Sympatry Predicts Spot Pigmentation Patterns and Female Association Behavior in the Livebearing Fish Poeciliopsis baenschi

In this study, we explored the possibility that differences in pigmentation patterns among populations of the fish Poeciliopsis baenschi were associated with the presence or absence of the closely related species P. turneri. If reproductive character displacement is responsible, spotting patterns in these two species should diverge in sympatry, but not allopatry. We predicted that female P. baenschi from sympatric sites should show a preference for associating with conspecifics vs. heterospecific males, but females from allopatric sites should show no such preferences. To evaluate these predictions, we compared spotting patterns and female association behaviors in populations of P. baenschi from Central Mexico. We found that both of our predictions were supported. Poeciliopsis baenschi that co-occured with P. turneri had spotting patterns significantly different than their counterparts from allopatric sites. Using a simultaneous choice test of video presentations of males, we also found that female P. baenschi from populations that co-occured with P. turneri spent significantly more time with males of their own species than with P. turneri males. In contrast, females from allopatric populations of P. baenschi showed no differences in the amount of time they spent with either conspecific or heterospecific males. Together, our results are consistent with the hypothesis that reproductive character displacement may be responsible for behavioral and spotting pattern differences in these populations of P. baenschi.


Introduction
A variety of phenotypic cues can be used in species recognition, including visual, chemical, and auditory cues. When closely related species do co-occur, they often diverge from one another in species recognition traits relative to their conspecific counterparts that occur in allopatry. Such divergence helps minimize costly reproductive interactions between heterospecifics. This form of divergence is known as reproductive character displacement (hereafter referred to as RCD) [1][2][3][4][5][6][7][8]. a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 co-occurs with P. turneri, and in other locations it occurs alone. Populations of P. turneri do not occur without P. baenschi.
We collected P. baenschi and P. turneri from 11 localities in western Mexico during May and June 2007 (Fig 2, Table 1). Our samples included six localities where P. baenschi co-occurs with P. turneri and five localities where P. baenschi occurs alone (Fig 2). All sympatric localities were taken from the same drainage system where these species have come into secondary contact [53]. Each sympatric collection was made at least 3.5 km apart to ensure these were independent replicates. Although we do not know the extent to which gene flow occurs among these populations, our analyses of spotting patterns did not show a clinal gradient (S1 Fig), suggesting that each sympatric sampling locality is distinct.   Table 1 All fish were collected with a hand-held seine net (1.3 m x 5 m; 8 mm mesh size). Fish used for the spotting pattern analysis were euthanized in the field, preserved in ethanol, and transported to the laboratory for data collection. Live fish used in the behavioral analysis were collected from three locations: site three (sympatric); and sites seven and 11 (allopatric; see Fig 2;  Table 1). Live fish were transported to the laboratory at Brigham Young University where they were housed in 20 gallon tanks at 22˚C, a temperature typical for these fishes in the wild. Fish were isolated by population, fed twice daily, and kept on a 12:12 hour LD cycle. The Institutional Animal Care and Use Committee at Brigham Young University approved the use of live fish for this study (IACUC protocol 06-0104) and all guidelines and recommendations in this protocol were strictly followed.

Quantifying Spots
We collected spot data from ethanol-preserved samples. Although ethanol can diminish the intensity of melanin spots, we had no difficulty measuring the shape or number of spots in our samples. We quantified spotting patterns by measuring two primary characteristics of spots of reproductively mature males; (1) the number of spots; and (2) the total pigmented area of spots. Our focus on males was to complement the female association tests described below. Previous work suggests that spotting patterns can sometimes vary between different sides of the same fish [43]. However, we found that asymmetry in spot size and color between sides was low compared to differences between populations and species, and that neither side had an inherent bias in number or shape of spots (S2 Fig). Hence, we collected all of our data from the left side of each fish. We examined 98 specimens of P. baenschi from allopatric sites, 111 specimens of P. baenschi from sympatric sites, and 67 specimens of P. turneri from sympatric sites. From these samples, we quantified the number and total area of spots using ImageJ 1.41 (http://rsbweb.nih.gov/ij/). The spots of these species were conspicuous allowing us to count the total number of spots on each fish by eye. We measured the total area of spots covering the side of the fish bounded by the opening of the operculum at the anterior margin and the end of the vertebral column at the posterior margin. We quantified area (in mm 2 ) by transforming each image to black and white and classifying pixels as pigmented or non-pigmented using a color threshold between 30 and 50.

Behavioral Assays
Our behavioral assay allowed P. baenschi females from allopatric and sympatric locations to choose between males of their own species versus P. turneri males. We used a dichotomouschoice test with a video playback system [48,51,54,55] to determine if P. baenschi females derived from sympatric and allopatric sites showed a difference in their association times with conspecific versus heterospecific males.
To measure association time, a single P. baenschi female was simultaneously presented with two stimulus videos of males following previously published methods [56]. In brief, we created stimulus videos using a composite of video images of three males from each of our three focal populations (P. baenschi in sympatry, P. baenschi in allopatry, and P. turneri). The males used to create the stimulus tapes were of similar size (within 2 mm) and were typical in terms of spotting patterns for each population. We conservatively chose a 2 mm size difference threshold because it is smaller than the 3 mm difference shown to be necessary to have an effect on mate choice in similar studies of poeciliid fish mating preferences [57][58][59]. In addition, males used to make the videos were chosen to be as similar as possible, with the primary difference being the number and character of spots. The composite video provided a 10 minute looped segment of a single male swimming back and forth across the screen. Males in these looped videos did not show any courtship behavior, although the swimming behavior was typical of Poeciliopsis males that approach females from the side or behind prior to mating. Hence, females were simply given an opportunity to spend time with one video male or the other. Previous observations in other poeciliid fishes has shown that the male that a female associates with is frequently a good indicator of male mating success [56,[60][61][62][63]. We also made control videos showing only the background with no stimulus.
We conducted a total of 10 trials for sympatric females and 11 trials for allopatric females, sample sizes sufficient to detect differences in association time in our study (see below). Females included in the study were separated from males for at least two week before starting the trials [64][65][66]. In each trial, a female was introduced into the tank and allowed to acclimate for 10 minutes while empty tank control videos were shown on monitors abutting the opposite sides of the test tank. Following the 10-minute acclimation period, we started the 10-minute male stimulus videos. Poeciliopsis baenschi females were presented with a choice between a conspecific and heterospecific male. To control for side preferences, we randomly assigned the side to which the males were presented. A video camera placed one meter from the front glass recorded each trial. All recording was done remotely from an adjacent room. Association time was defined as the amount of time a female spent in the third of the tank closest to the stimulus video [67].

Statistical Analyses
We compared spotting patterns between groups using a general linear model framework. The number of spots and total area of spots were both analyzed by analysis of covariance (ANCOVA). In each model, we tested for differences between allopatric P. baenschi, sympatric P. baenschi, and sympatric P. turneri. The number of spots and total area of spots can covary with fish body size because larger fish have larger spots. Hence, we included "area of fish" as a covariate. We quantified "area of fish" by outlining the fish body in ImageJ over the same area for which spot pigment was measured. To meet the assumptions of the statistical models, 'number of spots' was square root transformed and 'area of fish' was natural log transformed. Our results were the same regardless of whether or not the data were transformed; hence, for ease of interpretation we present non-transformed results. The interaction term between the "groups" (allopatric P. baenschi, sympatric P. baenschi, and sympatric P. turneri) and "area of fish" was also included in each model to determine if spotting patterns changed among groups as a function of body size. Finally, we tested for differences in the two spot traits between sympatric and allopatric populations of P. baenschi (excluding P. turneri) using a one-way analysis of covariance (ANCOVA). These ANCOVA models used the same variables as those described above.
To analyze the association behavior data, we used a one-way analysis of variance (ANOVA). We compared the amount of time females spent associating with a conspecific male versus time spent with the heterospecific P. turneri male. Because females had the option of not interacting by remaining in the center of the tank, we treated the amount of time females spent with each male as an independent measure [56,62,68]. This statistical test was performed separately for the sympatric population and the allopatric population because we wanted to know if each population differed in their association with conspecifics versus heterospecifics. We also tested for a tank side-bias by comparing the amount of time spent on each side of the tank during our control treatments when the control videos were presented. In total, we present the results for four separate tests.
Statistical significance was evaluated at the P < 0.05 level. All statistical tests were conducted in R [69]. We report the least square means and standard error for the number of spots, total area of spots and association time from the behavioral assays in the results. For the number and total area of spots, the least square means were adjusted for the covariate (the differences in size of the groups), thus allowing us to compare the differences in number of spot between groups regardless of the differences in body size among individuals between the groups. For the behavioral data, we used least square means to account for differences in the number of replicates for each of the treatments (10 sympatric vs. 11 allopatric).

Spotting Pattern
Allopatric P. baenschi, sympatric P. baenschi and P. turneri differed significantly from one another in their spotting patterns (Table 2; Fig 3). Consistent with our predictions, we found that sympatric populations were more divergent from P. turneri than were allopatric populations for both the number of spots and total pigmented area (Fig 3). Poeciliopsis turneri had the greatest number of spots (mean ± 1 SE, 7.58 ± 0.21), sympatric populations of P. baenschi had the fewest number of spots (4.64 ± 0.10), and allopatric populations of P. baenschi had spot numbers intermediate between the other two groups (6.45 ± 0.13; Fig 3A). Total area of spots varied significantly between the three groups. Poeciliopsis turneri had the greatest area (mm 2 ) of spots (9.29 ± 0.30), sympatric populations of P. baenschi had the smallest area of spots (3.36 ± 0.15), and allopatric populations of P. baenschi were intermediate, although closer to sympatric P.baenschi than P. turneri (4.26 ± 0.18; Fig 3B). We also found a significant interaction between group and area of fish (fish size) for both the number of spots and total area of spots (Table 2; Fig 4), indicating that spot area and number of spots scaled with body size, but did so in different ways for each of the groups. At smaller body sizes, the total area of spots was similar between the three groups, but at larger body sizes, P. turneri had a significantly greater area of spots than both sympatric and allopatric populations of P. baenschi (Table 2; Fig 4). Area of fish was not a good predictor of number spots, as shown by the low goodness of fit of the models (R 2 McF = 0.004, 0.004, 0.002; Fig 4; [70]). Our second set of analyses focused on the comparison between sympatric and allopatric populations of P. baenschi. We found that sympatric populations of P. baenschi had significantly fewer spots (4.55 ± 0.10) than allopatric populations (6.41 ± 0.11; F 1,205 = 145.45, P < 0.001). We also found that the total area of spots was smaller in sympatric populations (3.01 ± 0.10 mm 2 ) than in allopatric populations (3.81 ± 0.11 mm 2 ; F 1,205 = 11.17, P < 0.001). Interestingly, total spot area and number of spots were only weakly correlated (R = 0.24, P < 0.001) in P. baenschi, suggesting the potential for these traits to function somewhat independently.  Fig 5). In contrast, allopatric P. baenschi females showed no significant difference in their association time between their own species versus P. turneri (conspecific: 197.77 ± 45.69 s; heterospecific: 257.87 ± 45.69 s; ANOVA F 1,20 = 0.86, P = 0.36; Fig  5). To check whether our non-significant results were due to lack of statistical power, we conducted a post hoc power analysis using G Power 3 [71] and found that for the effect size observed in the present study (d = 0.198) with an alpha of 0.05, a sample of approximately 202 would be needed to obtain statistical power at the recommended 0.80 level [72]. Moreover, our controls revealed no evidence for a tank side effect because individuals were equally likely to spend time on either side of the tank in the absence of the male stimulus (sympatric female, Again, power analysis revealed that for the effect size observed here for both sympatric (d = 0.334) and allopatric females (d = 0.131) with an alpha of 0.05, a sample of approximately 74 and 458 would be needed to obtain statistical power at the recommended 0.80 level [72]. All power tests showed that to achieve a level of 0.80, sample sizes that are prohibitively large for most behavioral studies are requiered. Given that our non-significant results greatly overlapped, it is reasonable to conclude that no differences were observed.

Discussion
Our results are consistent with the reproductive character displacement hypothesis. Spotting patterns differed more between Poeciliopsis baenschi and P. turneri where these species cooccur than when P. baenschi occurs alone. Also, P. baenschi females from sympatric sites showed preferences for males of their own species, but no such preference was observed in P. baenschi from allopatric sites. Why selection should favor reproductive character displacement in these fishes? Two plausible explanations are reinforcement and reproductive interference. Although some species in the genus Poeciliopsis are known to hybridize [73,74], most taxa in the genus maintain distinct species boundaries. To date, no evidence of hybridization between P. baenschi and P. turneri exists, so we cannot rule out the possibility that reinforcement has led to our observed differences, but we consider it unlikely. In contrast, sexual interference is a plausible explanation given the overall phenotypic similarity between these two taxa and the similar male mating tactics of forcing copulations. Under these circumstances, mistaken mating attempts by males can be costly to females [75][76][77][78][79]. Costs such as energy investment to avoid males or actual injuries caused by males have caused females to change their association behavior to avoid harassment in closely related species [75][76][77][78][79]. Our data are consistent with this explanation. It is possible that the female association preferences observed here are not completely related to spotting patterns, but could be explained by confounding traits such as shape. However, we consider this unlikely given that we matched stimulus males to control for such differences. This matching accounted for coloration and male size traits that can influence visual preference in poeciliid female [9,62,[80][81][82].
Even though our data are consistent with an explanation of RCD, other factors could contribute to the observed patterns including confounding ecological variables [4,12,83] and differential fusion, where species that come into secondary contact either fuse or are maintained depending on the strength of mating discrimination present before secondary contact.
Ecological variables that affect the presence or discrimination of a phenotype could lead to changes in mating cues or association preference [18,82,[84][85][86]. It is possible that the relationship between spot number and species co-occurrence is driven by an unidentified common ecological factor. For example, both predation pressure and resource availability have been shown to affect pigmentation in poeciliid fishes [73,74,[87][88][89][90][91][92][93]. Predation can cause either an increase or decrease in pigmentation, depending on the degree to which it affects the conspicuousness of an individual. Unfortunately, there are no known differences in predation pressure among our collecting localities [94]. Environmental resources can also affect fish pigmentation where pigments are directly obtained from the diet [48,54,55,[75][76][77][78][79]. Unfortunately, this is not the case with melanin, a pigment that is not diet derived but is instead synthesized internally. Previous work [94] has demonstrated that the local environments of P. baenschi evaluated here -both sympatric and allopatric-do not differ in resource availability.
The hypothesis of differential fusion suggests that when species come in secondary contact, they fuse or persist as distinct species depending on the strength of mating discrimination that existed in allopatry [12]. Our results show that allopatric P. baenschi females had no association preference and thus had no effect on male mating potential. Therefore differential fusion is an Spotting Pattern and Female Association Behavior unlikely explanation. Our best explanation is that the presence or absence of heterospecific P. turneri has influenced spotting patterns and female association preferences in P. baenschi.
In conclusion, our results are consistent with the explanation that species recognition behavior has evolved in sympatric sites where there are potential fitness risks caused by the presence of a heterospecific, but is absent in allopatric sites where there are no such risks. Consistent with our findings, several other studies suggest that barring or spotting patterns on fish can be used as visual cues in species recognition [44]. Differences in spotting patterns and species association behaviors observed here suggest that sympatric populations of P. baenschi express phenotypes that reduce the possibility of reproductive interference. Whether these differences are genetically based or environmentally induced remains unknown. More research is needed to directly determine what fitness benefits sympatric populations of P. baenschi achieve in their habitat relative to allopatric populations. Nonetheless, our study points to the presence of the closely related fish P. turneri as an important factor for shaping phenotypic divergence in both spotting patterns and association behavior among P. baenschi populations.