Florida State University Libraries

Electronic Theses, Treatises and Dissertations The Graduate School

2003 The Evolution of Color Patterns and Color Vision in the Bluefin Killifish, Lucania Goodei Rebecca C. Fuller

Follow this and additional works at the FSU Digital Library. For more information, please contact [email protected]

THE FLORIDA STATE UNIVERSITY

COLLEGE OF ARTS AND SCIENCES

THE EVOLUTION OF COLOR PATTERNS AND COLOR VISION IN

THE BLUEFIN KILLIFISH, LUCANIA GOODEI

By

REBECCA C. FULLER

A Dissertation submitted to the Department of Biological Science in partial fulfillment of the requirements for the degree of Doctor of Philosophy

Degree Awarded: Spring Semester, 2003

The members of the Committee approve the dissertation of Rebecca C. Fuller defended on

______Joseph Travis Professor Directing Dissertation

______David Gruender Outside Committee Member

______Jim Fadool Committee Member

______William Herrnkind Committee Member

______Don Levitan Committee Member

Approved:

______

Thomas M. Roberts, Chair, Department of Biological Science

The Office of Graduate Studies has verified and approved the above named committee members.

To my parents, Bert and Nancy Fuller,

who taught me how to write.

iii

ACKNOWLEDGEMENTS

First and foremost, I thank Joe Travis for his mentoring and support throughout this project. In many ways, Joe has been the perfect advisor giving me scientific freedom to pursue this project and also giving me tremendous amounts of help and support when needed. In addition, Joe continues to be a role model for me for the manner in which he achieves an exceedingly high level of excellence in all aspects of his work.

My committee members, David Gruender, Jim Fadool, Doc Herrnkind, Don

Levitan, and Ted Williams, have also been very helpful throughout this project. In addition to these people, I would like to thank Thomas Hansen, David Houle, Tom

Miller, and Alice Winn for their help and advice and for making FSU a stimulating place to pursue science. In this same spirit, I also thank the area 3 graduate students for their fellowship and scientific interactions.

My lab mates have been especially great throughout this project. These people include Matt Aresco, Charlie Baer, Andria Beeler, Taimi Hoag, Nate Jue, Paul Richards,

Jean Richardson, Matt Schraeder, Angie Shelton, and Brian Storz. Undergraduates helping with this project include Jessica Draughon, Meghan McNeilly, Amber Polvere,

Josh Shramo, Tommy Waltzek, and Eric Wheeler. Special thanks go to Margaret

Gunzburger, Becca Hale, Lisa Horth, and Sheryl Soucy for help in the field and for help with life in general.

iv Much of this work has been a collaboration. Chapter 5 was a collaborative study with Leo Fleishman and Manuel Leal (Union College), Ellis Loew (Cornell University), and Joe Travis. Chapters 6 and 7 were collaborative studies with Karen Carleton and

Tyrone Spadey (University of New Hampshire), Jim Fadool, and Joe Travis. I thank them all for helping me to bridge the fields of physiology, molecular genetics, and evolution.

This work was funded by a fellowship from Graduate Women in Science (to R. C.

Fuller), a National Science Foundation Dissertation Improvement Grant (to J. Travis and

R.C. Fuller), and a National Science Foundation Grant (to J. Travis). In addition, my stipend was funded by the National Science Foundation and Florida State University.

Finally, I thank my beloved husband, Jeff, who, in addition to reading countless manuscripts and catching lots of fish in the Everglades, makes me a happier and healthier person.

v

TABLE OF CONTENTS

List of Tables...... vii List of Figures...... ix Abstract...... xii

1. Introduction...... 1

2. Lighting environment predicts relative abundance of male color morphs in bluefin killifish populations...... 5

3. Multiple mating events reduce female choosiness: a model and its implications for experimental design...... 29

4. Genetics, lighting environment, and heritable responses to lighting environment affect male color morph expression in bluefin killifish, Lucania goodei...... 53

5. Intraspecific variation in ultraviolet cone production and visual sensitivity in the bluefin killifish, Lucania goodei...... 85

6. Relative opsin expression reflects population differences in vision physiology in Lucania goodei: a real-time PCR study...... 106

7. Variable sensory systems in the bluefin killifish, Lucania goodei...... 126

8. Conclusions...... 138

Literature cited...... 142

Biographical sketch...... 159

vi

LIST OF TABLES

Table 2.1. Data for male color morphs across the 30 sampled populations...... 18

Table 2.2. Summary of results for each color morph. Factors significant in initial model refers to effects that are statistically significant in the presence of all the other variables. 'Robust to removal of outlier?' refers to whether or not the significant factors in either the initial or final model still have statistically significant effects upon removal of the outlier...... 19

Table 3.1. Regression analyses between female preference and male quality for two measures of preference, latency to spawn and interspawn interval...... 44

Table 4.1. Number of animals at various life-stages for each family in the greenhouse experiment...... 65

Table 5.1. Mean and coefficient of variation (CV) of λmax calculated across individuals for the spring and swamp populations. N = 11 for spring. N = 10 for swamp. Note that CV λmax is approximately 1% for all opsin classes...... 98

Table 5.2. Mean frequencies for each cone class in each of the two populations. N = 11 for spring. N = 10 for swamp. Values in bold indicate statistically significant differences after bonferonni adjustment at P < 0.01...... 101

Table 6.1. Primers and probes...... 113

Table 6.2. Efficiencies for the spring and swamp populations. Means and standard errors are shown. N=3...... 116

Table 6.3. Repeatabilities and coefficients of variation (CV). Coefficients of variation are calculated without the far outlier. N=6 for CV (across individual means). N=18 for CV (all measures). Average CV (within samples) is the average CV across the six individuals. Average CV (within samples) represents experimental error whereas CV (individual means) represents the true coefficient of variation among individuals...... 118

vii Table 6.4. Relative cone frequency and relative opsin expression for each opsin/cone type. Means and standard errors are listed for both populations plus the grand mean. Relative cone frequency: spring N=11, swamp N=10, grand mean N=21. Relative opsin expression: spring N=10, swamp N=10, grand mean N=20...... 121

Table 7.1. Effects of sires, dams within sires, environment and the interaction between sires and environment...... 131

viii

LIST OF FIGURES

Figure 2.1. Location of sampled populations. Only sampled drainages are shown...... 10

Figure 2.2. Schematic diagram of apparatus used to measure light transmission through the water. (A) Reflectance probe in contact with test cap. (B) Reflectance probe 20 mm away from the test cap...... 11

Figure 2.3. (A) Component loading of PC1 and PC2 onto light transmission of each wavelength between 360 - 800 nm. (B) Means and standard errors of UV/blue transmission (PC2) across drainages...... 16

Figure 2.4. Factors accounting for significant amounts of variation in blue and red morph abundance. Each data point represents a population. Arrows point to outliers with high leverage. (a) Relationship between the abundance of males with solid blue anal fins and UV/blue transmission (PC2). (b-c) Proportion of males with solid red anal fins as a function of (a) UV/blue transmission (PC2) and (c) drainage. Means and standard errors are shown...... 20

Figure 2.5. Factors affecting the abundance of yellow morphs. The proportion of males with solid yellow anal fins versus (A) UV/blue transmission (PC2) and (B) overall transmission (PC1)...... 22

Figure 3.1 The relationship between adopting a threshold of quality for mate choice and (A) offspring survival, (B) time spent searching for a mate, and (C) predation risk to females...... 34

Figure 3.2. (A) Model predictions for mate choice threshold across multiple spawns. (B) Model predictions for average mate choice threshold across clutch splitting strategies...... 41

Figure 3.3. Average threshold vs. the initial threshold for mate choice after holding females prior to spawning...... 42

ix Figure 3.4. The relationship between latency to spawn and male quality in two different one-way preference tests. The left-hand panel shows the results from simulations where the initial time since ovulation at the start varied among replicates. 100 simulations were conducted for each graph. In the right-hand panel, the initial time since ovulation was set at 0 at the start of each replicate...... 45

Figure 3.5. The relationship between interspawn interval and male quality in two one- way preference tests. In the left-hand panel, time since ovulation varied at the start of each replicate. In the right-hand panel, time since ovulation was set at 0 at the start of each replicate. 100 simulations were conducted for each graph...... 46

Figure 4.1. (A) Transmission of light across filters in the laboratory experiment. (B) Transmission of light through clear vs. tea-stained water in the greenhouse experiment...... 67

Figure 4.2. Average frequency of F1 males expressing a red or yellow anal fin in laboratory crosses. Bars are standard errors. B/B-yellow pelvics sire (n = 8), Y/Y-yellow pelvics sire (n = 8), R/R-red pelvics sire (n = 8), R/B-red pelvics sire (n = 9)...... 68

Figure 4.3. Distribution of clutches expressing some element of red or yellow in laboratory crosses. Each dot is a clutch. Grey filling indicates clutches with inadequate sample sizes (< 3). The predicted female hidden phenotype is indicated. Stippled dots indicate females that were predicted to carry the red phenotype but actually expressed as yellow. Hashed dots indicate females that were predicted to carry the yellow phenotype but actually expressed as red...... 71

Figure 4.4. The proportion of males expressing (A) solid blue anal fins and expressing (B) any element of blue on the anal fin in greenhouse crosses. Open dots indicate clutches raised in clear water. x's indicate clutches raised in tea-stained water...... 75

Figure 4.5. The proportion of males expressing some element of yellow or red in greenhouse crosses. The data are pooled across lighting treatments for each combination of sire and dam...... 78

Figure 5.1. (A) Absolute sensitivity for the spring and swamp populations. Units are 1/radiance at criterion with radiance in units of µmol m-2 sr-1 s-1. (B) Relative sensitivity for the spring and swamp populations. Means and standard errors are shown. Open symbols denote spring values. Filled symbols denote swamp values. N = 8 for all spring values. N = 5 for all swamp values except for wavelength 640 nm where N = 4. * significantly different with ANOVA, p < 0.05. + significantly different with Kruskal- Wallis, p < 0.05. No comparisons are statistically significant after a sequential bonferroni adjustment. Points are jittered for the purpose of display...... 96

x Figure 5.2. Cone profiles for animals from the swamp and spring population. Each graph is a histogram of λmax cone values from a single animal. Arrows indicate missing cone classes. The total number of measured cones is indicated in each graph...... 100

Figure 6.1. Average relative expression for UV, violet, blue, yellow, and red opsins for a spring (open bars) and a swamp (dark bars) population. Means and standard errors are shown. N=10 for each population...... 119

Figure 7.1. A. Relative expression of opsins expressed in animals from a spring and a swamp population. N = 10 for each population. B. Relative expression of opsins expressed in animals raised in tea-stained and clear water treatments. N = 78 for tea- stained water. N = 80 for clear water. Means and standard errors are shown...... 130

Figure 7.2. Relative expression of (A) yellow and (B) violet opsins expressed across dams nested within sires. Means and standard errors are shown...... 133

xi

ABSTRACT

The degree to which signals and sensory systems are heritable has implications for sexual

selection. Coevolutionary models predict heritable variation in both signals and sensory

systems. Constraint models predict that signals are heritable but that sensory systems are

invariant. I test these predictions by examining variation in color patterns and vision physiology in the bluefin killifish, Lucania goodei. Males are polymorphic in anal fin coloration . Anal fins are either solid blue, solid red, solid yellow, a combination of red and blue, or a combination of yellow and blue. Lighting conditions range from crystal, clear springs (high transmission UV/blue wavelengths) to tea-stained swamps (low transmission UV/blue wavelengths). In a census study, blue morphs were more common in populations with low transmission of UV/blue. Red and yellow morphs were more common in populations with high transmission of UV/blue. A genetics study revealed that a simple epistatic interaction accounts for the majority of variation in color patterns.

Red-versus-yellow is controlled by a single locus to which blue expression is orthogonal.

Males are more likely to express blue when raised in tea-stained water. There is heritable variation in the propensity of males to express blue as well as in their response to the environment.

Across populations, vision physiology covaries with lighting conditions.

Electroretinogram readings indicated that swamp animals were less sensitive to UV/blue

xii than spring animals. Microspectrophotometry data suggested that UV and violet cones were more abundant and that yellow and red cones were less abundant in spring animals as compared to swamp animals. To examine variation within populations, I used real- time PCR to measure the expression of opsin genes that determine cone sensitivity. In a breeding experiment, I found both genetic and environmental variation in relative opsin expression. Yellow opsin varied significantly across sires, and violet opsin varied significantly across dams. Both ultravioet and violet opsin expression was higher for animals from clear-water, whereas yellow and red opsin expression was higher for animals from tea-stained water. Thus, sensory systems are dynamic, readily evolvable traits in contrast to the invariant systems assumed by the constraint models of sexual selection.

xiii

CHAPTER 1

INTRODUCTION

Animal communication (where one animal emits a signal which is detected by the sensory system of another) provides the information upon which decisions in mating, parental provisioning, predator-prey interactions, and social interactions are made. From the signaler's perspective, a signal is a means of manipulation, but from the receiver's perspective, it is a potential source of information (Johnstone 1997). For receivers not to be exploited, they must be able to respond to selection. This raises the question: is there extant genetic variation in sensory systems?

Various models of sexual selection and animal communication assume different answers to this question. Within the context of sexual selection, male secondary sex characters are signals that are detected by the sensory systems underlying female mating preferences and male/male assessment rules. As originally delineated, sensory exploitation assumes that sensory systems are conserved within clades (Basolo 1990,

Ryan1990). Males evolve specific traits that exploit female sensory systems (Ryan and

Wagner 1987, Ryan and Keddy-Hector 1992, Basolo 1995, Ryan 1998). In contrast, sensory drive includes the possibility that sensory systems evolve in relation to environmental lighting conditions (Endler 1992, 1993). Selection is ongoing and favors

1 sensory systems that can best discern signals (whether they be from conspecifics, food items, or predators) from background noise (Endler 1992, 1993). The relevance of each of these models depends on the levels of genetic variation in sensory systems.

The presence (or absence) of variation in sensory systems has important implications for the handicap signaling/good genes models of sexual selection. In contrast to the two previous models which emphasize signal design, handicap signaling models emphasize signal content (i.e. the message sent by the signal). Handicap models suggest that preferences evolve because they cause females to mate with males that increase female fitness (Zahavi 1975, Pomiankowski 1988, Iwasa and Pomiankowski

1991). Many studies have tested this hypothesis by examining the fitness consequences of mate choice (Norris 1993, Møller 1994, reviews in Andersson 1994, Johnstone 1995).

However, the fundamental distinction between this model and sensory exploitation is whether mating preferences (and their underlying neurobiology) are heritable.

The critical issue in understanding the evolution of female mating preferences is assessing whether there are appreciable levels of genetic variation in sensory systems.

Below, I review the evidence concerning the evolvability of sensory systems and present results from a series of experiments designed to ascertain the degree to which signals and sensory systems are heritable and the manner in which they covary with the environment in the bluefin killifish, Lucania goodei.

The Evidence from Vision Physiology

Vertebrate vision is accomplished when light passes into the eye through the cornea and lens, is absorbed by photopigments, and stimulates ganglion nerves. These nerve

2 impulses are carried to the brain and processed. For animal vision to evolve, there must

be genetic variation within a population in at least one of these components. Vision physiologists hold that all variation among species and populations is due to genetic

effects, and all variation within populations is due to phenotypic plasticity (Thorpe et al.

1993, Douglas and Marshall 1999, Partridge and Cummings 1999). This view may be an

artifact of methods. Most studies only consider animals from one population, usually use

a small number of animals, and/or use animals from pet stores or inbred laboratory stocks

in an effort to reduce variation in vision physiology. As a result, it is not surprising that little variation is found.

Is there reason to expect that there is detectable genetic variation in vision physiology? Obviously genetic variation in vision physiology must have existed at some point in time. Differences in vision physiology have been documented repeatedly between animals living in different sensory environments (Lythgoe 1979, Bowmaker

1990, Thorpe et al. 1993, Harosi 1996). Such differences have been found among closely related species and even between populations within a species (Lythgoe et al. 1994,

McDonald and Hawryshyn 1995, Hunt et al. 1996, Huber et al. 1997). How much genetic variation exists within populations is unclear. A few studies have shown variation in photoreceptors within populations (for an example in primates see Shyue et al. 1995). In a study of 15 guppies, Archer et al. (1987) demonstrated a polymorphism in photopigments in the 529-579 nm range. Molecular evidence suggests that large changes in photopigment sensitivity can be achieved with a small number of amino acid substitutions in opsin proteins (Yokoyama and Yokoyama 1996, Yokoyama and

Radlwimmer 1998, Yokoyama et al. 1999).

3 Despite much discussion of "adaptive mechanisms of vision" (see Archer et al.

1999), there are no quantitative measurements of how selection acts on vision physiology and little understanding of which components of vision physiology are most responsive to selection. The research outlined below attempts to address this problem by assessing the degree to which vision physiology traits are heritable and the degree to which they covary with male color patterns and lighting environment in the bluefin killifish, Lucania goodei.

4

CHAPTER 2

LIGHTING ENVIRONMENT PREDICTS RELATIVE ABUNDANCE OF MALE COLOR MORPHS IN BLUEFIN KILLIFISH POPULATIONS

Abstract

Animal communication occurs when an animal emits a signal, the signal is transmitted through the environment, and then detected by the receiver. The environment in which signaling occurs should govern the efficacy of this process. In this study, I examine the relationship of lighting environment (light transmission and tree cover), location, and the relative abundances of male color morphs across 7 drainages and 30 populations in the bluefin killifish, Lucania goodei. I found that males with blue anal fins were more common in populations with low transmission of UV and blue wavelengths. In contrast, males with red anal fins (and to a lesser extent, males with yellow anal fins) were more common in populations with high transmission of UV and blue wavelengths. High

UV/blue light transmission should create a blue visual background and may make blue males less conspicuous and red males more conspicuous to conspecifics. Color contrast with the visual background may be more important than total brightness of the color

5 pattern. These results suggest that natural selection for effective intraspecific

communication drives the relative abundance of male color morphs in different lighting

habitats.

Introduction Animal communication occurs when a signaler emits a signal, the signal is transmitted through the environment, and then detected by a receiver (Endler 1992, 1993a). The environment in which signaling occurs should have large effects on the efficacy of this process. Signals should be most easily detected when they can be fully transmitted through the environment and when they differ from the background noise against which they are emitted. In terms of visual communication, the wavelengths of a given color pattern must be transmitted through the environment and the color pattern must differ from the color pattern of the visual background. Based on these general principles, we can predict that animals living in habitats with drastically different lighting environments and signal transmission properties should use different color patterns. Two different approaches have been adopted for studying the relationship between color pattern and lighting environment. The first approach is to describe the lighting environment in which males of a given species display and ask whether those lighting conditions are a specialized subset of available lighting conditions. These studies require thorough descriptions of color, background, pattern, ambient light spectra, light transmission, and visual capabilities of signallers and receivers (Endler 1990, 1993b, Bennett et al. 1994, Grill and Rush 2000). Endler and Théry (1997) showed that lekking birds in French Guiana only display in a subset of available lighting conditions. Similarly, Endler (1991) showed that male guppies engage in courtship in lighting environments that maximize their conspicuousness to females but reduce their conspicuousness to predators.

6 The second method is the comparative approach where comparisons among species (or populations) are made with reference to the lighting environment. Interspecific comparisons seek to demonstrate differences in color pattern between species inhabiting habitats with different lighting characteristics. For example, Price (1996) showed that finch species occurring in dark, tropical forests tend to be sexually monomorphic in color pattern whereas species in lighter, temperate forests are more likely to be sexually dimorphic in color pattern. In contrast, Marchetti (1993) showed that warbler species occurring in darker habitats tended to have more brightly colored males. The comparative method can also be applied to populations. Seehausen et al. (1997) took this approach in a study of two cichlids (Haplochromis neyerei and Neochromis "velvet- black"/"blue scraper"). Across populations of H. neyerei, the redness of breeding males increases with light transmission. Similarly with Neochromis, the blueness of breeding males increases with light transmission. Hence, males of both species are more colorful in populations where such coloration is visible. In this study, I take a comparative approach and examine the relationship between lighting environment and male color patterns across several populations within a species. Specifically, I examine the correlates between light transmission, tree cover, location, and relative abundance of various male color morphs for the bluefin killifish, Lucania goodei. L. goodei is an excellent system for examining the effect of lighting habitat on signals because both the lighting habitats and the signals are highly variable. L. goodei occurs in a variety of habitats ranging from crystal clear springs to more tea-stained swamps and from populations with total canopy cover to populations with absolutely no tree cover. The color pattern among L. goodei males is highly polymorphic and the abundance of color patterns varies widely among populations.

Breeding biology of the bluefin killifish, Lucania goodei

7 The bluefin killifish, Lucania goodei, is a freshwater fundulid found throughout peninsular Florida with a few populations occurring in southeastern Georgia and South Carolina (Page and Burr 1991). The breeding system is promiscuous. Males guard patches of vegetation that serve as substrates for females to attach eggs (Fuller 2001). Gravid females spawn 10-20 eggs each day for 7-10 days. Females visit and inspect males both singly and in groups. Females release one egg per spawn and are assumed to spread their eggs among several males. There is no evidence for male parental care (Fuller and Travis 2001). Males use their dorsal and anal fins when fighting other males and also when courting females. In fights, males flare their dorsal and anal fins and engage in circle fights (Fuller 2001). One male usually wins and chases the other away. Males also use their fins in the initial stages of courting females. Females approach males. Males then swim circle loops in front of the female and also around her. If the female remains, then the male assumes a position beneath the female where he rapidly twitches his head making clicking sounds (Foster 1967). If the female remains, then she and the male swim into the vegetation and simultaneously press their bodies against the vegetation at which time an egg is released, fertilized, and deposited on the vegetation. The color pattern is dimorphic between the sexes. Males have a red dot at the base of the caudal fin. The dorsal, anal, and pelvic fins are colored in males but lack color in females. The anterior 3/4 of the dorsal fin is blue on all males. The posterior 1/4 of the dorsal fin, the pelvic fins, and the anal fin are polymorphic among males. The polymorphism on the pelvic fins and posterior dorsal fin is relatively simple. The posterior 1/4 of the dorsal fin can be blue, red, or yellow. The pelvic fins are either red or yellow. The polymorphism on the anal fin is much more complex. There are five main categories of anal fin color patterns: solid red, solid yellow, solid blue, combination of

8 red and blue, and combination of yellow and blue. The degree to which these patterns have a genetic basis is not yet known. However, in the laboratory, most adult males retain the same pattern throughout their life suggesting that not all variation in this color pattern is associated with age and/or condition.

Methods I sampled fish across 7 drainages and 30 populations in Florida (Figure 2.1). I initially relied on accounts of L. goodei populations from the University of Florida Natural History Museum, but also sampled from other sites. I focused on drainages from which there appeared a high probability of obtaining 3 or more populations. At each population, I collected L. goodei using seines and dipnets. I collected until I had obtained a reasonably large number of males (30-60) or until I could no longer catch any fish. For each male, I recorded its color pattern. I also recorded male body length (standard length in mm). For the first two populations sampled (Guaranto Springs and Bradford Springs), I only recorded body length for a fraction of the males (11 of 33 and 39 of 58 males measured respectively). For the remaining populations, I recorded body length for nearly all males. All fish were released after being processed. Fish were collected between May 21 - July 17, 1999. At each population, I took measurements of the lighting conditions beginning with tree cover. Three categories of tree cover were as follows: none (0%), low (10- 50%), high (>50%). I measured the transmission of light through the water using an Ocean Optics S2000 spectrophotometer, a tungsten light source (LS1 lamp), and a reflectance probe. I built a sampling chamber that consisted of a flat, white, plastic test cap and a white PVC circular end cap. I drilled a hole into the end cap that matched the diameter of the reflectance probe and helped keep it at a 90o angle for measurements (Figure 2.2A). The reflectance probe emits a beam of light that strikes an object and then

9

Figure 2.1. Location of sampled populations. Only sampled drainages are shown.

10 patchcord to spectrophotometer and light source

a. PVC end cap reflectance probe in contact with test cap (white) test cap (white)

patchcord to spectrophotometer and light source reflectance probe 20 mm away from test cap PVC end cap b. (white) test cap (white)

Figure 2.2. Schematic diagram of apparatus used to measure light transmission through the water. (A) Reflectance probe in contact with test cap. (B) Reflectance probe 20 mm

away from the test cap.

11 bounces back and is measured by the probe.

To measure light transmission, I set the circular PVC end cap on the white flat test

cap in a bucket of water from the habitat. I then inserted the reflectance probe through

the hole in the circular end cap until the probe touched the flat test cap (Figure 2.2A).

Here, I took a reference measurement so that I knew the maximum amount of light

coming from the light source and being detected by the probe. I then moved the

reflection probe so that it was 20 mm away from the end cap and measured the proportion

of light detected by the probe (Figure 2.2B). I made all measurements under shade to

reduce the influence of any light leaking into the sampling chamber. Although this is not

a standard measure of light transmission, it provides a relative measure of light

transmission and allows for population comparisons in environmental lighting conditions.

I measured light transmission at each population from July 12-17, 1999.

For each light transmission reading, the spectrophotometer records the proportion of light detected between 360-800 nm in 0.32 nm increments. Hence, each curve was comprised of 1363 points. I reduced the number of points in each curve and only considered the transmission at every 2 nm (360, 362, ... 800 nm). This resulted in light transmission curves with 221 points each; each location is an observation of data for these 221 variables (transmission at each of 221 wavelengths). I examined these data with a principal components analysis on correlations in order to derive a smaller number of variables that could account for most of the variance among locations in light transmission. Running the principal components analysis on covariates produced nearly identical results. Finally, I calculated the overall light transmission across all wavelengths for each population; this is the average of transmission across all of these wavelengths.

12 Usually when performing principal components analysis, one wants to have many more observations than variables going into the analysis. If the associations among the variables are somewhat loose, then the analysis can produce a spurious structure due to low sample size (Stevens 1986). However in this case, the 221 variables are tightly correlated. In fact, we are not really measuring 221 separate variables but are instead describing a continuous function with 221 data points. To confirm that the high number of variables was not a problem, I smoothed the light transmission curves in two separate manners. I created one set of light transmission curves described by 14 data points where each point on the curve was 32 nm apart. I then created a second set of light transmission curves described by 7 data points where each point on the curve was 64 nm apart. I conducted two separate principal components analyses to ascertain whether or not the results of the original analysis were affected by the number of variables. I returned to 29 of the 30 populations in December 2001 - January 2002 and measured light transmission through the water using an alternate method. The reason for measuring the light transmission again was to assure that the method described above produced similar results to the beam transmission method (Endler 1990). In the beam transmission method, a light source is placed at various distances away from a sensor which records the total amount of light detected. Regressing the log of the amount of light detected by the sensor on distance provides an attenuation coefficient which is another measure of light transmission. The beam transmission method produced data similar to the light transmission measured by the reflectance probe method used in this study. Hence, the method I used in this study adequately captured the major components of variation in light transmission across populations. To simplify the analysis, I considered only the anal fin color pattern. I described each male as having either a solid red, solid blue, solid yellow, red-blue combination, or yellow-blue combination anal fin. For Newport Springs, three males could not be placed

13 in any of these categories. For each color (red, blue, yellow), I calculated the proportion of males with a solid colored anal fin (e.g. # males with solid blue anal fin / # total males). In addition, I calculated the proportion of males with red-blue combination anal fins and the proportion with yellow-blue combination anal fins. For each of the five color morphs, I performed a backwards step-wise analysis of a general linear model to determine which variables accounted best for the variation in the relative abundance of that color morph among populations. The original model included six variables: drainage, tree score, latitude, longitude, and two principal components. Latitude and longitude plus the two principal components were continuous variables. Drainage and tree score were categorical variables and were treated as dummy variables in the model (Draper and Smith 1981, p. 241). Independent variables were removed from the model if the F-ratio for their effect was less than one. I continued removing terms until all terms had an F-ratio greater than one. This occasionally resulted in a situation where some effects were left in the model even though the p-value did not reach statistical significance (p < 0.05). My rationale for this approach is that effects should be dropped from the model if there is more variation within than among groups (F < 1). When the F-ratio is greater than 1, this suggests that the effect is accounting for some variation even if the effect is not statistically significant. In this case, it is possible that effect really does contribute to the pattern (but there is a lack of power) or that the F- ratio is greater than one simply due to random chance. In either case, leaving the effect in the model makes subsequent tests for other effects more conservative. Type three sums of squares was used in all analyses. I visually examined the residuals from the models for departures from normality. In all cases, the residuals appeared to be normally distributed. Hence, no transformations of the raw color morph frequencies were necessary. Finally, I compared male body length between drainages, populations, and color

14 morphs using a nested ANOVA. All factors were treated as fixed effects. I nested individuals within populations within drainages. Furthermore, I examined the interaction between color morph and drainage and the interaction between color morph and population nested within drainage. Provided that they were not statistically significant, I

removed the interactions from the model. For body length data, I report means + standard errors in mm. All analyses were performed with SAS statistical package version 8. All tests were considered significant at p < 0.05.

Results

Two principal components summarized over 90% of the total variation in light transmission. The first component accounted for 77.87% of variation and correlated positively onto all wavelengths (Figure 2.3A). PC1 was highly correlated with the overall light transmission for each population (r = 0.9952, p < 0.0001, n = 30) as well as with the light transmission at each measured wavelength (p < 0.01 for all correlations). For the remainder of this paper, I refer to PC1 as overall transmission. The second component accounted for 13.3% of the variation and loaded positively onto UV and blue wavelengths (Figure 2.3A). PC2 was significantly correlated with light transmission from 360 - 478 nm, but not with light transmission from 480 - 800 nm. For the remainder of this paper, I refer to PC2 as UV/blue transmission. The third and fourth principal components together accounted for most of the remaining variation (9%) but failed to predict the relative abundance of male color morphs. Principal components analyses on both the 14 data point transmission curves and the 7 data point transmission curves produced results nearly identical to the original analysis. For both analyses, the first two components accounted for > 90% of the 15 1

PC1 0.5

PC2 0 component loadings

-0.5 a. 300 400 500 600 700 800 wavelength 2

1

0

-1 UV/blue transmission (PC 2) -2

b. StJohns Wakulla Suwanee Oklawaha Everglades Okeechobee Withlacootchee

Figure 2.3. (A) Component loading of PC1 and PC2 onto light transmission of each wavelength between 360 - 800 nm. (B) Means and standard errors of UV/blue transmission (PC2) across drainages.

16 variation; the first component loaded positively onto all wavelengths, and the second component loaded positively onto the UV/blue wavelengths and negatively (although weakly) onto the longer wavelengths. Hence, the results of the principal components analyses are robust to changes in the number of variables used to describe the light transmission curves.

UV/blue transmission differed among drainges (Figure 2.3B, F6,23 = 3.012, p = 0.025). UV/blue transmission was lower in the Everglades drainage than in the Wakulla drainage. Tree cover was lower in the Everglades, Okeechobee, Oklawaha, and St. John’s drainages than in the Suwanee, Wakulla, and Withlacootchee drainages (Kruskal- wallis, X2 = 13.57, d.f. = 6, P = 0.0348; Mann-Whitney U for multiple comparisons, P < 0.05). In addition, Withlacootchee sites had higher tree cover than Suwanee and Wakulla sites (Mann-Whitney U for multiple comparisons, P < 0.05) (Siegel and Castellan 1988, p. 213). Male anal fin color patterns were highly polymorphic (Table 2.1). At least two different male anal fin color patterns were present in all populations, and most populations had three or more color patterns. Bradford Springs contained the fewest color patterns and had only males with solid red anal fins and males with solid yellow anal fins. All populations contained both males with some type of yellow anal fin (solid yellow or yellow-blue combination) and males with some type of red anal fin (solid red or red-blue combination) with the exception of Arbuckle Creek. In Arbuckle Creek, I only obtained males with solid blue, solid red, or red-blue combination anal fins. Males with solid blue anal fins were more common in populations where UV/blue light transmission was low (Table 2.2, Figure 2.4A). Throughout the step-wise process, UV/blue light transmission was statistically significant in all models (p < 0.0064 in all models, final model: F1,21 = 18.023, p = 0.0004, r = -0.742). Hence, UV/blue transmission had a large effect even when all other independent variables were included

17

Table 2.1. Data for male color morphs across the 30 sampled populations. population drainage total solid solid solid red-blue yellow- blue red yellow combo blue combo East Canal Everglades 23 16 3 1 1 2 26-Mile Bend Everglades 73 30 9 6 16 12 I-75 Reststop Everglades 33 16 3 1 7 6 Payhayokee Everglades 67 12 8 12 22 13 Tamiami Canal Everglades 50 22 9 8 8 3 Arbuckle Cr. Okeechobee 34 23 1 0 10 0 L. Marion Okeechobee 31 13 0 0 14 4 L. Okeechobee Okeechobee 46 15 8 0 17 6 L. Tohopekalagia Okeechobee 48 16 15 4 12 1 West Canal Okeechobee 16 5 0 2 5 4 Delk's Bluff, Oklawaha 34 4 10 4 8 8 Oklawaha R. L. Eaton Oklawaha 14 10 0 0 2 2 Orange L. Oklawaha 35 14 3 0 9 9 Mill Dam L. Oklawaha 9 2 4 2 1 0 Alexander Cr. St.Johns 24 1 10 10 0 3 Crow's Bluff, St.Johns 20 5 9 1 5 0 St. Johns R. Melbourne St.Johns 28 12 5 1 6 4 Ponce de Leon St.Johns 32 6 16 5 5 0 Springs Salt Springs St.Johns 26 1 21 3 1 0 Blue Springs Suwanee 43 1 25 15 2 0 Bradford Springs Suwanee 58 0 42 16 0 0 Duncans Landing, Suwanee 7 1 2 4 0 0 Santa Fe R. Guaranto Springs Suwanee 33 1 22 10 0 0 Hart Springs Suwanee 26 3 11 9 2 1 Newport Springs Wakulla 7 0 1 3 0 0 St.Marks Refuge Wakulla 33 0 11 5 15 2 Wakulla R. Wakulla 38 1 9 26 1 1 Dunnellon, Withlacootchee 23 5 7 6 2 3 Rainbow R. Hernando, Withlacootchee 25 6 6 4 6 3 Tsala Apopka L. K-P Park, Withlacootchee 49 1 10 28 3 7 Rainbow R.

18 Table 2.2. Summary of results for each color morph. Factors significant in initial model refers to effects that are statistically significant in the presence of all the other variables. 'Robust to removal of outlier?' refers to whether or not the significant factors in either the initial or final model still have statistically significant effects upon removal of the outlier.

Color Factors significant Factors significant Additional Robust to morph in initial model in final model factors removal of remaining in outlier? final model solid UV/blue UV/blue overall yes blue transmission transmission transmission drainage solid red UV/blue UV/blue overall yes transmission transmission transmission drainage drainage tree score longitude solid nothing UV/blue tree score UV/blue yellow transmission transmission overall drops from transmission model, overall transmission not significant in final model yellow- longitude longitude nothing longitude blue drainage drainage and drainage combo marginally significant in all models p≈0.08-0.05 red-blue nothing nothing UV/blue drainage combo transmission becomes drainage significant in tree score final model latitude only

19 1

0.75

0.50

0.25 solid blue anal fins blue solid anal 0 a. -4 -3 -2 -1 0 1 2 UV/blue transmission (PC2) 1

0.75

0.50

0.25 solid red anal fins anal red solid 0 -4 -3 -2 -1 0 1 2 b. UV/blue transmission (PC2) 1

0.75

0.50

0.25 solid red anal fins anal red solid

0

c. StJohns Wakulla Suwanee Oklawaha Everglades Okeechobee Withlacootchee

Figure 2.4. Factors accounting for significant amounts of variation in blue and red morph abundance. Each data point represents a population. Arrows point to outliers with high leverage. (a) Relationship between the abundance of males with solid blue anal fins and UV/blue transmission (PC2). (b-c) Proportion of males with solid red anal fins as a function of (a) UV/blue transmission (PC2) and (c) drainage. Means and standard errors are shown. 20 in the model. UV/blue transmission, overall transmission, and drainage were the only terms in the final model although the latter two terms were not statistically significant

(drainage: F6,21 = 1.806, p = 0.1465; overall transmission: F1,21 = 3.407, p = 0.0791). Removal of the leftmost outlying point did not alter the results nor the effect of UV/blue transmission (F1,20 = 9.123, p = 0.0068). Longitude, latitude, and tree score dropped from the model. The abundance of males with solid red anal fins was affected by UV/blue transmission and drainage (Table 2.2, Figure 2.4 b-c). Both terms were statistically significant in both step-wise models. Therefore, both terms had strong effects in the presence of all other independent variables. Figure 2.4B shows that males with solid red anal fins were more common in populations with high transmission of UV/blue wavelength (F1,18 = 9.94, p = 0.0055, r = 0.406). Figure 2.4C shows the mean and standard error for each drainage (F6,18 = 6.55, p = 0.0008). There were significantly higher proportions of males with solid red anal fins in the Suwanee and St. John’s drainages than there were in the Everglades and Okeechobee drainages. In addition, the proportion of males with solid red anal fins was significantly higher in the Suwanee than in the Oklawaha drainage. Overall transmission, tree score, and longitude remained in the model although none were statistically significant (overall transmission: F1,18 = 2.88, p = 0.1071; tree score: F2,18 = 2.41, p = 0.1185; longitude: F1,18 = 3.48, p = 0.0783).

Again, removal of the outlier did not alter the results for UV/blue (F1,18 = 8.15, p =

0.0105) or drainage (F6,18 = 7.90, p = 0.0003), although overall transmission and longitude did drop from the model. Both UV/blue transmission and overall transmission had positive, significant effects on the abundance of males with solid yellow anal fins. There was a positive relationship between UV/blue transmission and abundance of males with solid yellow anal fins (Table 2.2, Figure 2.5A, F1,25 = 6.965, p = 0.0141, r = 0.539). Similarly, Figure 21

1

0.75

0.50

0.25 solid yellow anal fins fins anal yellow solid

0 a. -4 -3 -2 -1 0 1 2 UV/blue transmission (PC2)

1

0.75

0.50

0.25 solid yellow anal fins fins anal yellow solid 0 -2 -1 0 1 2 3 b. overall transmission (PC1)

Figure 2.5. Factors affecting the abundance of yellow morphs. The proportion of males with solid yellow anal fins versus (A) UV/blue transmission (PC2) and (B) overall transmission (PC1).

22 2.5B shows that males with solid yellow anal fins were more common in populations with high overall light transmission (F1,25 = 10.337, p = 0.0036, r = 0.480). However, neither of these terms accounted for significant amounts of variation in the first three models. Hence, these results should be interpreted with caution. Tree score remained in the model but was not statistically significant (F2,25 = 2.712, p = 0.0859). Drainage, longitude, and latitude dropped from the model. Removal of the far outlier had a large effect on this analysis. Without the far outlier, UV/blue dropped from the model. Longitude and latitude also dropped from the model. Overall transmission, tree score, and drainage remained in the model although none accounted for significant amounts of variation. Longitude and drainage accounted for significant variation in the abundance of males with yellow-blue combination anal fins. Both factors were significant in all models, and no other factors remained in the model. Males with combination yellow- blue anal fins were less common in populations in the northern part of their range as shown by the effect of longitude (F1,22 = 6.85, p = 0.0158, r = -0.472). The abundance of yellow-blue males also differed across drainages (F6,22 = 2.80, p = 0.0356) being most common in the Oklawaha and least common in the Suwannee. Removal of the outlier resulted in a similar model selection, but the effects of longitude and drainage were marginally significant in all models (0.08 < p < 0.05). The abundance of males with red-blue combination anal fins was not significantly predicted by any of the independent variables. The step-wise analysis resulted in a model containing drainage, tree score, latitude, and UV/blue light transmission, but none of these variables accounted for a significant amount of variation, although the overall model was statistically significant. Removal of the outlier did result in a final model where drainage was statistically significant (F6,20 = 2.96, p < 0.031). Male body length differed between color morphs, drainages, and populations

23 nested within drainages (color morphs: F4,903 = 3.55, p = 0.0069; drainages: F6,903 =

117.29, p < 0.0001; populations (drainages): F23,903 = 17.79, p < 0.0001). None of the interactions were statistically significant. Hence, both were removed from the model. Males with yellow anal fins were significantly larger than males with red, red-blue combination, or blue anal fins (yellow: 29.949 + 0.284, n = 177; red: 29.101 + 0.232, n = 247; red-blue 28.805+ 0.284, n = 182; blue: 28.585 + 0.275, n = 240; yellow-blue: 29.181 + 0.455, n = 91). However, these differences were quite small. Males with yellow anal fins were only 4.7% larger than males with blue anal fins and 2.9% larger than males with red anal fins. The differences in body length among drainages were much larger. For example, Withlacootchee males were significantly larger than males from all other drainages and were 29% larger than Everglades males.

Discussion The main result to emerge from this study is that environmental lighting conditions are strongly associated with the relative abundance of male color morphs between populations. This effect occurs despite the effect of drainage, and presumably does not reflect a historical constraint. Furthermore, the fact that all populations had at least two color morphs and that all drainages had all, five color morphs suggests that a lack of genetic variation in color pattern does not cause these patterns. The most striking pattern is the relationship between UV/blue transmission and the relative abundance of male color morphs. Males with blue anal fins are more common in turbid south Florida sites with low UV/blue transmission, whereas males with red anal fins (and to a lesser extent males with yellow anal fins) are more common in clear spring populations with high UV/blue transmission (Table 2.2). At first consideration, these results seem paradoxical because they suggest that there are more

24 males with blue anal fins in populations where UV/blue wavelengths do not travel very far. Using blue in such an environment should reduce the total brightness of the color pattern (Endler 1993b). There are two potential explanations for these findings. First, blue males may actually be less conspicuous in populations with high UV/blue transmission. Although total brightness may be greater in clear water populations, the contrast between the color pattern and the background (brightness contrast) may be low (Endler 1993b). The water in populations with high UV/blue transmission frequently takes on a blue tint (e.g. Blue Springs). While males are undoubtedly at times viewed against a background of green vegetation, they are likely also to be viewed with the water column serving as a background. Blue males may not be very conspicuous in such populations because they do not differ from their visual background. In comparison, red males should be very conspicuous in such populations. Provided that conspicuousness leads to higher mating success, males with red anal fins would have a selective advantage, while males with blue anal fins would have a disadvantage in populations with high UV/blue transmission. Similar patterns have been found among two color morphs in the 3-spined stickleback, Gasterosteus aculatus. Reimchen (1989) showed that males with red throat patches were more common in clear waters while males with black throat patches were more common in stained water populations on the Queen Charlotte Archipelago, Canada. That study suggested that a red throat viewed against a blue water column in clear-water populations maximizes contrast and presumably breeding success. In stained waters, this contrast may be reduced. This pattern has been subsequently documented for other G. aculatus populations (Boughman 2001, Scott 2001). Milinski and Bakker (1991) later showed that lighting environment affects male mating success. Females preferred males with red throat patches when viewed under white light, but not when viewed under green light. This presumably occurs due to a lack of red reflectance under green light. Whether

25 or not such visual habitats occur in nature is unclear. In a slightly different aquarium experiment on guppies, Long and Houde (1989) showed that females do not exert preferences for males with orange spots when viewed under orange light. This effect occurs due to a lack of contrasting reflectance between orange spots and other body parts. Hence, high transmission of orange wavelengths do not make orange males more conspicuous (Houde 1997). The other potential explanation for the abundance of blue morphs in populations with low UV/blue transmission is that blue color elements serve as a private channel for communication. Males use their fins in signaling to conspecific males and females over short distances (< 0.5m). Blue signals may be effective over such short distances but may attenuate over longer distances such as those from which predators would view them. If this were true, the survival of males with blue anal fins would be higher in populations with low UV/blue transmission. Endler (1987, 1991) has suggested a similar phenomenon in guppies (Poecilia reticulata). Guppies modify their behavior to reduce the conspicuousness to predators and to increase their conspicuousness to females. By displaying at certain times and places, males use wavelengths that travel far enough to be effective in attracting females but should attenuate over longer distances. While this system is similar to that found in guppies (freshwater fish species, highly polymorphic color pattern), there are important differences between these two species. In guppies, the majority of variation in color patterns has been explained in terms of differences in predation regime (Endler 1978, 1980, 1982). In the guppy system, there are entire suites of predators that are present in some populations, but not in others (Reznick et al. 1990, Reznick and Travis 1996). In L. goodei, differences in predation communities most likely cannot independently account for the observed differences in color morph abundance. Large fish (bass), insects (dragonflies), and avian predators (kingfishers, herons) are present in all of the sampled populations (for fish records see

26 University of Florida Museum of Natural History database). The most divergent predator communities most likely occur in the Everglades and Okeechobee drainages where exotic species are quite common (Loftus and Kushlan 1987). Admittedly, relative predator abundance could explain some residual variation in color morph abundance. Predators in different lighting environments probably differ in which male color morphs they see most easily and can prey upon. However, differences in predator communities most likely do not independently drive the pattern demonstrated in this study. In comparison to the guppy system, the lighting environment has a much larger effect on the relative abundance of color morphs in L. goodei. Guppies may use certain lighting environments preferentially at certain times of the day (Endler 1991, Reynolds et al. 1993). Endler (1980) has demonstrated increased predation on guppies whose size of spots differs from the average size of cobble. In addition, light transmission does affect the absolute measures of conspicuousness (Endler 1991). Still, there is little research demonstrating that color morph abundances in guppies (e.g. red, blue, yellow, black, etc.) differ dramatically across populations in relation to lighting environment independent of predation regime. The large amount of variation within L. goodei populations remains unexplained. If these morphs have a genetic basis, then there must be a mechanism allowing for the maintenance of such high genetic variation. Heterogeneity in the lighting environment is one potential explanation. The transmission properties of light change with water depth (Endler 1990, 1991). Also, fishes occurring close to the bottom may be viewed against a different visual background. Females can ovoposit eggs on floating, vertical, or bottom vegetation. If females have consistent preferences for such substrates, then this would suggest that females also differ consistently in the background against which they observe males. Alternatively, temporal variation in lighting environment could play a role, although this hypothesis requires that the environment varies over a smaller time

27 scale than the lifespan / generation of the organism (Ellner and Hairston 1994). Lighting environment could vary between wet and dry season especially in non-spring populations. Populations occurring in irrigation ditches and canals would be particularly prone to temporal variation in lighting environment. In addition other mechanisms such as frequency dependent selection through female mate choice or male/male competition (Andersson, 1994), mutation-selection balance, and/or heterozygote advantage could be important in the maintenance of this polymorphism (Lynch and Walsh, 1998). These alternatives remain to be tested. Finally, these results raise the obvious question of whether the vision physiology of L. goodei also varies among populations. Sensory exploitation and sensory bias hypotheses would suggest that these differences in color morphs represent the solutions adopted by males to stimulate the same receiver physiology under different environmental conditions (Basolo 1990, 1991, Ryan 1991). In contrast, coevolutionary models of female mating preferences and male traits rely on evolvable preferences and, therefore, evolvable sensory physiology (Iwasa and Pomiankowski 1994, Widemo and Saether 1999). It is possible that vision physiology has been adapted to local conditions and male color patterns evolve to match vision physiology. In this scenario, the signaling and receiving systems co-evolve to maximize communication under a wide variety of lighting conditions. I am currently investigating whether population and genetic variation is present in L. goodei sensory systems to resolve these issues.

28

CHAPTER 3

MULTIPLE MATING EVENTS REDUCE FEMALE CHOOSINESS: A MODEL

AND ITS IMPLICATIONS FOR EXPERIMENTAL DESIGN

Abstract

In recent years, the evolution of female mating preferences has received considerable theoretical attention. Most models focus on mate choice (and the associated costs) at a single mating event. However, most animals mate more than once, and females must exercise mate choice multiple times. Here, I examine the effects of multiple mating events on intersexual selection using a dynamic programming model. In this model, I manipulate the number of mating events while keeping lifetime fecundity constant. The model predicts that the number of decisions females make should be negatively correlated with overall choosiness provided that choice is costly. Females should be particularly risk-averse early in the mating sequence. If correct, these dynamics could make the experimental detection of preferences a challenging task. In particular, dichotomous choice tests should drastically underestimate female choosiness for females that must make many choices. These results suggest that experiments that measure

29 female choice over multiple mating events and that do not disrupt the timing of mating will provide the most accurate measurements of preference.

Introduction

In recent years, the evolution of female mating preferences has received considerable empirical and theoretical attention (Andersson 1994, Johnstone 1995, Widemo and

Sæther 1999). Two different groups of models have developed for predicting female choosiness. A large group of models focuses on female behavior at individual mating events (Janetos 1980, Real 1990, 1991, Getty 1995, Luttbeg 1996, Wiegmann and

Mukhopadhay 1998, Wiegmann et al. 1999). By examining the effects of various costs and benefits, models can predict how choosy females should be and which search rules they should use at a single mating event. In contrast, only a few models have taken a life- history approach and considered multiple mating events (Crowley et al. 1990, Warner

1998). How choosy should a female be at a mating event given the potential for future mating opportunities?

A life-history approach to mate choice was first adopted by Crowley et al. (1990) who modeled female choice across three mating cycles (which they envisioned as seasons) and found that choosiness increased as the seasons progressed. Based on life- history theory, Warner (1998) proposed in a verbal model that highly iteroparous species should not engage in mate assessment to the same degree as less iteroparous species.

Despite these fine endeavors, a number of theoretical issues remain unresolved. First,

Crowley et al. (1990) only examined female choice across three mating seasons. This is a small number of choices. Whether or not the temporal pattern of female choice differs

30 across a larger number of choices is unknown. Second, rigorous theoretical comparisons of mate choice between animals required to make different numbers of choices have not been made. Crowley et al. (1990) did not compare female choice between animals that disperse their reproductive effort across multiple mating events (hereafter referred to as clutch splitting) with animals that release their eggs in one mating event (hereafter referred to as clutch clumping). Finally, whether or not standard assays of mate choice are effective in measuring preference for females that must make multiple decisions is completely unknown. In this chapter, I explore the theoretical implications of multiple mating events (or in other words, having to choose a mate several times) on the strength of female preferences by combining a threshold model of mate choice (Real 1990, 1991) with a life-history model to predict optimal female choosiness across several mating events.

The number of mating decisions that females must make differs widely across populations and species. This variation has two main sources: variation in reproductive lifespan (i.e. number of breeding seasons) and variation in how reproductive effort is dispersed within a breeding season. Variation in reproductive lifespan is well documented (Roff 1992, Stearns 1992), and ranges from animals with a single breeding season (i.e. salmon) to long-lived animals with several breeding seasons (i.e. cod, turtles).

Animals also differ in how they allocate reproductive effort within a breeding season. Many animals with sperm storage need only mate once each season (i.e. lekking birds) (Höglund and Alatalo 1995). In contrast, some animals need to mate multiple times. This is particularly true for some external fertilizers that disperse their clutches among multiple spawnings. For example, in the rainbow darter, Etheostoma caeruleum,

31 females become gravid multiple times over the breeding season. Each time they become gravid they ovulate 14-60 eggs (Heins et al. 1996). Females can spawn as many as 300 eggs over the breeding season, but only release 8 eggs per spawn (Fuller 1998a). Thus, a female must spawn multiple times to fertilize and disperse her eggs. Fractional spawning is quite common among fishes (Breder and Rosen 1966) and drastically increases the number of mating decisions that females must make.

The design of the following model is based on a scenario common to many animals and is parameterized with the life-history of the bluefin killifish (Lucania goodei). L. goodei is a fractional spawner. Females in spawning condition will spawn every day for approximately two weeks (Breder and Rosen 1966). Each day females will ovulate 10-16 eggs. Females usually release one egg per spawn (occasionally two).

Females must spawn multiple times each day to distribute their eggs. Females also have the opportunity to choose their mates. Under field conditions, females visit males who court them (Fuller 2001). Females frequently reject males (or at least swim away).

There appears to be no male harassment or coercion of females, and there have been no reports of sneakers in this system (Foster 1967, Fuller 2001).

The Model

This model asks what is the optimal level of female choosiness for a female at a given time and mating event? I use a threshold model of mate choice developed for optimal choice at a single mating event (Real 1990, 1991) and link the costs and benefits of choices across multiple mating events using a dynamic programming model. I assume that females choose among males following a threshold model of choice in which

32 females set a threshold for the quality of a mate and reject all males below the threshold. Above the threshold, they accept all males and spawn. Females have complete knowledge of the distribution of male qualities and accurate perception of quality for individual males. Females can increase the survival of their offspring by choosing high quality males. Males vary in quality (q) ranging from 0 to 10 which translates into offspring survival by a factor of 10 (offspring survival = q/10). Hence, quality 10 males confer offspring survival of 1, and quality 0 males confer offspring survival of 0. In comparison to good-genes models (Pomiankowski 1987, Iwasa and Pomiankowski 1999), I do not assume that male quality has a genetic basis, only that males differ in their suitability as mates. This phenotypic effect is well known in natural systems and can result from variation in filial cannibalism rates (Petersen 1990, Kvarnemo et al. 1998), variation in male ability to guard eggs (Downhower et al. 1987), and variation in male ability to acquire a quality territory (Alatalo et al. 1986). In L. goodei, such high variation in offspring survival is a reasonable assumption because filial egg cannibalism rates are quite high (Fuller and Travis 2001). The distribution of male qualities is normal and sums to one (mean=5, standard deviation = 1.67. This creates a situation where 99.7% of the probability mass is in the range of 0 to 10 (Sokal and Rohlf 1995). A female that sets her threshold at a given level receives the average offspring survival expected from mating with males above that level (Figure 3.1A). For example, a threshold of zero corresponds to offspring survival of 0.5, ) which is the average across all males. The benefits of adopting a threshold of q are expressed as q=10 [( f (q))(q /10)] ∑) ) =qq b q)( = q=10 (1) f q)( ∑) =qq

33 1 0.9 0.8 0.7

0.6

0.5 0.4 Offspring survival 0246810 Quality threshold

50

40

30

20

Time search 10

0 0246810 Quality threshold

1

0.8

0.6

0.4

0.2 Predation risk 0 0246810 Quality threshold

Figure 3.1 The relationship between adopting a threshold of quality for mate choice and (A) offspring survival, (B) time spent searching for a mate, and (C) predation risk to females.

34 where f(q) represents the relative frequency of males of quality q in the population.

There are two costs to being choosy. Increased choosiness entails increased predation risk on the female and decreased egg viability. Both of these affects are mediated through the time spent searching for males. Following Real (1990, 1991), the time spent searching (ts) for a mate is the inverse of the proportion of acceptable males in

) q=10 the population ( ts q)( = /1 ) f q)( ). For a threshold of quality 0, the time spent ∑ =qq searching is 1 because 100% of the males are acceptable (Figure 3.1B). For a threshold of quality 5, the time spent searching is 1.64 because 61% of the males in the population are acceptable (1.64 =1/0.61). High quality males are very rare. The time spent searching for the very best males approaches infinity (Figure 3.1B).

I assume that predation risk increases as search times increase. This is a common assumption (Real 1990, 1991) with a modicum of direct empirical support (Reynolds and

Côté 1995, Grafe 1997). Many studies have demonstrated that females (and males in sex-role reversed systems) curtail search behaviors under increased predation risk suggesting that prolonged search increases mortality (Forsgren 1992, Magurran and

Seghers 1994, Fuller and Berglund 1996). There have been no experimental demonstrations of costly mate search in killifish, although Lafferty and Morris (1995) showed that parasitized pacific killifish (Fundulus parvipinnis) engage in more conspicuous behaviors and this apparently increases their risk to avian predators.

Anecdotally, I have observed bass (Micropterus) striking at L. goodei while females were inspecting males (Fuller, pers obs). For the purposes of this model, the predation risk (p) to females at each spawning interval equals a constant (0.02) multiplied by the time spent

35 ) ) searching given the threshold of quality she has adopted (p( q ) = 0.02ts( q )) (Figure

2.1C). The maximum predation risk is one.

In this model, egg viability decreases as the time since ovulation increases. This is a common phenomenon for many external fertilizers (Bry 1981, Stacey 1984).

Females in many species will release their entire clutch (which is usually inviable) if forced to hold the clutch too long following ovulation (killifish: Foster 1967; seahorses:

Vincent 1994; darters: Fuller 1998b). In this model, egg viability (v) equals a constant

(0.01) multiplied by the time since ovulation. Time (t) is treated as a state variable. At the start of any given spawning interval, the time since ovulation equals the current time ) plus the additional time needed to find a mate(t + ts(q)) . Viability varies with time as follows: ) qtv ),( = 1− (01.0 t + ts(q)) (2).

I allow time since ovulation to vary between 0 and 100. Hence, viability varies between 1 and 0.

Model structure

This model uses a dynamic program to determine the optimal level of female choosiness for a female at a given time and spawning interval. At each spawning interval, females must choose mates in a way that maximizes fitness. Note that in contrast to most dynamic programming models that use time to mark each step in the model (Mangel and

Clark 1988, Clark and Mangel, 2000), this model uses spawning intervals as each step.

Females that mate at random use one time unit (t = 1). A female that sets her threshold at

36 q = 9 will spend on average sixty-one time units (t = 61) finding a mate (Figure 1B). The amount of time that passes at each spawning interval varies depending on the threshold criterion. Each interval encompasses finding and mating with a male. The final step in this model is the last spawning interval. Female fitness (F) at the last spawning interval

(S) is simply the fitness benefit she obtains from that mating expressed mathematically as ) ) ) ) F qtS ),,( = 1( − p(q))Nb qtvq ),()( (3) where the first term on the right-hand side of the equation equals the probability of ) survival, N equals the number of eggs to be released per spawn, b(q ) represents the ) ) mating benefit from setting the threshold of choosiness at q , and ,( qtv ) represents egg viability at time t since ovulation. For each value of t, the model determines which value ) of q (threshold for choice) maximizes female fitness. The model then uses a backwards iteration to determine the optimum level of choice at each spawning interval given the current time since ovulation (t). This is done using the following dynamic programming equation:

) ) ) ) F qts ),,( = 1[( − p(q))Nb ,()( qtvq )] + F(s + ,1 t + ts(q′), q′) (4) where q′ represents the optimal threshold of choosiness at the next spawning interval.

By comparing the consequences of various thresholds, the model determines which level of choice confers the highest fitness benefits given the spawning interval and the time since ovulation. The dynamic programming equation differs from the equation for fitness at the last spawning interval in that it incorporates the expected future reproduction as a consequence of the current choice. The level of threshold adopted at one spawning

37 interval affects the time spent searching which affects time since ovulation (t) for the next spawning interval.

To compare the implications of clutch splitting, I ran the program for five different clutch splitting strategies. For each strategy, the female had 16 eggs to spawn.

In the extreme clutch splitting strategy, the female could only spawn one egg per spawn

(16-spawns: 1 egg/spawn). I ran the program for all combinations between extreme clutch splitting and extreme clutch clumping (16-spawns: 1 egg/spawn, 8-spawns: 2 eggs/spawn, 4-spawns: 4 eggs/spawn, 2-spawns: 8 eggs/spawn, 1-spawn: 16 eggs/spawn). The rationale for this approach is that the effects of clutch splitting can be examined while fecundity is kept constant.

Simulating natural behavior

By iterating the model forward, I determined the sequences of decisions that females make across spawning intervals under “natural conditions” (Mangel and Clark 1988,

Clark and Mangel 2000). The main assumption was that females begin spawning as soon as they ovulate their eggs (t=0). For each simulation, females began at t=0 and followed the optimal decisions determined from the backwards iterations of the dynamic programming model. The time since ovulation at each spawning interval was the sum of all search times for all previous spawns. From this simulation, I determined the optimal level of choice at each spawning interval as well as the average choice across all spawns for each clutch splitting strategy.

38 Implications for dichotomous choice tests

I next examined how common assays of female choice affect measures of choosiness and how this relates to the average choice made under natural conditions. Researchers frequently hold gravid females in isolation from males and then later offer them a dichotomous choice between two males (Houde 1997). This has two potentially important implications. First, researchers are measuring female choice at the first spawning interval. Second, researchers are increasing the time since ovulation by holding gravid females in isolation. To simulate these effects, I examined choosiness for different values of t, time since ovulation. I compared the initial choosiness for females at t=0, t=40, and t=80 along with the average choice made under natural conditions for each clutch splitting strategy.

One-way mate choice tests

I next examined the performance of one-way choice tests. I simulated the effects of holding gravid females in isolation and then pairing them with a single male. The latency until the initial spawning and the time interval between subsequent spawns serve as measures of preference (Hatfield and Schluter 1996, Houde 1997). I simulated this protocol in two ways. In the first simulation, I randomly varied both time since ovulation and male quality. For each replicate, I set the time since ovulation at the first spawning interval for a random value between 0 and 100. I then paired a female with a male chosen at random from the distribution of male qualities. I ran the model forward 100 times for each clutch splitting strategy. In the second simulation, I assumed that preference tests began as soon as females ovulated (t=0). Thus, only male quality varied. 39 Again, males were chosen randomly from the distribution of qualities. I ran the model forward 100 times for each clutch splitting strategy. For both sets of simulations, I compared the performance of latency to spawn and interspawn interval as measures of female choice. These programs were conducted using the Macintosh version of True

Basic. Systat (version 7) was used for all subsequent statistical analyses.

Results

Running the model forward with t=0 at the first spawning simulates what females should do under natural conditions. Females should be least choosy at their first spawning and most choosy at their last spawning (Figure 3.2A). Furthermore, the optimum mate choice threshold at the last spawning should not vary with clutch splitting strategy.

The model also predicted that overall choosiness decreases with increases in clutch splitting. The initial threshold for mate choice was lowest for females that release one egg per spawn (16 spawns) (Figure 3.2A). As a result, the average threshold for mate choice also decreased with the number of required spawns (Figure 3.2B). These qualitative results were robust to changes in model constants (changes in predation risk and egg viability with time). Increases in choice costs increase the differences between clutch splitting strategies.

Dichotomous choice tests underestimated the average strength of female mating preferences for clutch splitters because preference was measured when females were least choosy (Figure 3.2A, 3.3). There was no effect for the one-spawn strategy because the first choice is also the last choice where the female should be most choosy. The effects of assaying female preference prior to any spawning were most dramatic for the extreme

40 8

7

6 1 spawn 5 2 spawns

4 4 spawns 8 spawns 3 16 spawns 2 0481216 Sequential spawns Optimal threshold mate choice

8

6

4

2 Average threshold

0 124816 Number of spawns

Figure 3.2. (A) Model predictions for mate choice threshold across multiple spawns. (B) Model predictions for average mate choice threshold across clutch splitting strategies.

41

average - hold 0

8 initial - hold 0 initial - hold 40

initial - hold 80 6

4

2 Mate threshold

0 124816 Number of choices/spawns

Figure 3.3. Average threshold vs. the initial threshold for mate choice after holding females prior to spawning.

42 clutch splitting strategies where initial preferences were 35% lower than the average preference for that strategy (Figure 3.3).

Holding females in isolation prior to choice tests further decreased female choosiness (Figure 3.3). These effects were particularly significant for the extreme clutch splitting strategy (16 spawns: 1 egg/spawn). Note that the threshold of mate choice was 1 for the 16-spawns strategy after the female has been held until t = 80. Conversely, the threshold for the clumping strategy (one-spawn: 16 eggs/spawn) only decreased a small amount (14%) after the female was held for t = 40.

Latency to spawn was a relatively good measure of mate choice for all clutch splitting strategies (Table 3.1, Figure 3.4). Holding females for variable amounts of time induced variability in this measure of preference, but a correlation between latency to spawn and male quality was detectable for all clutch splitting strategies. For the one and two spawns strategies, there was variability in female preference across a fairly large range in male qualities (Figure 3.4). However, for the more extreme clutch splitting strategies, there was little variability in preference for the majority of male qualities. This means that preference will only be detected when a large range of male qualities is included in the test.

In contrast, interspawn interval was not a good measure of mate choice particularly for the 4-spawns and 2-spawns strategies (Figure 3.5, Table 3.1). Only the

16-spawns and 8-spawns strategies showed a significant correlation between male quality and female mating preference. The interspawn intervals obtained when females were not held (t=0) indicated a hump-shaped distribution between interspawn interval and male quality for all clutch splitting strategies. The peak of the hump is most sharp in the 2-

43

Table 3.1. Regression analyses between female preference and male quality for two measures of preference, latency to spawn and interspawn interval.

Clutch Splitting Strategy R2 R2

(latency to spawn) (interspawn interval)

16-spawns 0.1295* 0.1713*

8-spawns 0.2794* 0.1357*

4-spawns 0.3925* 0.0013

2-spawns 0.2912* 0.0032

1-spawn 0.3752*

*p < 0.01

44 100 100 one spawn - 16 eggs

75 75

50 50

25 25

0 0 0246810 0246810 100 100 eight spawns - 2 eggs

75 75

50 50

25 25

0 0 0246810 0246810 100 100 four spawns - 4 eggs

75 75

50 50

25 25

0 0 0246810 0246810

Latency to spawn 100 100 two spawns - 8 eggs 75 75

50 50

25 25

0 0 0246810 0246810 100 100 sixteen spawns - 1 egg 75 75

50 50

25 25

0 0 0246810 0246810 Male quality

Figure 3.4. The relationship between latency to spawn and male quality in two different one-way preference tests. The left-hand panel shows the results from simulations where the initial time since ovulation at the start varied among replicates. 100 simulations were conducted for each graph. In the right-hand panel, the initial time since ovulation was set at 0 at the start of each replicate.

45 40 40 two spawns - 8 eggs 30 30

20 20

10 10

0 0 0246810 0246810

40 40 four spawns - 4 eggs 30 30

20 20

10 10

0 0 0246810 0246810

15 15 eight spawns - 2 eggs

10 10

5 5 Interspawn interval

0 0 0246810 0246810

15 15 sixteen spawns - 1 egg

10 10

5 5

0 0 0246810 0246810 Male quality

Figure 3.5. The relationship between interspawn interval and male quality in two one- way preference tests. In the left-hand panel, time since ovulation varied at the start of each replicate. In the right-hand panel, time since ovulation was set at 0 at the start of each replicate. 100 simulations were conducted for each graph.

46 spawns strategy. As the number of spawns increases, the peak of the hump shifts to lower values of male quality. The reason for this shape is that latency to spawn for a given male quality affects the subsequent spawn interval. When paired with low quality males, females wait until the very end of the time frame to commence spawning with the male. As a result, their subsequent interspawn intervals are short. Females should always spawn all of their eggs (except when paired with male quality 0). When paired with a male of intermediate quality, females will readily spawn because those males meet the female’s initial thresholds. After the initial few spawnings, the threshold increases, and females wait for longer period of time before spawning again.

Strange dynamics occurred when females were paired with quality 0 males.

Females engaged in all but the final spawning when paired with quality 0 males. As a result, females in the 2-spawns strategy did not have an interspawn interval when paired with quality 0 males. This results in n < 100 for the 2-spawns strategy (Table 3.1).

Discussion

The main result of this study is that female choosiness is lowest early in the mating sequence, and this causes a decrease in the average choosiness exhibited over all spawns.

This phenomenon is particularly pronounced when females must engage in many matings to disperse their eggs. These findings match results from previous theoretical investigations of mate choice (Crowley et al. 1990) and life-history (Warner 1998). The reason for risk aversion early in the mating sequence is that increased predation risk early in life drastically reduces reproductive value and lifetime reproductive success (Roff

1992, Stearns 1992). While a female may obtain an increase in egg survival by

47 increasing her threshold for an early spawn, she increases the risk of not spawning her remaining eggs because she may be preyed upon.

Experimental evidence supports the prediction that female choosiness increases within breeding seasons. In a field study of sand gobies, Forsgren (1997) found that females visited more males and traveled longer distances in the latter part of the mating season. Similarly, Qvarnström et al. (2000) report that female collared flycatchers

(Ficedula albicollis) exhibit a stronger preference for male color patterns in the latter portion of the breeding season than they do earlier in the breeding season. In the guppy

(Poecilia reticulata), Dugatkin and Godin (1993) found that young females are more likely to copy the mate choice of older females than vice versa and suggested that young females engage in mate copying because it lowers the cost of mate choice. In addition, studies in life-history have repeatedly found that reproductive effort increases with age

(Roff 1992, Stearns 1992). Furthermore, similar theoretical and empirical results have been obtained in studies of oviposition preferences in insects (reviewed in Clark and

Mangel 2000).

We also find evidence to support this model from comparisons among species.

Fishes that have to make fewer mate choices over their lifespan tend to have pronounced mate choice. Poeciliid fishes provide some of the best evidence for the importance of mate choice in sexual selection (guppies: Endler and Houde 1995, Houde 1997; mollies:

Travis and Woodward 1980, Ptacek and Travis 1997; swordtails: Basolo 1990, 1995,

Rosenthal and Evans 1998). Mate choice should be quite high in this group as females need only mate a few times because they can store sperm for several broods (Constantz

1989). Similarly, mate choice has been demonstrated in external fertilizers where

48 females spawn most of their clutch with one male (sand gobies: Forsgren et al. 1996,

Forsgren 1997; sticklebacks: Östlund and Ahnesjö 1998). In contrast, mate choice is less prevalent in groups where females split their clutches and mate numerous times over the course of their lifespan (orangethroat darters: Pyron 1995, 1996; rainbow darters: Fuller, in press; blue-headed wrasse: Warner 1987, 1988). Blue-headed wrasse (Thalassoma bifasciatum) are particularly interesting because females have the potential to switch sexes and spawn as males late in life (Warner 1984). An individual in this species can easily be expected to spawn hundreds of time over the course of its lifespan (Warner

1998). Thus female mate choice should be very low because latter spawning events will be experienced as males. Although females care little about the quality of prospective mates, they will exercise preferences so as to decrease their susceptibility to predation

(Warner and Dill 1999).

The results of this model have important implications for the design of female mate choice tests. Using dichotomous choice tests employed for poeciliids will most likely underestimate female mating preference. Furthermore, if females are held for significant periods of time prior to testing, mate choice tests may erroneously indicate no mating preference. One-way mate choice tests may be more reliable than dichotomous choice tests. Latency to spawn seems to be a good indicator of preference although it performs better for strategies with fewer spawns. One drawback to this measure of preference is that it might easily be confounded with acclimation. Researchers frequently worry about an animal’s reaction to being placed in a new aquarium, new lighting, etc., and allow for an acclimation period before beginning a mate choice trial (Ptacek and Travis 1996, 1997). If animals really are traumatized by such events, then it may be difficult to determine when acclimation is over and latency to spawn begins.

49 Perhaps the best way to measure preference for clutch splitters is to allow them to interact naturally with males and females in an environment where they can be observed over multiple days. This ensures that females can begin spawning as soon as they ovulate their clutch. In addition, male mating success (the variable of ultimate interest) can be measured (Côté and Hunte 1989). Correlations between female behaviors (i.e. approaches) and male mating success would provide evidence for female choice as the mechanism of sexual selection (Houde and Endler 1995, Houde 1997). Arguments can be made against this practice because it introduces the possibility of male/male competition affecting male mating success. However, this can be dealt with experimentally. Using a female biased sex ratio should reduce the amount of male/male competition (Owens and Thompson 1994, Kvarnemo et al. 1995). For animals that are territorial around oviposition sites, adequate breeding substrate can be provided to minimize male/male competition (Forsgren et al. 1996). Finally, if male/male competition cannot be reduced under any circumstances, then this is informative because it indicates that female choice may not be an important determinant of male mating success.

This model also has implications for measuring the heritability of female mating preferences. Many studies use the repeatability of individual female preferences as an upper-limit of the heritability of female mating preferences (Boake 1989, Jennions and

Petrie 1997). The implicit assumption is that an individual in a constant environment faced with the same choice of mates should exhibit the same level of choosiness.

Jennions and Petrie (1997) thoroughly discuss factors that may cause repeatability estimates to underestimate heritability including the effect of female condition (i.e. body

50 size). I suggest that future reproduction and time since ovulation should also be considered. For clutch splitters or animals facing a decrease in egg viability with time since ovulation, the internal state changes with each mating event and with time. This results in varying levels of choosiness. For systems with extreme clutch splitting, using the repeatability of mating preferences as an upper-estimate of heritability could easily underestimate the phenotypic variation in preferences if spawning sequence and time since ovulation are not taken into account. Cross-over designs, such as those used by

Ptacek and Travis (1997), may prove useful in such experiments.

Finally, it is worth considering in what types of mating systems clutch splitting should initially evolve and the implications for mate choice. One possibility is that large clutches may have higher mortality than small clutches. Males and other egg predators may be more likely to find and eat large clutches. In L. goodei, male filial cannibalism rates correlate positively with clutch size (Fuller and Travis 2001). Another possibility is that clutch splitting is essentially a type of bet-hedging. Theory suggests that bet-hedging should evolve in situations where there is large random variation in clutch mortality

(Gillespie 1974, 1977). Variation in mortality risk would be reduced by dispersing eggs across different males and/or oviposition sites resulting in an increase in the geometric mean fitness across generations. Both scenarios suggest that large variation in predictable male benefits may not exist. This would further reduce the potential for mate choice in clutch splitters. While females may not prefer specific male traits, they still could exert preference for males with whom they have not previously spawned.

In summary, this model predicts that clutch splitting (i.e. spreading eggs across multiple mating events) should be correlated with a decrease in overall female choice

51 provided that choice is costly. Females should be particularly risk-averse early in the spawning sequence. These dynamics could make the experimental detection of preferences a challenging task. In particular, dichotomous choice tests should drastically underestimate female choosiness. These results suggest that experiments that measure preference over multiple mating events and that do not disrupt the timing of mating will provide the most accurate measurements of preference.

52

CHAPTER 4

GENETICS, LIGHTING ENVIRONMENT, AND HERITABLE RESPONSES TO

LIGHTING ENVIRONMENT AFFECT MALE COLOR MORPH EXPRESSION

IN BLUEFIN KILLIFISH, LUCANIA GOODEI

Abstract

Determining the degree to which variation in traits is controlled by genetics and/or environment is a fundamental step to understanding adaptation. In this study, I examine the genetic and environmental influences on color pattern expression in male bluefin killifish, Lucania goodei. This is a compelling system because both male color patterns and vision physiology are correlated with basic properties of the environment. Across populations, males with blue anal fins are more abundant in waters with low transmission of UV and blue wavelengths, whereas males with red anal fins (and to a lesser extent, males with yellow anal fins) are more abundant in populations with high transmission of

UV and blue wavelengths. Here, I present results from two paternal half-sib breeding

53 experiments (one performed in the laboratory, one performed in the greenhouse) where offspring were raised under various lighting treatments. In the laboratory experiment, I found that red-versus-yellow expression is controlled by a single, autosomal locus where yellow (Y) is dominant over red (y). There was little blue expression in the laboratory.

In contrast, in the greenhouse experiment, I found higher expression of blue anal fin morphs when males were raised in tea-stained water than when raised in clear water. I also found strong effects of sire and an interaction between sire and lighting environment.

Hence, there is heritable variation in male propensity to express blue anal fins and also in male response to the environment (i.e. heritable plasticity). These results show that a relatively simple epistatic interaction can produce a large amount of phenotypic variation in male color patterns.

Introduction

Studies of adaptation typically involve three steps: (1) comparative studies demonstrating correlations between trait means and environmental conditions; (2) genetic studies demonstrating a genetic component in trait variation; and (3) experimental studies demonstrating the connection between trait means and fitness (Reznick and Travis 1996,

Travis and Reznick 1998). Determining the degree to which trait expression is controlled by genetics and/or environment is a critical step for two reasons. First, genetic variation is obviously a key requirement for the evolution of traits via natural selection (Fisher

1958, Endler 1986). Second, across population correlations between trait means and environmental conditions can reflect either genetic differentiation and/or simple phenotypic plasticity where differences in trait means are caused by differential

54 expression in different environments (Reznick and Travis 1996, Lynch and Walsh 1998).

This dichotomy can quickly become quite complicated because some genotypes may respond differently to the same environmental gradient. Hence, the basic "genetics and/or environment" question often has a complicated answer. Complex sources of variation, such as environmental effects that differ by genotype or countergradient patterns, can generate patterns of phenotypic variation that are underlain by more genetic variation than one might suspect (Trussell 2000).

In this study, I examine the genetic and environmental influences on color pattern expression in male bluefin killifish, Lucania goodei. This is a compelling system because both male color patterns (Fuller 2002) and vision physiology (Fuller et al. in press) are correlated with basic properties of the environment indicating either some adaptive variation or some very striking environmental influence on phenotypes. In a comparative study across 30 Florida populations, Fuller (2002) found that males with blue anal fins were more abundant in waters with low transmission of UV and blue wavelengths (i.e. tea-stained swamps and lakes). In contrast, males with red anal fins

(and to a lesser extent, males with yellow anal fins) were more abundant in populations with high transmission of UV and blue wavelengths (i.e. clear springs). These differences in color morph frequencies are correlated with differences in vision physiology where animals in a spring population express more UV and violet cones than do animals from a swamp population (Fuller et al. in press). These results beg the question of whether simple genetic and/or environmental effects contribute to phenotypic variation or whether more complex sources of variation create these patterns.

55 There is good reason to expect large genetic effects. Inheritance studies of color pattern have frequently found large genetic effects. In guppies, Poecilia reticulata, color patterns of many morphs are controlled by Y-linked genes (Winge 1922a,b, Houde 1992,

1997, Brooks and Endler 2001). Extensive research on swordtails (Xiphophorus) has demonstrated more than 100 different pigmentation types, many of which are controlled by sex-linked loci (Kallman 1975, Angus 1989). In the side-blotched lizard, Uta stansburiana, both males and females have dewlaps that are either yellow, orange, blue, a combination of yellow and blue, or a combination of orange and blue (Sinervo and Lively

1996, Sinervo et al. 2001). This variation has a large genetic component (Sinervo and

Lively, 1996, Sinervo et al. 2000, Zamudio and Sinervo 2000) and is hypothesized to be under the control of a single locus. Additional examples of genetic variants producing color polymorphisms include happy face spiders, Theridion grallator, (Gillespie and

Oxford 1998), swallowtail butterflies, Papilio sp.,(Scriber et al. 1996), and morning glory flowers, Ipomoea purpurea, (Rauscher and Fry 1993).

There is also good evidence that environmental conditions can have strong effects on color patterns. The production of carotenoid-based color elements has been linked to foraging ability and general health in both birds and fish (house finches, Carpodacus mexicanus: Hill and Montgomerie 1994, Thompson et al. 1996, Brawner et al. 2000, guppies, Poecilia reticulata: Kodric-Brown 1989, Houde and Torio 1992, Grether 2000, firemouth cichlids, Cichlasoma meeki: Evans and Norris 1996). Some recent evidence also suggests that similar phenomena may occur with structural colors such as blue and

UV (blue grosbeaks, Guiraca caerulea: Keyser and Hill 2000, pied flycatchers, Ficedula hypoleuca: Siitari and Huhta 2002, blue-black grassquit, Volatinia jacarina: Doucet

56 2002). In addition to condition-dependence, there is also evidence that seasonality can affect the expression of color pattern (satyrine butterflies: Brakefield and Larsen 1984, blue tits, Parus caeruleus: Ornborg et al. 2002) as can temperature (eastern mosquito fish, Gambusia holbrooki: Horth 2001; hoverfly, Eristalis arbustorum: Ottenheim et al.

1996).

In this paper, I present the results of two experiments that investigated the sources of variation among males in color expression. In each experiment, I crossed males of different colors with several females and raised the offspring in different lighting environments. Our results show that red and yellow anal fins on males are controlled by simple Mendelian effects, but the blue fin will be expressed only by some genotypes and only when individuals are raised in a particular lighting environment. These effects illustrate that even a very few sources of variation can be combined to create substantial phenotypic diversity.

Study System

The bluefin killifish, Lucania goodei, is a freshwater fundulid found throughout peninsular Florida with a few populations occurring in southeastern Georgia and South

Carolina (Page and Burr 1991). The breeding system is promiscuous. Males guard patches of vegetation that serve as substrates for females to attach eggs (Fuller 2001).

There is no evidence for male parental care (Fuller and Travis 2001). Males use their dorsal and anal fins when fighting other males and also when courting females. In fights, males flare their dorsal and anal fins and engage in circle fights (Fuller 2001). Males also

57 use their fins in the initial stages of courting females. Males swim circle loops in front and around the female while flashing their anal and dorsal fins.

The color pattern is dimorphic between the sexes. Males have a red dot at the base of the caudal fin. The dorsal, anal, and pelvic fins are colored in males but lack color in females. The anterior 3/4 of the dorsal fin is blue on all males. The posterior 1/4 of the dorsal fin, the pelvic fins, and the anal fin are polymorphic among males. The polymorphism on the pelvic fins and posterior dorsal fin is relatively simple. The posterior 1/4 of the dorsal fin can be blue, red, or yellow. The pelvic fins are either red or yellow.

The polymorphism on the anal fin is much more complex. There are five main categories of anal fin color patterns: solid red, solid yellow, solid blue, combination of red and blue, and combination of yellow and blue. In a state-wide Florida census, over

99% of males expressing coloration could be placed into these categories. However, in one population, males with orange anal fins were found in addition to the yellow and red anal fin morphs (Fuller, 2002). In this study, I do not consider the orange color morph.

In addition, the red-blue combination and yellow-blue combination anal fin color patterns can take a variety of forms. In our study, I use only males with solid-colored anal fins as sires and score male offspring as belonging to one of the five categories.

Methods

I collected L. goodei at the 26-Mile Bend boat ramp in the Everglades (Broward, Co., FL,

USA) using dipnets and seines in January 2000 and August 2001. This is a high density population with good representation of most color patterns (Fuller 2002). I transported

58 the animals to Florida State University and housed them in 76-liter aquaria until the experiments began. Animals were fed daily with frozen chironomids and adult Artemia.

Laboratory Experiment

I chose four sires with distinctly different color patterns. One male was blue on the posterior region of the dorsal fin, blue on the anal fin, and yellow on the pelvic fins, and is referred to as the "B/B-yellow pelvics" sire hereafter. The second male was yellow on the posterior region of the dorsal fin, yellow on the anal fin, and yellow on the pelvic fins, and is referred to as the "Y/Y-yellow pelvics" sire hereafter. The third male was red on the posterior region of the dorsal fin, red on the anal fin, and red on the pelvic fins, and is referred to as the "R/R-red pelvics" sire hereafter. The fourth male was red on the posterior region of the dorsal fin, blue on the anal fin, and red on the pelvic fins, and is referred to as the "R/B-red pelvics" sire hereafter.

I crossed each of the four sires with nine randomly chosen females resulting in a total of 36 females (hereafter referred to as dams). For each sire, the nine dams were divided among three environmental lighting treatments (UV filter, UV/blue filter, grey filter). For each pairing, I placed a sire with a dam in an aquarium filled with clear well water and covered by one of the light filters. Each aquarium contained yarn mops that served as spawning substrate. I checked the mops daily for the presence of eggs. The sire and dam remained in the aquarium until a minimum of 20 healthy eggs was obtained, at which time the two fish were removed. I began crossing sires and dams in January 2000 and finished in July 2000.

59 In this experiment, offspring were raised under the lighting treatments from conception. I carefully removed the eggs from the mops and placed them in small plastic tubs that I floated in the aquarium. Upon hatching, fry were fed Artemia nauplii. Once all of the fry were eating and the parents had been removed, I released the fry into the aquarium. At approximately 2-3 months of age, offspring were switched to a diet of frozen brine shrimp and frozen bloodworms. Throughout the experiment, room temperature was kept at 22C.

The environmental lighting treatments were created by attaching plastic theater gels with known spectral properties to wooden frames that sat on top of the aquaria.

Light from flourescent bulbs (that mimic sunlight) had to pass through the filters before entering the aquaria. The no-UV treatment was created by attaching two layers of a plastic UV filter to the frame. This treatment reduced all wavelengths below 400 nm.

The no-UV/blue treatment combined one sheet of plastic UV filter layered with one sheet of blue filter that together reduced wavelengths below 500 nm. The grey filter treatment consisted of two layers of clear filter that removed roughly 20% of all the wavelengths and controlled for the presence of a filter. Lights were maintained on a 14L:10D cycle.

To ascertain the effects of our treatments, I measured transmission through the filters using a reflectance probe, an Ocean Optics S2000 spectrophotometer, and an LS-1 light source (Ocean Optics, Dunedin, FL). I determined the maximum light coming from the light source using a white standard. I then placed the filters on top of the white standard and measured the amount of light that transmitted through the filters, reflected off of the standard, and transmitted again through the filters. Because the light had to

60 pass through the filter twice, I measured transmission as the square root of the proportion of light detected by the reflectance probe.

I censused the aquaria once every 2-3 months from August 2000-August 2001.

During each census, I recorded the number of males, females, and juveniles, the standard length of males and females, and the color pattern of males. To increase the number of males expressing coloration, I occasionally removed the largest, most brightly colored males. At the end of the experiment, I calculated the proportions of males bearing various color. In addition to males present at the last census, I included males that had been removed and males that had apparently died.

After obtaining offspring, I treated dams with androgens to induce them to express the male color pattern. Our objective was to determine whether dams also carried genes for the male color pattern, and, if so, to determine the phenotype each dam expressed. After obtaining adequate eggs for the breeding experiment, I placed each dam in a 1.9 liter glass jar containing well water aerated with a small airstone. Each day, I added 20 µl of a 17α-methyltestosterone solution (concentration: 3 mg/ml ethanol). To prevent a build-up of ethanol (which is toxic to fish), I replaced the well-water twice weekly and recorded the female color pattern.

These data were analyzed using a general linear model to determine whether sires, lighting environment, or the interaction between sires and lighting environment accounted for a significant amount of variation in the relative abundances of males with solid red, solid yellow, and solid blue anal fins. I treated sires and lighting environment as fixed effects. In this experiment, there was no replication within individual dams.

Hence, the effect of sire was tested over the mean-square error term. I used the arcsine

61 transformation of the square-root of the proportion of males in a clutch bearing a given color pattern as our dependent variables. In the text, I report means and standard errors calculated on the untransformed proportions. I tested the residuals from all models to determine whether the assumption of normality was upheld. Finally, I calculated a chi- square statistic to test whether the hidden phenotype of females met with the predictions of a Mendelian model. Results were considered significant at P < 0.05. All analyses were performed using SAS V.8 (SAS Institute).

Greenhouse Experiment

I chose four sires with distinctly different color patterns. One male was yellow on the posterior region of the dorsal fin, blue on the anal fin, and yellow on the pelvic fins, and is referred to as the "Y/B-yellow pelvics" sire hereafter. The second male was yellow on the posterior region of the dorsal fin, yellow on the anal fin with a slight tinge of blue at the base, and yellow on the pelvic fins, and is referred to as the "Y/Y-yellow pelvics" sire hereafter. The third male was red on the posterior region of the dorsal fin, red on the anal fin, and red on the pelvic fins, and is referred to as the "R/R-red pelvics" sire hereafter.

The fourth male was red on the posterior region of the dorsal fin, blue on the anal fin, and red on the pelvic fins, and is referred to as the "R/B-red pelvics" sire hereafter.

I crossed each sire with 3-4 randomly chosen dams in the laboratory. I then divided each clutch between two environmental lighting treatments (clear vs. tea-stained water). Crosses were made using a similar protocol as in the laboratory experiment.

Upon hatching (approximately 10-14 days post-fertilization), I fed animals Artemia nauplii in the laboratory for approximately 2-4 weeks. I then transported them to the

62 greenhouse, divided each clutch so that there were equal numbers of similarly aged fry, and placed them in 114-liter tanks containing either clear or tea-stained water. I began crosses in August 2001 and finished March 2002. Again, I determined the hidden phenotype for a fraction of the dams by treating them with 20 µl of a 17α- methyltestosterone solution (concentration: 3 mg/ml ethanol).

For each combination of dam and sire, I set up two 114-liter tanks containing either clear or tea-stained water (26 tanks total). I filled tanks with well water and stocked them with mops and plastic plants. All water was treated with a buffer to keep the pH above 7. In addition, I set up filters to remove excess wastes and algae from the tanks. For the tea-stained water treatment, I added a small amount of instant, decaffinated tea to the water 2-3 times each week. Whenever algae began to grow in the water column, I drained the water, added new water, re-established the treatment, and added a slow-release algicide ("Green Water Control") to the tanks to keep the water column clear. Filamentous algae were allowed to grow on aquarium walls. This experiment ran from August 2001-December 2002.

To quantify treatment differences, I measured light transmission through the water in the UV-red wavelengths (350-700 nm). I placed a small sample of water in a

1.5ml Eppendorf tube, put the sample on ice, and returned to the lab where I measured relative transmission using a Beckman Coulter DU 640 spectrophotometer. The spectrophotometer records the transmission of light through 5mm of water from 350-700 nm in 0.5 nm intervals resulting in 701 data points for each curve. I smoothed these data by averaging the transmission every 2 nm resulting in curves with 175 data points. I

63 compared these data between sires, lighting environment, and the interaction between sires and lighting environment.

I planned a balanced breeding design where each sire was crossed with three dams. However, unforeseen complications resulted in an unbalanced design. Because one dam paired with the Y/B-yellow pelvics sire died after producing a small clutch, I paired this sire with another dam resulting in 4 dams for the Y/B-yellow pelvics sire

(Table 4.1). In addition, I accidentally lost all but one female offspring from one clutch during one census (Table 4.1, R/R, dam 3, clear water). Finally, I inadvertently contaminated a clutch with unrelated animals. Luckily, I still had both the dam and sire, and could spawn another clutch (Table 4.1, Y/B, dam 3, clear water, 45 fry). These complications resulted in an unbalanced experimental design.

I analyzed this experiment using a general linear model to determine whether sires, lighting environment, or the interaction between sires and lighting environment accounted for a significant amount of variation in the proportion of male offspring with solid red, solid yellow, solid blue, or any blue on the anal fins. I treated sire and lighting environment as fixed effects and dams nested within sires as random effects. Because the design was unbalanced, SAS used the Sattherwaite approximation to determine the error and degrees of freedom in the denominator of the sire F-test (Sokal and Rohlf 2000).

This results in an error term equal to 0.9608 * mean-square of dam (sire) + 0.0392 * mean-square error. All other F-tests were calculated using the mean-square error in the denominator. All other details of the analysis followed the procedures used for the GXE

- laboratory experiment.

64

Table 4.1. Number of animals at various life-stages for each family in the greenhouse experiment.

sire dam eggs fry fry fry in fry in males males males males females females eat clear tea w/ color w/ w/o w/o (clear) (tea) water water (clear) color color color (tea) (clear) (tea) R/B 1 236 198 105 53 52 7 10 0 0 12 8 R/B 2 271 251 135 63 62 9 6 0 0 6 5 R/B 3 191 186 130 66 64 15 10 2 0 17 10 R/R 1 218 182 69 34 35 5 12 0 0 11 10 R/R 2 222 181 127 59 59 8 9 9 0 12 8 R/R 3 310 233 168 83 85 0 15 1 0 0 15 Y/B 1 261 222 119 58 58 9 7 0 0 8 6 Y/B 2 101 74 46 17 17 4 3 0 0 2 2 Y/B 3 295 249 167 45 67 13 18 4 0 13 13 Y/B 4 152 128 66 27 27 8 7 0 2 7 3 Y/Y 3 286 234 177 85 85 17 14 6 0 19 9 Y/Y 4 54 26 20 8 10 6 4 0 0 3 6 Y/Y 5 369 304 210 104 106 10 13 1 0 9 17

65

Results

Laboratory Experiment

Our treatments created different lighting conditions. The no-UV treatment reduced transmission of light from 360-420 nm and presumably decreased the downwelling light at these wavelengths. The no-UV/blue treatment reduced transmission from 360-500 nm.

The grey filter reduced overall transmission (Figure 4.1A).

Few males expressed the blue anal fin coloration in the lab. Across all 36 families, 324 males expressed coloration on the anal fin (346 males total - 22 died before expressing color). Most males expressed either a red anal fin (142 males – 44%) or a yellow anal fin (163 males – 50%). Only fourteen males expressed a solid blue anal fin

(4%). Five males (2%) expressed either a yellow-blue or red-blue combination anal fin.

On average, there were 9.00 males expressing color (0.747 SE, n = 36) in each clutch.

Two clutches had only 2 males expressing color and were excluded from the analysis.

The proportion of male expressing red anal fins varied significantly among sires

(Figure 4.2, F3,22 = 19.37, P < 0.0001). Males with red anal fins were most abundant in the clutches of the R/R-red pelvics and R/B-red pelvics sires. In addition, lighting treatments also accounted for a significant amount of variation (F2,22 = 4.78, P = 0.019).

Males with red anal fins were more abundant in the grey filter treatment (mean = 0.573 +

0.125 SE, n = 12) than in the no-UV filter treatment (mean = 0.347 + 0.101 SE, n = 11).

There was no effect of the interaction between sire and environment.

66

1.0

0.8

0.6

0.4 transmission grey 0.2 no uv no uv+blue A. 0.0 350 400 450 500 550 600 650 700 wavelength

1.00

0.95

0.90 transmission clear tea

0.85 B. 350 400 450 500 550 600 650 700 wavelength

Figure 4.1. (A) Transmission of light across filters in the laboratory experiment. (B) Transmission of light through clear vs. tea-stained water in the greenhouse experiment.

67

1.0 red 0.8 yellow

0.6

0.4 frequency

0.2

0.0 B/B Y/Y R/R R/B yellow yellow red pelvics red pelvics pelvics pelvics

sires

Figure 4.2. Average frequency of F1 males expressing a red or yellow anal fin in laboratory crosses. Bars are standard errors. B/B-yellow pelvics sire (n = 8), Y/Y-yellow pelvics sire (n = 8), R/R-red pelvics sire (n = 8), R/B-red pelvics sire (n = 9).

68

The proportion of males with yellow anal fins was similarly affected by sire

(Figure 4.2, F3,22 = 12.22, P < 0.0001). Males with yellow anal fins were most abundant in the clutches sired by the B/B-yellow pelvics and Y/Y-yellow pelvics sires. In contrast to the pattern with red males, yellow males also tended to be more abundant in the no-UV treatment (mean = 0.638 + 0.102 SE, n = 12) than in the grey filter treatment (mean

=0.399 + 0.117, n = 12), but the differences attributable to lighting environment were only marginally significant (F2,22 = 3.38, P = 0.0523). The interaction between sire and environment did not account for a significant amount of variation.

Neither sire, lighting environment, nor the interaction between sire and environment accounted for a significant amount of variation in the relative abundance of males with blue anal fins (sire: F3,22 = 0.35; lighting environment F2,22 = 0.82; sire*environment F6,22 = 0.30).

A test for Mendelian Ratios

Almost all males expressed some component of either yellow or red. Of the 14 males with solid blue anal fins, four had either red pelvic fins or a red posterior region on the dorsal fin. Seven males with solid blue anal fins had either yellow pelvic fins or a yellow posterior region of the dorsal fin. Three males were not diagnosable for red or yellow because they were removed from the aquaria prematurely or died. Of the 5 males with combination anal fins, one expressed as yellow and blue, and four expressed as red and blue. This phenomenon allows us to categorize 99% of the males as expressing some component of either yellow or red.

69 An examination of the relative abundance in each clutch of males with any yellow vs. males with any shows a pattern of expression that is consistent with Mendelian inheritance where there is one autosomal locus controlling the expression of yellow and red with the yellow allele (Y) being dominant over the red allele (y) (Figure 4.3). The

B/B-yellow pelvics sire had 100% yellow offspring in all of his clutches with the exception of one clutch where seven male offspring were yellow and one was red. This pattern is consistent with this male being homozygous dominant (YY). The Y/Y-yellow pelvics sire had clutches that varied between 50% and 100% yellow male offspring with the exception of one clutch where one male offspring was yellow, and three were red.

This is consistent with this male being heterozygous (Yy). Mating with a female carrying the homozygous dominant genotype (YY) should produce 100% yellow male offspring.

A mating with a female carrying the heterozygous genotype (Yy) should produce 75% yellow and 25% red male offspring. A mating with a female carrying the homozygous recessive phenotype (yy) should produce 50% yellow and 50% red offspring.

The red sires also produced offspring in patterns consistent with the Mendelian hypothesis (Figure 4.3). The R/R-red pelvics sire produced clutches with 100% red offspring and clutches with roughly 50% red and 50% yellow offpring. The R/B-red pelvics sire produced a similar distribution of clutches (100% red, 50% red) with the notable exception of one clutch with 100% yellow offspring (8/8 males expressed as yellow). These ratios are consistent with these sires being homozygous recessive (yy).

Matings with homozygous recessive females (yy) should produce 100% red male offspring. Matings with heterozygous females (Yy) should produce 50% red and 50%

70

predicted yellow dam pro 1.0 0

por proportionoffspring red tio pro n por yell tio ow n off 0.5 0.5 red spri off ng spri ng roportion yellow offspring p 0 1.0

predicted red dam

B/B Y/Y R/R R/B yellow pelvics yellow pelvics red pelvics red pelvics Sires

Figure 4.3. Distribution of clutches expressing some element of red or yellow in laboratory crosses. Each dot is a clutch. Grey filling indicates clutches with inadequate sample sizes (< 3). The predicted female hidden phenotype is indicated. Stippled dots indicate females that were predicted to carry the red phenotype but actually expressed as yellow. Hashed dots indicate females that were predicted to carry the yellow phenotype but actually expressed as red.

71 yellow male offspring. Matings with homozygous dominant females (YY) should produce 100% yellow male offspring.

I determined the hidden phenotype for 32 dams to test the Mendelian hypothesis.

Based on the relative abundance of F1 male phenotypes, I predicted the phenotype of the dam (Figure 4.3). I could make no predictions for the B/B-yellow pelvics sire because I posited that he was a homozygous dominant male. Hence, the genotype of the dam should have no effect on the F1 male phenotypes. For the Y/Y-yellow pelvics sire, I predicted that clutches with roughly 50% red and 50% yellow male offspring would have dams carrying the red phenotype (yy) whereas clutches with roughly 75%-100% yellow offspring would have dams carrying the yellow phenotype (YY or Yy). For both the

R/R-red pelvics and R/B-red pelvics sires, I predicted that clutches with 100% red offspring would have dams carrying the red phenotype. For clutches with roughly 50% to 100% yellow male offspring, I predicted that dams would carry the hidden yellow phenotype (YY or Yy).

The predictions were met in 18/23 cases (Figure 4.3, X2 = 7.34, df = 1, P =

0.0067). Inclusion of the two clutches with only two male offspring increases the number of predictions upheld to 20/25. Four of the five cases where the predictions were not upheld can be attributed to sampling error. For the Y/Y-yellow pelvics sire, three of our predictions were not upheld, but in all of these clutches the proportion of males with yellow anal fins was between 50-75%. In one clutch, 9/16 male offspring were yellow

(56.25%) leading us to predict a red dam. In the second clutch, 11/20 male offspring were yellow (55%) leading us to predict a red dam. In the third clutch, 5/7 male offspring were yellow (71%) leading us to predict a yellow dam. The R/B-red pelvics

72 sire had one clutch where 4/4 male offspring were red (100%) leading to the prediction of a red dam. In all of these cases, the addition of one or two male offspring with the appropriate phenotype changes the frequencies to the point where the predicted dam phenotype matches the expressed dam phenotype. However, in one case the deviation from the predicted dam phenotype cannot be attributed to sampling error. The R/B-red pelvics sire had one clutch where 5/10 offspring were red (50%) leading to the prediction of a yellow dam. The expressed phenotype was red. I cannot explain this as only red offspring should result form the pairing of two animals carrying the homozygous recessive (yy) phenotype.

Greenhouse Experiment

Tea-stained and clear water treatments had dramatic effects on the transmission of light through the water (Figure 4.1B). UV and blue transmission was significantly higher in the clear water treatment than in the tea-stained treatment (ANOVA, P < 0.05 for 350-

537mm) with the largest differences occurring in the shortest wavelengths. In addition, transmission was slightly higher in the red wavelengths (641-700nm) in tea-stained water than in clear water. At first consideration, these differences in transmission may seem small (100% vs. 90%). However, these transmission values were measured after transmission of light across a small distance (5mm).

Blue expression was higher in male offspring raised in the greenhouse. Out of

239 male offspring expressing coloration, 28 had solid blue anal fins (12%), and 16 had either a yellow-blue (8 males - 3%) or red-blue (8 males - 3%) combination anal fin. Still, the vast majority of males had either solid red (95 males - 40%) or solid yellow (100

73 males - 42%) anal fins. There were an average of 11.0 (1.13 SE) males expressing color across the remaining 25 clutches (91% of all males). In addition, there was no evidence for differential survival across sires, dams, or lighting environments (see Table 4.1 for raw data).

The relative abundance of males with solid blue anal fins was significantly affected by sire (F3,9.8 = 8.46, P = 0.0045), lighting environment (F1,8 = 32.33, P =

0.0005), and the interaction between sire and lighting environment (F3,8 = 4.21, P =

0.0462) (Figure 4.4A). There was no statistically significant effect of dam (F9,8 = 0.92, P

= 0.554). An examination of Figure 4a shows that male offspring with solid blue anal fins were most abundant in the clutches of the Y/B-yellow pelvics sire (mean = 0.203 +

0.070 SE, n = 8) followed by the Y/Y-yellow pelvics sire (mean = 0.139 + 0.072 SE, n =

6) followed by the R/R-red pelvics sire (mean = 0.074+ 0.041 SE, n = 5) followed by the

R/B-red pelvics sire (mean = 0 + 0 SE, n = 6). Across lighting environments, male offspring with solid blue anal fins were much more common in the tea-stained water treatment (mean = 0.198 + 0.048 SE, n = 13) than in the clear water treatment (mean =

0.021 + 0.021 SE, n = 12). The significant interaction between lighting environment and sire is attributable to the fact that male offspring from the R/B-red pelvics sire were not more likely to express solid blue anal fins when raised in tea-stained water (Figure 4.4A).

In contrast, the male offspring of the R/R-red pelvics sire, Y/Y-yellow pelvics sire, and especially the Y/B-yellow pelvics sire were more likely to express solid blue anal fins when raised in tea-stained water.

Sire, lighting environment, and the interaction between sire and lighting environment had similar effects on the proportion of males expressing any blue on their

74

1.00

0.75 clear tea 0.50

roportion solid blue solid roportion 0.25 p

0.00 A. R/B R/R Y/Y Y/B red red yellow yellow pelvics pelvics pelvics pelvics

1.00 clear tea 0.75

0.50

roportion any blue any roportion 0.25 p

0.00 R/B R/R Y/Y Y/B B. red red yellow yellow pelvics pelvics pelvics pelvics

Figure 4.4. The proportion of males expressing (A) solid blue anal fins and expressing (B) any element of blue on the anal fin in greenhouse crosses. Open dots indicate clutches raised in clear water. x's indicate clutches raised in tea-stained water.

75 anal fins. A significant effect of sire (Figure 4.4B, F3, 9.1 = 4.45, P = 0.0347) was driven by both the Y/B-yellow pelvics sire and the Y/Y-yellow pelvics sire having a higher proportion of male offspring with some component of blue on their anal fins than the

R/R- red pelvics sire and the R/B- red pelvics sire (Y/B-yellow pelvics sire mean = 0.424

+ 0.127 SE, n = 8; Y/Y-yellow pelvics sire mean = 0.290 + 0.084 SE, n = 6; R/R-red pelvics sire mean = 0.074 + 0.041, n = 5; R/B-red pelvics sire mean = 0.017 + 0.017 SE, n = 6). There was also a significant effect of dam nested within sire (F9,8 = 5.55, P =

0.0122). Again, males raised in tea-stained water were more likely to have some element of blue on their anal fins than were males raised in clear water (Figure 4.4B, F1,8

= 39.39, P = 0.0002; tea-stained water mean = 0.340 + 0.087 SE, n = 13; clear water mean = 0.099 + 0.049 SE, n = 12). The significant interaction between sire and lighting environment (F3,8 = 4.32, P = 0.0435) was driven by the fact that male offspring of the

Y/B-yellow pelvics, Y/Y-yellow pelvics, and R/R-red pelvics sires were more likely to express some element of blue on their anal fins when raised in tea-stained water than were offspring from the R/B-red pelvics sire (Figure 4.4B).

Neither sire nor the interaction between sire and lighting environment accounted for significant amounts of variation in the relative abundance of males with solid yellow anal fins nor in the relative abundance of males with solid red anal fins (yellow morphs: sire F3,9.2 = 0.65, sire x lighting environment F3,8 = 0.33; red morphs: sire F3,9.2 = 1.86, P

= 0.2058, sire x lighting environment F3,8 = 1.18, P = 0.3752). Lighting environment did not affect the relative abundance of males with solid yellow anal fins (F1,8 = 0.43), but did affect the relative abundance of males with solid red anal fins (F1,8 = 5.77, P =

0.0430) where males with solid red anal fins were more abundant in clear water (mean =

76 0.499 + 0.079 SE, n = 12) than in tea-stained water (mean = 0.326 + 0.084 SE, n = 13).

In addition, dams had a significant effect upon the relative abundance of both males with solid yellow anal fins (F9,8 = 3.98, P = 0.0323) and males with solid red anal fins (F9,8 =

4.77, P = 0.0192).

Again, over 99% of the animals could be diagnosed as expressing some element of either red or yellow (Figure 4.5). Examination of the data shows ratios roughly consistent with Mendelian inheritance where there is a single, autosomal locus controlling yellow versus red with the yellow allele (Y) being dominant over the red allele (y). There were no clutches where either the Y/B-yellow pelvics sire or the Y/Y- yellow pelvics sire had 100% red male offspring. In contrast, both the R/R-red pelvics and R/B-red pelvics males had clutches distributed at approximately 100% red and 50% red. In addition, the R/R-red pelvics male had one clutch with close to 100% yellow offspring (14/15 yellow).

I discerned the hidden female phenotype for six dams. Five out of the six females met the predicted phenotype based on Mendelian ratios. The one deviation came from the R/R-red pelvics sire where 82% of the males were red (18% yellow). Yellow offspring should only result from matings between a red phenotype (yy) and a yellow phenotype (Yy). I therefore predicted the dam would carry the yellow the phenotype, but she expressed as red.

Both the Y/Y-yellow pelvics and the Y/B-yellow pelvics sires appear to be heterozygotes as both produced clutches with both yellow and red offspring. Hence, the face that I did not find a strong sire effect on red versus yellow expression in male offspring is not surprising. Dams did have large effects on the production of yellow/red

77

1.00 0.00

0.75 0.25 proportion red

0.50 0.50

proportion yellow 0.25 0.75

0.00 1.00 R/B R/R Y/Y Y/B red pelvics red yellow yellow pelvics pelvics pelvics

Figure 4.5. The proportion of males expressing some element of yellow or red in greenhouse crosses. The data are pooled across lighting treatments for each combination of sire and dam.

78 offspring as I would expect when the four sires are either heterozygous for yellow (Yy) or homozygous for red (yy).

Discussion

In this study, I have shown that a relatively simple epistatic interaction can produce a large amount of phenotypic variation in male color patterns. In the laboratory experiment,

I found strong evidence for a single autosomal locus that has a large effect on whether males can develop yellow or red anal fins. There were strong effects of sire, and the expression of the hidden phenotype in dams met with the predictions of Mendelian inheritance. Mendelian inheritance was indicated when I determined the underlying yellow vs. red phenotype and disregarded blue expression. This indicated that blue expression is not governed by the same gene that governs yellow versus red expression.

In the greenhouse experiment, I detected an orthogonal genetic effect on whether males develop blue anal fins essentially covering the yellow/red phenotype. Furthermore, I detected a heritable, plastic effect where some males are more likely to develop blue anal fins when raised in tea-stained water.

Hence, a single locus determines whether males can express red or yellow, but this can be suppressed by expression of blue which depends on orthogonal sets of genes which determine a male's propensity to express blue as well as his response to the environment. These results are similar to those found in an African satyrine butterfly,

Bicyclus anyana, where several single gene mutants have large effects on eyespot color patterns, but overall expression of eyespots is dependent on temperature, and genotypes differ in their response to temperature (reviewed in Brakefield et al. 1996, Brakefield

79 1998, but see Wijngaarden and Brakefield 2001). In both systems (B. anynana and L. goodei), genes with large effects on color pattern that act orthogonally to the expression of phenotypic plasticity. Variation in each of these axes can create a large number of color patterns.

The fact that both alleles (Y and y) are maintained in nearly all L. goodei populations suggests the presence of a balanced polymorphism. In a state-wide Florida census, Fuller (2002) found males with red in all populations and males with yellow in

29/30 populations. A neutral model of evolution predicts fixation in more populations due to drift (Hartl 1988). Heterozygote advantage could maintain this variation if yellow heterozygous males (Yy) have higher relative fitness than either homozygous yellow males (YY) or red males (yy). This scenario seems unlikely, because I cannot distinguish yellow homozygotes from yellow heterozygotes. However, such an effect could occur if there are differences in reflectance spectra between the two genotypes that individual L. goodei can detect but that are undetectable to humans (Endler 1990, Bennett et al. 1994,

Andersson et al. 1998, Hunt et al. 1998, Cuthill et al. 1999). Negative frequency dependence could also maintain both alleles (as well as genes for blue morphs) within populations. Negative frequency dependence could occur if males compete more intensely with males bearing the same phenotype (Gross and Charnov 1980, Partridge

1984, Sinervo and Lively 1996) or if there are female mating preferences for rare males

(Farr 1977, Partridge 1984, Hughes et al. 1999). Finally, both alleles (as well as genes for blue morphs) could be maintained within populations by variation in environmental conditions. Light is filtered as it passes through the water column resulting in different lighting environments at different depths (Endler 1990, Loew and McFarland 1990,

80 Johnsen 2002). Variation in color morphs could be maintained within populations if each phenotype had the highest mating success under a fraction of the lighting conditions (see

Endler 1991, Endler and Théry 1996).

Plasticity and its heritable variation

The overall pattern in plasticity matches the pattern found across populations. Males were more likely to express blue anal fins when raised in tea-stained water where there was lower transmission of UV and blue wavelengths. In a state-wide Florida census,

Fuller (2002) found more males with blue anal fins in populations with low transmission of UV and blue wavelengths. The congruence between these two studies demonstrates the robustness of this somewhat counterintuitive pattern. Males are obviously not maximizing perceived brightness (total number of photons reflected off of the color pattern and detected by a receiver) by expressing blue anal fins under conditions where blue does not transmit well and where animals possess fewer UV and violet retinal cones

(Fuller et al. in press). The most likely scenario is that blue males create high contrast with the water column or with other color elements on the body. If no photons are being detected from the anal fin, then it will appear black, which will produce high contrast with the body and visual background. On the other hand, clear water has high transmission of UV/blue wavelengths causing the water column to have a bluish tint.

This environment should create high contrast for yellow and red color morphs, but lower contrast for blue morphs. Another possibility is that blue males really are more conspicuous in clear water due to high, perceived brightness but that they suffer high mortality costs due to predation. I acknowledge that these speculations would benefit

81 from a proper analysis of reflectance spectra (Endler 1990, Bennet et al. 1994). I am currently analyzing reflectance spectra to compute actual brightness and contrast of blue anal fins in tea-stained and in clear water using the methods of Endler (1990).

The existence of heritable plasticity in blue morph expression raises several intriguing issues. Why don’t all males respond to the environment in the same manner?

Variation in plasticity may be maintained if the reliability of the cue used for plasticity varies (Weinig 2000a, 2000b). High variation in environmental conditions due to dry versus rainy seasons, variation in managed hydrology through canals, or variation in storms (i.e. hurricanes) could conceivably result in variation in the predictability of future conditions (Trexler et al. 2001). A second possibility is that there are costs to plasticity resulting in the maintenance of variation in plasticity (Via 1993, Dorn et al. 2000, Relyea

2002). Intuitively, the cost/benefit argument relies on a balance between fitness components maintaining variation within a population. A third possibility is that negative frequency dependence maintains variation in plasticity. If a rare-male mating advantage exists (due either to female choice or male-male competition), then this could result in the maintenance of variation in plasticity. This explanation relies on blue, yellow, and red morphs being acceptable alternatives in tea-stained water, but only yellow and red morphs being acceptable alternatives in clear water.

The actual mechanism for the induction of blue morphs is unknown. Based on the pattern among populations, I hypothesized that a reduction of either UV and/or blue wavelengths would be likely to trigger blue expression if plasticity was important.

However, these treatments had no effect in the laboratory experiment. In the Greenhouse experiment, I placed an emphasis on accurately mimicking spring populations (with clear

82 water) and swamp populations (with tea-stained water) to determine if there was any effect of environment on male color morph expression. This increased the likelihood of finding an environmental effect, but diminished my ability to determine the exact cue triggering the plasticity. I presume that the cue is a light signal such as the amount of a particular wavelength or the relative amount of light of certain wavelengths (e.g.

400nm/500nm ratio) that is detected by a light sensing organ (e.g. pineal gland, inner eye, retina, etc.). However, I cannot rule out the possibility that the cue is some attribute of the environment that is correlated with a reduction in UV/blue transmission such as some element of water chemistry.

In conclusion, I found that genetics, environment, and an interaction between genetics and environment affect color pattern expression in male L. goodei. Specifically,

I found evidence for a single, autosomal locus that controls the expression of yellow versus red where yellow is dominant over red. Lighting conditions also played a large role in the expression of blue anal fins. More males expressed blue anal fins when raised in the greenhouse than when raised in the laboratory. Furthermore, more males expressed blue anal fins when raised in tea-stained water than when raised in clear water.

In addition, sires had strong effects on whether male offspring expressed blue anal fins and also on the expression of plasticity. These results raise two questions. First, how is all of this variation maintained (red vs. yellow, genes for blue expression, genes for plasticity)? Frequency dependence, heterozygote advantage, variation in the environment, and/or a balance in fitness components are all possibilities. Second, given that the direction of plasticity is in accordance with across-population patterns, how can I explain this counterintuitive pattern (i.e. more blue males in low UV- and blue-

83 transmittance environments)? Males with blue anal fins are most likely maximizing contrast with background lighting conditions and sacrificing perceived brightness. The next step in this research program is to determine the fitness correlates of these color morphs under various lighting conditions.

84

CHAPTER 5

INTRASPECIFIC VARIATION IN ULTRAVIOLET CONE PRODUCTION AND

VISUAL SENSITIVITY IN THE BLUEFIN KILLIFISH, LUCANIA GOODEI

Abstract

Studies of visual ecology have typically focused on differences among species while paying less attention to variation among populations and/or individuals. Here, I show that

UV-sensitivity varies between individuals from two populations of bluefin killifish,

Lucania goodei. Animals from a swamp population, which has low transmission of UV and blue light, have lower absolute sensitivity to UV and violet wavelengths (360-440 nm) than animals from a spring habitat, which has high transmission of UV and blue light. Differences in sensitivity appear to arise through the reduced representation of UV and violet cones in the retinas of individuals from the swamp population. In addition, the expression of yellow and red cones also varies across populations, but whether this leads to differential sensitivity is unclear. The results have two important implications. First, the tight conservation of functional regions of opsin genes across taxa does not imply that visual systems are tightly constrained in their evolution; substantial differential sensitivity

85 can arise through differential expression of cone classes within the retina. Second, intraspecific visual signals in this species may evolve to maximize contrast between the signaler and the background (as opposed to brightness); males with blue anal fins are most abundant in swamp habitats where animals are less sensitive to UV/blue wavelengths and express fewer UV and violet cones. Ongoing work will determine whether these differences in sensitivity and cone expression have a genetic basis.

Introduction

Sensory drive predicts that natural selection favors adaptations of the sensory system and/or signal design to habitat conditions for efficiency of the communication system

(Endler 1992, 1993). Within the field of visual ecology, this hypothesis has been addressed through comparative studies that seek to link differences in the visual properties of animals from different populations (or species) with differences in the environmental conditions of their habitats (Partridge and Cummings 1999, Cronin et al.

2001, Cummings and Partridge 2001) and/or differences in their behavior (Boughman

2001). However, what constitutes an efficient system of communication is not always clear. This has lead to both studies showing a positive correlation between visual sensitivity at a given wavelength of light with the abundance of that wavelength (Lythgoe et al. 1999, McDonald and Hawryshyn 1995) and studies showing a negative correlation

(Boughman 2001) being taken as evidence in support of sensory drive. How should visual systems covary with environmental conditions?

A positive correlation will be generated between visual sensitivity and lighting conditions if selection favors visual systems that match the available wavelengths of light

86 in the environment. A fair amount of evidence supports this notion. In a number of aquatic animals, the spectral sensitivities of rods and some classes of cones in the retina are greatest for the wavelengths of light that are transmitted best in the habitat (Lythgoe

1984, Lythgoe et al. 1994, Hunt et al. 1996, Partridge and Cummings 1999, Yokoyama et al. 1999, Shand et al. 2002). There is also evidence that the intraocular filters animals use vary among habitats so that visual sensitivity is maximized for the available wavelengths of light (Cronin et al. 2001, Cronin and Caldwell 2002). Furthermore, in some cases there is evidence of differences in overall spectral sensitivity that correlate with habitat light conditions (McDonald and Hawryshyn 1995, Leal and Fleishman 2002).

There is also evidence to suggest negative correlations between environmental conditions and visual sensitivity. Laboratory studies suggest that animals can adjust cones to maintain a constant rate of photon-catch (i.e. photostasis) (for a review see Penn

1998). Under this process, a reduction in intensity at a particular wavelength results in a numerical increase in photoreceptors as well as an increase in absolute absorbance of photoreceptors to that wavelength (Penn and Williams 1986, Kröger et al. 1999). In a study of albino rat, Penn and Williams (1986) demonstrated that rats raised under high light conditions had less rhodopsin and shorter rod outer segments than rats raised under low light conditions. Similarly, Kröger et al. (1999) showed that Aequidens pulcher

(Cichlidae) raised in blue light had a lower proportion of blue sensitive cones in their retina than animals raised in red or green light, suggesting decreased sensitivity to blue when reared in a predominantly blue environment. In a comparison of stickleback populations, Boughman (2001) found that females from red shifted lighting environments

87 were less sensitive to red light (as measured by the optomotor response) than females from more blue-shifted lighting environments.

Finally, there is evidence suggesting that visual systems do not evolve easily (but see Nilsson and Pelger 1994). Recent comparative reviews of retinal photoreceptor classes in insects (Briscoe and Chittka 2001) and lizards (Loew et al. 2002) found little or no evidence of evolutionary divergence in the spectral sensitivity of individual photoreceptor classes across species occupying distinctly different lighting habitats.

More variation is found in aquatic organisms, but even here there is evidence that photoreceptors do not evolve easily (Cronin et al. 2002). For example, large regions of opsins are strongly conserved (Archer 1999, but see Shimmin et al. 1997) and the plasticity of cone spectral sensitivity is limited to shifts induced by chromophore usage

(Archer 1999, Partridge and Cummings 1999). In an experiment, Kröger et al. (1999) found no appreciable plasticity of cones in a cichlid. Animals were reared under two different lighting environments, and no differences were found in the normalized spectral sensitivities of cones between environments. Similar results have been found in mantis shrimp (Cronin et al. 2002). These studies lend some credence to the idea that there is low variation in some aspects of vision physiology.

The goal of this study is to determine whether there is variation among populations in vision physiology in the bluefin killifish, Lucania goodei. L. goodei is a freshwater fundulid found in a variety of lighting environments throughout Florida (Page and Burr 1991) ranging from crystal clear springs to tea-stained, turbid swamps (Fuller

2001, 2002). Males are highly polymorphic in coloration, and the relative abundance of the anal fin color morphs varies predictably with the lighting environment (Fuller 2002).

88 Males with blue anal fins are more abundant in populations with low transmission of UV and blue wavelengths. In contrast, males with red anal fins (and to a lesser extent, males with yellow anal fins) are more abundant in populations with high transmission of UV and blue wavelengths. The fact that male color patterns vary predictably across lighting conditions has ramifications for the visual system. Does vision physiology vary concordantly across lighting environments or do these different color patterns represent different ways of stimulating an invariant visual system under different lighting conditions?

In this study, I address this question by comparing the spectral sensitivity of individuals and the types and abundances of cones employed in the retina between animals from a spring population (high light, high UV/blue wavelength transmission) with animals from a swamp population (low light, low UV/blue wavelength transmission). Specifically, I test the following three hypotheses: 1) there is no variation in vision physiology across habitat type; vision physiology is fixed; 2) there is a positive correlation between lighting transmission and vision physiology; increased UV/blue wavelength transmission leads to increased visual sensitivity at UV/blue wavelengths; 3) there is a negative correlation between lighting transmission and vision physiology; decreased UV/blue wavelength transmission leads to increased sensitivity at UV/blue wavelengths in order to maintain a constant photon-catch.

Methods

Spectral sensitivity based on electroretinographic (ERG) flicker photometry

89 Animals were collected by RCF from a swamp population (26-Mile Bend, Everglades,

Broward, Co., FL, USA) and from a spring population (Wakulla Upper Bridge, Wakulla

River, Wakulla Co., FL, USA) in February 2001 and transported to Union College,

Schenectady, NY, USA. All animals were adults. Animals were maintained on a

12L:12D light schedule and fed frozen brine shrimp twice daily. All electroretinogram

(ERG) readings were recorded in February-March, 2001.

ERG flicker photometry methodology is detailed elsewhere (Fleishman et al.

1997, Jacobs et al. 1996). Animals were immobilized with an intramuscular injection of curare. Each individual was placed in a small holder with a wet sponge and was intubated so that aerated water flowed over its gills. ERGs were recorded differentially.

The active electrode consisted of a small stainless steel tube placed in contact with the cornea after application of xylocaine gel to the surface of the eye. The indifferent electrode was a platinum wire placed around the nape. The active electrode was mounted at the tip of a quartz fiber optic light guide through which the stimulus light was delivered, such that the stimulus light passed through the small stainless steel tube. The fiber optic was bifurcated and received input from two different source lights: a colored test light and a broad-band white control. The stimulus consisted of alternating flashes of equal duration of the colored test light and the control, with an equal period of dark (no stimulus) between each flash. The test light stimulus consisted of a monochromatic light the intensity of which could be varied over 5 log units using a linearly -variable optical quartz density neutral-density filter. The colored test stimuli were created by passing the focused output from a 300 W xenon arc lamp through a 1/8 m monochrometer resulting in monochromatic stimuli (10 nm 1/2-energy pass-band). The control consisted of a dim

90 light from a 50 W QTH fiber optic illuminator passed through a neutral density filter.

The test and control stimuli were passed through a spinning chopper wheel and into the two ends of the bifurcated fiber optic leading to the eye. These were positioned so that they created the alternating control-off-stimulus-off pattern described above. The entire alternating pattern was presented at a frequency of 4 hz.

Spectral sensitivity was measured by determining the test light intensity for a given wavelength that elicited a response equal in magnitude to that of the control stimulus. To determine the test intensity at which the response to test and control light were equal, the output was passed through a narrow electronic bandpass filter and centered on the stimulus frequency (4 hz). The output from the filter was a nearly sinusoidal signal with a frequency equal to the stimulus frequency. The phase of the sinusoid depended on whether the ERG response was greater to the control or to the test.

By adjusting the intensity of the test light stimulus, I could adjust the ERG output until the phase was intermediate and amplitude was minimal. This occurred when responses to the test and control stimulus flashes were equal. In order to find this equal response point, I repeatedly collected and digitized a buffer of four complete stimulus cycles in duration, which was an average of 20 repetitions. This entire process was performed for each wavelength in 40 nm steps from 360 to 640 nm.

Absolute sensitivity at each wavelength was measured as the inverse of the radiance of the test light when responses to test and control light were equal (1/radiance at criterion, with radiance measured in units of µmol m-2 sr-1 s-1, where 1 µmol = 6.02 x

1017 quanta. The absolute sensitivity was computed for each subject for light from 360-

640 nm. I calculated the overall absolute sensitivity by averaging absolute sensitivity

91 across all wavelengths. I also calculated the relative sensitivity at each wavelength by dividing all values in the absolute sensitivity curves by the maximum value for each individual. Hence, the wavelength where each individual is maximally sensitive is scored as 1. I compared the absolute sensitivity between the two populations provided that there were no statistically significant differences in overall sensitivity. I also examined the relative sensitivities to ascertain whether this variable produced the same pattern across populations.

I compared the curves between two populations using analysis of variance provided that variances were not significantly heteroscedastic as indicated by a Barttlett’s test of homogeneity. Otherwise, I used a non-parametric Kruskal-Wallis test. I present the unadjusted p-values, but also consider the p-values after a sequential bonferroni adjustment (Sokal and Rohlf 1995) to control for the accumulation of type 1 error, where

I controlled for 16 tests (8 wavelengths, 2 variables).

Microspectrophotometry

Animals from a swamp population (26-Mile Bend) and a spring population (Wakulla) were collected and transported to Cornell University, Ithaca, NY in October 2001.

Animals were maintained on a 12L:12D ratio at 21C and fed twice daily.

Microspectrophotometry (MSP) readings were taken in October 2001. All animals were adults. In addition, one swamp and two spring animals were sent to Cornell for MSP analysis March 2001.

92 MSP measurements were performed using methods identical to those described in

Loew (1994) and Provencio et al. (1992). All procedures were carried out under infrared illumination using appropriate image converters and video cameras. Animals were dark adapted for a minimum of 1 h after which they were euthanized. Enucleated eyes were hemisected and pieces of retina were immersed in a simple Sorensen’s phosphate buffer

(pH 7.2) with 6% sucrose or dextran added. The retinas were carefully teased from the retinal pigment epithelium and macerated using razor blade fragments and tungsten needles. A drop of the dispersed retina was sandwiched between two cover slips and transferred to the stage of the MSP. The MSP itself has been described in detail elsewhere (Loew 1994, Loew et al. 2002). A 100 W tungsten-halogen lamp together with quartz and mirror optics allowed for accurate absorbance measurement down to 340 nm with a rectangular measuring aperture as small as 1.5µm2.

I used template fitting to determine λmax (the wavelength at maximum absorbance for a template-derived visual pigment best fitting the experimental data). The rationale for using template fitting (as opposed to simply using the peak absorbance as a measure of λmax) is that the entire absorbance curve is informative as to the true λmax. Hence, the precision can be increased by using all of the data. Determination of λmax involves six steps: (1) smooth the data, (2) determine the peak absorbance (Xmax), (3) normalize the absorbance curve, (4) fit the templates, (5) calculate the standard deviation (SD) of λmax,

(6) compare with the actual data and choose the best fit.

The raw data were first smoothed using a digital filter routine (“smooft” Press et al. 1989). The smoothed spectrum was overlaid on the raw data and checked by eye to make sure that over-filtering or spurious data points had not shifted the apparent

93 maximum. If this was the case, then the unsmoothed data were used. Next, the peak absorbance was determined and used in the normalization. The peak absorbance (Xmax) was the calculated maximum of the best fit Gaussian to the data points 20nm either side of the estimated-by-eye absorbance maximum of the alpha band. Using Xmax, the data were then normalized using the method of Mansfield (1985) as presented by MacNichol

(1986). Normalized data were then fit using the A1 and A2 templates of Lipetz and

Cronin (1988). These templates allowed us to generate 40 estimates of λmax from the long-wavelength limb (absorbance data for wavelengths slightly greater than λmax) and 30 estimates of λmax from the short-wavelength limb (absorbance data for wavelengths slightly less than λmax). Mean λmax + standard deviation (SD) was then determined using the short-wavelength limb estimates, the long-wavelength estimates, and the combined estimates. For each of these three λmax values, a template curve was fit to the original data. A decision as to which fit best was made by visual examination. The template fit having the lowest SD usually had the best visual fit. The curve was discarded if the SD of

λmax was greater than 7.5 nm (see Sillman et al. 1999, 2001 for similar criteria). This process was repeated for each microspectrophotometer curve, after which the λmax values for each curve of a spectral class were averaged to yield a final estimate of mean λmax +

SD.

I compared the presence and absence of cone types between individuals of the two populations using X2 tests. For each individual, I calculated the relative frequencies of all cone types (i.e. #uv cones recorded/# total cones recorded), the mean λmax + SD for each cone class, and the coefficient of variation in λmax (CV λmax). I used analysis of variance to compare populations provided that variances were not significantly

94 heteroscedastic as indicated by a Barttlett’s test of homogeneity. Otherwise, I used a non-parametric Kruskal-Wallis test. All probabilities are two-tailed and considered significant at p < 0.05. I also present results after considering a sequential bonferroni adjustment for 15 comparisons (5 cone classes, 3 variables). All analyses were performed with SAS V.8 statistical software (SAS Institute, Inc., Cary, NC, USA).

Results

ERG flicker photometry

Absolute sensitivity to ultraviolet (UV) and blue wavelengths was higher for spring animals than for swamp animals (Figure 5.1a). Specifically, animals from the spring population were more sensitive at 360 nm (F1,11 = 4.99, P= 0.0473), 400 nm (F1,11 =

6.09, P = 0.0312), and 440 nm (Kruskal-Wallis X2 = 3.927, df = 1, P = 0.0475).

However, none of the comparisons were statistically significant after a sequential bonferroni adjustment. There were no significant differences between populations in overall absolute sensitivity (F1,11 = 1.45, P = 0.2545). Examination of relative sensitivity shows that spring animals were more sensitive in the UV/blue wavelengths (360 - 440 nm) and swamp animals were more sensitive in long wavelengths (>560 nm), although these differences were not statistically significant (Figure 5.1b, P > 0.12 in all tests).

Microspectrophotometry

Five cone classes were present in both populations (UV λmax = 359 + 4 nm, n = 68, violet

λmax = 405 + 3 nm, n = 382, blue λmax = 455 + 6 nm, n = 86, yellow λmax = 539 + 6 nm, n

= 226, red λmax = 573 + 7 nm, n = 262). Most cones best fit an A1 template (88.6%) 95

A 10

1 absolute sensitivity absolute

* * + 320 360 400 440 480 520 560 600 640 680 wavelength (nm)

1 B

0.75

0.5

0.25 relative sensitivity

0 320 360 400440 480 520560 600 640 680 wavelength (nm)

Figure 5.1. (A) Absolute sensitivity for the spring and swamp populations. Units are 1/radiance at criterion with radiance in units of µmol m-2 sr-1 s-1. (B) Relative sensitivity for the spring and swamp populations. Means and standard errors are shown. Open symbols denote spring values. Filled symbols denote swamp values. N = 8 for all spring values. N = 5 for all swamp values except for wavelength 640 nm where N = 4. * significantly different with ANOVA, p < 0.05. + significantly different with Kruskal- Wallis, p < 0.05. No comparisons are statistically significant after a sequential bonferroni adjustment. Points are jittered for the purpose of display.

96 while a smaller proportion best fit A2 (11.43%) (see electronic appendix for individual data). Neither average λmax nor the coefficient of variation in λmax (CV λmax) differed between the two populations (Table 5.1, average λmax: population * cone class F4, 86 =

1.02, P < 0.403; CV λmax: population * cone class F4, 86 = 1.00, P < 0.399).

All violet and UV cones were single cones. All red cones were members of double cones. The vast majority of yellow cones were members of double cones. I detected one single, yellow cone that I presume was a double cone that became detached from its complementary cone. Similarly, the vast majority of blue cones were members of double cones. I detected 11 single blue cones (70 double blue cones, 5 blue cone tip fragments) that I presume were actually detached elements of double cones. There were no statistically significant differences between λmax of single and double blue cones

(single blue λmax = 454 + 7, double blue λmax = 454 + 5). Finally, I also measured λmax on

10 detached fragment tips (5 blue, 3 red, 2 violet). Obviously, I could not assign a cone type to these pigments. Blue fragment tips had a slightly higher λmax than either the single blue or double blue cones (F2,83 = 3.69, P < 0.029, blue fragment tip λmax = 462 +

11). Similarly, red fragment tips had a slightly higher λmax than red double cones (F1,260 =

10.92, P < 0.0011, red double cone λmax = 573 + 7, red fragment tip λmax = 586 + 4).

All animals had double cones. I obtained MSP readings for both cones from 146 sets of double cones. There were three distinctly different types of double cones. I found

121 double cones where a yellow cone was paired with a red cone. I also found 23 double cones where a blue cone was paired with a yellow cone. I found two twin double cones. One involved a pair of red cones (λmax 566/568 nm). Another involved a pair of

97

Table 5.1. Mean and coefficient of variation (CV) of λmax calculated across individuals for the spring and swamp populations. N = 11 for spring. N = 10 for swamp. Note that CV λmax is approximately 1% for all opsin classes.

opsin class spring swamp spring swamp

mean λmax (SD) mean λmax (SD) CV λmax (SD) CV λmax (SD)

ultraviolet 359.38 (2.39) 359.45 (7.14) 1.072 (0.408) ------

violet 405.13 (1.56) 404.73 (2.28) 0.716 (0.167) 0.806 (0.249)

blue 453.59 (3.25) 456.38 (7.95) 0.910 (0.313) 0.744 (0.620)

yellow 537.51 (2.71) 540.92 (3.80) 0.940 (0.369) 1.046 (0.240)

red 572.58 (2.43) 573.17 (1.96) 0.949 (0.234) 1.149 (0.364)

98 far-red sensitive cones (λmax 584/586). In addition to the 146 sets of double cones (where

I measured both cones), I also measured 262 cones (47 blue, 81 yellow, 134 red) that were members of double cones for which I was unable to measure the complementary cone.

UV cones were found much more readily in animals from the spring population

(11/11) than they were in animals from the swamp population (2/10) (Figure 5.2, Pearson

X2 = 14.22, df = 1, P = 0.0002). This result was not attributable to the fact that I measured more cones per animal for the spring population than for the swamp population

(total cones per animal: F1,19 = 6.33, P = 0.0210, spring: 55 + 11 , swamp: 41 + 13).

Assuming that the frequency of UV cones in the swamp is 0.11 (i.e. the same as in the spring), the probability of not finding an UV-cone in a single individual where a minimum of 25 cones were measured was 0.054 ( (1-0.11)^25). Out of 10 animals, the expected number of animals in which no UV cones would be 0.54 (0.054*10). Hence, I would expect to find UV cones in at least 9 individuals. This conservative analysis demonstrates that these results cannot be attributed to differences in sampling.

Finally, an analysis of the mean cone frequencies indicates that UV and violet cones were more abundant in the spring population and that yellow and red cones were more abundant in the swamp population (Table 5.2).

Discussion

The MSP analysis led us to conclude that there were differences in the numbers of different photoreceptor types between the two populations. The ERG-based spectral sensitivity curves largely reflect the relative numbers of photoreceptors covering different

99 swamp spring 6 16 8 20 26 cones 12 69 cones 6 27 cones 15 50 cones 4 8 4 10 2 4 2 5

0 0 0 0 300 400 500 600 300 400 500 600 300 400 500 600 300 400 500 600 8 8 12 20 53 cones 33 cones 40 cones 57 cones 6 6 9 15 6 4 4 10

2 2 3 5

0 0 0 0 300 400 500 600 300 400 500 600 300 400 500 600 300 400 500 600

15 12 25 40 53 cones 54 cones 20 9 59 cones 30 68 cones 10 15 6 20 10 5 3 5 10 frequency 0 0 0 0 300 400 500 600 300 400 500 600 300 400 500 600 300 400 500 600 6 8 15 20 25 cones 5 45 cones 6 60 cones 49 cones 4 10 15 3 4 10 2 5 2 5 1 0 0 0 0 300 400 500 600 300 400 500 600 300 400 500 600 300 400 500 600 8 16 20 20 30 cones 56 cones 39 cones 66 cones 6 12 15 15

4 8 10 10

2 4 5 5

0 0 0 0 300 400 500 600 300 400 500 600 300 400 500 600 300 400 500 600 16

wavelength (nm) 12 65 cones

8

4

0 300 400 500 600 wavelength (nm)

Figure 5.2. Cone profiles for animals from the swamp and spring population. Each graph is a histogram of λmax cone values from a single animal. Arrows indicate missing cone classes. The total number of measured cones is indicated in each graph.

100

Table 5.2. Mean frequencies for each cone class in each of the two populations. N = 11 for spring. N = 10 for swamp. Values in bold indicate statistically significant differences after bonferonni adjustment at P < 0.01.

opsin class spring mean swamp mean

frequency (SD) frequency (SD)

ultraviolet 0.111 (0.054) 0.006 (0.013)

violet 0.450 (0.072) 0.267 (0.102)

blue 0.080 (0.037) 0.088 (0.053)

yellow 0.177 (0.033) 0.279 (0.044)

red 0.182 (0.044) 0.360 (0.081)

101 wavelength ranges (Jacobs et al. 1996). This data was consistent with, and corroborated the qualitative differences in MSP. Absolute sensitivity was greater from 360-440 nm for the spring population that had a higher frequency of UV and violet cones. There was also evidence to suggest intraspecific variation in the relative frequency of yellow and red cones, but the link with ERG sensitivity is less strong. However, the match between

ERG and the quantitative differences in MSP is less precise. In the spring population, the most abundant photoreceptor was the single, violet cone. Yet, according to ERG, spring animals were most sensitive to light at 560 nm. I cannot explain this discrepancy. I believe the qualitative pattern found in this study is robust (i.e. more UV and violet cones in spring animals, more yellow and red cones in swamp animals). However, whether

MSP is a precise tool for measuring the exact retinal make-up of various cone-types has yet to be resolved.

The main conclusion from this study is that there is intraspecific variation in visual sensitivity in the UV/blue wavelengths in L. goodei. To the best of our knowledge, this is the first demonstration of variation in UV vision among populations within a species. In addition, our results also provide evidence for a positive matching between the available wavelengths of light and the relative abundance of photoreceptors.

In some fish, UV cones are express in juveniles but then lost in the adult stage

(Bowmaker and Kunz 1987, Bowmaker 1990, reviewed in Beaudet and Hawryshyn

1997). I doubt that the differences among L. goodei populations in this study were due to differential effects of ontogeny. First, all of the experimental animals were adults.

Second, animals from the spring population (which had more UV cones) are larger than animals from the swamp (Fuller, unpubl. data). Although the manner in which body size

102 translates into age is unknown in L. goodei, I seriously doubt that spring animals were significantly younger than the swamp animals.

The presence of both UV and violet cones in the same individual is striking because usually only one short-wavelength-sensitive I (UV) opsin is expressed in fish.

Many fish possess UV cones and lack violet ones (Cichlidae: van der Meer and

Bowmaker 1995, Carleton and Kocher 2001; Adrianichthyidae: Hisatomi et al. 1997;

Cyprinidae: Palacios et al. 1998, Hisatomi et al. 1996; Pleuronectidae: Helvik et al.

2001). In addition to L. goodei, the mummichog, Fundulus heteroclitus, the guppy,

Poecilia reticulata, and a cyprinid, Danio aequipinnatus, have both ultraviolet cones

(λmax 363, 389, 358 respectively) and violet cones (λmax 400, 408, 408 respectively)

(Archer et al. 1987, Archer and Lythgoe 1990, Flamarique and Hárosi 2000, Palacios et al. 1996). Molecular analyses suggest that violet cones in chicken and frog evolved from

UV sensitive opsins (SWS1) found in fish (Hisatomi et al. 1996, Yokoyama and

Yokoyama 1996, Hisatomi et al. 1997, Yokoyama 1997, Hunt et al. 2001). Either a gene duplication of the SWS1 opsin has occurred with subsequent divergence in function or the violet opsin in these fish has evolved de novo from a different opsin class.

Differences in UV/blue wavelength sensitivity appear to be driven by differential expression of UV and violet cones. Carleton and Kocher (2001) have recently shown in cichlids that closely related species expressing nearly identical opsins can vary by more than 10-fold in relative expression levels of opsins (and, therefore cones). The preponderance of variation in vision physiology and visual sensitivity across multiple populations and habitat types (McDonald and Hawryshwyn 1995, Boughmann 2001,

Cronin et al. 2001, Rodd et al. 2002, this study), along with the knowledge that closely

103 related species can vary greatly in opsin expression (Carleton and Kocher 2001) means that the possibility of readily available genetically based variation in vision systems should not be neglected.

A curious pattern is emerging with respect to the male color patterns employed by

L. goodei. There is an inverse relationship between visual sensitivity to UV/blue and the abundance of blue morphs within a population. Males with blue anal fins are most common in populations where UV/blue wavelengths attenuate quickly (Fuller 2002) and where animals are not very sensitive to those wavelengths (this study). Perceived brightness of blue anal fins can clearly not be greater in swamp populations. The most likely scenario is that blue males create high contrast with the water column or with other color elements on the body. If indeed no photons are being detected from the anal fin, then it will appear black, which will produce high contrast with the body and visual background. Such contrast is detected most effectively by those photoreceptors whose sensitivity matches the background illumination. Thus, the blue and UV fins in the swamp habitats may be effectively stimulating the long wavelength photoreceptors by contrasting with the bright background. On the other hand, spring populations have high transmission of UV/blue wavelengths causing the water column to have a bluish tint.

This environment should create high contrast for yellow and red color morphs, but lower contrast for blue morphs, and in this case the short wavelength photoreceptors (whose sensitivity matches the backlight) may be effectively stimulated by the contrast between the backlight and animal color patterns. Similar patterns have been found across species in Anolis lizards (Leal and Fleishman 2002). Anolis cristatellus males are found in habitats with less UV background radiance yet have dewlaps that reflect strongly in the

104 UV, while A. cooki males are found in habitats with higher UV background radiance and have dewlaps that lack UV reflectance. Males appear to be optimizing contrast with background lighting conditions even if this results in a decrease in overall brightness.

In conclusion, L. goodei populations vary in their sensitivity to UV/blue wavelengths and in the relative expression of UV and violet cones. In addition, the expression of yellow and red cones also varies across populations, but whether this leads to differences in sensitivity is unclear. The results match previously described patterns of interspecific variation (more sensitivity to more commonly encountered wavelengths)

(Lythgoe 1984, Lythgoe et al. 1994, Hunt et al. 1996, Partridge and Cummings 1999,

Yokoyama et al. 1999) better than they match patterns of compensation found in other studies (wherein increases in sensitivity occur in response to reductions in specific wavelengths) (Penn and Williams 1986, Kröger et al. 1999, Boughman 2001). Our results imply that the tight conservation of opsins and photoreceptor spectral sensitivity across taxa does not imply that visual systems are tightly constrained in their evolution; substantial differential sensitivity can arise through differential expression of cone classes within the retina. I am currently determining the degree to which variation in UV and violet cone expression is determined by environmental and/or genetic factors.

105

CHAPTER 6

RELATIVE OPSIN EXPRESSION REFLECTS POPULATION DIFFERENCES

IN VISION PHYSIOLOGY IN LUCANIA GOODEI: A REAL-TIME PCR STUDY

Abstract

Studies in the evolution of vision have relied primarily on across species comparisons

(i.e. the comparative method). The use of quantitative genetic techniques (i.e. heritability studies, selection studies, etc.) has been precluded due to the difficulty in objectively measuring vision physiology in large numbers of individuals. In this study, I examine the effectiveness of a molecular technique, real-time PCR, in inferring components of vision physiology in the bluefin killifish, Lucania goodei, to determine whether differences in relative opsin expression correlate with differences in relative cone frequency. A previous microspectrophotometry study showed that animals from a spring population possess a higher relative frequency of UV and violet cones and a lower frequency of yellow and red cones than animals from a swamp population. In this study, I found a good qualitative match between the relative opsin expression and relative cone frequency.

Spring animals expressed higher amounts of UV and violet opsins, whereas swamp animals expressed higher amounts of yellow and red opsins. However, as a quantitative measure, relative opsin expression is not a valid measure of actual cone frequency (e.g.

106 relative blue opsin expression does not equal blue cone frequency). Still, using real-time

PCR to measure vision physiology for quantitative genetic studies will allow researchers to answer the questions of whether there is heritable variation in relative opsin expression and relative cone abundance. Real-time PCR represents a strong tool for studies requiring measurements of vision physiology in large numbers of individuals.

Introduction

Understanding how natural selection acts on vision physiology is an important endeavor not only for understanding the evolution of vision itself (Hisatomi 1994, Yokoyama &

Yokoyama 1996, Yokoyama 1997, Hunt et al. 2001), but also for understanding the evolution of mating preferences (Bennett et al. 1994, Macias Garcia & de Perera 2002), sensory biases (Endler 1992, 1993, Endler & Basolo 1998, Ryan & Keddy-Hector 1992), foraging ecology (Regan et al. 2001, Garamszeg et al. 2002, Martin & Prince 2002), and even speciation (Endler & Houde 1995, Seehausen et al. 1997, Thorpe & Richard 2001,

Servedio 2001). However, studies of the evolution of vision typically rely on the comparative method -- across species comparisons of vision physiology (Lythgoe et al.

1994, van der Meer & Bowmaker 1995, Yokoyama & Yokoyama 1996, Fleishman et al.

1997, Cummings & Partridge 2001). To date, there are few estimates of selection differentials (or gradients) and few estimates of heritabilities for most aspects of vision physiology (but see Endler et al. 2001).

The lack of quantitative genetic studies in vision physiology is due in large part to our inability to thoroughly measure large numbers of individuals (as required in quantitative genetic and selection studies) (Falconer & MacKay 1996, Lynch & Walsh

107 1998). Endler et al. (2000) did perform a selection study on optomotor response where they selected for increased sensitivity to blue and red wavelengths. Similarly,

Boughmann (2001) measured optomotor response for animals from multiple populations and found a negative correlation between sensitivity to red light and the degree to which water color is red-shifted. The limitation of the optomotor response is that it primarily measures the sensitivity to long wavelengths (Schaerer & Neumeyer 1996). In addition, as a behavioral measure of sensitivity, the optomotor response provides little information on how the underlying vision physiology varies. Microspectrophotometry (MSP) has effectively been used to determine which types of cones animals possess in their retinas.

This has been a strong tool in species comparisons (Lythgoe et al. 1994, van der Meer &

Bowmaker 1995, Cummings & Partridge 2001) and also in studies of plasticity (Cronin et al. 2001). The drawback of MSP is that it mainly provides data on the presence/absence of cone types, but is less effective in measuring the relative abundance of various cone types. In addition, MSP is very time-consuming. Other researchers have relied on electroretinogram methods (ERG) where the sensitivity of animals to various wavelengths of light is measured across the entire spectrum (Leal and Fleishman 2002).

This method is also exceedingly time consuming. In addition, the measures of sensitivity rely on comparing the individuals response to a wavelength of interest with its response to a control. As a consequence, choosing the control stimulus can have an effect in the interpretation of results (Jacobs et al. 1996). Finally, some researchers have studied the filtering properties of the lens, cornea, oil droplets, and ellipsosomes (Thorpe & Douglas

1993, Flamarique & Harosi 2000, Cronin & Caldwell 2002). These measures may be

108 more amenable to quantitative genetic studies. However, to date, no such studies have been conducted.

In this study, I examine the effectiveness of a molecular technique, real-time

PCR, in inferring components of vision physiology in the bluefin killifish, Lucania goodei. Real-time PCR measures relative gene expression. The rationale is as follows.

The spectral properties of cones depend on the photopigments that they contain.

Photopigments consist of pairing a vitamin A molecule (either A1 or A2) with an opsin protein. While shifts between A1 and A2 may induce a slight change (Bridges 1972,

Munz & McFarland 1977), the opsin properties are the main determinants of cone spectral properties (Yokoyama 1996, Yokoyama & Yokoyama 1997, Yokoyama et al.

1999, Partridge & Cummings 1999). Each opsin is a product of a single gene. Therefore, by measuring the relative expression of opsin genes, one hopes to make inferences concerning the relative abundance of cone types. Carleton & Kocher (2001) used real- time PCR to measure relative opsin expression in four cichlid species and found that different cichlids use different subsets of genes. They suggested that gene regulation (as opposed to opsin differentiation) may create different spectral sensitivity in these species.

The objective of this study was to determine whether the relative expression of opsin mRNAs (measured with real-time PCR) matches the qualitative and quantitative pattern of relative cone frequencies found in bluefin killifish, Lucania goodei, from a spring and a swamp population. Is there higher expression of UV and violet opsins in spring individuals than in swamp individuals? Is there lower expression of yellow and red opsins in spring animals than in swamp animals? Does the average proportion of the

109 various classes of opsins match the average proportion of various cone classes in the retina?

Study System

The bluefin killifish, Lucania goodei, is an excellent system to study the evolution of vision. Animals live under a variety of lighting conditions ranging from springs where the water is crystal clear with high transmission of UV and blue wavelengths to tea- stained swamps where the transmission of UV and blue wavelengths is greatly reduced

(Fuller 2002). We have a good basic understanding of vision physiology in this species.

L. goodei has five main cone types: UV(λmax 359 nm), violet (λmax 405 nm), blue (λmax

455 nm), yellow (λmax 539 nm), red (λmax 573) (Fuller et al. in press). The relative abundance of cones in the retina varies between populations with different lighting environments. In an MSP study comparing a spring and a swamp population, spring animals were much more likely to possess UV cones (chapter 5). In addition, the relative abundance of cones varied such that animals from the spring population had a higher frequency of UV and violet cones and a lower frequency of yellow and red cones.

Finally, there is good reason to suspect that visual communication is important in this species. The sexes are dimorphic in coloration where males are conspicuously colored and females are cryptic (Foster 1967, Page & Burr 1991, Fuller 2001). In addition, males are polymorphic in dorsal, anal, and pelvic fin coloration (Fuller 2001, 2002). The relative abundance of these morphs varies across populations in relation to the lighting environment (Fuller 2002). Males with blue anal fins are much more abundant in tea- stained swamps where UV and blue wavelengths do not transmit well. In contrast, males 110 with red anal fins (and to a lesser extent, males with yellow anal fins) are more abundant in populations where there is high transmission of UV and blue wavelengths.

Methods

I collected L. goodei from a spring population (Wakulla River, Upper Bridge, Wakulla

Co., FL, USA) and a swamp population (26-Mile Bend, Broward Co., FL, USA)

September 5-9, 2002. Animals were then transported to the University of New

Hampshire. Twelve animals from each population were euthanized and their eyes were removed for RNA isolation. In addition, their sex and standard length (hereafter referred to as size) were recorded. Two animals from each population were used for sequencing and 10 animals were used for real-time PCR.

Both eyes of each individual were homogenized in 1 ml of Trizol solution. The

RNA was then removed, precipitated with isopropanol, and resuspended in 50 ul of

DNAse/RNAse free water. The RNA was later stored at -80C. cDNA was obtained by reverse transcribing the RNA. For each individual, 1 ug of total RNA was used in a 25 ul reaction. Messenger RNA was isolated using a poly-T primer which provided the binding site for the reverse transcriptase (SuperScript II).

Real-time PCR

Data on opsin sequences were obtained from Blows and Yokoyama (unpublished data).

Because Blows and Yokoyama obtained sequences only from spring animals, I also obtained sequences from a small portion of the opsins for 2 animals from each population to ascertain that there were no differences in sequence at the primer and probe sites.

111 Primer express software (ABI version 1.5) was used to design the primers and probes (Table 6.1). Each probe covered an exon-exon boundary so that only cDNA could be amplified in RT-PCR. Three different loci (LWSA-1, LWSA-2, and LWSB) confer sensitivity in 560-580nm. All three loci are identical at one exon-exon boundary. Hence, a set of primers and probes were designed that were common to all three loci. Forward and reverse primers were located so that the product length was relatively short (87-109 bp). Sequencher was used to verify that the primers and probes were unique to each opsin.

Our experimental protocol followed Carleton & Kocher (2001). For each real- time PCR (RT-PCR), 0.2 ul of cDNA was placed in a 30 ul reaction with the appropriate primers, probes, and taqman mix. Probes were 5'-labeled with 6' Fam and 3'-labeled with

TAMRA. During PCR, the primer disrupted the probe and released the 3' TAMRA allowing it to fluoresce. The amount of fluorescence was monitored over 40 cycles

(94C-15s/55C-30s/65C-1 minute). RT-PCR was performed using the ABI Prism 7700

Sequence Detection System at Florida State University.

The relative abundance of each opsin in an individual cDNA mixture was based on its critical cycle number following the methodology of Carleton & Kocher (2001).

Critical cycle number was determined when the fluorescence exceeded a threshold set close to the background fluorescence. Genes having high expression have a smaller critical cycle numbers than genes with low expression. The relative opsin expression was calculated as a fraction of total opsin genes for an individual according to the following:

112 Table 6.1. Primers and probes.

opsin forward primer probe reverse primer

(λmax)

UV TTACACCTTGTG CCGTAGCAGGCCTG GGGTTTGCAGATGACC

(359 nm) TGCCTTGGAA GTGACGTCCT AGGTAC

violet GCTGCAAGATTG GGTGTTGGTGGCAT CCAACCATCTTTCGAA

(405 nm) AAGGATTTACTG GGTCAGCCTTTG TGCAA

blue CATGCAAGATTG ACACTAGGGGGTAT CCAGCCATCGTTCAAA

(455 nm) AAGGTTTCATTG GGTAAGCCTGTGGT AGCT

CTCT yellow CTTCTGCGGTAT AACACTCGGAGGTG AACAATATATCTCTCA

(539 nm) TGAGGGATTC AGGTTGCTCTCTGG ATAGCCAGAACAA

T red TGGTGTGCTCCT TGGAGCAGGTATTG TCTTCACTTCCACTGA

(573 nm) CCCATCTT GCCCCATGGAC ACACATCAG

113 1 T 1( + E )Cti i = i . Tall 1 Cti ∑ 1( + Ei )

Ti is the proportional gene expression for a given gene i. Ei is the PCR efficiency for Tall each primer/probe set, and Cti is the critical cycle number for each gene.

Efficiencies were calculated for 3 individuals from each population (6 animals total). For each individual, a dilution series of the cDNA mixture was performed (1 ul,

0.5 ul, 0.2ul, 0.1ul, 0.05ul, 0.01ul, 0.005ul, 0.001ul cDNA) and used in RT-PCR for each opsin. The slope of ln (concentration) on critical cycle number allows us to calculate E as ((exp(-slope))-1). Efficiencies were compared between the opsin primer/probe sets, populations, and opsin primer/probe sets among populations using ANOVA.

The repeatability of the measurements was calculated to assess measurement error. Three replicate real-time PCRs were performed using 0.2 ul cDNA for each opsin for 3 individuals from each population. In addition, all RT-PCR reactions deamed likely to be outliers were flagged and re-ran. Because the samples had generic codes, this analysis was performed blind to individual population of origin, sex, and size.

Repeatability was measured as the proportion of variation accounted for by individual effects. The coefficient of variation (CV) was calculated across individuals and across all pooled data, as well as among measures within each individual.

The effect of population of origin, size, and sex on relative gene expression was examined for each of the 5 opsins using analysis of covariance. In the original analysis, all interactions were examined. Provided that they were not statistically significant, they were removed them from the model. For individuals used in the repeatability analysis, 114 the mean of the three replicate measures was used. For individuals used to calculate efficiency, the values obtained from the reactions using 0.2 ul cDNA was used. Finally, relative opsin expression was compared with relative cone frequency. Means and standard erros were calculated for relative cone frequency based on raw data presented in an earlier MSP study (Fuller & colleagues, subm. man.). Results were considered significant at P < 0.05. All analyses were performed using SAS V.8.

Results

Efficiencies

Table 6.2 shows the efficiency for each of the five opsin primer/probe sets in both the spring and the swamp population. There were no statistically significant differences in efficiency among the different opsin primer/probe sets (F4,20 = 0.467, P = 0.760), nor among populations (F1,20 = 1.831, P = 0.191), nor between the interaction of population and opsin primer/probe set (F4,20 = 1.310, P = 0.300). The average efficiency was 1.109

(95% confidence limits 1.05791 - 1.162). In theory, the efficiency cannot be greater than

1, so I set all efficiencies for all opsin primer/probe sets equal to 1.

Outliers and Repeatability

We identified one of the real-time PCR reactions in the repeatability data as a probable outlier. Inclusion of this data point in the analysis bears this out. In the repeatability analysis of the blue opsin, the studentized residual was 1666.2. A value greater than 2 indicates a potential outlier. This data point was omitted from the analysis unless otherwise noted.

115

Table 6.2. Efficiencies for the spring and swamp populations. Means and standard errors are shown. N=3.

opsin (λmax) spring swamp

efficiency efficiency

UV (359 nm) 1.021 (0.015) 1.097 (0.045)

violet (405 nm) 1.053 (0.007) 1.125 (0.077)

blue (455 nm) 1.194 (0.090) 1.051 (0.172)

yellow (539 nm) 1.060 (0.048) 1.265 (0.033)

red (573 nm) 1.049 (0.116) 1.184 (0.044)

116 Table 6.3 shows the repeatability of our measurements both with and without inclusion of the far-outlier. When the outlier is removed, measures of opsin expression were significantly repeatable for all opsins except yellow (Table 6.3). The non- significance of yellow is most likely attributable to the fact that yellow had the lowest coefficient of variation across both individual means (Table 6.3). Blue opsin repeatability tended to be statistically significant. The marginal p-value for this opsin is most likely attributable to the large coefficient of variation within individuals.

Differences in gene expression

Population had a large effect on relative gene expression for all opsins except blue

(Figure 6.1). In general, the spring population was shifted towards greater relative gene expression of opsins that absorbed in the UV and violet range (350-410), whereas the swamp population was biased towards greater relative gene expression of opsins that absorbed in yellow in the red range (530-580) (Figure 6.1). The initial analyses of covariance found no significant interactions between size-by-population, size-by-sex, population-by-sex, or size-by-population-by-sex for relative gene expression of any of the five opsins (P>0.10 for all tests). Hence, the interaction terms were removed from the models. Both the UV and violet opsins had higher expression in fish from the spring population than those from the swamp population (UV: F1,16 = 28.52, P < 0.001; violet:

F1,16 = 12.92, P = 0.00243). In addition, for violet opsin expression, smaller animals and females tended to have slightly higher expression, but these effects were not statistically significant (size: F1,16 = 3.34, P = 0.0863; sex: F1,16 = 3.52, P = 0.0792). There were no such effects for UV. For the blue opsin, there were no effects of population, sex, or size

(P > 0.40 for all tests, spring average = 0.0013 (0.00022 SE), swamp average = 0.0012

117

Table 6.3. Repeatabilities and coefficients of variation (CV). Coefficients of variation are calculated without the far outlier. N=6 for CV (across individual means). N=18 for CV (all measures). Average CV (within samples) is the average CV across the six individuals. Average CV (within samples) represents experimental error whereas CV (individual means) represents the true coefficient of variation among individuals.

opsin (λmax) repeatability repeatability CV CV average

with far without far (individual (all CV

outlier outlier means) measures) (within

samples)

UV (359 nm) 81.2** 83.2** 41.5 42.8 15.6 violet (405 nm) 73.4** 72.1** 36.3 40.1 18.3 blue (455 nm) 20.8 53.6+ 48.7 62.6 34.5 yellow (539 nm) 37.0 43.0 10.6 15.2 12.2 red (573 nm) 37.2 93.6** 28.2 27.4 8.0

** P < 0.01

* P < 0.05

+ 0.1 < P <0.05

118

0.5 spring swamp 0.4

0.3

0.2

relative expression relative 0.1

0.0 UV violet blue yellow red

opsins

Figure 6.1. Average relative expression for UV, violet, blue, yellow, and red opsins for a spring (open bars) and a swamp (dark bars) population. Means and standard errors are shown. N=10 for each population.

119 (0.00020 SE)). In contrast, both the yellow and red opsins had higher expression in fish from the swamp population (yellow: F1,16 = 4.42, P = 0.052; red: F1,16 = 6.88, P =

0.0185). Neither sex nor size accounted for significant amounts of variation the expression of either yellow or red opsins.

Table 6.4 compares the relative opsin expression with relative cone frequencies based on data presented in an MSP study (chapter 5) The qualitative differences in opsin expression between the two populations match the direction of differences in relative cone frequency. The spring population had higher relative abundance of UV and violet cones and also had higher expression of UV and violet opsins. In contrast, the swamp population had a higher proportion of yellow and red cones and higher expression of yellow and red opsins. However, the quantitative patterns do not match. In other words, the average frequency of a given cone type does not match the average relative expression for the corresponding opsin.

Discussion

The objective of this study was to determine whether measurements of relative opsin expression (made with real-time PCR) reflect relative cone abundances in the retina.

With respect to qualitative patterns (Table 6.4), the answer is yes. Spring animals have higher frequencies of UV and violet cones and lower frequencies of yellow and red cones than animals from the swamp (Fuller et al. in press). These patterns match those found in this study. I found that spring animals have higher UV and violet opsin expression and lower yellow and red opsin expression than animals from the swamp.

120

Table 6.4. Relative cone frequency and relative opsin expression for each opsin/cone type. Means and standard errors are listed for both populations plus the grand mean. Relative cone frequency: spring N=11, swamp N=10, grand mean N=21. Relative opsin expression: spring N=10, swamp N=10, grand mean N=20.

population opsin / cone relative cone relative opsin type frequency expression

spring UV 0.111 (0.016) 0.278 (0.021) violet 0.449 (0.022) 0.065 (0.006) blue 0.080 (0.011) 0.0013 (0.0002) yellow 0.180 (0.010) 0.374 (0.028) red 0.181 (0.013) 0.281 (0.013)

swamp UV 0.006 (0.004) 0.129 (0.013) violet 0.266 (0.032) 0.044 (0.005) blue 0.089 (0.017) 0.0012(0.0002) yellow 0.278 (0.014) 0.448 (0.025) red 0.361 (0.026) 0.378 (0.028)

grand average UV 0.061 (0.015) 0.204 (0.021) violet 0.362 (0.028) 0.055 (0.004) blue 0.084 (0.010) 0.0012 (0.0001) yellow 0.227 (0.014) 0.411 (0.020) red 0.267 (0.024) 0.329 (0.018)

121 There were no differences among the populations in the frequency of blue cones nor in blue opsin expression. The qualitative match in the direction of differences between the populations appears robust.

Quantitatively the proportion of cone types measured in the MSP study do not match well with the relative opsin expression found in this study (Table 6.4). Note that there is no reason to assume that the MSP data are not without error (E. Loew, pers. comm.). MSP is actually a poor tool for measuring the relative abundance of cone types in the retina. The cones one measures with MSP may come from a small area of retina.

Whether this area is representative of the entire retina is unknown. In addition, the amount of photopigment packed into individual cone cells may easily vary among cone types (Flamarique Novales & Harosi 2000) which would create discrepancies between relative cone frequency and relative opsin expression. However, even with these assumed sources of variation, there is still an unresolvable problem with the blue opsin expression. According to the real-time data, blue opsin represents 0.14% of all the opsins expressed. In the MSP study, blue cones were 8.35% of the measured cones. This is an

80-fold difference! Indeed, if blue cones were as rare as suggested by the real-time data, then I should not have even detected them.

What can account for this discrepancy? In theory, this difference could be caused by a faulty primer/probe set. However, I can rule out this possibility due to the fact that the primer/probe sets did not differ in efficiency. In addition, I performed real-time PCR on PCR products containing the binding sites where I controlled for total copy number. I found no evidence for differential performance among primer/probe sets (Fuller unpubl. data). Another possibility is that there are multiple blue loci that contribute to blue

122 sensitivity, similar to the LWS loci. While I am unable to rule out this possibility, I find it doubtful. Blows and Yokoyama have sequenced over 250 degenerate clones and have found no evidence of an additional blue opsin (Blows, pers. comm.).

Assuming the blue opsin expression results are valid, what could cause the disparity between opsin expression and relative cone abundance? One possibility is differential opsin expression throughout the day (Chen et al. 1992). I controlled for variation in diurnal rhythms in opsin expression by euthanizing the animals over a relatively short time. If diurnal rhythms are pronounced, then isolating RNAs from animals at a different time of day might produce difference amounts of opsin RNAs.

Another possibility is that there is differential reverse transcription from RNA to cDNA among the opsins. Blue opsins could be more likely to degrade or may anneal less well to the polyT primer.

Regardless of the cause, the implications of this quantitative discrepancy between relative opsin expression and cone frequency is that cone frequencies cannot be inferred from relative opsin expression data, but that relative differences in cone frequencies can be inferred with relative opsin expression data in L. goodei.

Population differences

The results match a general pattern seen in visual ecology. The types of cones animals possess in their retinas generally match the available spectrum of light (Lythgoe 1984,

Lythgoe et al. 1994, McDonald and Hawryshyn 1995, Hunt et al. 1996, Partridge and

Cummings 1999, Yokoyama et al. 1999, Cronin & Caldwell 2002, Cronin et al. 2002, although see Penn and Williams 1986, Kröger et al. 1999, Boughmann 2001). Here, I

123 found higher UV and violet opsin expression in the spring population where there is higher transmission of UV and violet wavelengths. Animals from the swamp population

(where there is lower transmission of UV/blue light) had higher expression of yellow and red opsins.

In addition, this study further supports the counterintuitive pattern seen in male color pattern versus lighting environment and vision physiology. Males with blue anal fins are more abundant in populations where animals express fewer UV and violet opsins and where transmission of UV/blue wavelengths is low (Fuller 2002). How can such a counterintuitive pattern be explained? One option is that males are maximizing contrast while sacrificing brightness. If the retina is tuned to detect objects against a tea-stained background, blue anal fins may be stimulating the violet and UV cones. In spring populations, where the retina may be tuned to detect objects against a high UV/blue background, red and yellow anal fins may be stimulating the yellow and red cones. I am currently testing these hypotheses by quantifying visual environments and color patterns using the methods of Endler (1990).

In conclusion, I found that measuring relative opsin expression with real-time

PCR is a valid method for inferring qualitative differences in cone frequencies. Greater expression of a given opsin corresponds with a higher frequency of the given cone type.

However, as a quantitative measure, relative opsin expression is not a valid measure of actual cone frequency. Still, using real-time PCR to measure vision physiology for quantitative genetic studies (e.g. heritabilities, variance-covariance matrices, etc.) will allow researchers to answer the questions of whether there is heritable variation in relative opsin expression and relative cone abundance. Real-time PCR represents a

124 strong tool for studies requiring inferences of vision physiology in large numbers of individuals.

125

CHAPTER 7

VARIABLE SENSORY SYSTEMS IN THE BLUEFIN KILLIFISH, LUCANIA

GOODEI

The sensory bias and sensory exploitation models of sexual selection predict the evolution of male secondary sexual traits to match the sensory system characteristics of females (Basolo 1990, Ryan et al. 1990). In their early forms, they assumed that sensory systems were relatively static whereas male traits were variable and evolvable. Although this assumption has since been altered (Endler 1992, Ryan and Keddy-Hector 1992,

Endler and Basolo 1998), the question still remains: how variable are sensory systems within populations? Here, I show both environmental and genetic variation in opsin expression in a single population of bluefin killifish, Lucania goodei. I measured relative opsin expression (which is correlated with relative frequency of cones in the retina) using real-time PCR for offspring from a paternal half-sib breeding study where offspring were raised under different lighting conditions. Yellow opsin expression varied significantly across sires, and violet opsin expression varied significantly across dams. These results highlight the fact that sensory systems are dynamic, readily evolvable traits in contrast to the static, invariant systems assumed by early versions of the sensory bias and sensory exploitation models of sexual selection.

126

In their original forms, the sensory bias and sensory exploitation models of sexual selection were based on the idea that female sensory systems are largely invariant and represent an evolutionary constraint (Basolo 1990, Ryan et al. 1990). Invariant sensory systems result in selection favoring secondary sex characters in males that exploit the biases in the sensory systems of females. The use of terms such as "exploitation" implied that males coerce females into mating, possibly resulting in a reduction in fitness (Ryan et al. 1990). Yet if mating with such males entails a reduction in fitness, then selection should eliminate the bias provided that there is variation in sensory systems (Bradbury and Vehrencamp 2000). In contrast, if mating with such males entails an increase in fitness (or is neutral with respect to fitness) and if there is variation in bias, then further elaboration of the trait occurs via direct benefits, good-genes, and/or Fisherian processes

(Andersson 1994, Kokko et al. 2002). Indeed, in recent years, the assumption of invariant sensory systems has been relaxed and many theoretical models have paired the idea of sensory bias and sensory exploitation models with other models of sexual selection (Schluter et al. 1993, Holland and Rice 1998, Servedio 2001). Still, in order for male secondary sex characters to evolve solely via sensory bias and sensory exploitation, sensory systems must be constrained in their response to selection. Such a constraint can be achieved by either a lack of genetic variation or by pleiotropy.

How much variation is there in vision physiology within populations? At one level, we know that there must be variation in visual systems because different species

(and different populations within species) living under different environmental conditions differ in their visual properties (types of cones, types of ocular filters, etc.) (Lythgoe

127 1984, Cronin et al. 2001, Cronin and Caldwell 2002, Boughman 2002). Still, we have little understanding of the degree to which such variation is an effect of genetics and/or environment, and whether significant variation occurs within populations. This is surprising given that differences in sensory systems are often interpreted as evidence of divergence and are suggested to play a large role in speciation (Boughman 2002).

Hence, assessing the genetic and environmental components of variation in visual properties is critical to understanding both the degree to which sensory systems can act as static constraints in sexual selection and the degree to which they can act as diversifying agents in speciation.

In this study, I determine the extent to which variation in opsin expression is controlled by genetics and/or environment in the bluefin killifish, Lucania goodei, using a paternal half-sib breeding experiment. L. goodei is a small freshwater fundulid that occurs under a wide range of lighting environments ranging from tea-stained swamps that have reduced transmission of UV/blue wavelengths to crystal clear springs that have high transmission of UV/blue wavelengths (Fuller 2002). Both male coloration and sensory systems vary across populations. Males with blue anal fins are more abundant in tea- stained swamps, whereas males with red anal fins (and to a lesser extent, males with yellow anal fins) are more abundant in clear springs. In contrast, swamp animals are less sensitive to UV/blue wavelengths and possess fewer UV and violet cones than animals from swamp populations (chapter 5). These differences in cone frequency match differences in opsin expression (chapter 6). The opsin protein is the main determinant of the spectral sensitivity of a given cone type (Yokoyama and Yokoyama 1996). The UV opsin, when attached with a vitamin A chromophore, creates a UV sensitive

128 photopigment (maximum absorbance(λmax) = 359 nm). Similarly, violet, blue, yellow and red opsin loci correspond to violet (λmax = 405 nm), blue (λmax = 455 nm), yellow (λmax =

539 nm), and red (λmax = 573 nm) cone classes (chapter 5). Because of this relatively straightforward genotype-phenotype map, we can use differences in opsin expression to infer qualitative differences in cone frequency (Chapter 6, Carleton and Kocher 2001).

Indeed, animals from the spring population, which have a higher frequency of UV and violet cones, also have higher expression of UV and violet opsins (Figure 7.1a). In contrast, swamp animals, which have a higher frequency of yellow and red cones, also have a higher frequency of yellow and red opsins. While the qualitative match in opsin frequency versus cone frequency is good, the quantitative match is less precise. The proportion of opsins of a given class need not match the proportion of the corresponding cones. Hence, differences among individuals in opsin expression reflect qualitative differences in cone abundances (i.e. more versus less), but are uninformative on precise parameter estimates for actual cone abundances. In partaicular, the frequency of blues in the retina (8%) does not match the relative blue opsin expression (0.1%). In this study, I assessed the environmental and genetic effects on opsin expression by comparing offspring from different paternal half-sib families raised under different environments

(clear or tea-stained water).

I found both genetic and environmental effects on opsin expression, but no interaction between genetics and environment (Table 7.1). Environmental effects were manifested as overall shifts towards increased expression of either short wavelength sensitive opsins (UV and violet) or long-wavelength sensitive opsins (yellow and red).

Animals raised in tea-stained water expressed relatively more yellow and red opsins

129

0.5 spring a swamp 0.4

0.3

0.2

relative expression relative 0.1

0.0 UV violet blue yellow red

opsins

0.5 b

clear 0.4 tea

0.3

0.2

relative expression relative 0.1

0.0 UV violet blue yellow red

opsins

Figure 7.1. A. Relative expression of opsins expressed in animals from a spring and a swamp population. N = 10 for each population. B. Relative expression of opsins expressed in animals raised in tea-stained and clear water treatments. N = 78 for tea- stained water. N = 80 for clear water. Means and standard errors are shown.

130 Table 7.1. Effects of sires, dams within sires, environment and the interaction between sires and environment. opsin effect MS F P

UV sire 0.00808 2.07 0.1658 dam (sire) 0.00391 1.29 0.2392 environment 0.68701 227.24 0.0000 sire*environment 0.00170 0.56 0.6410 violet sire 8.40 E-5 0.11 0.9541 dam (sire) 7.98 E-4 6.08 0.0000 environment 9.31 E-4 7.09 0.0087 sire*environment 1.98 E-4 1.51 0.2144 blue sire 2.34 E-8 0.40 0.7562 dam (sire) 5.90 E-8 2.19 0.0215 environment 4.61 E-8 1.71 0.1927 sire*environment 1.39 E-8 0.52 0.6721 yellow sire 0.02584 8.72 0.0034 dam (sire) 0.00296 0.90 0.5360 environment 0.13660 41.55 0.0000 sire*environment 0.00079 0.24 0.8680 red sire 0.0064 1.13 0.3818 dam (sire) 0.00592 1.67 0.0942 environment 0.24048 67.73 0.0000 sire*environment 0.00329 0.93 0.4302

Initial analyses included the interaction between environment and dam (sire). This term was non-significant in all models and was subsequently dropped. All statistically significant effects in the subsequent models were also significant in the initial model unless noted otherwise (see text for blue opsin results). Degrees of freedom in the numerators were as follows: sire df=3, dam(sire) df=10, environment df=1, sire * environment df=3. All terms, except sire, were tested over the mean-square error (df=140). Due to the unbalanced design, the sire effect is tested using an error term estimated by the Satterthewaite approximation which is similar to the dam(sire) mean square. Degrees of freedom in the approximated error term are as follows: UV=10.3, violet=10.1, blue=9.7, yellow=10.4, red=10.2. The analysis shown for the blue opsin is calculated for the data after the removal of a far-outlier (mean-square error df=139).

131 whereas animals raised in clear water expressed more UV and violet opsins (Figure

7.1B). The striking similarity between within population plasticity (Figure 7.1B) and across population patterns (Figure 7.1A) is suggestive of strong uniform selection. In addition, the complete lack of genetic variation in plasticity also suggests strong uniform selection on plasticity.

I detected genetic variation in the expression of yellow and violet opsins (Figure

7.2 A,B). There was significant variation among sires in yellow opsin expression (Table

7.1). Offspring from the y/y sire expressed more yellow opsins than offspring from the y/b sire. I also detected significant variation among dams in the expression of violet opsins (Figure 7.2B). In theory, the dam effect can be influenced by both additive genetic, non-additive genetic (i.e. dominance and epistasis), and non-genetic environmental effects (i.e. maternal environment, parental care, etc.) (Falconer and

Mackay 1996, Lynch and Walsh 1998). I doubt that there are large non-genetic environmental effects that differ across dams in this system. Eggs are spawned externally and deposited on plants where they receive no parental care (Fuller and Travis 2001). I removed eggs from spawning substrates within 24 hours of fertilization. Half of the offspring were then raised in tea-stained water while the other half were raised in clear water. Hence, any maternal effects would have to occur early and persist in the face of drastic different environments.

The results for blue opsin expression are more tenuous. There is a far-outlier in this data (studentized residual = 20.62). Exclusion of this data point results in a trend for significant variation among dams (F10,10 = 2.78, P = 0.0612). If the interaction between

132

0.50

0.45 a

0.40

0.35 yellow opsin expression

0.30 dams dams dams dams r/b sire r/r sire y/y sire y/b sire

0.08

b 0.07

0.06

0.05 violet opsin expression

0.04 dams dams dams dams r/b sire r/r sire y/y sire y/b sire

Figure 7.2. Relative expression of (A) yellow and (B) violet opsins expressed across dams nested within sires. Means and standard errors are shown.

133 environment and dams nested within sires is dropped from the model (environment * dams(sire) F10,129 = 0.83), then the dam effect becomes statistically significant (Table

7.1). However, given that measurement error in blue expression is somewhat high, these results for blue expression should be taken with caution. All other results are extremely robust to removal of individual data points and inclusion of the interaction term between environment and dam nested within sire. To the best of our knowledge, this is the first demonstration of genetic variation in visual properties within a population.

These results raise a host of questions. First, genetic variation was present for the expression of some opsins (yellow, violet, perhaps blue), but not of others (UV, red).

The power to detect an effect (particularly due to sire) was somewhat low due to the number of sires in the design. In addition, I found genetic variation in opsin expression for the opsins that had lower environmental variation. Still, taking these results at face value, what are the implications of genetic variation in the expression of some opsins but not of others? One could argue that there is still the possibility for a constrained sensory system. If males evolve secondary sex characteristics that only stimulate the UV or red cones, then sensory bias alone may be able to result in the elaboration and/or fixation of a particular male trait. I am doubtful whether such a scenario could occur. The perception of color relies on the comparison of output from multiple cone types (Bowmaker and

Hunt 1999), and nearly all proposed measures of color incorporate the measurement of color across multiple axes corresponding to the various cone types (Endler 1990). Still, the question of whether genetic variation in some (but not all) components of vision results in the evolvability of the entire system needs to be addressed experimentally.

134 The second question raised by this experiment is whether the scale of variation in opsin expression creates meaningful variation in the visual experience of individuals.

The coefficient of variation (CV) in yellow opsin expression across sire means was 6.2%.

The average CV in violet opsin expression among dams averaged across sires was 15.8%.

Even if we assume that the CV in opsin expression results in a proportional change in cone frequency, does such a change create important differences in the visual experiences of individuals? Currently, I do not have the data to answer this question. Studies on sticklebacks have shown that animals living in different lighting environments differ in the excitability of retinal ganglion cells (McDonald and Hawryshyn 1995) and in optomotor responses (a behavioral assay of visual sensitivity) (Boughman 2001). These differences in visual properties are associated with differences in mating behavior

(Boughman 2001). Whether this variation is genetic or environmental is unknown, but it suggests that differences in vision physiology can result in differences in behavior.

While these issues need to be resolved, the main result remains: there is substantial genetic and environmental variation in sensory systems within populations.

Without a doubt, proponents of sensory bias and sensory exploitation have clearly demonstrated that female mating preferences can evolve prior to the evolution of favored male traits (Basolo 1990, Ryan et al. 1990, Burley and Samanski 1998). The question that remains is what happens when male traits initially arise within populations? The idea that the frequency and elaboration of sensory exploiting male trait increases over time to meet an invariant sensory system is most likely incorrect. Sensory systems are variable and will, therefore, coevolve with male traits. Furthermore, if male traits

135 exploiting sensory biases are costly for females, then biases themselves should be evolvable and should respond to selection.

Methods

Breeding Design

Lucania goodei were collected from the 26-Mile Bend Population in Broward, Co., FL,

USA. I chose four sires with distinctly different color patterns as sires (red/blue, red/red, yellow/yellow, and yellow/blue). Color codes refer to the coloration of the posterior portion of the dorsal fin and then to the coloration of the anal fin. I crossed each sire with

3-4 randomly chosen dams in the laboratory. I then divided each clutch between two environmental lighting treatments (clear vs. tea-stained water). A cross was made by placing a sire and a dam in an aquarium containing yarn mops that served as spawning substrates. Eggs were carefully removed from spawning mops and reared them in the laboratory for approximately 2-4 weeks. I then transported them to the greenhouse, divided each clutch so that there were equal numbers of similarly aged fry, and placed them in aquaria containing either clear or tea-stained water. All water was treated with a buffer to keep the pH above 7. For the tea-stained water treatment, I added a small amount of instant, decaffinated tea to the water 2-3 times each week. Analyses of light transmission verified that the tea treatment significantly reduced the transmission of UV and blue wavelengths through the water column relative to the clear water treatment.

This experiment ran from August 2001-December 2002.

136 Real-time PCR

I measured opsin expression for 4-6 animals from each treatment combination. For each individual, cDNA was obtained by reverse transcribing RNA isolated from eye tissue.

Primers and probes are described elsewhere (chapter 6). For each real-time PCR (RT-

PCR) reaction, 0.2 ul of cDNA mixture was placed in a 30 ul reaction with the appropriate primers, probes, and taqman mix. The amount of flourescence was monitored over 40 cycles (94C-15s/55C-30s/65C-1 minute) using the ABI Prism 7700

Sequence Detection System at Florida State University. The relative abundance of each opsin was based on its critical cycle number (Cti) which was determined when the flourescence exceeded a threshhold set close to the background flourescence (Carleton and Kocher 2001). Relative opsin expression was calculated as a fraction of total opsin genes for an individual according to the following:

1 T 1( + E )Cti i = i . Tall 1 Cti ∑ 1( + Ei )

Ti is the proportional gene expression for a given gene i. Ei is the PCR efficiency for Tall each primer/probe set, and Cti is the critical cycle number for each gene.

I used analysis of variance to examine the effects of sire, environment, dam nested within sires (dam(sire)), and the interactions between environment and sire and between environment and dam(sire). Dam(sire) and the interaction between environment and dam(sire) were treated as random effects. Due to an unbalanced design, I used the

Satterthwaite approximation of error and degrees of freedom (Sokal and Rohlf 1995).

137

CHAPTER 8

CONCLUSIONS

There is little evidence to support the idea of invariant sensory systems in the bluefin killifish, Lucania goodei. Visual sensitivity and the relative abundance of cones varied among populations (chapter 5) as did the expression of the opsin genes that determine the spectral sensitivities of cones (chapter 6). Within populations, both genetics and environmental lighting conditions affected the expression of opsins (and presumably the abundance of cones and visual sensitivity). Hence, vision physiology should readily respond to selection. If sensory biases lead animals to perform maladaptive behaviors, then selection should eliminate the bias.

Are there other mechanisms for the production of sensory bias? The answer is yes. Pleiotropy (i.e. one gene affects multiple traits) can also cause sensory bias. In this case, the bias arises, not because of a lack of genetic variation, but because sensory systems have multiple functions. Strong selection on sensory systems in one function

(e.g. finding food) has correlated effects on other behaviors (e.g. finding mates). Rodd and colleagues (2002) have recently suggested that the color patterns of male guppies

(Poecilia reticulata) are mimics of food items. This is based upon data showing that variation among populations in female mating preference for males with orange spots can 138 be explained by the attraction of both sexes to inanimate orange objects. Given that these animals frequently eat orange food items, the implication is that selection for preference/recognition of orange food items also creates a bias for orange males. The critical question for the sensory bias hypothesis is whether biases preclude the evolution of adaptive mate choice. Can male color patterns that mimic food colors spread through populations simply via selection on foraging ability (i.e., selection on foraging pulls along neutral/maladaptive mate choice) or do male food-mimic color patterns spread through populations only when their expression is correlated with fitness benefits to females (i.e., bias and adaptive mate choice coincide)? In my post-doc, I will be simulating natural selection on neural networks to answer this question.

The other striking result from this work is the consistent pattern between lighting environment, color pattern, and vision. Correlations between lighting environment, color pattern, and vision were consistent both within and among populations. Across populations, there were more males with blue anal fins in populations with low transmission of UV/blue wavelengths (chapter 2). Within a population, males were more likely to express blue anal fins when raised in tea-stained water (low transmission of

UV/blue wavelengths) than when raised in clear water (high transmission of UV/blue wavelengths) (chapter 4). In contrast, animals from a swamp population had a lower frequency of UV and violet cones and higher frequency of yellow and red cones than animals from a spring population. At one level, these patterns are consistent with the theory of sensory drive (Endler 1992) which predicts that signals and sensory systems should covary with environmental conditions. However, sensory drive makes no prediction about the expected direction of change. Hence, either a positive or a negative

139 correlation between color pattern and vision/light conditions is consistent with this theory. In this work, I found a negative relationship between the types of signals employed and the visual properties and lighting environments. Males with blue anal fins were more common in conditions where UV/blue wavelengths do not transmit well

(chapters 2,4) and where animals are likely to be less sensitive to UV/blue wavelengths

(chapters 5,6,7). This pattern may actually be quite common. In earlier work on rainbow darters, Etheostoma caeruleum, I noticed that males from deeper, tannin stained streams were more likely to have larger, darker patches of blue than animals from shallower, clear streams. Similarly, in sailfin mollies, Poecilia latipinna, large males from darker stained water appear to have larger areas of blue on their caudal fins than large males from clearer water (Fuller pers obs). Males are obviously not maximizing perceived brightness

(total number of photons reflected off of the color pattern and detected by a receiver) by expressing blue anal fins under conditions where blue does not transmit well and where animals possess fewer UV and violet retinal cones (Fuller et al. in press). There are two potential observations for such a pattern.

The most likely scenario is that blue males create high contrast with the water column or with other color elements on the body. If no photons are being detected from the anal fin, then it will appear black, which will produce high contrast with the body and visual background. On the other hand, clear water has high transmission of UV/blue wavelengths causing the water column to have a bluish tint. This environment should create high contrast for yellow and red color morphs, but lower contrast for blue morphs.

Another possibility is that blue males really are more conspicuous in clear water due to high, perceived brightness but that they suffer high mortality costs due to predation. I

140 acknowledge that these speculations would benefit from a proper analysis of reflectance spectra (Endler 1990, Bennet et al. 1994). I am currently analyzing reflectance spectra to compute actual brightness and contrast of blue anal fins in tea-stained and in clear water using the methods of Endler (1990).

The biggest weakness of this work is the lack of fitness correlates for the various male color morphs and also for individuals with differing components of their vision physiology. Does the color pattern function in female choice or in male/male competition? Is there negative frequency dependence? Do blue males have lower or higher mating success in clear water? Does the predation risk of blue males increase or decrease in clear water? Do spring populations harbor animals carrying genes for plasticity or does selection eliminate them because of the constant visual environment?

Do animals with different visual sensitivities vary in behavior and what are the fitness implications of such variation? I currently do not have the answer to these questions, but

I soon will.

141

LITERATURE CITED

Alatalo RV, Lundberg A, Glynn C (1986) Female pied flycatchers choose territory quality and not male characteristics. Nature 323: 152-153.

Andersson M. (1994) Sexual selection. Princeton: Princeton University Press.

Andersson A, Oernborg J, and Andersson.M (1998) Ultraviolet sexual dimorphism and assortative mating in blue tits. Proc R Soc Lond Ser B 265: 445-450.

Angus RZ (1989) A genetic overview of poeciliid fishes. In: Meffe GK and Snelson FF Jr (eds) Ecology and evolution of livebearing fishes (Poeciliidae). Prentice Hall, Englewood Cliffs, NJ, pp 51-68.

Archer SN (1999) Visual pigments and photoreception. In: Archer SN, Djamgoz MBA, Loew ER, Vallerga S (eds) Adaptive mechanisms in the ecology of vision. Kluwer, Dordrecht, pp 25-42.

Archer SN, Lythgoe JN (1990) The visual pigment basis for cone polymorphism in the guppy, Poecilia reticulata. Vision Res 30: 225-233.

Archer SN, Endler JA, Lythgoe JN, Partridge JC (1987) Visual pigment polymorphism in the guppy Poecilia reticulata. Vision Res 27: 1243-1252.

Basolo AL (1990) Female preference predates the evolution of the sword in swordtail fish. Science 250: 808-810.

Basolo AL (1991) Male swords and female preferences. Science 253: 1426-1427.

Basolo AL (1995) A further examination of a preexisting bias favoring a sword in the genus Xiphophorus. Anim Behav 50: 365-375.

Beaudet L, Hawryshyn CW (1999) Ecological aspects of vertebrate visual ontogeny. In: Archer SN, Djamgoz MBA, Loew ER, Vallerga S (eds) Adaptive mechanisms in the ecology of vision. Kluwer, Dordrecht, pp 413-437.

142 Bennett ATD, Cuthill IC, Norris KJ (1994) Sexual selection and the mismeasure of color. Am Nat 144: 848-860.

Boake CRB (1989) Repeatability – its role in evolutionary studies of mating-behavior. Evol Ecol 3: 173-182.

Boughman JW (2001) Divergent sexual selection enhances reproductive isolation in sticklebacks. Nature 411: 944-947.

Boughman JW (2002) How sensory drive can promote speciation. TREE 17:571-577.

Bowmaker JK (1990) Visual pigments of fishes. In: Douglas RH, Djamgoz MBA (eds) The visual system of fishes. London: Chapman & Hall, pp 63-81.

Bowmaker JK, Kunz YW (1987) Ultraviolet receptors, tetrachromatic colour vision and retinal mosaics in the brown trout (Salmo trutta): age dependent changes. Vision Res 27: 2101-2108.

Bowmaker JK, Hunt DM (1999) Molecular biology of photoreceptor spectral sensitivity. In: Archer SN, Djamgoz MBA, Loew ER, Vallerga S (eds) Adaptive mechanisms in the ecology of vision. Kluwer, Dordrecht, pp. 439-462.

Bradbury JW, Vehrencamp SL (2000) Economic models of animal communication. Anim Behav 59: 259-268.

Brakefield PM (1998) The evolution-development interface and advances with the eyespot patterns of Bicyclus butterflies. Heredity 80: 265-272.

Brakefield PM, Gates J, Keys D, Kesbeke F, Wijngaarden PJ, Monteiro A, French V, Carroll SB (1996) Development, plasticity and evolution of butterfly eyespot patterns. Nature 384: 236-242.

Brakefield PM, Larsen TB (1984) The evolutionary significance of dry and wet season forms in some tropical butterflies. Biol J Linn Soc 22: 1-12.

Brawner WR III, Hill GE, and Sunderman CA (2000) Effects of coccidial and mycoplasmal infections on carotenoid-based plumage pigmentation in male house finches. Auk 117: 952-963.

Breder CM, Rosen DE 1966. Models of reproduction in fishes. Neptune City, NJ: TFH Press.

Bridges CDB (1972) The rhodopsin-porphyropsinvisual system. In: Dartnall HJA (ed), Handbook of sensory physiology, VII/I. Berlin: Springer. pp 417-480.

143 Briscoe AD, Chittka L (2001) The evolution of color vision in insects. Annu Rev Entomol 46: 471-510.

Brooks R, Endler JA (2001) Direct and indirect sexual selection and quantitative genetics of male traits in guppies (Poecilia reticulata). Evolution 55: 1002-1015.

Bry C (1981) Temporal aspects of macroscopic change in rainbow trout (Salmo gairdneri) oocytes before ovulation and of ova fertility during the post-ovulation period: Effect of treatment with 17α-hydroxy-20β-dihydroprogesterone. Aquaculture 24: 153- 160.

Burley NT, Samanski R (1998) “A taste for the beautiful”: Latent aesthetic mate preferences for white crests in two species of Australian grassfinches. Am Nat 152:792- 802.

Carleton KL, Kocher TD (2001) Cone opsin genes of African cichlid fishes: tuning spectral sensitivity by differential gene expression. Mol Biol Evol 18: 1540-1550.

Chen D-M, Christianson JS, Sapp RJ, Stark WS (1992) Visual receptor cycle in normal and period mutant Drosophila: Microspectrophotometry, electrophysiology, and ultrastructural morphometry. Visual Neurosci 9: 125-135.

Clark CW, Mangel M (2000) Dynamic state variable models in ecology: methods and applications. Oxford: Oxford University Press.

Constantz GD (1989) Reproductive biology of poeciliid fishes. In: GK Meffe, Snelson FF Jr (eds), Ecology and evolution of livebearing fishes (Poeciliidae). pp. 33-50.

Côté IM, Hunte W (1989) Male and female mate choice in the redlip blenny – why bigger is better. Anim Behav 38: 78-88.

Cronin TW, Caldwell RL, Marshall J (2001) Tunable colour vision in a mantis shrimp. Nature 411: 547-548.

Cronin TW, Caldwell RL (2002) Tuning of photoreceptor function in three mantis shrimp species that inhabit a range of depths. II. Filter pigments. J Comp Physiol A 188: 187-197.

Cronin TW, Caldwell RL, Erdmann, MV (2002) Tuning of photoreceptor function in mantis shrimp species occupying a range of depths. J Comp Physiol A 188: 179-186.

Crowley PH, Travers SE, Linton MC, Cohn SL, Sih A, Sargent RC (1990) Mate density, predation risk, and the seasonal sequence of mate choices: a dynamic game. Am Nat 137: 567-596.

144 Cummings ME, Partridge JC (2001) Visual pigments and optical habitats of surfperch (Embiotocidae) in the California kelp forest. J Comp Physiol A 187: 875-889.

Cuthill IC, Bennett ATD, Partridge JC, and Maier EJ (1999) Plumage reflectance and the objective assessment of avian sexual dichromatism. Am Nat 153: 183-200.

Dorn LA, Pyle EH, Schmitt J (2000) Plasticity to light cues and resources in Arabidopsis thaliana: testing for adaptive value and costs. Evolution 54: 1982-1994.

Doucet SM (2002) Structural plumage coloration, male body size, and condition in the Blue-Black Grassquit. Condor 104: 30-38.

Downhower JF, Blumer LS, Brown L (1987) Seasonal-variation in sexual selection in the mottled sculpin. Evolution 41: 1386-1394.

Dugatkin LA, Godin J-GJ (1993) Female mate copying in the guppy (Poecilia reticulata): age-dependent effects. Behav Ecol 4: 289-292.

Draper NR, Smith H (1981) Applied regression analysis, 2nd ed. New York: John Wiley & Sons.

Ellner S, Hairston NG (1994) Role of overlapping generations in maintaining genetic variation in a fluctuating environment. Am Nat 143: 403-417.

Endler JA (1978) A predator’s view of animal color patterns. Evol Biol 11: 319-364.

Endler JA (1980) Natural selection on color patterns in Poecilia reticulata. Evolution 34: 76-91.

Endler JA (1982) Convergent and divergent effects of natural selection on color patterns in two fish faunas. Evolution 36: 178-188.

Endler JA (1987) Predation, light intensity and courtship behaviour in Poecilia reticulata (Pisces: Poeciliidae). Anim Behav 35: 1376-1385.

Endler JA (1986) Natural selection in the wild. Princeton University Press: Princeton, NJ.

Endler JA (1990) On the measurement and classification of color in studies of animal colour patterns. Biol J Linn Soc 41: 315-352.

Endler JA (1991) Variation in the appearance of guppy color patterns to guppies and their predators under different visual conditions. Vision Res 31: 587-608.

145 Endler JA (1992) Signals, signal conditions, and the direction of evolution. Am Nat S139: S125-S153.

Endler JA (1993a) Some general comments on the evolution ad design of animal communication systems. Phil Trans R Soc Lond B 340: 215-225.

Endler JA (1993b) The color of light in forests and its implications. Ecolog Monogr 63: 1-27.

Endler JA, Basolo AL (1998) TREE 13: 415-420.

Endler JA, Basolo A, Glowacki S, Zerr J (2001) Variation in response to artificial selection for light sensitivity in guppies (Poecilia reticulata). Am Nat 158: 36-48

Endler JA, Houde AE (1995) Geographic variation in female preferences for male traits in Poecilia reticulata. Evolution 49: 456-468.

Endler JA & Thery M (1996) Interacting effects of lek placement, display behavior, ambient light, and color patterns in three Neotropical forest dwelling birds. Am Nat 148: 421-452.

Evans MR, Norris K. (1996) The importance of carotenoids in signaling during aggressive interactions between male firemouth cichlids (Cichlasoma meeki). Behav Ecol 7: 1-6.

Falconer DS, Mackay TFC (1996) Introduction to Quantitative Genetics, 4th Ed. Longman, Harlow, Essex.

Farr JA (1977) Male rarity or novelty, female choice behavior and sexual selection in the guppy, Poecilia reticulata Peters (Pisces: Poeciliidae). Evolution 31: 162-168.

Fisher RA (1958) The genetical theory of natural selection. Dover, New York.

Flamarique IN, Harosi FI (2000) Photoreceptors, visual pigments, and ellipsosomes in the killifish, Fundulus heteroclitus: A microspectrophotometric and histological study. Vis Neurosci 17: 403-420.

Fleishman LJ, Bowman M, Saunders D, Millwer WE, Rury MJ, Loew ER (1997) The visual ecology of Puerto Rican anoline lizards: habitat light and spectral sensitivity. J Comp Physiol A 181: 446-460.

Forsgren E (1992) Predation risk affects mate choice in a gobiid fish. Am Nat 140: 1041-1049.

146 Forsgren E (1997) Mate sampling in a population of sand gobies. Anim Behav 53: 267- 276.

Forsgren E, Kvarnemo C, Lindström K (1996) Mode of sexual selection determined by resource abundance in two sand goby populations. Evolution 50: 646-654.

Foster NR (1967) Comparative studies on the biology of killifishes (Pisces: Cyprinodontidae). (PhD dissertation). Ithaca: Cornell University.

Fuller R, Berglund A (1996) Behavioral responses of a sex-role reversed pipefish to a gradient of perceived predation risk. Behav Ecol 7: 69-75.

Fuller RC (1998a) Fecundity estiamtes for rainbow darters, Etheostoma caeruleum, in southwestern Michigan. Ohio J Sci 98: 2-5.

Fuller RC (1998b) Sperm competition affects male behaviour and sperm output in the rainbow darter. Proc R Soc Lond Series B 265: 2365-2371.

Fuller RC (2001) Patterns in male breeding behaviors in the bluefin killifish, Lucania goodei: a field study (Cyprinodontiformes: Fundulidae). Copeia 2001: 823-828.

Fuller RC (2002) Lighting environment predicts relative abundance of male color morphs in bluefin killifish populations. Proc R Soc Lond Ser B 269: 1457-1465.

Fuller RC, Travis J (2001) A test for male parental care in a fundulid, the bluefin killifish, Lucania goodei. Environ Biol Fish 61: 419-426.

Fuller R., Fleishman LJ, Leal M, Travis J, Loew E (in press) Intraspecific variation in ultraviolet cone production and visual sensitivity in the bluefin killifish, Lucania goodei. J Comp Physiol A

Garamszegi LZ, Møller AP, Erritzøe J (2002) Coevolving avian eye size and brain size in relation to prey capture and nocturnality. Proc R Soc Lond B 269: 961-967.

Getty T (1995) Search, discrimination, and selection: mate choice by pied flycatchers. Am Nat 145: 146-154.

Gillespie JH (1974) Natural selection for within-generation variance in offspring number. Genetics 76: 601-606.

Gillespie JH (1977) Natural selection for variances in offspring numbers: a new evolutionary principle. Am Nat 111: 1010-1014.

147 Gillespie RG, Oxford GS (1998) Selection on the color polymorphism in Hawaiian happy-face spiders: Evidence from genetic structure and temporal fluctuations. Evolution 52: 775-783.

Grafe TU (1997) Costs and benefits of mate choice in the lek-breeding reed frog, Hyperolius marmoratus. Anim Behav 53: 1103-1117.

Grether GF (2000) Carotenoid limitation and mate preference evolution: a test of the indicator hypothesis in guppies (Poecilia reticulata). Evolution 54: 1712-1724.

Grill CP, Rush VN (2000) Analysing spectral data: comparison and application of two techniques. Biol J Linn Soc 69: 121-138.

Gross MR, Charnov EL (1980) Alternative male life histories in bluegill sunfish. PNAS 77: 6937-6940.

Hartl DL (1988) A primer of population genetics. 2nd ed. Sinauer, Sunderland, MA.

Hatfield T, Schluter D (1996) A test for sexual selection on hybrids of two sympatric sticklebacks. Evolution 50: 2429-2434.

Heins DC, Baker, JA, Tylicki DJ (1996) Reproductive season, clutch size, and egg size of the rainbow darter, Etheostoma caeruleum, from the Homochitto River, Mississippi, with an evaluation of data from the literature. Copeia 1996: 1005-1010.

Helvik JV, Drivenes O, Naess TH, Fjose A, Seo HC (2001) Molecular cloning and characterization of five opsin genes from the marine flatfish Atlantic halibut. Vis Neurosci 18: 767-780.

Hill GE, Montgomerie R (1994) Plumage color signals nutritional condition in the house finch. Proc R Soc Lond Ser B 258: 47-52.

Hisatomi O, Kayada S, Aoki Y, Iwasa T, Tokunga GAF (1994) Phylogenetic relationships among vertebrate visual pigments. Vis Res 34: 3097-3102.

Hisatomi O, Satoh T, Barthel LK, Stenkamp DL, Raymond PA, Tokunaga F (1996) Molecular cloning and characterization of the putative ultraviolet-sensitive visual pigment of goldfish. Vision Res 36: 933-939.

Hisatomi O, Satoh T, Tokunaga F (1997) The primary structure and distribution of killifish visual pigments. Vision Res 37: 3089-3096.

Höglund J, Alatalo RV (1995) Leks. Princeton, NJ : Princeton University Press.

148 Holland B, Rice WR (1998) Perspective: Chase-away sexual selection: Antagonistic seduction versus resistance. Evolution 52: 1-7.

Horth L (2001) Understanding the maintenance of a genetic body color polymorphism in male mosquitofish, Gambusia holbrooki. (PhD dissertation) Florida State University.

Houde AE(1992) Sex-linked heritability of sexually selected character in a natural population of Poecilia reticulata (Pisces: Poeciliidae) (guppies). Heredity 69: 229-235.

Houde AE (1997) Sex, color, and mate choice in guppies. Princeton: Princeton University Press.

Houde AE, Torio AJ (1992) Effect of parasitic infection on male color pattern and female choice in guppies. Behav Ecol 3: 346-351.

Hughes KA, Du L, Rodd FH, Reznick DH. (1999) Familiarity leads to female mate preference for novel males in the guppy, Poecilia reticulata. Anim Behav 58: 907-916.

Hunt DM, Fitzgibbon J, Slobodyanyuk SJ, Bowmaker JK (1996) Spectral tuning and molecular evolution of rod visual pigments in the species flock of cottoid fish in Lake Baikal. Vision Res 36: 1217-1224.

Hunt DM, Wilkle SE, Bowmaker JK, Poopalasundaram S (2001) Vision in the ultraviolet. Cell Mol Life Sci 58: 1583-1598.

Hunt S, Bennett ATD, Cuthill IC, Griffiths R (1998) Blue tits are ultraviolet tits. Proc R Soc Lond Ser B 265: 451-455.

Iwasa Y, Pomiankowski A 1994 The evolution of mate preferences for multiple sexual ornaments. Evolution 48: 853-867.

Jacobs GH, Neitz J, Krough K (1996) Electroretinogram flicker photometry and its applications. J Optical Soc Am A 13: 641-648.

Janetos AC (1980) Strategies of female choice: a theoretical analysis. Behav Ecol Sociobiol 7: 107-112.

Jennions MD, Petrie M (1997) Variation in mate choice and mating preferences: A review of causes and consequences. Biol Rev 72: 283-327.

Johnsen, S. (2000) Cryptic and conspicuous coloration in the pelagic environment. Proc R Soc Lond Ser B 269: 243-256.

Johnstone RA (1995) Sexual selection, honest advertisement and the handicap principle – reviewing the evidence. Biol Rev 70: 1-65.

149

Kallman KD (1975) The platyfish, Xiphophorus maculatus. In: King RC (ed) Handbook of Genetics, Vol. 4. Plenum Publ. Corp., New York, NY, USA. pp 81-132

Keyser AJ, Hill GE. (2000) Structurally based plumage coloration is an honest signal of quality in male blue grosbeaks. Behav Ecol 11: 202-209.

Kodric-Brown A (1989) Dietary carotenoids and male mating success in the guppy: An environmental component to female choice. Behav Ecol Sociobiol 25: 393-401.

Kokko H, Brooks R, McNamara JM, Houston AI (2002) The sexual selection continuum. Proc R Soc Lond B 269: 1331-1340.

Kröger RHH, Bowmaker JK, Wagner HJ (1999) Morphological changes in the retina of Aequidens pulcher (Cichlidae) after rearing in monochromatic light. Vision Res 39: 2441-2448.

Kvarnemo C, Forsgren E, Magnhagen C (1995) Effects of sex ratio on intra- and inter- sexual behaviour in sand gobies. Anim Behav 50: 1455-1461.

Kvarnemo C, Svensson O, Forsgren E (1998) Parental behaviour in relation to food availability in the common goby. Anim Behav 56: 1285-1290.

Lafferty KD, Morris AK (1995) Altered behavior of parasitized killifish increases susceptibility to predation by bird final hosts. Ecology 77: 1390-1397.

Leal M, Fleishman LJ (2002) Evidence for habitat partitioning based on adaptation to environmental light in a pair of sympatric lizard species. Proc R Soc Lond Ser B 269: 351-359.

Lipetz LE, Cronin TW (1988) Application of an invariant spectral form to the visual pigments of crustaceans - implications regarding the binding of the chromophore. Vision Res 28: 1083-1093.

Loew ER (1994) A third, ultraviolet-sensitive, visual pigment in the today-gecko (Gekko-gekko). Vision Res 34: 1427-1431.

Loew ER, McFarland WN (1990) The underwater visual environment. In: Douglas RH, Djamgoz MBA (eds) The Visual System of Fish. Chapman and Hall, London. pp 1-44.

Loew ER, Fleishman LJ, Foster RG, Provencio I (2002) Visual pigments and oil droplets in diurnal lizards: a comparative study of Caribbean anoles. J Exp Biol 205: 927-938.

Loftus WF, Kushlan JA (1987) Freshwater fishes of southern Florida. Bulletin of the Florida State Museum: Biological Sciences 31: 147-344.

150

Long KD, Houde AE (1989) Orange spots as a visual cue for female mate choice in the guppy (Poecilia reticulata). Ethology 82: 316-324.

Luttbeg B (1996) A comparative Bayes tactic for mate assessment and choice. Behav Ecol 7: 451-460.

Lynch M, Walsh B (1998) Genetics and analysis of quantitative traits. Sinauer, Sunderland, MA.

Lythgoe JN (1984) Visual pigments and environmental light. Vision Res 24: 1539- 1550.

Lythgoe JN, Muntz WRA, Partridge JC, Shand J, Williams DMcB (1994) The ecology of the visual pigments of snappers (Lutjanidae) on the Great Barrier Reef. J Comp Physiol A 174: 461-467.

Macias Garcia C, de Perera TF (2002) Ultraviolet-based female preferences in a viviparous fish. Behav Ecol Sociobiol 52: 1-6.

MacNichol EF Jr (1986) A unifying presentation of photopigment spectra. Vision Res 26: 1543-1556.

Magurran AE, Seghers BH (1994) Sexual conflict as a consequence of ecology: evidence from guppy, Poecilia reitculata, populations in Trinidad. Proc R Soc Lond B 255: 31-36.

Mangel M, Clark C (1988) Dynamic modeling in behavioral ecology. Princeton, NJ: Princeton University Press.

Mansfield RJW (1985) Primate photopigments and cone mechanisms. In: Fein A, Levine JS (eds) The Visual System. Liss, New York, pp 89-106.

Marchetti K (1993) Dark habitats and bright birds illustrate the role of the environment in species divergence. Nature 362: 149-152.

McDonald CG, Hawryshyn CW (1995) Intraspecific variation of spectral sensitivity in threespine tickleback (Gasterosteus aculeatus) from different photic regimes. J Comp Physiol A 176: 255-260.

Milinski M , Bakker TCM (1990) Female sticklebacks use male coloration in mate choice and hence avoid parasitized males. Nature 344: 330-333.

Munz FW, McFarland WN (1977) Evolutionary adaptations of fishes to the photic environment. In: Crescitelli F (ed) The visual system in vertebrates. Springer-Verlag, New York. pp 193-274.

151

Nilsson DE, Pelger S (1994) A pessimistic estimate of the time required for an eye to evolve. Proc R Soc Lond Ser B 256: 53-58.

Ornborg J, Andersson S, Girffith SC, Sheldon BC. (2002) Seasonal changes in a ultraviolet structural colour signal in blue tits, Parus caeruleus. Biol J Linn Soc 76: 237- 245.

Östlund S, Ahnesjö I (1998) Female fifteen-spined sticklebacks prefer better fathers. Anim Behav 56: 1177-1183.

Ottenheim NM, Volmer AD, Holloway GJ (1996) The genetics of phenotypic plasticity in adult abdominal colour pattern of Eristalis arbustorum (Diptera: Syrphidae). Heredity 77: 493-499.

Owens IPF, Thompson DBA (1994) Sex-differences, sex-ratios and sex-roles. Proc R Soc Lond Series B 258: 93-99.

Page LM, Burr BM (1991) Freshwater fishes: North America north of Mexico. Houghton Mifflin, Boston

Palacios AG, Goldsmith TH, Bernard GD (1996) Sensitivity of cones from a cyprinid fish (Danio aequipinnatus) to ultraviolet and visible light. Vis Neurosci 13: 411-421.

Palacios AG, Varela FJ, Srivastava R, Goldsmith TH (1998) Spectral sensitivity of cones in the goldfish, Carassius auratus. Vision Res 38: 2135-3146.

Partridge JC, Cummings ME (1999) Adaptations of visual pigments to the aquatic environment. In: Archer SN, Djamgoz MBA, Loew ER, Vallerga S (eds) Adaptive mechanisms in the ecology of vision. Kluwer, Dordrecht, pp 251-283.

Partridge, L. (1983) Non-random mating and offspring fitness. In: Bateson P (ed) Mate choice. Cambridge University Press, Cambridge. pp 227-255

Penn JS (1998) Early studies of the photostasis phenomenon. In: TP Williams, AB Thistle (eds) Photostasis and related phenomena. Plenum Press: New York. pp 1-16.

Penn JS, Williams TP (1986) Photostasis: regulation of daily photon-catch by rat retinas in response to various cyclic illuminances. Exp Eye Res 43: 915-928.

Petersen CW (1990) The occurrence and dynamics of clutch loss and filial cannibalism in two Caribbean damselfishes. J Exp Mar Biol 135: 117-133.

Pomiankowski A (1987) Sexual selection – the handicap principle does work sometimes. Proc R Soc Lond Series B 231: 123-145.

152

Press WH, Flannery BP, Teukolsky SA, Vetterling WT (1989) Numerical recipes in Pascal. Cambridge University Press, Cambridge.

Price T (1996) An association of habitat with color dimorphism in finches. Auk 113: 256-257.

Provencio I, Loew, ER, Foster RG (1992) Vitamin-A2-based visual pigments in fully terrestrial vertebrates. Vision Res 32: 2201-2208.

Ptacek MB, Travis J (1996) Inter-population variation in male mating behaviours in the sailfin mollie, Poecilia latipinna. Anim Behav 52: 59-71.

Ptacek MB, Travis J (1997) Mate choice in the sailfin molly, Poecilia latipinna. Evolution 51: 1217-1231.

Pyron M (1995) Mating patterns and a test for female mate choice in Etheostoma spectabile (Pisces, Percidae). Behav Ecol Sociobiol 36: 407-412.

Pyron M (1996) Male orangethroat darters, Etheostoma spectabile, do not prefer larger females. Environ Biol Fish 47: 407-410.

Qvarnström A, Pärt T, Sheldon BC (2000) Adaptive plasticity in mate preference linked to differences in reproductive effort. Nature 405: 344-347.

Rausher MD, Fry JD (1993) Effects of a locus affecting floral pigmentation in Ipomoea purpurea on female fitness components. Genetics 134: 1237-1247.

Real LA (1990) Search theory and mate choice. I. Models of single-sex discrimination. Am Nat 136: 376-404.

Real LA (1991) Search theory and mate choice. II. Mutual interaction, assortative mating, and equilibrium variation in male and female fitness. Am Nat 138: 901-907.

Reimchen, T.E. 1989 Loss of nuptial color in threespine sticklebacks (Gasterosteus aculeatus). Evolution 43: 450-460.

Relyea RA (2002) Costs of phenotypic plasticity. Am Nat 159: 272-282.

Reynolds JD, Cóté IM (1995) Direct selection on mate choice: female redlip blennies pay more for better mates. Behav Ecol 6: 175-181.

153 Reynolds JD, Gross MR, Coombs MJ (1993) Environmental conditions and male morphology determine alternative mating behaviour in Trinidadian guppies. Anim Behav 45: 145-152.

Reznick DA, Bryga H, Endler JA (1990) Experimentally induced life-history evolution in a natural population. Nature 346: 357-359.

Reznick DA, Travis J (1996) The empirical study of adaptation in natural populations. In: Rose MR, Lauder GV (eds) Adaptation. San Diego: Academic Press. pp 243-289.

Rodd HF, Hughes KA, Grether GF, Baril CT (2002) A possible non-sexual origin of mate preference: are male guppies mimicking fruit? Proc R Soc Lond Ser B 269: 475- 481.

Roff DA (1992) Evolution of life histories – theory and analysis. New York: Chapman and Hall.

Rosenthal GG, Evans CS (1998) Female preference for swords in Xiphophorus helleri reflects a bias for large apparent size. PNAS USA 95: 4431-4436.

Ryan MJ 1991 Sexual selection, sensory systems, and sensory exploitation. Oxf Surv Evol Biol 7: 157-195.

Ryan MJ, Fox JH, Wilczynski W & Rand AS (1990) Sexual selection for sensory exploitation in the frog Physalaemus pustulosus. Nature 343: 66-67.

Ryan M.J, Keddy-Hector A (1992) Directional patterns of female mate choice and the role of sensory biases. Am Nat 139: s4-s35.

Schaerer S, Neumeyer C (1996) Motion detection in goldfish investigated with the optomotor response is "color blind". Vision Res 36: 4025-4034.

Schluter D, Price T (1993) Honesty, perception and population divergence in sexually selected traits. Proc R Soc Lond B 253: 117-122.

Schmitt J, Wulff RD (1993) Light spectral quality, phytochrome and plant competition. TREE 8: 45-51.

Scott RJ (2001) Sensory drive and nuptial colour loss in the three-spined stickleback. J Fish Biol 59: 1520-1528.

Scriber JM, Hagen RH, Lederhouse RC (1996) Genetics of mimicry in the tiger swallowtail butterflies, Papilio glaucus and P. canadensis (Lepidoptera: Papilionidae). Evolution 50: 222-236.

154 Seehausen O, van Alphen JJM, Witte F (1997) Ciclid fish diversity threatened by eutrophication that curbs sexual selection. Science 277: 1808-1811.

Servedio MR (2001) Beyond reinforcement: The evolution of premating isolation by direct selection on preferences and postmating, prezygotic incompatibilities. Evolution 55: 1909-1920.

Shand J, Hart NS, Thomas N, Partridge JC (2002) Developmental changes in the cone visual pigments of black bream, Acanthopagrus butcheri. J Exp Biol 205: 3661-3667.

Shimmin LC, Mai P, Li W-H (1997) Sequences and evolution of human and squirrel monkey blue opsin genes. J Mol Evol 44: 378-382.

Shuster SM, Wade MJ (1991) Equal mating success among male reproductive strategies in a marine isopod. Nature 350: 608-610.

Siegel S, Castellan NJ (1988) Nonparametric statistics for the behavioral sciences. New York: McGraw-Hill.

Siitari H, Huhta E (2002) Individual color variation and male quality in pied flycatchers (Ficedula hypoleuca): a role of ultraviolet reflectance. Behav Ecol 13: 737-741.

Sillman AJ, Carver JK, Loew ER (1999) The photoreceptors and visual pigments in the retina of a boid snake, the ball python (Python regius). J Exp Biol 202: 1931-1938.

Sillman AJ, Johnson JL, Loew ER (2001) Retinal photoreceptors and visual pigments in Boa constrictor imperator. J Exp Zool 290: 359-365.

Sinervo B, Lively CM (1996) The rock-paper-scissors game and the evolution of alternative male strategies. Nature 380: 240-243.

Sinervo B, Bleay C, Adamopoulou C (2001) Social causes of correlational selection and the resolution of a heritable throat color polymorphism in a lizzard. Evolution 55: 2040-2052.

Sinervo B, Svensson E, Comendant T (2000) Density cycles and an offspring quantity and quality game driven by natural selection. Nature 406: 985-988.

Sokal RR, Rohlf FJ (1995) Biometry: the principles and practice of statistics in biological research. Freeman and Company: New York.

Stacey NE (1984) Control of the timing of ovulation by exogenous and endogenous factors. In: Potts GW, Woton RJ (eds) Fish reproduction: strategies and tactics. San Diego: Academic Press. pp 207-222.

155 Stearns SC (1992) The evolution of life histories. Oxford: Oxford University Press.

Stevens J (1986) Applied multivariate statistics for the social sciences. Hillsdale, NJ: Erlbaum.

Thompson CW, Hillgarth N, Leu M, McClure HE (1996) High parasite load in house finches (Carpodacus maxicanus) is correlated with reduced expression of a sexually selected trait. Am Nat 149: 270-294.

Thorpe A, Douglas RH, Truscott RJW (1993) Spectral transmission and short-wave absorbing pigments in the fish lens. 1. Phylogenetic distribution and identity. Vision Res 33: 289-300.

Thorpe RS, Richard M (2001) Evidence that ultraviolet markings are associated with patterns of molecular gene flow. PNAS 98: 3929-3934.

Travis J, Woodward BD (1989) Social-context and courtship flexibility in male sailfin mollies, Poecillia-latipinna. (Pisces, Poeciliidae). Anim Behav 38: 1001-1011.

Trexler JC, McElroy TC, Loftus WF, Jordan F, Chick JH, Kandl KL, Bass OL Jr (2001) Ecological scale and its implications for freshwater fishes in the Florida Everglades. In: The Everglades, Florida Bay, and Coral Reefs of the Florida Keys: An Ecosystem Sourcebook. CRC, Boca Raton, FL. pp 153-181.

Trussell GC (2000) Phenotypic clines, plasticity, and morphological trade-offs in an intertidal snail. Evolution 54: 151-166. van der Meer HJ, Bowmaker JK (1995) Interspecific variation of photoreceptors in four co-existing haplochromine cichlid fishes. Brain Behav Evol 45: 232-240.

Via S (1993) Adaptive phenotypic plasticity: target or by-product of selection in a variable environment? Am Nat 142: 352-365.

Vincent ACJ (1994) Seahorses exhibit conventional sex roles in mating competition, despite male pregnancy. Behaviour 128: 135-151.

Warner RR (1984) Mating systems and hermaphroditism in coral reef fish. Am Sci 72: 128-136.

Warner RR (1987) Female choice of sites versus mates in a coral reef fish, Thalassoma bifasciatum. Anim Behav 35: 1470-1478.

Warner RR (1988) Traditionality of mating-site preferences in a coral-reef fish. Nature 335: 719-721.

156 Warner RR (1998) The role of extreme iteroparity and risk avoidance in the evolution of mating systems. J Fish Biol 53: 82-93.

Warner RR, Dill LM (2000) Courtship displays and coloration as indicators of safety rather than of male quality: the safety assurance hypothesis. Behav Ecol 11: 444-451.

Weinig C 2000a. Differing selection in alternative competitive environments: shade- avoidance responses and germination timing. Evolution 54: 124-136.

Weinig C 2000b. Plasticity versus canalization: population differences in the timing of shade-avoidance responses. Evolution 54: 441-451.

Widemo F, Saether SA (1999) Beauty is in the eye of the beholder: causes and consequences of variation in mating preferences. TREE 14: 26-31.

Wijngaarden PJ, Brakefield PM (2001) Lack of response to artificial selection on the slope of reaction norms for seasonal polyphenism in the butterfly Bicyclus anynana. Heredity 87: 410-420.

Winge O (1922a) A peculiar mode of inheritance and its cytological explanation. J Genetics 12: 137-144.

Winge O (1922b) One-sided masculine and sex-linked inheritance in Lebistes reticulatus. J Genetics 12: 145-162.

Wiegmann DD, Mukhopadhay K (1998) The fixed sample search rule and use of an indicator character to evaluate mate quality. J Theor Biol 193: 709-715.

Wiegmann DD, Mukhopadhay K, Real L (1999) Sequential search and the influence of male quality on female mating decisions. J Math Biol 39: 193-216.

Yokoyama S (1997) Molecular genetic basis of adaptive selection: examples from color vision in vertebrates. Annu Rev Genet 31: 315-336.

Yokoyama S, Yokoyama R (1996) Adaptive evolution of photoreceptors and visual pigments in vertebrates. Annu Rev Ecol Syst 27: 543-567.

Yokoyama S, Zhang H, Radlwimmer FB, Blow NS (1999) Adaptive evolution of color vision of the Comoran coelacanth (Latimeria chalumnae). Proc Natl Acad Sci USA 96: 6279-6284

Zamudio KR, Sinervo B (2000) Polygyny, mate-guarding, and posthumous fertilization as alternative male mating strategies. PNAS 97: 14427-14432.

157 Zuk M, Thornhill R, Ligon JD, Johnson K, Austad S, Ligon SH, Thornhill NV, Costin C (1990) The role of male ornaments and courtship behavior in female mate choice of red jungle fowl. Am Nat 136: 459-473.

158

BIOGRAPHICAL SKETCH

REBECCA C. FULLER Dept. of Biological Science Tallahassee, Florida 32306-4340 [email protected]

Educational Background

Ph.D., Florida State University, Biological Science 4/03, Advisor: Joseph Travis M.S., Michigan State University, Zoology 5/98, Advisor: Tom Getty B.S., University of Nebraska, Major: Biology, Minors: English and Chemistry, Highest Distinction 9/89-5/93, Advisor: Anthony Joern Fulbright Scholar, Uppsala University, Sweden. 12/93-9/94, Advisor: Anders Berglund Institute for Ecosystem Studies - “Introduction to Ecosystem Ecology” 1/95 Organization for Tropical Studies - Fundamentals of Tropical Biology - 6/97-8/97

Fellowships & Scholarships

University Research Fellowship FSU, 2000 - 2003 National Science Foundation Dissertation Improvement Award, 1999-2001 National Science Foundation Graduate Fellowship, 1995-1998 Michigan State Distinguished Fellowship, 1994 Research Training Group Assistantship - supplemental research funds, 1994 Fulbright Scholarship - Uppsala University, Sweden , 1993 Rhoades Scholarship Semi-Finalist - Omaha, NE - November 1992 Regeant’s Scholarship - University of Nebraska-Lincoln - tuition waiver - 1989-1993 Sigma Xi - Grant in Aid of Research - Awarded January 1997 American Museum of Natural History - Theodore Roosevelt Funds - Awarded May 1997

Teaching Assistantships

Evolution Teaching Assistantship - Florida State University - 1/99-5/99, 8/99-12/99, 1/00-5/00 Lower Vertebrates Teaching Assistantship - Florida State University - 8/98-12/98 General Biology Teaching Assistantship - University of Nebraska - Lincoln - 8/92-5/93 Honors Biology Teaching Assistantship - University of Nebraska - Lincoln - 1/91-5/91 159

Publications

Fuller RC (2002) Lighting environment predicts relative abundance of male color morphs in bluefin killifish populations. Proc R Soc Lond Ser B. 269: 1457-1465.

Fuller RC (2001) Patterns in male breeding behaviors in the bluefin killifish, Lucania goodei: a field study (Cyprinodontiformes: Fundulidae). Copeia 2001: 823-828.

Fuller RC (1999) Costs of group spawning to dominant, primary males in the rainbow darter, Etheostoma caeruleum. Copeia 1994: 1084-1088.

Fuller RC (1998) Sperm competition affects male behavior and sperm output in the rainbow darter. Proc R Soc Lond B. 265: 2365-2371.

Fuller RC (1998) Fecundity estimates for rainbow darters, Etheostoma caeruleum, in southwestern Ohio. Ohio J Sci 98: 2-5.

Fuller RC (in press) Disentangling female mate choice and male competition in the rainbow darter, Etheostoma caeruleum. (Copeia).

Fuller RC, Berglund A (1996) Behavioral responses of a sex role reversed pipefish to a gradient of perceived predation risk. Behav Ecol 7: 69-75.

Fuller RC, Fleishman LJ, Leal M, Travis J, Loew E (in press) Intraspecific variation in ultraviolet cone production and visual sensitivity in the bluefin killifish, Lucania goodei. J Comp Physiol A.

Fuller RC, Houle D (2003) Inheritance of developmental instability. In: Polak M (ed) Developmental Instability: Causes and Consequences. Oxford University Press, Oxford. pp 427-290.

Fuller RC, Houle D (2002) Detecting genetic variation in developmental instability by artificial selection on fluctuating asymmetry. J Evol Biol: 15: 954-960.

Fuller RC, Joern A (1996) Grasshopper susceptibility to predation in response to vegetation cover and patch area. J Orthopt Res 5: 175-183.

Fuller RC, Travis J (2001) A test for male parental care in a fundulid, the bluefin killifish, Lucania goodei. Environ Biol Fish 61: 419-426.

McCune AR*, Fuller RC*, Beck AA, Dawley JM, Fadool JM, Houle D, Travis J, Kondrashov AS (2002) A low genomic number of recessive lethals in natural populations of bluefin killifish (Lucania goodei) and zebrafish (Danio rerio). Science 2002: 2398-2401.

160 *these authors contributed equally to this work

Rettig J, Fuller RC, Corbett A, Getty T (1997) Fluctuating asymmetry indicates competition in leaves of an even-aged poplar clone. OIKOS 80: 123-127.

Fuller RC Multiple mating events reduce female choosiness: a model and its implications for experimental design. (In revision).

Presentations

American Society of Ichthyology and Herpetology. Kansas City, MO, 7/2002. Society for the Study of Evolution, Knoxville, TN, 6/2001. Animal Behavior Society. Atlanta, GA: 8/2000. American Society of Ichthyology and Herpetology. LaPaz, Mexico: 6/2000. Ecology, ethology, and evolution of fishes. Athens, FA: 5/2000. Joint meeting: British and American Ecological Societies. Orlando, FL: 2/2000. (poster) Fuller RC American Society of Ichtyology and Herpetology. PennState PA:7/1999 Fuller RC International Society for Behavioral Ecology. Asilomar, CA USA: 7/1998 Michigan State University Fifth Annual Research Day. East Lansing, MI USA: 4/1998. Fuller RC Darterfest. Toshimongo, MS USA: 3/1998. Animal Behavior Society Midwest Regional Conference. Indiana University, Bloomington, IN USA: November 22-24, 1996. (poster) XXIV International Ethological Conference. Honolulu, HI USA: 8/1995 Annual West Coast Marine Biology Summer Meeting: Fiskebäckskil, Sweden: 7/1994. Nebraska Academy of Sciences. Lincoln, NE USA: 4/1993. University of Nebraska Biological Sciences Student Symposium. Lincoln, NE USA: 4/1993. Invited Speaker - University of Illinois, Illinois Natural History Survey. Champaign- Urbana, IL USA: November 1, 1996 Invited Speaker - Alma College, Biology Department Seminar. Alma, MI USA: January 1996.

Professional Affiliations, Memberships, and Honoraries

Phi Beta Kappa, 5/93-present International Society of Behavioral Ecology, 1994 - present Animal Behavior Society, 1995 - present Society for the Study of Evolution, 1996 - present Sigma Delta Epsilon - Graduate Women in Science - Omega Chapter, 1996 – present American Society of Ichthyology and herpetology, 1999 - present

161

Journals Reviewed

Behavioral Ecology Environmental Biology of Fishes Oecologia Proceedings of the Royal Society London: Series B Biological Review

162