Introduction

In viviparous organisms, the transfer of nutrients between parent and offspring shows different gradations along a matrotrophy continuum. Lecithotrophy is found at one of the extremes of the matrotrophy continuum, in which the nutrients available for the development of the embryos are entirely provided by the yolk generated within the ovarian follicles. At the other extreme is matrotrophy, where nutrients are transferred exclusively by maternal tissues during gestation (Marsh-Matthews 2011; Kalinka 2015). Some viviparous species are strictly lecithotrophic or exclusively matrotrophic. However, between these two extremes, numerous lecithotrophic species show varying degrees of matrotrophy (Skalkos et al. 2023). Placentotrophy is a particular type of matrotrophy, in which mothers actively provide nutrients to developing embryos via complex placental structures (apposition or fusion of maternal and fetal tissues for physiological exchange) (Mossman 1937; Blackburn 2015). Viviparous species that exhibit strict lecithotrophy have very simple placentas, whereas species with complex placental structures are capable of extensive matrotrophy (Kwan et al. 2015; Olivera-Tlahuel et al. 2019).

In lecithotrophic females, the entire reproductive investment occurs prior to fertilization in the form of yolk as the only source of embryo nutrition (Marsh-Matthews 2011), which creates the need to mate with good-quality males to secure that their investment will culminate in good-quality offspring. However, there is also the possibility that females may mistakenly select a poor-quality male, based on dishonest signals of individual quality, jeopardizing their investment (Arnqvist and Rowe 2005). Given the high pre-copulatory investment, females of lecithotrophic species must be careful in their choice of mate (pre-copulatory sexual selection processes). This mate choice is generally based on phenotypic signals that females are attracted to and that could reflect the quality of the male (e.g. size of ornaments, body color, courtship quality; Andersson 1994). In contrast, females of matrotrophic species reduce the risks of poor pre-copulatory mate choice caused by false signals, because (i) placentas give females the physiological possibility of asymmetric nutrient assignment among the offspring of males of different qualities, and (ii) placentotrophy implies the production of large numbers of small-sized eggs that require less pre-fertilization investment (i.e. minimal or no yolk reserves) (Zeh and Zeh 2000; Banet and Reznick 2008). Thus, the evolution of advanced degrees of placentotrophy results in a shift from pre- to post-copulatory sexual selection (Haig 1990; Crespi and Semeniuk 2004; Pollux et al. 2014; Kalinka 2015).

The shift from pre-copulatory to post-copulatory sexual selection gives rise to an interesting conflict between females and males (i.e. when sexes have incompatible fitness optima in relation to the reproductive process; Arnqvist and Rowe 2005) of promiscuous species (Eberhard 1996; Arnqvist 1998; Kalinka 2015; Furness et al. 2015, 2019). In particular, the evolution of complex placentas allows females to select males “cryptically” (e.g. lower allocation of nutrients to embryos fertilized by low-quality males or cryptic female choice to selectively use the best sperm; Eberhard 2009). In addition to cryptic female choice, elaborate placentas can provide the physical environment for sperm competition to occur (Evans and Pilastro 2011; Fitzpatrick 2020). Under these cryptic selection conditions that provide greater control to females, males should respond evolutionarily with counteradaptations that allow them to increase their chances of success in fertilization (e.g. secretion of substances that block the female reproductive tract to prevent copulation with other males, development of morphological structures that facilitate forced copulation, increased investment in the gonads to produce a greater number and/or more competitive sperm; Gage 1994; Eberhard 1996, 2015).

In this context, Pollux et al. (2014) demonstrated in fishes of the family Poeciliidae (all but one species give birth to live individuals; most of these through lecithotrophy, but some through different degrees of placental matrotrophy; Turner 1940) that the evolution of elaborate placentas in females is negatively related to male investment in sexual selection traits (e.g. size, coloration, ornaments, reproductive behavior). Thus, males of many placental species have evolved coercive strategies based on traits that help them force copulation with females and that increase their chances of success in the sperm competition processes that take place inside the female reproductive tract. Examples of such traits include longer gonopodia that allow males to coercively inseminate females without the need to be in close proximity to them (a gonopodium is the modified anal fin that males use to transfer their sperm directly into the female gonopore; Rosen and Gordon 1953), harassment behavior, small body size that increases their chances of going unnoticed by females during sneak copulation attempts, and potentially harmful structures at the tip of the gonopodium, which presumably are intended to hurt females, reduce their receptivity, and hence prevent their mating with other males (Rosen and Gordon 1953; Pollux et al. 2014). Some studies have suggested that gonopodial hooks and spines may serve to decrease the insemination success of rival males by injuring the female genitalia (Constantz 1984; Greven 2005; Langerhans 2011; Cummings 2018).

This pattern has been confirmed in comparative studies carried out at the interspecific level, which demonstrate that indeed, in this family of viviparous fishes, males of many lecithotrophic species exhibit elaborate courtship, large body sizes, smaller gonopodia, and colorful ornaments. In contrast, males of many matrotrophic species generally tend to be smaller than females, show no ornamentation or courtship, increased sneaking behaviors, and force females to copulate by means of longer gonopodia (Pollux et al. 2014; Furness et al. 2019, 2021; Reznick et al. 2021). Still, there is wide variation among both lecithotrophic and placentotrophic species in male ornamentation (body coloration and size of dorsal or caudal fins) as well as in the relative importance of sneaking and courtship behaviors for obtaining copulations, in such a way that males from lecithotrophic species may also exhibit sneaking behaviors and coercively copulate with females (e.g. see Sato et al. 2011). However, there is solid evidence that male traits associated with intersexual conflict and forced copulations are more prevalent in placental species (Pollux et al. 2014; Furness et al. 2019, 2021; Reznick et al. 2021).

Despite the strong support in the literature, the relationship, if any, between placental complexity and male reproductive traits at the intraspecific level, however, is unknown. Therefore, it is necessary to test, at the microevolutionary scale, whether the evolution of higher degrees of placentotrophy across populations of a given species promotes greater sexual conflict, which in turn promotes the evolution of enhanced strategies to ensure paternity, similar to the pattern that has been detected at the interspecific scale. The genus Poeciliopsis is characterized by variations between and within species in the degrees of female placentotrophy, where a positive relationship between the amount of maternal supply to developing embryos after fertilization and the degree of placental complexity has been confirmed (Turner 1940; Thibault and Schultz 1978; Reznick et al. 2002; Kwan et al. 2015; Olivera-Tlahuel et al. 2019; Molina-Moctezuma et al. 2020; Saleh-Subaie et al. 2021). Therefore, it is reasonable to hypothesize that, within Poeciliopsis species, males have also experienced phenotypic changes associated with greater post-copulatory sexual selection and sperm competition in populations in which females have evolved higher levels of placentotrophy. Our main objective was to test this hypothesis using data from several populations of three species of the genus Poeciliopsis. We predicted that males from populations where females exhibit greater degrees of placentotrophy will make a greater investment in two traits that could increase their probabilities of reproductive success. First, gonad size, because larger testes imply greater sperm production (Grier and Uribe 2009; Montgomerie and Fitzpatrick 2009). Second, gonopodium length, because a longer gonopodium facilitates forced copulations (Evans et al. 2011).

Materials and methods

Study species and quantification of reproductive traits

We used preserved males from 11 populations of Poeciliopsis gracilis, 11 populations of P. infans, and three populations of P. prolifica (Table 1). All these populations correspond to streams and rivers isolated from each other and, therefore, the possibility of gene flow among them is rather low (Supplementary material, Fig. S1). All preserved specimens were obtained from the collection of fishes of the Laboratorio de Ecología Evolutiva y Demografía Animal, Facultad de Ciencias, Universidad Nacional Autónoma de México. Males were processed as follows: we first photographed each fish with a graph paper in the background to measure both gonopodium length (0.01 cm) and standard length of the male using the software ImageJ® v. 1.52. Before photographing each male, we verified that the gonopodium was fully extended and not damaged (Fig. 1). We then dissected each male to extract the testes. We dried the testes as well as the somatic body of the fish (i.e. digestive tract removed) at 55 °C for 24 h in a desiccating oven. We weighed the dry body using an Ohaus scale (0.0001 g) and the dry testes using a Mettler Toledo XP6 micro-scale (0.000001 g). Given that larger males may have longer gonopodia and heavier testes and that male size (standard length) differed statistically among populations of both P. gracilis and P. infans (Supplementary material, Fig.S2-S4), we used size-corrected measures of both testes mass and gonopodium length. To calculate such size-corrected measures, we implemented linear regressions between the dry mass of the testes and male dry mass as well as between gonopodium length and standard length of the male (separately for each species but pooling data from all populations). We then calculated the residuals from these regressions, which we considered size-corrected measures of testes mass and gonopodium length, respectively.

Table 1 Geographic location, matrotrophy index (MI) obtained from Saleh-Subaie et al. (2021), and sample size (n = number of males) for the populations that we studied of three fish species of the genus Poeciliopsis
Fig. 1
figure 1

Photograph of a Poeciliopsis gracilis male showing the extended gonopodium. The circles and continuous line depict how we measured gonopodium length. The arrows and dashed line depict how we measured standard length of the male

To test our hypothesis, we used data on the degree of placentotrophy previously obtained by Saleh-Subaie et al. (2021) for the same populations that we examined in this study. The females used by Saleh-Subaie et al. (2021) and the males that we examined here were collected simultaneously from each population, between 2012 and 2013. The degree of placentotrophy is commonly quantified by means of the matrotrophy index (MI), which is calculated as the dry mass of the offspring at birth divided by the dry mass of the egg at fertilization (Skalkos et al. 2023). Values of the MI between 0.5 and 0.7 correspond to lecithotrophic species (Reznick et al. 2002; Furness et al. 2019, 2021). Such values indicate that embryos of lecithotrophic species lose between 30% and 50% of their dry mass due to the metabolic costs derived from development. Values closer to 1 correspond to species that exhibit incipient matrotrophy because females are capable of compensating for the mass loss caused by metabolic costs by providing small amounts of nutrients during development. Values greater than 1 indicate extensive matrotrophy because embryo mass increases substantially during development as a result of the active transfer of nutrients by the mother throughout development (Reznick et al. 2002; Marsh-Matthews 2011). The degree of placentotrophy (i.e. the MI) of each species/population, sample sizes (number of males), and geographic locations of our study populations are shown in Table 1.

We chose these three species because they represent three distinct modes of maternal provisioning: P. gracilis is considered as lecithotrophic, P. infans exhibits incipient matrotrophy, and P. prolifica is matrotrophic (Saleh-Subaie et al. 2021). Therefore, we were able to examine differences in the potential for microevolution of male reproductive traits among these distinct modes of maternal provisioning. The median degree of matrotrophy of each species as well as the distribution of MI values across populations within species can be seen in Supplementary material, Fig.S5. In addition, there is wide variation among populations within these three species in the matrotrophy index. In P. gracilis, the MI in some populations is more than twice as large (1.14) as that of the population with the smallest MI value (0.48). Similarly, in P. infans the largest MI value (2.19) is close to three times that of the population with the smallest MI value (0.79). In P. prolifica, there is an almost 4-fold difference in MI values between populations 1 and 3 (Table 1). This implies that in some populations, females have little control over the amount of nutrients that embryos receive during development, whereas in other populations females are able to transfer significant amounts of nutrients during development, which provides them with some degree of physiological control and, hence, potential mechanisms of cryptic choice. Therefore, we expected that even in species that have been considered as lecithotrophic or incipient matrotrophic, an increase in the degree of female control over nutrient provisioning to individual embryos would promote sexual conflict and, in consequence, males must evolve larger testes and longer gonopodia.

Statistical analysis

To test the hypothesis that testes mass and gonopodium length evolve in response to the degree of placentotrophy of females at the intraspecific level, we implemented linear and mixed models in the R statistical software v. 4.2.1 (R Core Team 2022), using the package ‘lme4’ (Bates et al. 2015). In all cases, we used our size-corrected individual values of testes mass and gonopodium length as response variables. We conducted our analyses separately for each species and for each male trait. We analyzed both traits separately because in all three species, our size-corrected individual values of testes mass and gonopodium length were not significantly correlated (P. gracilis: r = -0.06, P = 0.56; P. infans: r = -0.02, P = 0.81; P. prolifica: r = -0.02, P = 0.93). For both P. gracilis and P. infans, we used the degree of placentotrophy (i.e. the MI per population) as fixed effect and the distinct populations as random effect (population-specific intercepts). We fitted four competing models separately for size-corrected testes mass and gonopodium length: (i) a model including only the fixed effect of the MI values, (ii) a model including this fixed effect as well as the random effect of the distinct populations, (iii) a model that only included the random effect (i.e. differences among populations), and (iv) an intercept-only model in which neither the MI values nor the distinct populations affected these male reproductive traits.

Given that we only have data for three populations of P. prolifica (Table 1), we simply compared size-corrected individual values of testes mass and gonopodium length among the three populations of this species. According to our hypothesis, we expected to observe statistical differences among populations in male reproductive traits, with the population with highest MI (4.86; Table 1) having the largest average values of testes mass and gonopodium length. Similarly, we expected the smallest average values of these traits in the population with lowest MI (1.26; Table 1). Therefore, we fitted and compared only two models for this species (separately for testes mass and gonopodium length): (i) a model including differences among populations (as fixed effect in this case), and (ii) an intercept-only model.

We used the Akaike information criterion adjusted for small sample sizes (AICc; Akaike 1973; Burnham and Anderson 2002) to assess model fit. The model with lowest AICc score was identified as the model that best explained our observed data. However, in those cases where two or more models had similarly low AICc scores (i.e. models differing by less than 2 AICc units from the top model [ΔAICc < 2]) we selected the simpler model. We also calculated Akaike weights (w), which are estimates of the relative support for each competing model in the data (Burnham and Anderson 2002). Based on these Akaike weights, for both P. gracilis and P. infans, we calculated model-averaged estimates of the effect of the degree of placentotrophy on testes mass and gonopodium length (i.e. the model-averaged regression coefficient for the fixed effect of the MI; as instructed by Burnham and Anderson 2002). Model-averaged estimates take into account model uncertainty and are more robust than those derived from any single model alone (Johnson and Omland 2004). An evident effect of the MI values would be indicated by a model-averaged regression coefficient with a 95% confidence interval that does not include zero.

Finally, we verified the assumptions of normality and homoscedasticity of residuals by visually inspecting quantile-quantile plots as well as scatter plots of model residuals against the values of the main predictor variable. We verified assumptions only from the top model for each species and male trait.

Results

In P. gracilis, the model that provided the best fit to testes mass included differences among populations with no effect of the matrotrophy index (Table 2). The second best-fitting model also had strong support in the data (ΔAICc = 1.89) and included the effect of population as well as a small positive effect of the MI values on testes mass (Fig. 2a). We selected the former model because it is simpler than the latter. The model-averaged regression coefficient that represents the effect of the MI was positive (0.08) but its 95% confidence interval included zero (-0.20–0.36). Thus, evidence of an influence of the degree of placentotrophy on testes mass of P. gracilis was weak (Fig. 2a).

Table 2 Model selection results for two male reproductive traits (testes mass and gonopodium length) of two fish species of the genus Poeciliopsis. For these analyses we used size-corrected measures of both traits. Competing models represent that testes mass and gonopodium length may vary as a linear function of the average matrotrophy index of each population (the fixed effect, denoted as ‘matrotrophy’) as well as among distinct populations (the random effect, denoted as ‘population’). For each model we show the difference in values of the sample-size-adjusted Akaike information criterion with respect to the top model (ΔAICc) and the Akaike weight (w). The best-fitting model is indicated by ΔAICc = 0 and models with ΔAICc < 2 also have strong support. For each trait, models are listed according to their fit to the data, from best to worst. We also show the effect of matrotrophy (regression coefficient with its 95% confidence interval within parentheses) for those models that included this fixed effect
Fig. 2
figure 2

Statistical relationships between the average degree of female placentotrophy per population (quantified by the matrotrophy index) and size-corrected measures of testes mass and gonopodium length of males of Poeciliopsis gracilis (a, b) and P. infans (c, d). The fitted (dashed) lines represent the estimated effects of the matrotrophy index on each trait. Individual data points are colored by population

Fig. 3
figure 3

Comparisons of size-corrected measures of testes mass (a) and gonopodium length (b) of males of Poeciliopsis prolifica among three populations that differ in the average degree of female placentotrophy (quantified by the matrotrophy index). The asterisk indicates the population that was significantly different from the other two populations

Greater model uncertainty was observed in gonopodium length of P. gracilis (Table 2). Three of the four competing models had strong support in the data (ΔAICc < 2). The top model included both MI and population, but the confidence interval of the regression coefficient that represents the effect of the MI included zero (Table 2). The second best-fitting model (ΔAICc = 0.02) only included differences among populations. The third model with strong support (ΔAICc = 1.46) also included the effect of the MI with no differences among populations. In this case, the estimated effect of the MI was positive and its confidence interval did not include zero (Table 2; Fig. 2b). The model-averaged regression coefficient for the effect of the MI values on gonopodium length was also positive (0.59), but its 95% confidence interval included zero (-0.11–1.30). Thus, even though there is a trend towards longer gonopodia in populations of P. gracilis with higher MI values (Fig. 2b), the statistical support for this trend was not strong enough to be conclusive.

Regarding testes mass of P. infans, a single model had strong support in the data (Table 2). This top model only included differences among populations, with no effect of the MI values. Consistent with this result, the model-averaged regression coefficient that represents the effect of the MI on testes mass was notably small (0.01) and its 95% confidence interval included zero (-0.05–0.08). Thus, we found no evidence of an influence of the degree of placentotrophy on testes mass of P. infans (Fig. 2c).

High model uncertainty also occurred in gonopodium length of P. infans (Table 2). Three models had strong support (ΔAICc < 2). The models that ranked first and second included a negative effect of the MI. In the top model, which only included the effect of matrotrophy with no differences among populations, the confidence interval for the effect of the MI did not include zero. However, in the second best-fitting model, which included both MI and population, the confidence interval for the effect of the MI did include zero (Table 2). The third model with strong support only included differences among populations, with no effect of the MI values. The model-averaged regression coefficient that represents the effect of the MI on gonopodium length was also negative (-0.22), with a confidence interval that barely included zero (-0.46–0.02). Thus, we detected a negative trend towards shorter gonopodia in populations of P. infans with higher MI values (Fig. 2d). However, also in this case, the statistical support for this negative trend was not strong enough to be conclusive.

Even though we observed the heaviest testes of males of P. prolifica in the population where females exhibit the greatest degrees of placentotrophy (Fig. 3a), the intercept-only model provided the best fit to the data and the model including differences among populations had weaker support (ΔAICc = 3.42; Table 3). Therefore, testes mass did not vary substantially among populations of this species.

Table 3 Model selection results for two male reproductive traits (testes mass and gonopodium length) of Poeciliopsis prolifica. For these analyses we used size-corrected measures of both traits. We compare the fit of two competing models: one including differences among populations (denoted as ‘population’) and the other in which the trait does not vary among populations (intercept-only model). For each model we show the difference in values of the sample-size-adjusted Akaike information criterion with respect to the top model (ΔAICc) and the Akaike weight (w). The best-fitting model is indicated by ΔAICc = 0 and models with ΔAICc < 2 also have strong support. For each trait, models are listed according to their fit to the data, from best to worst

Regarding gonopodium length of P. prolifica, we found evidence of statistical differences among populations, as indicated by the top model for this trait (Table 3). In this case, the intercept-only model had considerably weaker support (ΔAICc = 5.79). However, the differences among populations in gonopodium length were not consistent with the observed differences among populations in the MI. Males from the population with the intermediate value of the MI had substantially shorter gonopodia than males from the other two populations (Fig. 3b).

All the models that provided the best fit to both traits of all three species met the assumptions reasonably well. In no case did we observe substantial deviations from normality or a noticeable lack of homoscedasticity (Supplementary material, Fig.S6).

Discussion

Here we tested for the first time at the intraspecific level, the potential relationship between the degree of female placentotrophy and the evolution of male reproductive traits, which has already been demonstrated by Pollux et al. (2014) at the interspecific level. We detected in males of Poeciliopsis gracilis a trend towards longer gonopodia in populations where females exhibit higher degrees of placentotrophy, which is consistent with the hypothesis that increased female control over the amount of resources that are actively transferred to embryos during development should intensify an intersexual conflict and, consequently, males should evolve traits that increase their probability of mating success, such as larger testes and longer gonopodia (Pollux et al. 2014; Furness et al. 2019; Reznick et al. 2021). However, the statistical support for this positive association was relatively weak and, therefore, this finding does not represent conclusive evidence. We also found larger testes in the population of P. prolifica with the highest degree of placentotrophy, but we lacked statistical power (we only have three populations) to provide more conclusive evidence for this species. Interestingly, we found shorter gonopodia in populations of P. infans with high levels of placentotrophy, which is contrary to what we originally expected. However, this relationship did not have sufficient statistical support either. In summary, despite the observed trends in the data, our findings fail to provide convincing support for an effect of placentotrophy on intraspecific variation in male reproductive traits.

The evolution of elaborate placentas allowed females to switch from visually assessing a potential mate based on his ornamentation and courtship to internally “selecting” ejaculates of the best quality by means of physiological and/or morphological mechanisms (Eberhard 1996; Kalinka 2015). Therefore, in addition to their primary function, which is to nourish developing embryos, placentas can serve as a medium for post-copulatory sexual selection (Furness et al. 2019). In this context, we originally hypothesized that more advanced degrees of placentotrophy in some populations could promote an increase in testes mass as a counteradaptive response. The size of the testes is determined by sperm production rate, sperm storage capacity, and the size of the ejaculate that is necessary for successful fertilization (Grier and Uribe 2009; Montgomerie and Fitzpatrick 2009). Sperm competition theory predicts an increased investment in spermatogenic components under scenarios of increased sperm competition (as documented in a wide range of taxa; Gage 1994; Hosken 1997, 2001). Under these circumstances, sperm number per ejaculate will be the primary determinant of individual reproductive success (Parker 1970). Our data from P. gracilis and P. infans are not consistent with this hypothesis. Despite wide variation among populations of both species in the degree of female placentotrophy, testes mass was completely unrelated to such interpopulation variation. In P. prolifica, the heaviest testes occurred in males from the population with the highest degree of placentotrophy, but this is only a pattern with weak statistical support. Therefore, data from other populations of this species are needed to provide more conclusive evidence of this potential relationship.

The selective effect of the degree of female placentotrophy might occur on other characteristics of the male gonads that do not affect their size. In other words, post-copulatory sexual selection might exert a selective pressure on sperm quality rather than on sperm quantity. Size and motility of individual spermatozoa have a direct influence on fertilization success (Smith and Ryan 2010; Ramm and Schärer 2014). Therefore, males from populations where females are more placentotrophic may produce sperm of better quality (larger and faster) without the need to increase their quantity. Further examination of the morphological and physiological characteristics of the testes and spermatozoa (e.g. amount of spermatogenic tissue, size and motility of spermatozoa, and sperm production rate; Ramm and Schärer 2014) are needed to test this hypothesis.

At the interspecific level, there is now solid evidence in poeciliid fishes that the evolution of elaborate placentas has induced evolutionary changes in male morphology and behavior (Reznick et al. 2021). In particular, males from placentotrophic species have evolved a large and structurally complex gonopodium that allows them to copulate furtively with females without their consent (Evans et al. 2011; Pollux et al. 2014). We expected a similar microevolutionary response at the intraspecific level. In P. gracilis, we observed a trend towards longer gonopodia in populations where females are more placentotrophic, which is consistent with this hypothesis, but this quantitative evidence is not strong enough to be conclusive. Perhaps in our study species and populations gonopodium length is not the result of intersexual conflict but is instead under pre-copulatory sexual selection through female choice as has been documented in previous studies (Rosen and Tucker 1961; Langerhans et al. 2005; Kahn et al. 2010). This means that females from these and other Poeciliopsis species, in which female choice is apparently absent given the lack of male ornaments and courtship behavior (Reznick et al. 2021), might in fact prefer to copulate with males that have relatively long gonopodia. In this way, their offspring would inherit the fitness benefits of such long intromittent organs (Eberhard 1996).

Another non-mutually exclusive explanation, which arises from the work conducted by Evans et al. (2011) and Gasparini and Pilastro (2011) in guppies (Poecilia reticulata), is that males with longer gonopodia are more likely to reach and fertilize females through forced copulations in an environment where male-male competition is strong. In populations with male-biased sex ratios, competition for females is intense, and a longer gonopodium likely allows males to achieve a greater number of copulations. This advantage is independent of the degree of female placentotrophy. Therefore, we suggest that the social environment can exert strong directional selection on gonopodium length, as has been documented for the intromittent organ of insects (Frazee and Masly 2015; Dougherty and Shuker 2016). In this sense, the observed interpopulation variation in gonopodium length might be due to interpopulation differences in the intensity of male-male competition and not to interpopulation variation in the degree of female placentotrophy.

Natural selection could be another driver of the observed intraspecific variation in gonopodium size. Relatively long gonopodia of Gambusia affinis and G. hubbsi have been associated with reduced burst-swimming performance, suggesting that a reduced gonopodium size is favored by natural selection in environments that demand efficient locomotion such as sites with high predation or fast water currents (Langerhans et al. 2005; Evans and Meisner 2009). However, evidence suggests that female preferences for larger genitalia appear to counteract natural selection for shorter gonopodia, creating a balance between natural and sexual selection (Langerhans et al. 2005). The relative strength of both types of selection likely differs among populations, which can explain the observed intraspecific variation in male genitalia. Nevertheless, Kelly et al. (2000) found that in environments with intense predation, selection could favor longer gonopodia because it is more efficient in transferring sperm to the female tract during forced copulations, which is a more common breeding strategy in high-predation environments (Magurran and Seghers 1994; Evans et al. 2002).

Within populations, the degree of female placentotrophy may vary in the short term (e.g. among years) in response to environmental conditions (Trexler and DeAngelis 2003, 2010). For example, high and constant food availability have a positive influence on the amount of nutrients that females of P. infans actively transfer to developing embryos (Molina-Moctezuma et al. 2020). Hence, the degree of placentotrophy in this species may increase in years when food is abundant. Potentially transient levels of placentotrophy imply that males do not have enough time to evolve a counteradaptation, such as greater investment in gonopodium length and testes mass, in response to increased placentotrophy. Such short-term changes in the degree of placentotrophy within populations may explain why the association between male reproductive traits and degree of female placentotrophy can be detected at the interspecific level (Pollux et al. 2014; Reznick et al. 2021), but not at the intraspecific level.

Alternatively, a conflict between sexes that can promote an evolutionary change in male reproductive investment may only occur at higher levels of female placentotrophy (i.e. in placental lineages where the matrotrophy index is substantially greater than 1). Species that are predominantly lecithotrophic or with incipient matrotrophy exhibit relatively simple placental tissues that may not provide the appropriate physical environment for sperm selection or sperm competition to occur, not even in populations where females have increased their degree of placentotrophy. In contrast, species with extensive matrotrophy exhibit elaborate placentas where these post-copulatory processes of sexual selection are more likely to take place (Kwan et al. 2015; Olivera-Tlahuel et al. 2019). In consequence, only in the latter species a male-female conflict may be strong enough to cause increases in testes mass and gonopodium length in response to site-specific increases in the degree of placentotrophy. This helps to explain why we observed larger testes in the population of P. prolifica with the highest value of the matrotrophy index (this species exhibits substantial matrotrophy), whereas this pattern was not evident in the other two species (P. gracilis is lecithotrophic and P. infans exhibits incipient matrotrophy; Saleh-Subaie et al. 2021).

According to Pollux et al. (2014) and Reznick et al. (2021), most species of the genus Poeciliopsis evolved predominantly male morphology and behavior oriented to forced copulations as the main mating tactic. However, there is evidence of different male phenotypes associated with distinct mating strategies within species (e.g. in P. occidentalis some males exhibit courtship behavior whereas others exhibit gonopodial thrusts; Constantz 1975). Therefore, when males exhibit different breeding tactics within a population, the observed degree of investment in phenotypic traits that provide males with some advantage during the processes of cryptic female choice (e.g. increased sperm production) should be proportional to the relative frequency of the tactics used for insemination in that population (Bisazza 1993). In other words, if a substantial number of males from a given population rely on courtship behavior to obtain copulations, then the averages of gonopodium length and testes mass for this population would be lower than expected for a species that is supposed to experience strong post-copulatory sexual selection. Hence, it is reasonable to expect interpopulation differences in the relative frequency of distinct mating strategies (courtship versus forced copulations) that in turn could promote interpopulation variation in average testes mass and gonopodium length. Currently, there is a lack of information about intraspecific variation in mating tactics and male phenotypes of the Poeciliopsis species that we studied.

Finally, we must notice that there is also wide variation among males within populations in both testes mass and gonopodium length. This interindividual variation likely arises from the combination of intrinsic (e.g. genetic) and extrinsic (e.g. differences among males in food intake) factors that could affect the expression of these reproductive traits. Such large variation among males also made it difficult to detect a selective effect of the degree of female placentotrophy on testes mass and gonopodium length. Even if our study does not provide strong support for the hypothesis of a positive association, at the interpopulation level, between placentotrophy and male investment in reproductive traits, the observed trends in the gonopodia of P. gracilis and in the testes mass of P. prolifica highlight the need for additional data from more males, populations, and species. Taken together, our findings indicate that there are multiple evolutionary forces involved in the microevolution of male reproductive traits of Poeciliopsis species. Such forces presumably result from pre- and post-copulatory sexual selection as well as from natural selection. Certainly, the strength and direction of post-copulatory sexual selection do not act similarly in all the populations that we studied and the magnitude of other pre-copulatory sexual selective forces also likely differs among populations.