waters, and Kenneth Sprague were instrumental in providing technical support and aiding in behavioral trials. We are grateful for additional help in the field and lab from

Heather Holzhauer. Jessica Monjaras and Anthony Folck also facilitated lab work.

Phillip San Miguel, Rick Westerman, Allison Sorg, Viktoria Krasnyanskaya, and Paul

Parker provided support in the Purdue Genomic Core for the preparation and sequencing of transcriptome libraries. Matt Hale, Garrett McKinney and Giles Goetz were helpful with differential gene expression analyses and bioinformatics. I also thank my committee members, Dr. Esteban Fernandez-Juricic, Dr. Rick Howard, and Dr. Greg Hunt for feedback during the development of this project. This work was supported by an NSF

Career Award to KMN (NSF-DEB-0845265) and an NSF Graduate Research fellowship to ACB (DGE1333468).

Table 2-1 Summary of the contrasts that were done to find differential gene expression associated with behavioral traits Contrast Values Used for Categorization Name Comparison individual estimates for the number of Aggression aggressive displays in a mirror stimulation more vs. less aggressive Intercepts test individuals Exploration individual estimates for the latency to exit an less vs. more exploratory Intercepts isolation box individuals Moving individual estimates for the time spent more vs. less active Intercepts moving in a mirror stimulation test individuals Dispersal individual estimates for the total number of Activity more vs. less dispersive chambers visited in a dispersal test intercepts individuals individual estimates for the magnitude of Aggression change in the number of aggressive displays more vs. less habituating in Slopes in a mirror stimulation test across four trials aggression individuals individual estimates for the magnitude of Exploration change in the latency to exit an isolation box more vs. less habituating in Slopes across four trials exploration individuals Dispersal individual estimates for the magnitude of Activity change in the total number of chambers more vs. less habituating in Slopes visited in a dispersal test across four trials dispersal individuals Dispersal family estimates for the total number of Activity by more vs. less dispersive chambers visited in a dispersal test Family families Upstream/ cross-type estimates for the magnitude of the Downstream change in the proportion of chambers visited Preference less vs. more habituating in that were upstream/downstream of the

Slopes by upstream/downstream 76 starting location in a dispersal test

Cross-type preference cross-types

Table 2-2 The number of differentially expressed genes with FDR<0.05 across the contrasts. The identities of the differentially expressed genes were compared, and genes present in two or more contrasts were designated as shared, while genes that occurred in a single contrast were designated as unique. up- down- Contrast Total Shared Unique %Shared %Unique regulated regulated Aggression 390 11 379 2.82 97.18 Intercepts 60 330 Exploration 0 0 0 0 0 Intercepts 0 0 Moving 2 2 0 100 0 Intercepts 2 0 Dispersal Activity 7 7 0 100 0 intercepts 7 0 Aggression 8 8 0 100 0 Slopes 0 8 Exploration 9 3 6 33.33 66.67 Slopes 3 6 Dispersal 12 8 4 66.67 33.33 Activity Slopes 11 1 Dispersal Activity by 98 18 80 18.37 81.63 Family 60 38 Upstream/Down stream Preference 722 242 480 33.52 66.48 Slopes by Cross- type 361 361

Female vs. Male 77 0 0 0 0 0 Anadromous 0 0 Female vs. Male 1 0 1 0 100 Resident 0 1 Anadromous vs. 217 196 21 90.32 9.68 Resident Female 103 114 Anadromous vs. 67 62 5 92.54 7.46 Resident Male 27 40

Table 2-3 Summary of the number of significant results for each contrast in the IPA analysis (Fisher’s exact p<0.05). Biofunction Contrast Biofunctions Pathways Categories Aggression Intercepts 84 276 31

Exploration Intercepts 45 35 5

Moving Intercepts 74 198 80

Dispersal Activity intercepts 9 5 0

Aggression Slopes 42 43 0

Exploration Slopes 78 229 8

Dispersal Activity Slopes 74 189 8

Dispersal Activity by Family 79 273 8

Upstream/Downstream Preference 13 Slopes by Cross-type 78 309

Female vs. Male Anadromous 50 55 5

Female vs. Male Resident 74 249 26 78

Anadromous vs. Resident Female 81 274 10

Anadromous vs. Resident Male 80 281 7

a)# 9 b)# c)# d)#160000 e)#0.08 3 400 8 150000 0.06 350 140000 7 2 130000 0.04 6 300 1 120000 5 0.02 250 110000 4 0 200 100000 0 3 90000 -1 150 -0.02 2 80000

1 -2 100 70000 -0.04 f)# 0.5 g)# h)#60 i)# j)# 1 60 0.04 0 40 -0.5 40 0.02 0.5 -1 20 20 0 0 -1.5 0 0 -0.02 -2 -0.5 -20 -20 -0.04 -2.5

-3 -1 -40 -40 -0.06

Figure 2-1 Histograms and box plots of behavioral phenotypes in all individuals tested (n=108), with individuals chosen for gene expression work highlighted in gray (n=36): a) number of aggressive displays, b) latency to exit isolation box, c) moving during mirror stimulation test, d) total chambers visited during dispersal experiment, e) cross-type slopes for upstream preference, f) aggression slopes, g) exploration slopes, h) total chambers slopes, i) family intercepts for total chambers visited in dispersal experiment, j) cross-type slopes for downstream preference. The lines on the box plot correspond to the 79 median, first and third quantiles in the distribution. Whiskers drawn to the furthest point within 1.5 x interquartile range (IQR) from the box (or the upper or lower data point values if that is shorter), with potential outliers represented as disconnected points. The confidence diamond contains the mean and the upper and lower 95% of the mean. Red brackets define the densest region for the distribution.

80 80

CHAPTER 3. GENOME-WIDE ASSOCIATION STUDY (GWAS) OF ECOTYPES IN LAKE SUPERIOR BROOK TROUT, SALVELINUS FONTINALIS

3.1 Abstract

Understanding patterns of extant genetic diversity is crucial for understand not only the evolutionary history of populations, but also how populations should be managed and conserved. On a genomic level, we test the hypothesis that differences within and across polymorphic populations of brook trout (Salvelinus fontinalis) in Nipigon Bay, Ontario,

Canada, are heritable and linked to allelic variation in the genome. We also evaluate population structure using the same genetic markers. In this study, we find that life history variation in Lake Superior brook trout appears to be highly heritable, but genome wide association analysis revealed only a few single nucleotide polymorphism (SNP) loci associated with ecotype. Moreover, the population genetic signature from 900 SNP markers distributed across the genome shows existing population structure across the tributaries of Nipigon Bay that should be considering when making stocking decisions. In particular, the Cypress River seems to be the main source of the sampled coaster ecotype 81 captured in Lake Superior. Our study confirms that there is little to no genetic differentiation between life history types on a genome-wide level. By and large, the preponderant genomic signature suggests that coaster brook trout are an ecotype derived from resident populations, rather than a distinct stock. Understanding whether there is

82 genetic variation for ecotypic divergence and existing population structure has important implications for the conservation and restoration of the declining coaster ecotype.

3.2 Introduction

The brook trout (Salvelinus fontinalis) is native to eastern North America, although in most of the historically occupied habitats brook trout populations have been greatly reduced if not completely extirpated (EBJV 2006; Hudy et al. 2008). Throughout their range, brook trout are typically freshwater stream residents, but there is variation with examples of partial, facultative anadromy (Curry et al. 2010) and a potadromous

(migrating within fresh water) form in Lake Superior (Newman and Dubois 1996). These life history tactics (anadromy vs. residency) are heritable and genetically correlated with body size, even in sympatric populations (Theriault et al. 2007). Brook trout in Lake

Superior almost exclusively use shallow waters close to shores, earning them the name

“coasters” (Newman and Dubois 1996; Mucha and Mackereth 2008). Coaster brook trout are noted for their large size (Behnke and Tomelleri 2002). Comparisons of stable isotope signatures between fish caught in the lake and those captured in streams support the 82

hypothesis of relatively distinct types of brook trout that differ in habitat use and trophic ecology (Robillard et al. 2011a). There are no differences in age structure, size or condition between coasters and residents while in the stream (Kusnierz et al. 2009).

However, lake caught fish grow faster and live longer than stream caught fish and differences in length-at-age were apparent by the end of the first year of life (Robillard et al. 2011b). Stream-lake movements of coasters, likely driven by habitat requirements, may be flexible and facultative and not due to a priori growth differences (Kusnierz et al.

2009). It is thought, based on habitat availability, larger tributaries, which can hold large

83 numbers of resident brook trout, will produce fewer coasters due to potential population density pressures (D'Amelio 2004).

Across the entire native range, intact populations of brook trout exist in only 5% of watersheds and have been completely extirpated from over 20% of the subwatersheds

(EBJV 2006). Brook trout have declined and remain depressed due to overharvest, habitat degradation, and competition with introduced species (Hansen 1994). Historically found in Lake Huron, Michigan, and over a hundred stream tributaries to Lake Superior, only a handful of streams in Lake Superior currently support viable populations of coaster brook trout (Newman and Dubois 1996; Horns et al. 2003; Newman et al. 2003). Overharvest is particularly evident in the coasters of Lake Superior, which were taken by anglers in great numbers (Newman and Dubois 1996; Horns et al. 2003). Coaster brook trout were greatly reduced or eliminated from most areas of Lake Superior before scientific data about their populations could be collected (Newman and Dubois 1996). Extant populations that support the coaster ecotype currently exist only in Lake Nipigon,

Nipigon Bay and tributary streams, in streams and near shore areas of Isle Royale, and in 83

the Salmon-Trout River; this distribution is drastically reduced from the 119 tributaries to

Lake Superior that historically supported coasters (Newman and Dubois 1996; Newman et al. 2003). Current management strategies include stocking, rehabilitation of deteriorating stream habitats, and harvest regulation (Newman and Dubois 1996). Brook trout restoration effects have been largely unsuccessful in the past (Hansen 1994).

Effective methods for the conservation of the genetic diversity among the coaster brook trout under captive breeding have been proposed to improve the stocking programs

(Cooper et al. 2009).

Genetically, coaster brook trout appear to be a life history variant of brook trout derived from populations in riverine habitats (D'Amelio and Wilson 2008; Wilson et al.

2008). Across populations and studies, varying degrees of genetic similarity and divergence between resident and migratory ecotypes of brook trout have been reported

(Wilson et al. 2008). There is substantial genetic variation across Lake Superior, but not between coasters and residents living in sympatry, when looking at the mitochondrial genome (Wilson et al. 2008). In Nipigon Bay, microsatellite data suggest that sympatric migratory and resident forms have a high degree of gene flow and are more similar to each other genetically than coasters from different streams (D'Amelio and Wilson 2008).

Coasters appear to act as vectors for gene flow among riverine populations, providing genetic connectivity among allopatric tributaries (D'Amelio and Wilson 2008). This suggests that coaster and resident forms can and do readily interbreed within these populations, while maintaining regional structure among brook trout stocks (Wilson et al.

2008). However, neutral genetic markers (such as microsatellites) used in population genetic studies may not capture genetic variation linked to genes of functional 84

significance, which are more likely to be shaped by natural selection. In a meta-analysis comparing population genetic divergence with quantitative trait divergence across taxonomic groups, quantitative traits could be highly differentiated within and among populations with very low population genetic divergence (or historically high gene flow)

(Leinonen et al. 2008). Even in the presence of gene flow, there can be adaptation by natural selection on complex phenotypes.

To better understand the possible units for conservation and management, there is a need for genetic, ecological, and behavioral studies to describe past and present coaster

85 stocks in Lake Superior and compare with resident populations (Newman and Dubois

1996; Newman et al. 2003). This will help answer the question of whether coaster brook trout are genetically distinct from existing stream resident populations, or whether the genetic profile necessary for migration already exists within stream resident populations but is not being expressed (Newman et al. 2003). Understanding whether genetic variation exists for ecotypic divergence is critical in determining whether populations should be managed to explicitly conserve the genetic variation underlying coaster and resident ecotypes, or whether conservation should be focused on environmental conditions including habitat, non-native competitors, and other environmental influences.

Whether lake and stream ecotypes result from genetic polymorphism or phenotypic plasticity remains unexplored, but has important implications for the conservation and restoration of the declining coaster ecotype. The reintroduction and establishment of reproducing populations of genetically appropriate stocks in suitable areas is key to the conservation and enhancement of coaster populations on a broad scale in Lake Superior

(Newman et al. 2003). 85

We test the hypothesis that variation in ecotype within and across populations of brook trout in Nipigon Bay, Ontario, Canada, is linked to allelic variation at some loci in the genome. The brook trout used in this study have known phenotypes, coaster (lake migrant) or resident, due to previous stable isotope analysis (Coppaway 2011; Robillard et al. 2011b). We also evaluate population structure using the same markers.

Understanding the genetic differences and possible population structure of the extant populations of resident and coaster brook trout would enable better management decisions.

3.3 Materials and Methods

3.3.1 Sampling and DNA extraction

Samples used in this study were combined from three other studies (including one still in progress (Sicoly unpublished)) examining differences between coasters and residents, all conducted in Nipigon Bay (and tributaries), Lake Superior, Ontario

(Coppaway 2011; Robillard et al. 2011b; Sicoly unpublished). Nipigon Bay is located along the northernmost shore of Lake Superior and is one of the few areas within Lake

Superior where large coaster brook trout are still observed. Fish were captured in the lake as well as 10 different tributaries along a 100-km stretch of shoreline: Stillwater Creek,

Clearwater Creek, an un-named creek known as Mazukma Creek locally, Dublin Creek,

MacInnes Creek, Cypress River, Little Cypress River, Little Gravel River, Nishin Creek, and McLean’s Creek (Figure 2-1). All individuals were caught below barriers to migration to and from Lake Superior (Coppaway 2011; Robillard et al. 2011b; Sicoly unpublished). During the summers of 2005, 2006, 2008, 2009, and 2011, fish were caught through a combination of angling and electrofishing, measured for total length 86

(mm), as well as fork length (mm) and weight (g) (in some cases), and a fin clipping was taken (Coppaway 2011; Robillard et al. 2011b). The clipping was from the adipose fin, stored in an unbleached paper envelope, and was oven-dried, and used for stable isotope analysis (SIA) (Coppaway 2011; Robillard et al. 2011b). In some cases, a separate fin clip was stored in ethanol for use in genetic work. A subset of individuals also had ages calculated using otolith measurements as described in Robillard (2011a). For this study the samples used for DNA extraction consisted of a combination of these dried fin tissues, ground dried fin tissues, and fin clippings that were stored in ethanol. DNA was extracted

87 from the samples of a total of 173 fish using standard phenol:chloroform procedures

(Wasko et al. 2003).

3.3.2 RAD library construction

Restriction-site-associated DNA (RAD) libraries prepared for Illumina sequencing were produced as previously described (Miller et al. 2012; Hale et al. 2013).

A total of 500 ng of DNA was used for each sample (n=173 in total). Briefly, DNA from each individual was digested with SbfI and then barcoded with a unique 6-base sequence

5’ to the SbfI cut site. Each barcode sequence differed from any other barcode within the same library by at least two bases. Barcoded DNA from 10 or 12 samples was pooled and fragmented using the Sonic Ruptor 400 (Omni International, Kennesaw, GA) as previously described (Hale et al. 2013). Fragments were size selected between 400 and

600 bp, blunt-ended, and an A nucleotide was added to the 3’ end to which the Illumina

P2 sequencing primer was ligated. Using primers that targeted the adapter sequences, the library was amplified by polymerase chain reaction (PCR; 96°C for 3 min then 14 cycles of 96°C for 15 sec, 60°C for 15 sec, and 72°C for 30 sec followed by 5 min at 72°C), and 87 again size selected for fragment sizes 400-600 bp. Libraries were quantified using the

KAPA SYBR FAST qPCR kit (KAPA Biosystems, Woburn, MA) on the StepOne Plus

(Applied Biosystems, Foster City, CA) real-time PCR platform. Libraries were sequenced on an Illumina HiSeq 2000 or HiSeq2500 sequencer (Illumina Inc., San Diego,

CA, USA), in a single direction, with a read length of 100 bp.

3.3.3 Bioinformatics

Sequenced libraries were processed with Stacks v. 1.19 (Catchen et al. 2013).

Briefly, reads were de-multiplexed according to their barcode, trimmed to 75 bp, and

88 quality filtered with default settings in process_radtags. Single nucleotide polymorphisms (SNPs) were discovered using quality-filtered sequences from 24 individuals who had an overall quality filtered sequence depth of 2.5-3.5 million reads.

For the individuals used to build this SNP catalog, ustacks was initially run at default settings, with the exception of setting the minimum number of sequence reads to create a stack or an allele at m = 5, and using the bounded SNP model which bounded sequencing error between 0 and 0.01 for the calling of SNP genotypes. To build the catalog, these 24 individuals were used to identify polymorphic sites in the program cstacks, again using default settings with the exception of setting the number of mismatches allowed between samples to n = 2. To subsequently process data to call genotypes for all individuals in the dataset, ustacks was run for every individual at a minimum stack depth of m = 2. This allows Stacks to identify alleles within individuals that match the catalog to a minimum sequence depth of 2 sequences per allele, given the maximum likelihood for the observed genotype. The sstacks program module was then run to match individuals to the catalog for genotyping. Finally, the program populations was executed to return genotypes for 88

individuals and all loci that had a minimum depth of sequencing of 8 sequences. The populations program and vcftools (Danecek et al. 2011) were subsequently used also to filter the data for final genetic analyses, as described below.

3.3.4 SNP filtering

Polymorphic RAD loci were filtered to include those RAD-tags that only contained two haplotypes, since SNPs within a single 100 bp sequence are linked to one another.

Among the bi-allelic RAD-tags, SNPs were further filtered by removing loci that had a minor allele frequency <0.01, that had a genotyping success <60%, and that had an

89 observed heterozygosity >70% to remove over-merged paralogous loci. Finally, individuals with <60% scoring rate were removed from analyses.

3.3.5 Genome-wide association analyses

Previous stable isotope analyses (SIA) were used to confirm the existence of two types of brook trout (coasters and residents) differing in summer habitat use and trophic ecology (Robillard et al. 2011b). This was confirmed with the telemetry data from a subset of the samples used herein (Coppaway 2011). We performed a discriminant function analysis (DFA) using the stable isotope data (both δ13C and δ15N scores) for individuals to create a categorical phenotype and assign them as coasters or residents. We also performed an analysis using only the δ13C scores from the SIA as the phenotype, as these were previously shown to correspond to trophic differences (Robillard et al. 2011b).

To test for genome-wide associations between RAD-tag SNP genotypes and the phenotypes, we used the R package GAPIT with the model selection option (Lipka et al.

2012). We used a unified mixed linear model (MLM) approach, also known as a ‘Q-K’ model, which simultaneously accounts for both population structure (Q) and cryptic 89 familial relatedness (K) among individuals. Familial relatedness (K) was estimated by calculating a kinship matrix using the complete set of RAD-tag SNPs across all fish used for GWAS. The R package GAPIT (Lipka et al. 2012) was used to generate the kinship matrix using the EMMA algorithm (Kang et al. 2008). Linear mixed models can effectively improve the fit of the model by limiting the detection of false positives through the simultaneous correction for both population structure and cryptic genetic relatedness between every pair of individuals in the statistical model (Yu et al. 2006;

Fopa Fomeju et al. 2014). SNP markers with p< 0.001 were considered to be significantly

90 associated with the phenotypic trait. This same threshold has previously been used and is considered conservative in a study with a small sample size (Emebiri 2014). With a phenotype as complex as ecotype, it was more appropriate to identify plausible associations than to hold the analysis so stringent that only the largest effects can be detected.

3.3.6 Alignment of RAD-tags and annotation of genes

Polymorphic, filtered RAD-tags used in genetic analyses were aligned to both the rainbow trout (Berthelot et al. 2014) and Atlantic salmon genomes using bowtie2

(Langmead and Salzberg 2012). From the mapped location to either of the genomes, coding regions within 50 kb were identified and annotated by BLAST. In the case of the rainbow trout genome, coding region coordinates within scaffolds and their respective proteins are available at http://www.genoscope.cns.fr/trout-ggb/data/. Rainbow trout protein sequences translated from the coding sequences within 50 kb of the mapped

RAD-tag were annotated by blastp against the nr database in NCBI. In the case of the

Atlantic salmon genome, public data includes only scaffold sequences (NCBI project 90

AGKD00000000, sequence accession numbers AGKD02000001-AGKD02094355). For matches of brook trout RAD-tag sequences to these scaffolds, a 50 kb window of nucleotides on either side of the match was extracted. Genes were predicted in the full

100 kb region of the Atlantic salmon scaffolds of interest using AUGUSTUS (Stanke et al. 2008), and the training set available from the most closely related fish species,

Petromyzon marinus, from which to identify coding regions. Predicted protein sequence from the nearest gene within 50 kb was annotated using blastp against the nr database in

NCBI. In many cases, the best hit for the Atlantic salmon predicted proteins were to as

91 yet un-annotated rainbow trout genes, in which case the next closest and annotated match was reported.

3.3.7 Population genetic analyses

To evaluate population genetic diversity and structure using markers distributed widely across the genome, we further filtered data to include only markers that had a 75% scoring rate, or better, in the original dataset used for GWAS. Pairwise FST values were calculated between collections using the Weir and Cockerham (1984) method in Genepop v. 4 (Rousset 2008). To identify potential population structure in the data, FastStructure

(Raj et al. 2014) was run for values of K from 2 through 20, first using the ‘simple’ priors.

Model complexity, assessing the values of K that most likely fit the data, were evaluated with the chooseK.py program distributed with FastStructure, and then FastStructure was rerun using logistic priors in the range of K determined to best fit the data. FastStructure runs with logistic priors consisted of 100 replicate runs at each K value, and the 25 runs within each K with the best likelihood score were averaged for individual membership in clusters identified. Clumpak (http://clumpak.tau.ac.il/contact.html) was used to visualize 91

individual membership to each of these clusters at the range of K values. Finally, a subset of individuals in our study were captured in Lake Superior, and were of unknown original with respect to the stream from which they came. We used an individual assignment test (Rannala and Mountain 1997) executed in GENECLASS2 (Piry et al.

2004), to determine, among the streams that were sampled in this study, the possible assignments of fish collected in Lake Superior to their natal streams.

3.4 Results

3.4.1 Samples and phenotypes

A total of 173 fish from the 11 capture locations were sequenced at RAD-tag sites

(Table 3-1). However, 45 of these produced low-yield RAD-tag sequence data and were pruned from all analysis (Table 3-1), leaving 128 fish. The 45 individuals that were pruned had >40% missing data and thus were removed from GWAS and population genetic analyses. Of those remaining 128 fish, 4 were excluded from analysis for lack of stable isotope data, 61 were classified as resident brook trout and 63 were classified as coasters by the DFA based on stable isotope data. The phenotype assigned by the DFA was the same as the assigned ecotype of the fish caught in the field for 141 (89%) of the samples. The capture location of those 124 fish used in analyses is summarized in Table

3-1.

3.4.2 SNP filtering

Eight lanes of sequencing were conducted, resulting in a total of 630,191,985 quality-filtered reads. The average number of quality-filtered reads per individual was 92

4.36 million (range ~ 1.2 – 14.3 million) (Table C-1). Removing SNPs that with >40% missing genotypes, minor allele frequency <0.01, and >70% observed heterozygosity reduced the number of polymorphic haplotypes or tags to 4,676.

3.4.3 GWAS

Association tests for phenotypes based on SI data included analyses with two expressions of ecotype: 1) using the binary migratory vs. resident ecotype designated from the DFA, and 2) using the δ13C measures for each individual. Under the Bayesian information criterion (BIC)-based forward model selection, the model that included the

93 first principal component to account for population structure was selected in the analysis

(Figure 3-2), as well as the kinship matrix. Of the 4,676 markers analyzed, 5 were significant (p<0.001) for the analysis of the categorical phenotype based on the DFA

(Figure 3-3 and Table 3-2). Four markers were significant (p<0.001) for the analysis using the stable isotope analysis of δ13C as the phenotype and the two analyses share a single significant marker (Figure 3-4 and Table 3-2). There was one marker that was significant in both analyses. When analyzing the phenotype of ecotype (coaster versus resident, as determined by the DFA of the SI data), 89.2% of the variance is additive genetic and is therefore an estimate of heritability of this binary ecotype.

3.4.4 Alignment and annotation of RAD-tags

Of the 4,676 RAD-tags, 2,940 (62.87%) were successfully aligned to the rainbow trout genome: 1,436 (30.71%) aligned to more than a single location, 1,504 (32.16%) aligned exactly once, and 1,736 (37.13%) did not align. Of those successfully aligned

RAD-tags, 2,188 aligned to within 50 kb of a coding region (at least one gene within a scaffold). 1,197 RAD-tags mapped to locations within the coding region of genes (within 93 the boundaries of start and stop codons, inclusive of both introns and exons). Using the

Atlantic salmon genome, there was a 69.59% overall alignment rate, 1,422 (30.41%) did not align, 1,587 (33.94%) aligned only once, and 1,667 (35.65%) aligned more than once.

Since gene predictions from the Atlantic salmon genome were not available a priori, we predicted and annotated genes only for those that were of interest from the GWAS analysis from the Atlantic salmon genome. The mapping and annotation information for the RAD-tags that were significant in the GWAS are in Table 3-2. The one locus that was significant in both association tests was unable to be annotated.

3.4.5 Analysis of population structure

Based on the FastStructure results, using 906 SNP loci that passed filtering, there are between two and five populations (or clusters, K) of brook trout in these collections

(Figure 3-5). The fish captured in Lake Superior seem most similar to fish from the

Cypress River, and this is confirmed by individual assignment tests (see Table C-2).

Seventy percent of the fish captured in Lake Superior were assigned to the Cypress River, while 25% were assigned to MacInnes Creek, and 5% (n = 1 individual) was assigned to

Dublin Creek. The FastStructure plots seem to reflect the patterns seen in the FST calculations (Table 3-3), whereby those samples collected in Lake Superior seem most similar to collections made in the Cypress River (i.e. low FST). The FastStructure results and pairwise FST also indicate that the two most easterly collections from Nishin and

McLean’s Creeks were quite different from all other collections. Stillwater and

Clearwater Creeks, the two most westerly collection sites, also had notably different cluster memberships compared to most other sample sites. Table C-3 contains the pairwise FST values for all individuals sequenced across the eleven capture locations 94

using 906 loci. Table 3-3 contains the same information for the subset of individuals used in the GWAS analyses.

3.5 Discussion

The degree to which extant genetic diversity is shaped by natural selection, genetic drift, and phenotypic plasticity is an important question not only in understanding the evolutionary history of populations, but also in understanding how populations should be managed and conserved (Waples and Hendry 2008). In this study, we find that ecotypic variation appears to be highly heritable, but genome wide association analysis

95 revealed only a few loci associated with ecotype. Moreover, the population genetic signature from 900 SNP markers distributed across the genome is consistent with prior results of little to no genetic differentiation between life history types (D'Amelio and

Wilson 2008). Our study suggests, as previously described, that coaster brook trout are an ecotype derived from resident populations, rather than a distinct stock, and are possibly acting as vectors of gene flow. Although we only found a few markers significantly associated with ecotype, this study serves as a good exploratory look for regions of the genome associated with life history type and a comparison of population structure that had been previously only been examined using neutral markers.

The fact that few loci were significantly associated with ecotype could be an issue of power (both in marker number and in sample number), as well as in the genetic architecture of the trait. Though ecotype, designated by δ13C signature, has a reported heritability of 0.89 estimated from markers in this study, only four of the 4,676 loci tested were found to be significant. There were five SNPs significantly associated with being a coaster or resident (as defined by a DFA of the SIA data), and of those, we were able to 95 map three to the rainbow trout genome. When looking for the closest genes to where the

RAD-tags mapped in the rainbow trout genome and their annotations, one interesting association was found. The closest gene (~21 kb away) to one of the tags associated with ecotype is a potassium voltage-gated channel subfamily KQT member 1-like (KCNQ1), annotated from Mexican tetra (Astyanax mexicanus). Though the distance to the nearest gene may seem large with respect to possible influence of SNP variation on expression of the gene, it has been demonstrated that loci greater than 500 kb away can in linkage disequilibrium with a gene (Jung et al. 2004). When deciding the parameters used in

96 association studies, it is helpful to think about the possible linkage disequilibrium patterns around a loci. However, there can be significant variation in the distance of linkage disequilibrium depending on where you look in the genome or even within subpopulations of the same species (Lu et al. 2011). A paralog of KCNQ1 is KCNQ3, which plays a critical role in the regulation of neuronal excitability (Wang 1998).

Interestingly, there is a regulatory transcription factor binding site in the KCNQ3 gene promoter for egr-1. The immediate early gene egr-1 is a transcription factor implicated in neuronal plasticity (St-Cyr and Aubin-Horth 2009). It has been associated with behavioral transformation, such as the transition to social dominance in cichlid fish,

Astatotilapia burtoni. In this species, egr-1 expression was a product of an environmental change, in particular social opportunity, with higher expression observed in ascending males than in stable subordinate and dominant males (Burmeister et al.

2005). It is possible that a similar mechanism is responsible for behavioral differences between coaster and resident brook trout and the plasticity necessary for coasters to transition between habitats. Of course all eight of the loci identified here, especially the 96

one that was significant in both association tests, deserve further investigation as they may be linked to a gene of interest that was missed due to lack of mapping or annotation due to our short sequences, the contiguity of available reference resources, or a combination thereof.

The high heritability (89%) observed for ecotype (as defined by a DFA of the SIA data) in this study suggests that there is significant additive genetic variance within

Nipigon Bay brook trout for variation in ecotype. It is possible that it is not a single gene, but rather several regions of small effect that contribute to this complex phenotype. It has

97 been observed that the gene expression patterns of brook trout break down in the hybrid offspring of anadromous and resident sympatric parents (Mavarez et al. 2009). Perhaps a similar divergence in regulatory networks can account for the differences between coasters and residents. It is also possible that we just did not have enough samples or enough loci to find more loci associated with ecotype. If there are many loci of small effect, our sample sizes may not have the power to detect that. RAD-tag sequencing, while allowing us to sample thousands of markers across the genome, reduces the sampling of the whole genome and does not allow us to sample every possible linkage disequilibrium block in the genome. Life history traits, including migration, are complex.

Although the propensity to migrate is heritable (Theriault et al. 2007), the decision to migrate is also influenced by environmental factors (Dingle 1996). Migration is affected by habitat productivity and thus is sensitive to habitat disturbance (Näslund et al. 1993) and may be density dependent (Westley et al. 2008).

The population genetic signature observed using nearly 1000 SNP loci in this study are largely consistent with prior studies using very small subsets of loci, but this 97

study offers the additional insight into where lake-captured fish are likely to be originating from. Our population genetic analyses of 10 collection streams, and a small number of lake captured fish suggests that there are between two and five populations. A novel finding in this study is that the fish captured in Lake Superior are most similar to fish from Cypress River, suggesting that this is the main source of coasters in Nipigon

Bay. There is also an effect of geography, with the eastern most sampled creeks (Nishin and McLean’s Creeks) being different from the other sampled locations, and the western most creeks (Stillwater and Clearwater Creeks) also cluster distinctly. One reason

Clearwater Creek may be clustering so uniquely is that prior to stream rehabilitation

(about ten years ago) there were brook trout living in the stream that were isolated due to stream conditions at the mouth, which have now been fixed to allow movement in and out of the lake (Mackereth pers. comm.). Stillwater is notable because it has some stocked lakes within its catchment (Mackereth pers. comm.). Introgression of hatchery genes into wild populations appears to vary regionally and may be related to local population size, habitat integrity, and anthropogenic pressures (Wilson et al. 2008).

Despite slight differences in the divergence estimates between streams (relative FST values) from previous findings, our overall conclusions are the same, when considering all markers there appears to be no divide between coasters and residents (D'Amelio and

Wilson 2008). This confirms that coasters are not a genetically distinct lineage, but rather an ecotype derived from resident brook trout populations. Though our study suggests there is not genetic differentiation between coasters and residents, larger sample sizes and providing more coverage of markers in the genome would increase the power for such tests. 98

Understanding the genetic and environmental contributions to the phenotypic diversity observed in Lake Superior brook trout is highly relevant to the selection of management options for conservation and restoration, such as translocations or stocking in the case of heritable polymorphism versus habitat restoration if expression of the coaster ecotype is driven by environmentally induced phenotypic plasticity alone.

Although life history variation in Lake Superior brook trout does appear to be heritable, there were very few alleles associated with it in this study. There is some existing population structure across the tributaries of Nipigon Bay and this should be considering

99 when making stocking decisions. Although coaster brook trout are not genetically distinct from existing stream resident populations, it is still yet to be determined whether lake and stream ecotypes result from phenotypic plasticity, but this has important implications for the conservation and restoration of the declining coaster ecotype. With the confirmation of coasters as interdependent on resident brook trout, conservation efforts should be focused on rehabilitation of tributary systems, with focus on factors that could be limiting their production in the streams and their survival in the lake.

3.6 Acknowledgements

I would like to acknowledge the input and collaboration of my coauthors Rob

McLaughlin, Rob Mackereth, Chris Wilson, and Krista Nichols. I thank Melissa

Robillard, Clayton Coppaway, and Bill Sloan for field collections of samples. Phillip San

Miguel, Rick Westerman, Allison Sorg, Viktoria Krasnyanskaya, and Paul Parker provided support in the Purdue Genomic Core for the preparation and sequencing of

RAD libraries. Krista Nichols was invaluable for her assistance with the bioinformatics and population structure analysis. Giles Goetz facilitated annotations of RAD tags to the 99

rainbow trout and Atlantic salmon genomes. Matt Hale and Garrett McKinney were helpful with questions about library preparation and bioinformatics. This work was funded in part by the Great Lakes Fisheries Commission, and an NSF pre-doctoral fellowship to ACB.

100

Capture >60% Location Sequenced GSR Used in Analyses Clearwater 16 11 11

Cypress 30 27 27

Dublin 23 21 20

Mazukma 1 1 1

Lake Superior 33 20 17

Little Cypress 4 0 0

Little Gravel 2 0 0

McLean’s 17 7 7

MacInnes 29 25 25

Nishin 3 2 2 100

Stillwater 15 14 14

Grand Total 173 128 124

101

Table 3-2 The loci that were significantly (p<0.001) associated with δ13C values from stable isotope analysis (SIA) and/or a categorical variable created to categorize fish as lake or stream based on a discriminant function analysis (DFA) of the δ13C and δ15N SIA data. The mapping information, using both rainbow trout and Atlantic salmon, including the scaffold, chromosome, number of genes within 50 kilobases, minimum distance (bases) to a gene, the closest gene, the accession, annotation, and e-value. GWAS BKT_2130 BKT_3851 BKT_42710 BKT_47144 BKT_51132 BKT_64986 BKT_70042 BKT_78382 loci Trait δ13C δ13C DFA, δ13C DFA DFA δ13C DFA DFA RBT scaffold scaffold scaffold scaffold scaffold scaffold _6901 _609 _1405 _564 _1408 RBT chrUn chrUn22 chrUn24 chrUn5 chrUn chrom genes within 50 2 3 1 1 1

kb min dist. 1677 0 3923 28798 20893 to gene closest GSONMT00 GSONMT00 GSONMT00 GSONMT0005 GSONMT0002 gene 053176001 061700001 028320001 7796001 6870001 GB CDQ90952 CDQ73126 CDQ80926 CDQ72477 CDQ80941 accession blastx XP_007544 XP_00829769 XP_00723165 XP_007234018 accession 924 2 3 Fanconi potassium centrosom anemia voltage-gated al protein group J MOB kinase blastx channel of 164 kDa- no good protein activator 3C annotatio subfamily KQT like annotation homolog [Stegastes n member 1-like [Astyanax isoform X1 partitus] [Astyanax mexicanus] [Poecilia mexicanus] formosa] blastx 1.00E-95 8.00E-43 3.00E-154 1.00E-137 evalue SSA jcf10009635 jcf1000413 ccf1000000 jcf1000700 jcf100079106 ccf100000202 scaffold 31_0-0 715_0-0 007_0-0 925_0-0 1_0-0 9_0-0 genes

within 50 5 1 4 3 2 1 101

min dist. 0 0 71 1583 7545 13303 to gene hit XP_009290 XP_004071 CDQ58856 CDQ59553 CDQ94629 CDQ80941 accession 024 607 unconventi C-type unnamed onal unnamed unnamed unnamed mannose Ssalar hit protein myosin- protein protein protein receptor 2- descripti product XVIIIa product product product like on [Oncorhync isoform X1 [Oncorhync [Oncorhynchu [Oncorhynchu [Oryzias hus mykiss] [Danio hus mykiss] s mykiss]. s mykiss]. latipes]. rerio]. evalue 3.00E-85 0 0 0 9.00E-46 4.00E-116

rho guanine potassium putative nucleotide voltage-gated helicase exchange channel XP_007234018 mov-10-B.1 next best factor 10-like subfamily KQT with evalue of [Stegastes hit protein-like member 1-like 4e-54 partitus] [Astyanax [Astyanax XP_008278 mexicanus].XP mexicanus].XP 315 _007228553 _007231653

102

Table 3-3 Pairwise population FST value for individuals used in the GWAS analyses using their nine capture locations and 906 loci

Lake Clearwater Cypress Dublin Mazukma McLean’s MacInnes Nishin Superior Clearwater 0.0553

Cypress 0.0018 0.0789

Dublin 0.0687 0.0440 0.0893

Mazukma 0.1329 0.0901 0.1752 0.0221

McLean’s 0.1399 0.0894 0.1654 0.0339 0.0546

MacInnes 0.0121 0.0391 0.0226 0.0440 0.0453 0.0938

Nishin 0.2350 0.1603 0.2659 0.0462 0.1637 0.0424 0.1644

Stillwater 0.0588 0.0291 0.0799 0.0188 0.0351 0.0465 0.0353 0.0796

102

103

Figure 3-1 Capture locations within Nipigon Bay, Lake Superior of brook trout used in GWAS. The Unnamed creek is also known locally as Mazukma Creek.

103

104

PC2 vs. PC1 Capture Location 20 Clearwater Cypress Dublin Lake Superior 15 MacInnes Mazukma McLean's

10 Nishin Stillwater PC2 5

-5

-10 -20 -15 -10 -5 0 5 10 15 20 25 PC1

104

105

-Log Base 10 p-value 2

105

106

4.5

3.5

2.5

-Log Base 10 p-value 1.5

0.5

106

107

%Creek% %Creek% %Creek%

Dublin%Creek% Nishin Lake%Superior% Cypress%River% S,llwater%Creek% MacInnes McLean’s%Creek% Clearwater%Creek%Mazukama

108

LIST OF REFERENCES

108

LIST OF REFERENCES

Adriaenssens B, Johnsson JI. 2010. Shy trout grow faster: exploring links between personality and fitness-related traits in the wild. Behav Ecol. 22:135-143.

Aschenbach WG, Suzuki Y, Breeden K, Prats C, Hirshman MF, Dufresne SD, Sakamoto K, Vilardo PG, Steele M, Kim JH, Jing SL, Goodyear LJ, DePaoli-Roach AA. 2001. The muscle-specific protein phosphatase PP1G/R(GL)(G(M))is essential for activation of glycogen synthase by exercise. The Journal of biological chemistry. 276:39959.

Ashley PJ, Ringrose S, Edwards KL, Wallington E, McCrohan CR, Sneddon LU. 2009. Effect of noxious stimulation upon antipredator responses and dominance status in rainbow trout. Anim Behav. 77:403-410.

Aubin-Horth N, Letcher BH, Hofmann HA. 2005. Interaction of rearing environment and reproductive tactic on gene expression profiles in Atlantic salmon. J Hered. 96:261-278.

Aykanat T, Thrower FP, Heath DD. 2011. Rapid evolution of osmoregulatory function by modification of gene transcription in steelhead trout. Genetica. 139:233-242.

Barton M, Bond CE. 2007. Bond's biology of fishes. 3rd ed. Belmont, CA: Thomson. 108

Behnke RJ, Tomelleri JR. 2002. Trout and salmon of North America. Chanticleer Press ed. New York: Free Press.

Bell AM. 2007. Future directions in behavioural syndromes research. Proceedings of the Royal Society B-Biological Sciences. 274:755-761.

Bell AM, Sih A. 2007. Exposure to predation generates personality in threespined sticklebacks (Gasterosteus aculeatus). Ecol Lett. 10:828-834.

109

Bell AM, Stamps JA. 2004. Development of behavioural differences between individuals and populations of sticklebacks, Gasterosteus aculeatus. Anim Behav. 68:1339- 1348.

Benjamini Y, Hochberg Y. 1995. Controlling the false discovery rate: a practical and powerful approach to multiple testing. Journal of the Royal Statistical Society Series B (Methodological). 289-300.

Berthelot C, Brunet F, Chalopin D, Juanchich A, Bernard M, Noel B, Bento P, Da Silva C, Labadie K, Alberti A, Aury JM, Louis A, Dehais P, Bardou P, Montfort J, Klopp C, Cabau C, Gaspin C, Thorgaard GH, Boussaha M, Quillet E, Guyomard R, Galiana D, Bobe J, Volff JN, Genet C, Wincker P, Jaillon O, Roest Crollius H, Guiguen Y. 2014. The rainbow trout genome provides novel insights into evolution after whole-genome duplication in vertebrates. Nature communications. 5:3657.

Björnsson BT. 1997. The biology of salmon growth hormone: from daylight to dominance. Fish Physiol Biochem. 17:9-24.

Bjornsson BT, Stefansson SO, McCormick SD. 2011. Environmental endocrinology of salmon smoltification. Gen Comp Endocrinol. 170:290-298.

Blumstein DT, Daniel JC. 2007. Quantifying behavior the JWatcher way. Sinauer Associates. 109 Bradford MJ, Taylor GC. 1997. Individual variation in dispersal behaviour of newly emerged Chinook salmon (Oncorhynchus tshawytscha) from the upper Fraser River, British Columbia. Can J Fish Aquat Sci. 54:1585-1592.

Brown C, Laland KN, Krause J. 2006. Fish cognition and behavior. Oxford ; Ames, Iowa: Blackwell Pub.

Brunelli JP, Wertzler KJ, Sundin K, Thorgaard GH. 2008. Y-specific sequences and polymorphisms in rainbow trout and Chinook salmon. Genome. 51:739-748.

Burmeister SS, Jarvis ED, Fernald RD. 2005. Rapid behavioral and genomic responses to social opportunity. PLoS Biol. 3:e363.

110

Cai YQ, Wang W, Hou YY, Zhang Z, Xie J, Pan ZZ. 2013. Central amygdala GluA1 facilitates associative learning of opioid reward. J Neurosci. 33:1577-1588.

Carter AJ, Feeney WE, Marshall HH, Cowlishaw G, Heinsohn R. 2013. Animal personality: what are behavioural ecologists measuring? Biological reviews of the Cambridge Philosophical Society. 88:465-475.

Catchen J, Hohenlohe PA, Bassham S, Amores A, Cresko WA. 2013. Stacks: an analysis tool set for population genomics. Mol Ecol. 22:3124-3140.

Charles K, Guyomard R, Hoyheim B, Ombredane D, Baglinière J-L. 2005. Lack of genetic differentiation between anadromous and resident sympatric brown trout (Salmo trutta) in a Normandy population. Aquat Living Resour. 18:65-69.

Clarke WC, Withler RE, Shelbourn JE. 1994. Inheritance of Smolting Phenotypes in Backcrosses of Hybrid Stream-Type X Ocean-Type Chinook Salmon (Oncorhynchus-Tshawytscha). Estuaries. 17:13-25.

Cloues RK, Sather WA. 2003. Afterhyperpolarization regulates firing rate in neurons of the suprachiasmatic nucleus. The Journal of neuroscience : the official journal of the Society for Neuroscience. 23:1593.

Cole SW, Galic Z, Zack JA. 2003. Controlling false-negative errors in microarray differential expression analysis: a PRIM approach. Bioinformatics. 19:1808-1816. 110

Cooper AM, Miller LM, Kapuscinski AR. 2009. Conservation of population structure and genetic diversity under captive breeding of remnant coaster brook trout (Salvelinus fontinalis) populations. Conserv Genet. 11:1087-1093.

Coppaway C, W., 2011. Dynamics of Residency for Brok Charr (Salvelinus fontinalis) in Lake Superior Tributaries: University of Guelph.

Croy MI, Hughes RN. 1991. The role of learning and memory in the feeding behaviour of the fifteen-spined stickleback, Spinachia spinachia L. Anim Behav. 41:149-159.

111

Csillik B, Schwaller B, Mihaly A, Henzi T, Losonczi E, Knyihar-Csillik E. 2010. Upregulated expression of oncomodulin, the beta isoform of parvalbumin, in perikarya and axons in the diencephalon of parvalbumin knockout mice. Neuroscience. 165:749-757.

Curry RA, Bernatchez L, Whoriskey F, Audet C. 2010. The origins and persistence of anadromy in brook charr. Rev Fish Biol Fish. 20:557-570.

D'Amelio S. Coaster Brook Trout Stream Habitat Evaluation. Trout Unlimited Canada; 2004. Trout Unlimited Canada.

D'Amelio S, Wilson C. 2008. Genetic population structure among source populations for coaster brook trout in Nipigon Bay, Lake Superior. Trans Am Fish Soc. 137:1213-1228.

Danecek P, Auton A, Abecasis G, Albers CA, Banks E, DePristo MA, Handsaker RE, Lunter G, Marth GT, Sherry ST, McVean G, Durbin R, Genomes Project Analysis G. 2011. The variant call format and VCFtools. Bioinformatics. 27:2156-2158.

Davie A, Minghetti M, Migaud H. 2009. Seasonal variations in clock-gene expression in Atlantic salmon (Salmo salar). Chronobiol Int. 26:379-395.

Dingemanse NJ, Both C, Drent PJ, Tinbergen JM. 2004. Fitness consequences of avian

personalities in a fluctuating environment. Proceedings of the Royal Society of 111 London Series B-Biological Sciences. 271:847-852.

Dingemanse NJ, Both C, van Noordwijk AJ, Rutten AL, Drent PJ. 2003. Natal dispersal and personalities in great tits (Parus major). Proceedings of the Royal Society of London Series B-Biological Sciences. 270:741-747.

Dingemanse NJ, Reale D. 2005. Natural selection and animal personality. Behaviour. 142:1159-1184.

Dingle H. 1991. Evolutionary Genetics of Animal Migration. Am Zool. 31:253-264.

Dingle H. 1996. Migration : The Biology of Life on the Move. Cary, NC, USA: Oxford University Press, Incorporated.

112

Drouin J, Sun YL, Chamberland M, Gauthier Y, De Léan A, Nemer M, Schmidt TJ. 1993. Novel glucocorticoid receptor complex with DNA element of the hormone- repressed POMC gene. EMBO J. 12:145-156.

EBJV. Eastern Brook Trout: Status and Threats. 2006.

Elliott JM. 1993. The pattern of natural mortality throughout the life cycle in contrasting populations of brown trout, Salmo trutta L. Fisheries Research. 17:123-136.

Emebiri LC. 2014. Genetic variation and possible SNP markers for breeding wheat with low-grain asparagine, the major precursor for acrylamide formation in heat- processed products. J Sci Food Agric. 94:1422-1429.

Ensminger AL, Westneat DF. 2012. Individual and Sex Differences in Habituation and Neophobia in House Sparrows (Passer domesticus). Ethology. 118:1085-1095.

Farwell M, McLaughlin RL. 2009. Alternative foraging tactics and risk taking in brook charr (Salvelinus fontinalis). Behav Ecol. 20:913-921.

Fitzpatrick MJ, Ben-Shahar Y, Smid HM, Vet LEM, Robinson GE, Sokolowski MB. 2005. Candidate genes for behavioural ecology. Trends Ecol Evol. 20:96-104.

Fitzpatrick MJ, Feder E, Rowe L, Sokolowski MB. 2007. Maintaining a behaviour polymorphism by frequency-dependent selection on a single gene. Nature. 447:210-U215. 112

Flicek P, Ahmed I, Amode MR, Barrell D, Beal K, Brent S, Carvalho-Silva D, Clapham P, Coates G, Fairley S, Fitzgerald S, Gil L, Garcia-Giron C, Gordon L, Hourlier T, Hunt S, Juettemann T, Kahari AK, Keenan S, Komorowska M, Kulesha E, Longden I, Maurel T, McLaren WM, Muffato M, Nag R, Overduin B, Pignatelli M, Pritchard B, Pritchard E, Riat HS, Ritchie GR, Ruffier M, Schuster M, Sheppard D, Sobral D, Taylor K, Thormann A, Trevanion S, White S, Wilder SP, Aken BL, Birney E, Cunningham F, Dunham I, Harrow J, Herrero J, Hubbard TJ, Johnson N, Kinsella R, Parker A, Spudich G, Yates A, Zadissa A, Searle SM. 2013. Ensembl 2013. Nucleic Acids Res. 41:D48-55.

Folmar LC, Dickhoff WW. 1980. The Parr-Smolt Transformation (Smoltification) and Seawater Adaptation in Salmonids - a Review of Selected Literature. Aquaculture. 21:1-37.

113

Fopa Fomeju B, Falentin C, Lassalle G, Manzanares-Dauleux MJ, Delourme R. 2014. Homoeologous duplicated regions are involved in quantitative resistance of Brassica napus to stem canker. BMC Genomics. 15:498.

Froy O, Gotter AL, Casselman AL, Reppert SM. 2003. Illuminating the Circadian Clock in Monarch Butterfly Migration. Science. 300:1303-1305.

Fuiman LA, Cowan JH. 2003. Behavior and recruitment success in fish larvae: Repeatability and covariation of survival skills. Ecology. 84:53-67.

Gosling SD. 2001. From mice to men: What can we learn about personality from animal research? Psychological Bulletin. 127:45-86.

Grahame NJ, Low MJ, Cunningham CL. 1998. Intravenous self-administration of ethanol in beta-endorphin-deficient mice. Alcoholism, clinical and experimental research. 22:1093.

Grant JWA, Noakes DLG, Jonas KM. 1989. Spatial Distribution of Defence and Foraging in Young-off-the-Year Brook Charr, Salvelinus fontinalis. J Anim Ecol. 58:773-784.

Gress CJ, Jacenko O. 2000. Growth Plate Compressions and Altered Hematopoiesis in Collagen X Null Mice. The Journal of Cell Biology. 149:983-993.

Hale MC, Thrower FP, Berntson EA, Miller MR, Nichols KM. 2013. Evaluating adaptive 113

divergence between migratory and nonmigratory ecotypes of a salmonid fish, Oncorhynchus mykiss. G3. 3:1273-1285.

Hansen MJ. The State of Lake Superior in 1992. Ann Arbor, MI; 1994. p. 110.

Harding LB, Schultz IR, Goetz GW, Luckenbach JA, Young G, Goetz FW, Swanson P. 2013. High-throughput sequencing and pathway analysis reveal alteration of the pituitary transcriptome by 17alpha-ethynylestradiol (EE2) in female coho salmon, Oncorhynchus kisutch. Aquatic toxicology. 142-143:146-163.

Hecht BC, Campbell NR, Holecek DE, Narum SR. 2013. Genome-wide association reveals genetic basis for the propensity to migrate in wild populations of rainbow and steelhead trout. Mol Ecol. 22:3061-3076.

114

Hecht BC, Thrower FP, Hale MC, Miller MR, Nichols KM. 2012. Genetic architecture of migration-related traits in rainbow and steelhead trout, Oncorhynchus mykiss. G3. 2:1113-1127.

Hoar WS. 1976. Smolt Transformation - Evolution, Behavior, and Physiology. Journal of the Fisheries Research Board of Canada. 33:1233-1252.

Horns WH, Bronte CR, Busiahn TR, Ebener MP, Eshenroder RL, Gorenflo T, Kmiecik N, Mattes W, Peck JW, Petzold M, Schreiner D, R. Fish-community objectives for Lake Superior. . Ann Arbor, MI; 2003. p. 78.

Hudy M, Thieling TM, Gillespie N, Smith EP. 2008. Distribution, Status, and Land Use Characteristics of Subwatersheds within the Native Range of Brook Trout in the Eastern United States. N Am J Fish Manage. 28:1069-1085.

Jung M, Ching A, Bhattramakki D, Dolan M, Tingey S, Morgante M, Rafalski A. 2004. Linkage disequilibrium and sequence diversity in a 500-kbp region around the adh1 locus in elite maize germplasm. Theor Appl Genet. 109:681-689.

Kang HM, Zaitlen NA, Wade CM, Kirby A, Heckerman D, Daly MJ, Eskin E. 2008. Efficient control of population structure in model organism association mapping. Genetics. 178:1709-1723.

Kappeler P, (ed.). 2010. Animal behaviour: evolution and mechanisms. Berlin, London,

New York: Springer. 114

Keeley ER. 2001. Demographic Responses to Food and Space Competition by Juvenile Steelhead Trout Ecology. 82:1247-1259.

Keenleyside MHA, Yamamoto FT. 1962. Territorial Behaviour of Juvenile Atlantic Salmon (Salmo salar L.). Behaviour. 19:139-169.

Kitano J, Lema SC, Luckenbach JA, Mori S, Kawagishi Y, Kusakabe M, Swanson P, Peichel CL. 2010. Adaptive divergence in the thyroid hormone signaling pathway in the stickleback radiation. Curr Biol. 20:2124-2130.

115

Kusnierz PC, Stimmell SP, Leonard JBK. 2009. Movement and growth indicators in resident and adfluvial coaster brook trout (Salvelinus fontinalis) in the Hurricane River, Lake Superior, Michigan, USA. J Gt Lakes Res. 35:385-391.

Lahti K, Laurila A, Enberg K, Piironen J. 2001. Variation in aggressive behaviour and growth rate between populations and migratory forms in the brown trout, Salmo trutta. Anim Behav. 62:935-944.

Langmead B, Salzberg SL. 2012. Fast gapped-read alignment with Bowtie 2. Nature methods. 9:357-359.

Leinonen T, O'hara RB, Cano JM, Merila J. 2008. Comparative studies of quantitative trait and neutral marker divergence: a meta-analysis. J Evol Biol. 21:1-17.

Li B, Dewey CN. 2011. RSEM: accurate transcript quantification from RNA-Seq data with or without a reference genome. BMC bioinformatics. 12:323.

Liedvogel M, Akesson S, Bensch S. 2011. The genetics of migration on the move. Trends Ecol Evol. 26:561-569.

Lipka AE, Tian F, Wang Q, Peiffer J, Li M, Bradbury PJ, Gore MA, Buckler ES, Zhang Z. 2012. GAPIT: genome association and prediction integrated tool. Bioinformatics. 28:2397-2399.

Lohse M, Bolger AM, Nagel A, Fernie AR, Lunn JE, Stitt M, Usadel B. 2012. RobiNA: a 115

user-friendly, integrated software solution for RNA-Seq-based transcriptomics. Nucleic Acids Res. 40:W622-627.

Lominac KD, Oleson EB, Pava M, Klugmann M, Schwarz MK, Seeburg PH, During MJ, Worley PF, Kalivas PW, Szumlinski KK. 2005. Distinct roles for different Homer1 isoforms in behaviors and associated prefrontal cortex function. J Neurosci. 25:11586-11594.

Lu Y, Shah T, Hao Z, Taba S, Zhang S, Gao S, Liu J, Cao M, Wang J, Prakash AB, Rong T, Xu Y. 2011. Comparative SNP and haplotype analysis reveals a higher genetic diversity and rapider LD decay in tropical than temperate germplasm in maize. PLoS ONE. 6:e24861.

116

Lucas MD, Drew RE, Wheeler PA, Verrell PA, Thorgaard GH. 2004. Behavioral differences among rainbow trout clonal lines. Behav Genet. 34:355-365.

Mackney PA, Hughes RN. 1995. Foraging Behaviour and Memory Window in Sticklebacks. Behaviour. 132:1241-1253.

Magnhagen C. 2008. Fish behaviour. Enfield, NH: Science Publishers.

Manzon LA. 2002. The role of prolactin in fish osmoregulation: a review. Gen Comp Endocrinol. 125:291-310.

Matthews EA, Linardakis JM, Disterhoft JF. 2009. The fast and slow afterhyperpolarizations are differentially modulated in hippocampal neurons by aging and learning. J Neurosci. 29:4750-4755.

Mavarez J, Audet C, Bernatchez L. 2009. Major disruption of gene expression in hybrids between young sympatric anadromous and resident populations of brook charr (Salvelinus fontinalis Mitchill). J Evol Biol. 22:1708-1720.

McKinney GJ, 2013. Selection and Differentiation Between Populations and Sexes in Brook Trout (Salvelinus fontinalis) and Rainbow Trout (Oncorhynchus mykiss): Purdue University.

Metcalfe NB. 1986. Intraspecific Variation in Competitive Ability and Food-Intake in Salmonids - Consequences for Energy Budgets and Growth-Rates. J Fish Biol. 116

28:525-531.

Metz JR, Peters JJM, Flik G. 2006. Molecular biology and physiology of the melanocortin system in fish: A review. Gen Comp Endocrinol. 148:150-162.

Miller MR, Brunelli JP, Wheeler PA, Liu S, Rexroad CE, 3rd, Palti Y, Doe CQ, Thorgaard GH. 2012. A conserved haplotype controls parallel adaptation in geographically distant salmonid populations. Mol Ecol. 21:237-249.

Moseley AE, Williams MT, Schaefer TL, Bohanan CS, Neumann JC, Behbehani MM, Vorhees CV, Lingrel JB. 2007. Deficiency in Na,K-ATPase alpha isoform genes alters spatial learning, motor activity, and anxiety in mice. J Neurosci. 27:616- 626.

117

Mucha JM, Mackereth RW. 2008. Habitat use and movement patterns of brook trout in Nipigon Bay, Lake Superior. Trans Am Fish Soc. 137:1203-1212.

Mueller JC, Pulido F, Kempenaers B. 2011. Identification of a gene associated with avian migratory behaviour. Proceedings Biological sciences / The Royal Society. 278:2848-2856.

Nakagawa S, Schielzeth H. 2010. Repeatability for Gaussian and non-Gaussian data: a practical guide for biologists. Biological reviews of the Cambridge Philosophical Society. 85:935-956.

Näslund I, Milbrink G, Eriksson LO, Holmgren S. 1993. Importance of Habitat Productivity Differences, Competition and Predation for the Migratory Behaviour of Arctic Charr. Oikos. 66:538-546.

Näslund J, Aarestrup K, Thomassen Søren T, Johnsson Jörgen I, Post J. 2012. Early enrichment effects on brain development in hatchery-reared Atlantic salmon (Salmo salar): no evidence for a critical period. Can J Fish Aquat Sci. 69:1481- 1490.

Newman LE, Dubois RB. Status of Brook Trout in Lake Superior. Ann Arbor, Michigan; 1996.

Newman LE, DuBois RB, Halpern T, N., (EDS.). A brook trout rehabilitation plan for

Lake Superior. Ann Arbor, MI; 2003. 117

Nichols KM, Edo AF, Wheeler PA, Thorgaard GH. 2008. The genetic basis of smoltification-related traits in Oncorhynchus mykissi. Genetics. 179:1559-1575.

Niemela PT, Dingemanse NJ, Alioravainen N, Vainikka A, Kortet R. 2013. Personality pace-of-life hypothesis: testing genetic associations among personality and life history. Behav Ecol. 24:935-941.

O'Malley KG, Banks MA. 2008. A latitudinal cline in the Chinook salmon (Oncorhynchus tshawytscha) Clock gene: evidence for selection on PolyQ length variants. Proceedings Biological sciences / The Royal Society. 275:2813-2821.

118

O'Malley KG, Camara MD, Banks MA. 2007. Candidate loci reveal genetic differentiation between temporally divergent migratory runs of Chinook salmon (Oncorhynchus tshawytscha). Mol Ecol. 16:4930-4941.

O'Malley KG, Ford MJ, Hard JJ. 2010a. Clock polymorphism in Pacific salmon: evidence for variable selection along a latitudinal gradient. Proceedings Biological sciences / The Royal Society. 277:3703-3714.

O'Malley KG, McClelland EK, Naish KA. 2010b. Clock genes localize to quantitative trait loci for stage-specific growth in juvenile coho salmon, Oncorhynchus kisutch. J Hered. 101:628-632.

Overli O, Harris CA, Winberg S. 1999. Short-term effects of fights for social dominance and the establishment of dominant-subordinate relationships on brain monoamines and cortisol in rainbow trout. Brain Behav Evol. 54:263-275.

Overli O, Sorensen C, Pulman KGT, Pottinger TG, Korzan W, Summers CH, Nilsson GE. 2007. Evolutionary background for stress-coping styles: Relationships between physiological, behavioral, and cognitive traits in non-mammalian vertebrates. Neurosci Biobehav Rev. 31:396-412.

Petraschka M, Li S, Gilbert TL, Westenbroek RE, Bruchas MR, Schreiber S, Lowe J, Low MJ, Pintar JE, Chavkin C. 2007. The absence of endogenous beta-endorphin selectively blocks phosphorylation and desensitization of mu opioid receptors following partial sciatic nerve ligation. Neuroscience. 146:1795-1807. 118

Piry S, Alapetite A, Cornuet JM, Paetkau D, Baudouin L, Estoup A. 2004. GENECLASS2: a software for genetic assignment and first-generation migrant detection. J Hered. 95:536-539.

Pottinger TG, Carrick TR. 2001. Stress responsiveness affects dominant-subordinate relationships in rainbow trout. Horm Behav. 40:419-427.

Quinn TP, Myers KW. 2004. Anadromy and the marine migrations of Pacific salmon and trout: Rounsefell revisited. Rev Fish Biol Fish. 14:421-442.

Raj A, Stephens M, Pritchard JK. 2014. fastSTRUCTURE: Variational Inference of Population Structure in Large SNP Data Sets. Genetics. 197:573-589.

119

Raleigh RF. 1971. Innate Control of Migrations of Salmon and Trout Fry from Natal Gravels to Rearing Areas. Ecology. 52:291-&.

Rankin CH, Abrams T, Barry RJ, Bhatnagar S, Clayton DF, Colombo J, Coppola G, Geyer MA, Glanzman DL, Marsland S, McSweeney FK, Wilson DA, Wu CF, Thompson RF. 2009. Habituation revisited: an updated and revised description of the behavioral characteristics of habituation. Neurobiol Learn Mem. 92:135-138.

Rannala B, Mountain JL. 1997. Detecting Immigration by Using Multilocus Genotypes. Proc Natl Acad Sci U S A. 94:9197-9201.

Reale D, Festa-Bianchet M. 2003. Predator-induced natural selection on temperament in bighorn ewes. Anim Behav. 65:463-470.

Reale D, Martin J, Coltman DW, Poissant J, Festa-Bianchet M. 2009. Male personality, life-history strategies and reproductive success in a promiscuous mammal. J Evol Biol. 22:1599-1607.

Reale D, Reader SM, Sol D, McDougall PT, Dingemanse NJ. 2007. Integrating animal temperament within ecology and evolution. Biological Reviews. 82:291-318.

Rhodes JS, Quinn TP. 1998. Factors affecting the outcome of territorial contests between hatchery and naturally reared coho salmon parr in the laboratory. J Fish Biol. 53:1220-1230. 119

Riddell BE, Swain DP. 1991. Competition Between Hatchery and Wild Coho Salmon (Oncorhynchus-kisutch) - Genetic-variation for Agonistic Behavior in Newly- emerged Wild Fry Aquaculture. 98:161-172.

Robillard MM, Casselman JM, McLaughlin RL, Mackereth RW. 2011a. Alternative growth histories in populations of Lake Superior brook trout: Critical support for partial migration. Biol Conserv. 144:1931-1939.

Robillard MM, McLaughlin RL, Mackereth RW. 2011b. Diversity in Habitat Use and Trophic Ecology of Brook Trout in Lake Superior and Tributary Streams Revealed Through Stable Isotopes. Trans Am Fish Soc. 140:943-953.

120

Robinson MD, McCarthy DJ, Smyth GK. 2010. edgeR: a Bioconductor package for differential expression analysis of digital gene expression data. Bioinformatics. 26:139-140.

Robinson MD, Oshlack A. 2010. A scaling normalization method for differential expression analysis of RNA-seq data. Genome Biol. 11:R25.

Rounsefell GA. Anadromy in North American Salmonidae. 1958. p. 171-185.

Rousset F. 2008. genepop'007: a complete re-implementation of the genepop software for Windows and Linux. Molecular ecology resources. 8:103-106.

Sakai DD, Helms S, Carlstedt-Duke J, Gustafsson JA, Rottman FM, Yamamoto KR. 1988. Hormone-mediated repression: a negative glucocorticoid response element from the bovine prolactin gene. Genes Dev. 2:1144.

Schjolden J, Winberg S. 2007. Genetically determined variation in stress responsiveness in rainbow trout: Behavior and neurobiology. Brain Behav Evol. 70:227-238.

Schneider T, Thalau H-P, Semm P, Wiltschko W. 1994. Melatonin is crucial for the migratory orientation of pied flycatchers (Ficedula hypoleuca Pallas). The Journal of experimental biology. 194:255-262.

Sicoly M, Mackereth, Wilson, and Nichols. unpublished. 120

Sih A, Bell A, Johnson JC. 2004a. Behavioral syndromes: an ecological and evolutionary overview. Trends Ecol Evol. 19:372-378.

Sih A, Bell AM, Johnson JC, Ziemba RE. 2004b. Behavioral syndromes: An integrative overview. Q Rev Biol. 79:241-277.

Sloman KA, Wilson RW, Balshine S. 2006. Behaviour and physiology of fish. Amsterdam ; Boston: Elsevier Academic Press.

Sokolowski MB. 2001. Drosophila: Genetics meets behaviour. Nat Rev Genet. 2:879- 890.

121

St-Cyr S, Aubin-Horth N. 2009. Integrative and genomics approaches to uncover the mechanistic bases of fish behavior and its diversity. Comparative Biochemistry and Physiology a-Molecular & Integrative Physiology. 152:9-21.

Stanke M, Diekhans M, Baertsch R, Haussler D. 2008. Using native and syntenically mapped cDNA alignments to improve de novo gene finding. Bioinformatics. 24:637-644.

Sutherland BJ, Hanson KC, Jantzen JR, Koop BF, Smith CT. 2014. Divergent immunity and energetic programs in the gills of migratory and resident Oncorhynchus mykiss. Mol Ecol. 23:1952-1964.

Szumlinski KK, Dehoff MH, Kang SH, Frys KA, Lominac KD, Klugmann M, Rohrer J, Griffin W, 3rd, Toda S, Champtiaux NP, Berry T, Tu JC, Shealy SE, During MJ, Middaugh LD, Worley PF, Kalivas PW. 2004. Homer proteins regulate sensitivity to cocaine. Neuron. 43:401-413.

Theriault V, Garant D, Bernatchez L, Dodson JJ. 2007. Heritability of life-history tactics and genetic correlation with body size in a natural population of brook charr (Salvelinus fontinalis). J Evol Biol. 20:2266-2277.

Thrower FP, Hard JJ, Joyce JE. 2004. Genetic architecture of growth and early life- history transitions in anadromous and derived freshwater populations of steelhead. J Fish Biol. 65:286-307. 121 Vollestad LA, Quinn TP. 2003. Trade-off between growth rate and aggression in juvenile coho salmon, Oncorhynchus kisutch. Anim Behav. 66:561-568.

Wada H. 2008. Glucocorticoids: Mediators of vertebrate ontogenetic transitions. Gen Comp Endocrinol. 156:441-453.

Wang H. 1998. KCNQ2 and KCNQ3 Potassium Channel Subunits: Molecular Correlates of the M-Channel. Science. 282:1890-1893.

Waples RS, Hendry AP. 2008. EDITORIAL: Special Issue: Evolutionary perspectives on salmonid conservation and management. Evolutionary Applications. 1:183-188.

122

Wasko AP, Martins C, Oliveira C, Foresti F. 2003. Non-destructive genetic sampling in fish. An improved method for DNA extraction from fish fins and scales. Hereditas. 138:161-165.

Weir BS, Cockerham CC. 1984. Estimating F-Statistics for the Analysis of Population Structure. Evolution. 38:1358-1370.

Westley PAH, Hilborn R, Quinn TP, Ruggerone GT, Schindler DE. 2008. Long-term changes in rearing habitat and downstream movement by juvenile sockeye salmon (Oncorhynchus nerka) in an interconnected Alaska lake system. Ecol Freshw Fish. 17:443-454.

Westneat DF, Hatch MI, Wetzel DP, Ensminger AL. 2011. Individual variation in parental care reaction norms: integration of personality and plasticity. Am Nat. 178:652-667.

Wilson ADM, Stevens ED. 2005. Consistency in context-specific measures of shyness and boldness in rainbow trout, Oncorhynchus mykiss. Ethology. 111:849-862.

Wilson CC, Stott W, Miller L, D'Amelio S, Jennings MJ, Cooper AM. 2008. Conservation genetics of lake superior brook trout: Issues, questions, and directions. N Am J Fish Manage. 28:1307-1320.

Wolf M, van Doorn GS, Leimar O, Weissing FJ. 2007. Life-history trade-offs favour the

evolution of animal personalities. Nature. 447:581-584. 122

Yelvington DB, Weiss GK, Ratner A. 1985. Habituation of the prolactin response in rats to psychological stress. Psychoneuroendocrinology. 10:95-102.

Yu J, Pressoir G, Briggs WH, Vroh Bi I, Yamasaki M, Doebley JF, McMullen MD, Gaut BS, Nielsen DM, Holland JB, Kresovich S, Buckler ES. 2006. A unified mixed- model method for association mapping that accounts for multiple levels of relatedness. Nature genetics. 38:203-208.

APPENDICES 123

123

Appendix A Supplementary Material for Chapter 1

Table A-1 Summary statistics for size measurements of fish used in Chapter 1. Length (mm) and weight (g) taken before and after behavioral experiments. Growth in length and weight were also calculated for each individual. Columns include sample size, mean, standard deviation, minimum and maximum for each measure.

Variable N Mean Std Dev Minimum Maximum

initial_length 108 37.41 1.81 33.00 42.00

initial_weight 108 0.45 0.08 0.20 0.60

final_length 103 47.76 2.41 43.00 53.00

final_weight 103 1.13 0.18 0.80 1.60 growth_length 103 0.01 0.00 0.01 0.01 growth_weight 103 0.03 0.01 0.02 0.05

123

124

Table A-2 Correlation matrix for the size measurements of fish used in Chapter 1. Pearson correlations were conducted, with the correlation coefficient placed in upper cell, p-value for correlation in center, and sample size for comparison in lower cell. initial_length initial_weight final_length final_weight growth_length growth_weight

initial_length 1 0.84 0.66 0.67 -0.38 -0.30

<.0001 <.0001 <.0001 <.0001 0.00

108 108 103 103 103 103

initial_weight 1 0.52 0.59 -0.35 -0.57

<.0001 <.0001 0.00 <.0001

108 103 103 103 103

final_length 1 0.94 0.38 0.27

<.0001 <.0001 0.01

103 103 103 103

final_weight 1 0.27 0.25

0.01 0.01

103 103 103

growth_length 1 0.75

<.0001

103 103

growth_weight 1

124 103

125

Appendix B Supplementary Material for Chapter 2

Table B-1 Raw number of paired reads, total bases, and quality filtered reads generated through Illumina sequencing for each of the 36 individuals Individual Raw reads Total Bases Quality & adapter clipped reads 1-A 45229596 4568189196 44060292 1-B 44147174 4458864574 43097566 1-F 43351980 4378549980 42598012 1-H 47363440 4783707440 46193452 1-I 64567130 6521280130 63058384 10-D 46392232 4685615432 45331634 10-E 40472396 4087711996 39620970 10-F 43510436 4394554036 42375802 11-F 48348302 4883178502 47406538 11-G 46037108 4649747908 45357728 11-I 35569162 3592485362 34717342 12-G 37202606 3757463206 36313224 2-C 57936622 5851598822 56493702 2-D 43063352 4349398552 42082248 2-E 43586786 4402265386 42658274 3-B 44416786 4486095386 43294940 3-H 43137938 4356931738 42017754 3-I 44143556 4458499156 43101524 4-A 50002976 5050300576 48795730 4-E 57104740 5767578740 55627752 4-F 43340780 4377418780 42101624 4-G 43584944 4402079344 42487958 5-G 48520394 4900559794 47303756 6-A 50662848 5116947648 49607498 6-B 45775140 4623289140 44954060 125 6-E 42333160 4275649160 41643444

7-C 43151172 4358268372 42122474 7-E 73985208 7472506008 72343618 7-G 45782810 4624063810 44833566 7-I 50200256 5070225856 49594400 8-A 45116178 4556733978 44512682 8-B 41575852 4199161052 40736604 8-F 42587446 4301332046 41650472 9-B 46172694 4663442094 45057004 9-D 46928782 4739806982 45930122 9-E 48709634 4919673034 47719098

126

Table B-2 Number of paired sequence reads, aligned reads, percent alignment, and number of genes aligned to the transcriptome for each sample in Chapter 2 Individual Reads Aligned Reads Percent Alignment Genes 1-A 22030146 12967071 59 117371 1-B 21548783 12313512 57 129568 1-F 21299006 12523429 59 117108 1-H 23096726 13634441 59 108597 1-I 31529192 18503064 59 136625 10-D 22665817 12696691 56 119054 10-E 19810485 11640939 59 112860 10-F 21187901 12581253 59 93050 11-F 23703269 14679944 62 82438 11-G 22678864 13659211 60 90923 11-I 17358671 10228886 59 70349 12-G 18156612 11090430 61 75831 2-C 28246851 16810698 60 116058 2-D 21041124 12158203 58 120218 2-E 21329137 12513770 59 114968 3-B 21647470 12268061 57 117469 3-H 21008877 12044848 57 109318 3-I 21550762 12442092 58 117855 4-A 24397865 13579626 56 136325 4-E 27813876 15898776 57 134130 4-F 21050812 12110535 58 119099 4-G 21243979 11934146 56 121440 5-G 23651878 12962751 55 131820 6-A 24803749 14304515 58 135669 6-B 22477030 13092141 58 123603 126 6-E 20821722 12517530 60 116041 7-C 21061237 11970164 57 115066 7-E 36171809 21090824 58 147551 7-G 22416783 13495532 60 101898 7-I 24797200 15129578 61 123482 8-A 22256341 13128432 59 127779 8-B 20368302 11766937 58 124918 8-F 20825236 11971259 57 118632 9-B 22528502 13300791 59 121416 9-D 22965061 13440253 59 113995 9-E 23859549 13578382 57 119779

127

Appendix C Supplementary Material for Chapter 3

Table C-1 Summary information for each sample sequenced at RAD-tag sites (n=173) from Chapter 3, including: year captured, capture location (0=Clearwater, 1= Cypress, 2=Dublin, 3=Mazukma, 4=Lake Superior, 5=Little Cypress, 6=Little Gravel, 7= McLean’s, 8= MacInnes, 9=Nishin, 10=Stillwater), δ13C and δ15N scores from stable isotope analysis (SIA), categorical variable based on a discriminant function analysis of the δ13C and δ15N SIA data (0=stream, 1=lake). Lightly shaded samples were pruned from analysis for low yield sequence data and darker shaded samples were excluded for lack of stable isotope data. Sample CaptureYear CaptureLocation δ13C d15N categoricalDFA quality filtered reads BKT_1_001dil 2005 1 -21.15 9.28 1 2134633 BKT_1_430 2006 10 -24.41 5.93 0 4099736 BKT_10_338 2006 4 -21.13 6.73 1 2010846 BKT_1044_MEL190R 2005 4 -21.38 9.68 1 567902 BKT_160_444dil 2006 10 -24.6 7.37 0 1819972 BKT_161_445dil 2006 4 NaN NaN NaN 55580 BKT_163_447dil 2006 4 NaN NaN NaN 46585 BKT_18_004dil 2005 2 -23.65 6.57 0 1911295 BKT_2_434 2006 10 -25.6 6.18 0 10418797 BKT_247dil 2005 2 NaN NaN NaN 4491474 BKT_251 2005 4 NaN NaN NaN 3012558 BKT_283dil 2005 4 -19.78 9.74 1 2978011 BKT_284dil 2005 4 -20.56 9.23 1 4087889 BKT_3_340 2006 4 -20.11 8.38 1 3045324 BKT_336dil 2006 4 -18.66 7.48 1 2976707 BKT_337dil 2006 4 -19.56 8.17 1 1137966 BKT_343dil 2006 4 -20.04 8.9 1 6750395 BKT_344dil 2006 4 -21.18 7.18 1 1821267 BKT_347dil 2006 4 -20.31 9.25 1 4105866 BKT_351dil 2006 4 -18.33 8.88 1 3317808 BKT_352dil 2006 4 -17.93 8.43 1 4688690 BKT_353dil 2006 4 -20.25 8.2 1 1987228 BKT_354dil 2006 4 -19.7 8.33 1 2949341 BKT_4_335 2006 4 -19.08 8.3 1 4793280 BKT_425dil 2006 10 -25.24 5.96 0 2351375 BKT_426dil 2006 10 -26.36 7.96 0 3908738 127 BKT_428dil 2006 10 -24.74 6.88 0 1494681 BKT_429dil 2006 10 -25.75 6.33 0 2604872 BKT_432dil 2006 10 -26.38 6.49 0 1220829 BKT_433dil 2006 10 -24.43 6.38 0 3652451 BKT_435dil 2006 10 -25.83 6.56 0 1239006 BKT_436dil 2006 10 -27.15 7.26 0 936828 BKT_437dil 2006 10 -27.13 6.41 0 2478010 BKT_6_438 2006 10 -23.76 6.86 0 4083195 BKT_440dil 2006 10 -23.08 8.41 0 2293731 BKT_46_330dil 2005 4 NaN NaN NaN 333589 BKT_48_332dil 2005 4 NaN NaN NaN 234926 BKT_49_333dil 2005 4 NaN NaN NaN 3948438 BKT_5_003dil 2005 1 -22.25 8.75 0 2255991 BKT_5_5 2009 4 NaN NaN NaN 11218722

128

BKT_50_334dil 2005 4 NaN NaN NaN 527864 BKT_7_4 2009 4 NaN NaN NaN 14007339 BKT_8_2 2009 4 NaN NaN NaN 3238641 BKT_9_1 2009 4 NaN NaN NaN 11853670 BKT_Cf124 2011 1 -20.30 8.63 1 2567802 BKT_Cf127 2011 1 -16.69 9.68 1 1092681 BKT_Cf224 2011 1 -20.75 8.31 1 6235416 BKT_Cf227 2011 1 -18.18 8.68 1 1888094 BKT_Cf324 2011 1 -19.90 9.82 1 5110225 BKT_Cf327 2011 1 -18.31 8.87 1 2768058 BKT_Cf424 2011 1 -19.12 8.49 1 4512200 BKT_Cf427 2011 1 -18.80 8.95 1 2068805 BKT_Cm1027 2011 1 -20.01 9.00 1 5505410 BKT_Cm124 2011 1 -16.51 8.90 1 7417821 BKT_Cm127 2011 1 -26.61 8.85 0 5593119 BKT_Cm224 2011 1 -17.17 9.19 1 6827224 BKT_Cm227 2011 1 -19.60 10.44 1 1963783 BKT_Cm324 2011 1 -18.67 9.49 1 5192355 BKT_Cm327 2011 1 -18.08 9.06 1 2055769 BKT_Cm424 2011 1 -19.01 8.97 1 4432036 BKT_Cm427 2011 1 -19.07 9.05 1 2758932 BKT_Cm524 2011 1 -19.75 9.60 1 6022684 BKT_Cm527 2011 1 -19.97 8.85 1 4577008 BKT_Cm624 2011 1 -19.54 9.21 1 3912035 BKT_Cm627 2011 1 -18.99 9.02 1 4737724 BKT_Cm727 2011 1 -20.53 8.69 1 3955363 BKT_Cm827 2011 1 -19.91 9.22 1 2777030 BKT_Cm927 2011 1 -18.30 8.59 1 2686926 BKT_Df125 2011 2 -21.62 8.43 1 4746785 BKT_Df225 2011 2 -20.64 8.13 1 5093923 BKT_Df325 2011 2 -18.91 7.31 1 5351265 BKT_Dm125 2011 2 -22.83 5.98 0 4466626 BKT_Dm225 2011 2 -22.56 8.09 0 3202401 BKT_Dm325 2011 2 -24.85 8.00 0 5688683 BKT_Dm425 2011 2 -21.98 7.52 1 3110528 BKT_Dm525 2011 2 -23.57 7.27 0 2754125 BKT_Dm625 2011 2 -22.21 8.72 0 4700486 BKT_Dm725 2011 2 -18.95 9.50 1 3050669 BKT_Dm825 2011 2 -22.47 9.66 0 5065456 BKT_FM050 2008 0 -25.99 8.04 0 2571318 BKT_Km126 2011 3 -22.98 6.10 0 5249911

BKT_MEL007 2005 2 -23.99 6.12 0 4051138 128 BKT_MEL011 2005 2 -22.61 8.87 0 2840850 BKT_MEL044 2005 7 -23.13 7.71 0 1880893 BKT_MEL050 2005 7 -23.74 7.77 0 819932 BKT_MEL051 2005 7 -24.63 8 0 1598184 BKT_MEL052 2005 7 -22.21 8.17 0 2098063 BKT_MEL053_144 2005 7 -24.69 7.79 0 1098330 BKT_MEL054_146 2005 7 -22.63 8.12 0 2079481 BKT_MEL063_501 2005 1 -21.48 9.12 1 1839176 BKT_MEL064_503 2005 1 -21.58 8.13 1 703499 BKT_MEL092 2005 4 -21.86 9.78 1 3762009 BKT_MEL093_1047 2005 4 -21.98 10.74 1 353260 BKT_MEL099 2005 4 -19.39 8.06 1 1558862 BKT_MEL106_31 2005 2 -24.05 6.46 0 719390 BKT_MEL110_43 2005 2 -24.63 6.71 0 959570 BKT_MEL148_137 2005 7 -21.54 8.8 1 925837 BKT_MEL149_139 2005 7 -24.03 8.97 0 3271432 BKT_MEL150_141 2005 7 -22.32 8.79 0 841156 BKT_MEL151_143 2005 7 -24.36 8.01 0 1646802 BKT_MEL152_145 2005 7 -23.78 7.95 0 120760 BKT_MEL161_500 2005 5 -23.8 6.74 0 3741778 BKT_MEL162_502 2005 1 -21.85 9.41 1 10170912 BKT_MEL163_504 2005 1 -23.23 7.57 0 388592

129

BKT_MEL245DILdust 2005 4 -21.62 9.2 1 101839 BKT_MEL349DIL 2006 4 -19.96 8.78 1 14292980 BKT_MEL350DIL 2006 4 -19.5 8.8 1 9273149 BKT_MEL355DIL 2006 4 -18.73 9.04 1 13903713 BKT_MEL379_95 2006 9 -24.43 5.41 0 9845118 BKT_MEL380_96 2006 10 -26.09 6.8 0 7341496 BKT_MEL381_97 2006 8 -23.76 7.77 0 1360551 BKT_MEL382_98 2006 5 -24.72 6.85 0 197645 BKT_MEL383_99 2006 5 -24.39 7.95 0 280224 BKT_MEL384_100 2006 5 -23.15 7.86 0 801056 BKT_MEL385_101 2006 8 -24.37 8.01 0 75737 BKT_MEL386_102 2006 8 -23.7 8.23 0 19444 BKT_MEL387_103 2006 8 -23.24 8.05 0 577368 BKT_MEL388_104 2006 6 -20.71 8.17 1 38691 BKT_MEL389_105 2006 6 -22.39 7.82 0 31921 BKT_Mm126 2011 8 -23.86 8.36 0 3836066 BKT_Mm226 2011 8 -24.40 7.29 0 6108538 BKT_RM014DIL 2008 9 -23.54 5.59 0 5307234 BKT_RM015DIL 2008 9 -21.97 6.58 1 838542 BKT_RM016 2008 0 -24.79 5.96 0 4003222 BKT_RM017DIL 2008 0 -24.26 5.44 0 5559524 BKT_RM018 2008 0 -24.87 6.99 0 1568036 BKT_RM019 2008 0 -24.07 6.26 0 3918740 BKT_RM020 2008 8 -20.5 7.29 1 5100738 BKT_RM021 2008 8 -20.83 7.88 1 5495498 BKT_RM022 2008 8 -19.88 7 1 4994663 BKT_RM023 2008 8 -19.27 6.67 1 8138412 BKT_RM024 2008 8 -21.05 6.54 1 6815914 BKT_RM025 2008 8 -23.07 6.5 0 4080176 BKT_RM026 2008 8 -21.02 7.1 1 5407508 BKT_RM027 2008 8 -20.24 6.98 1 7113679 BKT_RM028 2008 8 -20.68 6.3 1 7632196 BKT_RM029 2008 8 -20.46 6.79 1 7882529 BKT_RM030 2008 8 -19.91 7.43 1 3291547 BKT_RM032 2008 8 -21.08 6.33 1 5214366 BKT_RM033 2008 8 -19.55 6.78 1 9458545 BKT_RM034 2008 8 -22.42 5.56 1 4804292 BKT_RM035 2008 8 -24.38 7.35 0 3898483 BKT_RM037 2008 8 -19.59 7.89 1 7990199 BKT_RM038 2008 8 -21.48 8.85 1 9244566 BKT_RM039 2008 8 -23.7 6.42 0 3785434

BKT_RM040 2008 2 -23.5 6.77 0 4467325 129 BKT_RM041 2008 2 -21.44 7.28 1 2085936 BKT_RM042 2008 2 -22.64 7.08 0 3343507 BKT_RM043 2008 2 -22.52 6.31 0 6941901 BKT_RM044 2008 2 -22.77 5.64 0 5080754 BKT_RM045 2008 2 -22.92 6.84 0 4596565 BKT_RM046 2008 7 -22.94 8.39 0 1622636 BKT_RM047 2008 8 -24.06 6.65 0 2342514 BKT_RM048 2008 0 -24.54 7.34 0 1876780 BKT_RM049 2008 0 -23.4 6.28 0 3745691 BKT_RM055 2008 8 -23.45 7.14 0 2484031 BKT_RM056 2008 7 -23.07 7.08 0 2643315 BKT_RM057 2008 8 -23.62 5.73 0 4572585 BKT_RM059R 2008 7 -22.49 7.55 0 1181428 BKT_RM060 2008 0 -24.21 5.17 0 767977 BKT_RM061 2008 7 -23.31 7.02 0 6652548 BKT_RM062 2008 8 -24.47 6.23 0 2787340 BKT_RM063 2008 7 -24.54 5.96 0 2130431 BKT_RM064 2008 7 -23.79 6.55 0 2351395 BKT_RM065 2008 8 -23.06 6.68 0 2261149 BKT_RM068 2008 0 -24.42 7.63 0 3340482 BKT_RM069 2008 0 -26.35 7.04 0 3565897 BKT_RM071 2008 0 -24.42 6.76 0 1879299

130

BKT_RM072 2008 0 -25.97 5.57 0 1467514 BKT_RM074 2008 0 -25.92 6.18 0 1970241 BKT_RM077 2008 0 -24.49 7.06 0 770386 BKT_RM079 2008 0 -24.77 6.89 0 3065275 BKT_RM080 2008 0 -27.06 7.68 0 2294876

130

131

Table C-2 Individual assignment tests of the fish caught in Lake Superior in Chapter 3

rank assignments assigned sample ID 1 % 2 % 10_338 Cypress River 100 251 Dublin Creek 100 283dil Cypress River 100 284dil Cypress River 100 336dil Cypress River 100 343dil MacInnes Creek 100 344dil Cypress River 100 347dil Cypress River 100 351dil MacInnes Creek 100 352dil MacInnes Creek 100 353dil Cypress River 100 354dil Cypress River 100 3_340 Cypress River 99.999 MacInnes Creek 0.001 4_335 MacInnes Creek 100 5_5 Cypress River 100 8_2 Cypress River 100 MEL092 MacInnes Creek 100 MEL349DIL Cypress River 100 MEL350DIL Cypress River 100 MEL355DIL Cypress River 100

number assigned % assigned Cypress River 14 70 Dublin Creek 1 5 131 MacInnes Creek 5 25

132

Table C-3 Pairwise population FST value for all individuals sequenced using their eleven capture locations and 906 loci from Chapter 3 Lake Little Little Superior Clearwater Cypress Dublin Mazukma Cypress Gravel McLean’s MacInnes Nishin

Clearwater 0.06

Cypress 0.0002 0.0793

Dublin 0.0690 0.0462 0.0856

Mazukma 0.1280 0.0735 0.1643 0.0164

Little Cypress 0.0890 0.0994 0.0977 0.0883 0.3662

Little Gravel 0.1294 0.0385 0.2314 0.0806 0.5000 0.25

McLean’s 0.1301 0.0804 0.1494 0.0347 0.0095 0.1174 0.0244

MacInnes 0.0130 0.0436 0.0201 0.0443 0.0460 0.0750 0.0877 0.0875

Nishin 0.1989 0.1230 0.2230 0.0396 0.0840 0.1863 0.0000 0.0270 0.1375

Stillwater 0.0527 0.0312 0.0681 0.0206 0.0321 0.0849 0.0959 0.0469 0.0307 0.0683

132

VITA

133

VITA

Ashley Chin-Baarstad Department of Biological Sciences, Purdue University West Lafayette, IN 47907

Education

B.S., Integrative Biology, 2008, University of Florida, Gainesville, Florida Ph.D., Biology, 2014, Purdue University, West Lafayette, Indiana

Work and Research Experience

2010-2013 NSF Graduate Research Fellow; Ph.D. Advisor: K. Nichols 2008-2010 Purdue Doctoral Fellow; Ph.D. Advisor: K. Nichols 2006-2008 Lab Assistant, University of Florida; supervisor: H. Klug 2007 Field Assistant, Tvärminne Zoological Station, Finland

Teaching Experience 133 2013- present Graduate Teaching Assistant, Department of Biological Sciences, Purdue

University, BIOL111, Fundamentals of Biology II 2009-2011 Teacher, Super Saturday Program, Gifted Education Resource Institute, Purdue University, Endangered (grades 1-2), Weird Science (grades 1-2) and Trip Through the Rainforest (PreK-K) 2010 Graduate Teaching Assistant, Department of Biological Sciences, Purdue University, BIOL111, Fundamentals of Biology II

134

2009 Graduate Teaching Assistant, Department of Biological Sciences, Purdue University, BIOL110, Fundamentals of Biology I

Research Presentations

Talk “Individual behavioral variation in juvenile rainbow trout, Oncorhynchus mykiss.” Animal Behavior Society Meeting, University of Colorado, July 2013 Poster “Individual behavioral variation in juvenile rainbow trout, Oncorhynchus mykiss.” Animal Behavior Society Meeting, University of New Mexico, June 2012 Poster “Individual variation for aggression in juvenile rainbow trout (Oncorhynchus mykiss). “ Gordon Research Conference: Genes & Behavior, Galveston, TX, March 2012 Poster “Should you eat your young before someone else does? Effect of Predator Presence on Filial Cannibalism in the Sand Goby.” Department of Zoology Undergraduate Symposium, University of Florida, April 2008 Poster “Should you eat your young before someone else does? Effect of Predator Presence on Filial Cannibalism in the Sand Goby.” 16th Annual Animal

Behavior Conference, Indiana University, April 2009 134

Grants

2013 Travel Grant, Women In Science Program (WISP), Purdue University 2012 Travel Grant, Midwest Crossroads Alliance for Graduate Education and the Professoriate (AGEP), Purdue University 2011 Research Grant, Midwest Crossroads Alliance for Graduate Education and the Professoriate (AGEP), Purdue University 2010-2013 National Science Foundation Graduate Research Fellowship 2008-2010 Purdue Doctoral Fellowship, Graduate School Purdue University 2009 Purdue Graduate Fellowship Incentive Award

135

2007 UF-SEAGEP International Research Experience for Undergraduates Award 2007-2008 University Scholars Award, University of Florida

Service and Mentoring

2011-2013 Biology Graduate Student Representative for the Graduate Women in Science Program, Purdue University 2010-2011 Member of the Biology Graduate Student Council, Purdue University Ecology, Evolution, and Population Biology Representative Biology Graduate Student Retreat Planner 2010-2013 Undergraduate Mentor in the Nichols Lab, Purdue University 2008-2014 Women in Science Program participant, Purdue University 2009-2011 Purdue Women’s Network Participant, Purdue University

135

PUBLICATION 136

PUBLICATION

Chin-Baarstad, A., Klug, H, & Lindström, K. 2009. Should you eat your offspring before someone else does? Effect of an egg predator on filial cannibalism in the sand goby. Animal Behaviour. 78: 203-208.