Variation in predation regime drives sex-specific differences in mosquitofish foraging behaviour

nearby predators and females. Females of this species are larger than males, bear live young and show higher foraging rates in the wild than males. On the other hand, males spend more time pursuing females in the wild, and may exhibit greater flexibility in foraging behaviours based on the immediate context. Our results show that varying


Introduction
Predation can act as a strong selective agent on prey phenotypes, shaping the evolution of prey traits such as life history, morphology and behaviour (Reznick et al. 1990, Swaddle and Lockwood 1998, Bell and Sih 2007, Langerhans 2007).Major ecosystemlevel impacts of predation not only derive from direct lethal effects that alter the mortality regimes of prey populations, but can also occur through indirect trait-mediated effects (Peacor andWerner 2001, Preisser et al. 2005).That is, perceived predation risk can induce changes in prey traits, irrespective of lethal effects, such as altered reproductive output, activity levels or habitat use (Valeix et al. 2009, Zanette et al. 2011, Sha et al. 2020).In the absence of predators, or in relatively low-risk environments, an increase in prey population densities can result in reduced resources per capita, which should select for individuals with high competitive potential (Palkovacs et al. 2011, Reznick et al. 2019).There is indeed support for organisms trading off predator defences for competitive ability, with the absence of predators favouring individuals with relatively high growth rates (Yoshida et al. 2003, 2004, Araujo et al. 2017) or higher survival in competitive environments (Kraaijeveld and Godfray 1997).Despite this, the majority of studies addressing how predator-mediated selection can drive adaptive shifts in prey traits have focused on traits known to confer survival benefits in high-risk environments, whereas traits associated with competitive advantage have received less attention.
The relative effects of both predation and competition can also be sex specific.For example, in many systems, females seem to experience higher predation risk as a result of either their larger body size or gestation state (Magurran and Nowak 1991, Svensson 1997, Plath et al. 2011).Additionally, as a consequence of anisogamy, males usually face stronger competition for mating opportunities, whereas female fitness tends to be more limited by resource competition (Perrin andMazalov 2000, Lehtonen et al. 2016).In other words, if females are both more susceptible to predation and more affected by resource competition, they should face a stronger tradeoff between predator avoidance and competition.
Foraging is a key behavioural trait that can directly influence reproductive output, but is also susceptible to the tradeoff between predator avoidance and competitive ability.That is, increased foraging effort can make the individual more susceptible to predation (Cowlishaw 1997, Cooper Jr. 2000), but can also be favoured by selection via resource competition (Mitchell et al. 1990).Thus, behavioural traits associated with foraging capacity should be under divergent selection across areas with high and low predation risk.The consistent selective pressures in different predation regimes can therefore result in different phenotypes that persist even when the immediate cues (predator presence or conspecific density) change, as opposed to individuals adjusting their behaviour based on current perceived predation risk (Brown andKotler 2004, Verdolin 2006) or population density (Davidson and Morris 2001).We still lack a strong understanding of how predation regimes shape such foraging behaviours in prey across systems (but see Fraser et al. 2004 andElvidge et al. 2016) and whether this effect depends on prey sex.
In this study, we used the model system of the post-Pleistocene radiation of Bahamas mosquitofish Gambusia hubbsi to examine how naturally contrasting predation environments have shaped foraging behaviours of prey populations.The Bahamas mosquitofish is a sexually dimorphic live-bearing fish inhabiting inland blue holes on Andros Island, The Bahamas.Some blue holes completely lack piscivorous fish (low-predation regime) whereas other populations (highpredation regime) coexist with a highly piscivorous fish, the bigmouth sleeper Gobiomorus dormitor (Langerhans et al. 2007, Björnerås et al. 2020).The elevated extrinsic mortality rates in high-predation blue holes has resulted in marked differences in Bahamas mosquitofish population density between regimes, with low-predation populations having ca five times higher density than high-predation counterparts (Heinen et al. 2013).These two contrasting predation regimes have driven adaptive phenotypic divergence in numerous traits and performance capabilities in Bahamas mosquitofish (Langerhans et al. 2007, Langerhans 2018, Riesch et al. 2020).For example, mosquitofish originating from high-predation environments have evolved a body shape more suitable for fast-start swimming bursts, while low-predation fish have evolved a body shape better suited for energy-efficient prolonged swimming (Langerhans et al. 2007, Langerhans 2009, Araujo et al. 2017).In low-predation populations, we expect high levels of foraging effort to be advantageous because of stronger resource competition and the absence of predation threat.In high-predation environments, on the other hand, it should be advantageous to avoid unnecessary activity in order to avoid the attention of predators, as well as to conserve energy for potential predator escapes.Unlike males, female mosquitofish grow indeterminately and their weight positively correlates with their fecundity (Riesch et al. 2013), while the larger body size and gestation may potentially make them more favourable prey.Consequently, they may simultaneously be more at risk to predation and have higher motivation for gathering resources.In the wild, Bahamas mosquitofish in low-predation populations tend to show higher foraging rates than those in high-predation populations (Heinen et al. 2013), but whether differences persist in the absence of immediate predation risk and under controlled density remains unknown.
We hypothesize that natural selection will more strongly favour high foraging effort in low-predation environments, and that a difference in foraging effort between predation regimes will persist even when fish from either regime forage in the absence of immediate predation risk.We test two predictions in this study.First, we predicted that mosquitofish from low-predation regimes would show higher foraging rates and food consumption rates compared to individuals from high-predation regimes, while foraging efficiency (food consumption per foraging attempt) might remain similar between predation regimes.Second, owing to the stronger hypothesized tradeoff in females, we predicted that differences between predation regimes would be stronger in females.

Material and methods
Blue holes in the Bahamas are vertical water-filled caves that have been colonized by fish since the ocean levels rose after the last interglacial period ~15 000 years ago (Fairbanks 1989, Mylroie et al. 1995).Inland blue holes have a surface freshwater layer and are without inlets or outlets except for some underground tunnels filled with anoxic saline groundwater (Bottrell et al. 1991).These small but vertically deep lakes are stable environments with simple fish and plankton communities (Heinen et al. 2013, Björnerås et al. 2020).In addition to Bahamas mosquitofish, each blue hole used in this study harbours one additional fish species (Table 1): the predatory bigmouth sleeper was always present in the high-predation sites, whereas the low-predation sites had either the sheepshead minnow Cyprinodon variegatus or the crested goby Lophogobius cyprinoides.The latter two species are similarly sized to Bahamas mosquitofish and represent potential competitors of G. hubbsi but unlikely pose any meaningful predation threat (Langerhans et al. 2007, Heinen et al. 2013, Langerhans 2018).Because the fish communities of blue holes are dominated by Bahamas mosquitofish, intraspecific resource competition is expected to play a much greater role in their foraging levels and abilities than interspecific competition (Heinen et al. 2013).Prior analyses of population densities and age structure suggest that all age classes of Bahamas mosquitofish suffer higher mortality in the presence of bigmouth sleepers, and thus strongly elevated mortality risk occurs for all Gambusia hubbsi in high-predation blue holes (Heinen et al. 2013, Riesch et al. 2020).Most Bahamas mosquitofish populations in blue holes are also highly isolated from one another, and their genetic relatedness does not correlate with predation regime (Schug et al. 1998, Riesch et al. 2013).Thus, the Bahamas mosquitofish system provides a natural experiment with multiple populations with minimal gene flow at early stages of speciation (Langerhans et al. 2007, Heinen-Kay and Langerhans 2013, Langerhans and Riesch 2013).The mosquitofish diet is dominated by small invertebrates, such as zooplankton and insect larvae (Gluckman andHartney 2000, Araujo et al. 2014).While foraging propensity typically depends on mosquitofish body size (Rehage et al. 2005), the average body length and weight do not differ between predation regimes (Langerhans et al. 2007, Langerhans 2018).
Foraging trials were used to evaluate the feeding behaviour of individuals from six mosquitofish populations (three high-predation (HP) and three low-predation (LP) populations, Table 1).Fish were caught in the wild using hand nets and transported to a field laboratory on the day of capture.The fish were kept in mixed-sex 45-l tanks with water from their respective blue holes, filtration pumps and artificial plant-like plastic structures.The fish were acclimated to laboratory conditions for at least one day, and all fish were tested within five days of capture.Fish were fed flake food and dried Chironomidae larvae in the evenings and starved overnight to standardize hunger levels among individuals.Because Bahamas mosquitofish typically swim and forage in small groups in the wild, and because we explicitly wished to quantify foraging behaviour within a competitive scenario, we used two individuals in each trial.Each experimental replicate consisted of two similarly sized fish of the same sex and population of origin to avoid sexual interactions and population-and size-specific effects on behaviours.The individuals of each pair were placed together in a transparent arena (22 × 10 × 14 cm) filled with water from their own tank and left to acclimate for 15 min before the start of each trial (n = 14 (HP females), 19 (HP males), 12 (LP females), 16 (LP males) (see the Supporting information for more detailed sample sizes).The sex of mature mosquitofish is easily identified, as the male anal fin develops into a gonopodium upon maturation.At the start of the trial, we initiated video recording from a lateral view and added 15 small (2-3 mm) cut pieces of Chironomidae larvae into the centre of the tank.Trials lasted 15 min, starting when food entered the tank.We recorded foraging behaviours from the videos using the Behavioral observation research interactive software BORIS (Friard and Gamba 2016).Two behaviours were manually recorded from the 15-min trials: 1) foraging rate: number of foraging attempts performed by the fish, defined as an individual's mouth contacting or distinctly nudging towards the bottom of the tank, whether a food item was consumed or not, and 2) consumption rate: number of times a fish successfully consumed a food item.A third behaviour, foraging efficiency, was calculated from these two measures as the proportion of successful foraging instances out of all foraging attempts during the 15-min trial (food consumed/foraging attempts).Because we were interested in average foraging behaviours during competitive interactions, we pooled foraging behaviours for the two fish in each trial.Following the trial, each fish was photographed from above, with a standardized scale on the bottom of the container for the measurement of standard length (SL).Standard length was calculated from the photographs using the ImageJ ver.5.1 program (Schneider et al. 2012).The average SL was 27.7 mm (± 5.1 SD) for females and 23.0 mm (± 2.2 SD) for males.
We used a generalized linear model (GLM) with a Poisson distribution (log link function) to examine foraging rate and consumption rate, and a general linear model to examine foraging efficiency.Each model included the factors sex, predation regime, the interaction between sex and predation regime, the covariates average SL (log 10 -transformed body length) of fish in each trial, the size difference between individuals in pairs (absolute value of the difference between log 10 -transformed SL values), and number of days in captivity, along with population of origin as a random variable nested within predation regime.To prevent type 1 errors caused by overdispersion of data, the standard errors of the Poisson GLMs were scaled to quasipoisson distribution (Hilbe 2014) when analysing foraging rate and consumption rate.We did not transform foraging efficiency, as data met assumptions of normality of residuals.All statistical analyses were performed using R ver.3.5.1 (<www.r-project.org>).In the case of significant interaction terms, we inspected least-squares means to interpret patterns in the data.Owing to relatively low statistical power, we did not perform post hoc tests, but rather examined effect sizes to draw conclusions only from clear patterns of observed differences.

Results
Foraging rate was significantly affected by the interaction term between predation regime and sex (Table 2).Inspection of least-squares means revealed that females more strongly differed between predation regimes than males, with LP females showing higher foraging rates than HP females, while males showed only a weak trend in the opposite direction (Fig. 1a).This resulted in a pattern of sexual dimorphism in HP populations (higher foraging rate in males than females), but only weak evidence of any sex differences in LP populations.Similarly, consumption rate was also significantly affected by the interaction term between predation regime and sex (Table 2), and showed a similar pattern of group differences (Fig. 1b).LP females showed a higher consumption rate than HP females, while males again showed little evidence of differences between predation regimes (only a weak trend and in the opposite direction; Fig. 1b).This resulted in a similar pattern of sexual dimorphism for consumption rate as for foraging rate.No model terms were significant for foraging efficiency (Table 2, Fig. 1c).

Discussion
Using controlled trials, where fish competed for a constant number of prey items under a common density and in the absence of immediate predation risk, we found that the predation regime of origin affected foraging behaviours of a live-bearing fish in a sex-specific manner.Female foraging behaviours matched our initial prediction, with female mosquitofish from low-predation regime showing higher foraging and consumption rates, as compared to females originating from high-predation environments.However, males did not exhibit this pattern, showing weak trends in the opposite direction from females.Below, we argue that these sex-and predation-specific results are likely due to natural selection on foraging behaviours (expressed through plasticity or genetically-based differences) and life-history strategies in the wild.
Long-term exposure to different predator and competition regimes has apparently shaped foraging behaviours of female Bahamas mosquitofish.In the wild, female Bahamas mosquitofish show elevated foraging rates in low-predation populations (Heinen et al. 2013).We found that in a controlled laboratory setting, with constant density and the absence of any immediate predation threat, females from low-predation populations still showed higher foraging and consumption rates than counterparts in high-predation populations.Considering that the quality and quantity of resources are not known to differ between predation regimes (Heinen et al. 2013), this means that when competing for food low-predation females likely show higher rates of energy acquisition than high-predation females.This matches our prediction based on divergent selection between predation regimes: high population densities and the absence of chronic predation risk should select for higher competitive ability in low-predation blue holes.The stronger differences in foraging behaviours between predation regimes in females compared to males support our prediction that female mosquitofish face a stronger tradeoff between predator avoidance and resource competition.In multiple systems, females and males experience different sexspecific predation risk, such as egg-carrying females being preferred as prey (Svensson 1997, Plath et al. 2011) or males with conspicuous sexual signals and ornaments being more susceptible to predators (Zuk andKolluru 1998, Godin andMcDonough 2003).In Bahamas mosquitofish, larger females may be more favourable prey for the bigmouth sleeper, and as live-bearers, their slower breeding potential may cause them to avoid risks to a greater extent than males.Female mosquitofish are indeed more risk averse than males, showing lower boldness and reduced exploration of novel environments (Heinen-Kay et al. 2016), similar to patterns observed in other taxa, such as guppies (Magurran and Nowak 1991, Piyapong et al. 2010, Lucon-Xiccato et al. 2016).On the other hand, the indeterminate growth linked to fecundity of female mosquitofish can result in a higher motivation for gathering resources in females (Boehlert et al. 1991, Hayward and Gillooly 2011, Barneche et al. 2018).
While the female mosquitofish in our study maintained foraging patterns similar to those previously observed in the wild, the same was not true for males.When observed in blue holes, wild male mosquitofish from low-predation sites show slightly higher foraging rates than those in high-predation sites (Heinen et al. 2013).In the present study, we observed similar foraging and consumption rates for males from low-and high-predation populations, with the trend in the opposite direction as that observed in the wild.This difference between in situ observations and controlled laboratory trials may be due to sex-specific plasticity in foraging behaviours.In contrast to females, male poeciliid fishes typically allocate more time to sexual behaviour than to feeding (Magurran 2005, Heinen et al. 2013).In high-risk environments, male mosquitofish must trade off time for courting, predator avoidance and foraging.Thus, males from high-predation populations may take advantage of both the absence of predators (i.e.lower perceived risk) and the absence of females (i.e.no need for mate searching or sexual behaviours) in the experimental context to increase their foraging behaviours.Similar compensatory foraging has been observed in other poeciliid fish, where individuals from high-predation sites resume foraging faster and at elevated rates after predatory encounters in comparison to low-predation individuals (Elvidge et al. 2014).The more risk-averse females may also show behavioural plasticity in the opposite direction, reducing their foraging behaviours in the experiment following a generally elevated level of caution in an unfamiliar situation.This may explain why females did not show higher average foraging rates than males in our study.
Our study suggests that the loss of predation pressure in the low-predation blue holes may have selected for changes in foraging behaviours, and that these effects may differ between females and males.The persisting divergent foraging behaviours between wild populations can in turn have implications for population differentiation, whether they reflect phenotypic plasticity, fixed genetic differences or both.For example, similarly to how migrating or introduced predator-naïve individuals may face higher mortality in areas with high predation risk (Langerhans 2009, Ingley andJohnson 2016), individuals from high-predation areas can be outcompeted in low-predation areas where population densities are much higher.Such immigrant inviability can act to reduce gene flow between populations.Sex-specific costs and benefits of foraging traits may also lead to different foraging strategies between males and females (Lewis et al. 2006, Pichegru et al. 2013) that can also be involved in population differentiation (Berner et al. 2008).A stronger tradeoff between resource competition and predator avoidance in females compared to males could therefore lead to females diverging more between populations, or potentially to male-biased dispersal rates if other factors do not additionally select against male immigrants.Loss of top predators is a major current global issue (Myers and Worm 2003, Strong and Frank 2010, Estes et al. 2011, He et al. 2019), so studying the effects of predation on prey behaviour and community composition is an important task (Heithaus et al. 2008).Still, a majority of the studies estimating the effects of altered predation pressure mainly focus on prey traits that may increase survival in the presence of predators (Fowler et al. 2018), which may lead to underestimating the importance of traits that are selected for in low-risk but high-competition environments.It has been shown that after major predator loss, populations adapted to high-predation environments can start to resemble low-predation populations both morphologically and behaviourally relatively fast (Palkovacs et al. 2011).Combined with the large number of traits that consistently differ among prey populations living with or without predators (Langerhans 2018), this indicates that there may be strong selection for particular phenotypes in the absence of major predators.Considering our results, studies should therefore carefully evaluate whether the phenomenon of predator loss can be fully understood without testing competition-associated traits and, in addition, potential sex effects.
We conclude that low predation pressure, which also results in higher population densities and thus stronger within-population resource competition, seems to have strong sex-specific effects on foraging behaviour.We suggest that these differences may be caused by differences in lifehistory and mating strategies, which in turn can affect motivation for risk-taking and acquiring resources.Sex-specific effects should be considered more often when studying population differentiation, adaptation and predator loss.

Figure 1 .
Figure 1.Variation among sexes and predation regimes in (a) foraging rate, (b) consumption rate and (c) foraging efficiency.Values depict estimated marginal means and error bars denote standard error.HP = high predation, LP = low predation.

Table 1 .
Characteristics and location of the blue holes in our study.Predation regime is either high predation (HP) or low predation (LP).

Table 2 .
Results of quasipoisson GLM on foraging rate and consumption rate, and GLM on foraging efficiency.