Introduction

Viviparous fishes of the family Poeciliidae have been the subject of numerous studies on a broad range of topics. From research on these organisms, we have gained substantial understanding of population genetics (Barson et al. 2009), animal cognition (Cummings 2018), the origin of unisexual species (Vrijenhoek 1994), social behavior (Krause et al. 2011), pre- and post-copulatory sexual selection (Evans and Pilastro 2011; Rios-Cardenas and Morris 2011), biological invasions in freshwater ecosystems (Santi et al. 2020), the effects of environmental pollutants (Gomes-Silva et al. 2020), and cancer etiology (Schartl 2014; Lu et al. 2018). A particular field of knowledge that has grown remarkably from studies on poeciliid fishes is the evolution of life histories (Johnson and Bagley 2011). We now know how distinct ecological factors, such as food availability, predation intensity, and population density can cause drastic intraspecific divergence in life-history traits such as age and size at maturity, number and size of offspring, and total reproductive investment (Johnson 2001; Moore et al. 2016; Gorini-Pacheco et al. 2018; Roth-Monzón et al. 2021). Research on life histories of poeciliids has also provided insight into the causes and consequences of the evolution of complex reproductive strategies and associated morphological structures such as placentas, placentotrophy, multiple paternity, sperm retention, and superfetation (Olivera-Tlahuel et al. 2017; Furness et al. 2019, 2021; Dekker et al. 2022; García-Cabello et al. 2022).

Despite the large body of knowledge on life history evolution that has arisen from studies on this group of viviparous fishes, some aspects still deserve attention. Here, we focus on two of these aspects. First, recent comparative studies that analyzed the evolution of reproductive modes are based on extensive datasets that contain species-specific values of several life-history traits (Pollux et al. 2014; Furness et al. 2019, 2021; Reznick et al. 2021). Even though these datasets compiled information on numerous species, they are not yet complete. Currently, the family Poeciliidae includes almost 300 species (Zúñiga-Vega et al. 2022) and, according to these datasets, life-history information is available for less than 180 species (approximately 60% of the species in the family). Furthermore, some life-history traits are underrepresented in these datasets whereas others, such as the matrotrophy index that quantifies the amount of post-fertilization provisioning to developing embryos, have been quantified for a large number of poeciliid species (Furness et al. 2021). Specifically, number of offspring produced per brood, size of each individual offspring, and total reproductive allotment (proportion of the female mass that is devoted to offspring production) are three key life-history traits that have yet to be quantified for the majority of poeciliids (Olivera-Tlahuel et al. 2015). Additional data on these life-history traits will allow future comparative studies on their evolution, similar to those that have been conducted on other reproductive traits such as placentotrophy and superfetation (Pollux et al. 2014; Furness et al. 2019, 2021; García-Cabello et al. 2022).

Second, poeciliid fishes represent an excellent model system to understand the mechanisms underlying the life-history trade-off between number and size of offspring. Studies on some poeciliids have provided empirical evidence of this life-history trade-off (Schrader and Travis 2012; Frías-Alvarez et al. 2014; O’Dea et al. 2015). However, analyses of the factors that cause this trade-off or, even more interesting, of the circumstances under which this trade-off does not occur are still needed. Pregnant females of viviparous animals have finite space within their reproductive tract that they can devote to offspring production (Sun et al. 2012). Therefore, an overall increase in reproductive effort should entail a reduction in either the number or the individual size of their offspring (i.e., they produce either numerous small or few large offspring; Warne and Charnov 2008). In this case, the trade-off may arise from a morphological constraint. However, in some animals with indeterminate growth, such as poeciliid fishes, the volume of the reproductive tract increases hyperallometrically as they age, which implies less space limitations in older and larger females (Somarakis et al. 2004; Gomes and Monteiro 2008; Kharat et al. 2008; Barneche et al. 2018). Thus, relatively old and large females should be able to allocate energy and resources into both number and size of young, whereas younger and smaller reproductive females likely experience a morphological constraint that imposes a strong offspring size-number trade-off. To date, less than a handful of studies in fishes have examined if this trade-off depends on the size or age of females (Quinn et al. 2011; O’Dea et al. 2015; Lasne et al. 2018).

In this study, we contribute to the knowledge of life histories of viviparous fishes by addressing the following objectives. (1) To provide quantitative information on number of offspring produced per brood, size of each individual offspring, and total reproductive allotment for some members of the family Poeciliidae for which data on these life-history traits are still missing. (2) To test the hypothesis that the trade-off between number and size of offspring is stronger in small reproductive females because of the morphological constraints associated with a small body size. Conversely, reaching larger sizes should relieve such space restrictions, making this trade-off undetectable in larger females.

Materials and methods

Study species. We examined pregnant females of the following six species of the family Poeciliidae: Gambusia sexradiata, Poeciliopsis latidens, Poeciliopsis viriosa, Priapella intermedia, Pseudoxiphophorus jonesii, and Xiphophorus hellerii. With the exception of G. sexradiata, these species are underrepresented in the literature of poeciliid life histories (Riesch et al. 2010). Data on number of offspring (brood size) is available for G. sexradiata, Poeciliopsis viriosa, and X. hellerii (Reznick and Miles 1989; Riesch et al. 2010; Pires et al. 2011), but lacking for the remaining three species (Po. latidens, Priapella intermedia, and Pseudoxiphophorus jonesii). Regarding offspring size at birth, there are quantitative reports for G. sexradiata, Poeciliopsis latidens, Po. viriosa, and Pseudoxiphophorus jonesii (Riesch et al. 2010; Pires et al. 2011; Olivera-Tlahuel et al. 2015), whereas the value of this life-history trait is unknown for Priapella intermedia and X. hellerii. In turn, total reproductive allotment (RA) has been quantified for G. sexradiata, Poeciliopsis viriosa, and X. hellerii (Reznick and Miles 1989; Riesch et al. 2010; Pires et al. 2011), and no information of this trait is available for the remaining three species (Po. latidens, Priapella intermedia, and Pseudoxiphophorus jonesii).

All the preserved specimens that we examined came from the National Collection of Fishes (Instituto de Biología, Universidad Nacional Autónoma de México). The particular regions (Mexican states) where females of our six focal species were collected were (collection years and sample sizes within parentheses): Gambusia sexradiata in southern Veracruz (2013, n = 42), Poeciliopsis latidens in central Sinaloa (2014 and 2018, n = 158), Poeciliopsis viriosa in southern Nayarit (2013, n = 35), Priapella intermedia in eastern Oaxaca (2012 and 2013, n = 78), Pseudoxiphophorus jonesii in southern Veracruz (2012 and 2013, n = 42), and Xiphophorus hellerii in eastern Oaxaca (2012, n = 82).

Quantification of life-history traits. We dissected preserved gestating females of all six species and extracted both reproductive and digestive tracts. Before dissection, we used a digital caliper to measure the standard length (SL) of each female. We counted the number of developing embryos per female and identified their stage of development according to the classification proposed by Haynes (1995). We considered embryos from stages 4 (recently fertilized egg) to 11 (mature embryo). We excluded stages 1–3 because they correspond to unfertilized ova. In the case of Poeciliopsis latidens and Po. viriosa, which are the two species that exhibit superfetation (ability of females to bear simultaneously two or more groups of embryos at different developmental stages; Turner 1937), we counted the number of simultaneous broods and recorded the number of embryos contained in each brood (i.e., the number of embryos that shared the same developmental stage). All embryos and the female body were desiccated at 55 °C during 24–48 h in a drying oven. Then, the dry mass of each embryo was estimated by dividing the dry mass of an entire brood by the number of embryos contained in that brood. Reproductive allotment (RA) was calculated as the proportion of the total dry mass of the female that consisted of developing embryos (across all broods in the case of superfetating species):

$$ {\text{RA}} = \frac{{{\text{dry}}\;{\text{mass}}\;{\text{of}}\;{\text{all}}\;{\text{embryos}}}}{{{\text{female}}\;{\text{somatic}}\;{\text{dry}}\;{\text{mass}} + {\text{dry}}\;{\text{mass}}\;{\text{of}}\;{\text{all}}\;{\text{embryos}}}}. $$

Data analyses. We searched for a size-dependent trade-off between number and size (dry mass) of embryos by means of a multi-model inference approach. We implemented a set of linear models in which we used individual embryo dry mass as response variable (transformed to natural logarithm). For the two species that exhibit superfetation (Po. latidens and Po. viriosa), we used only the estimated embryo dry mass from a single brood, which we chose randomly from each superfetating female. In this way, we did not include more than one data point per female in our models, thereby avoiding non-independence in the data (Zúñiga-Vega et al. 2007; Frías-Alvarez and Zúñiga-Vega 2016). However, to examine if our results for these two species are robust regardless of the choice of a particular brood, we also implemented all the models that we describe below using the estimated embryo mass from all the broods that were present in each superfetating female (i.e., including more than one data point per female; as per Zúñiga-Vega et al. 2007) as well as choosing a different brood.

Competing models differed in the predictor variable(s) that could affect embryo dry mass. The main predictor variable that we considered was brood size (i.e., number of embryos per brood) and, given our hypothesis of a trade-off, we expected a negative effect on embryo mass. A size-dependent trade-off (stronger in small females and weaker or non-existent in large females) was modeled by means of an interaction between female SL and brood size affecting embryo mass. From this model, we expected a strong negative relationship between number of embryos and individual embryo mass in small females and a weaker relationship (a less steep negative slope or a slope close to zero) in larger females.

Given that developmental stage may also affect individual embryo mass (embryo dry mass either decreases or increases as development progresses depending on whether the species is lecithotrophic or matrotrophic, respectively; Marsh-Matthews 2011), we built models in which stage was included either as the only predictor variable or in combination with other predictors (both as an additive and interactive effect). In addition to the aforementioned interactive effect of female SL and brood size, we also built a model in which female SL was the only predictor (large females may produce large embryos regardless of brood size and developmental stage), models that included female SL in combination with stage (additive and interactive effects), and a model with an additive effect of female SL and brood size. This latter model represents a similar offspring size-number trade-off for both small and large females, with larger females producing larger embryos overall. Finally, we also fitted an intercept-only model in which embryo mass was not affected by any predictor variable. In total, we fitted 11 competing models separately for each species. The list of models can be seen in Table 1.

Table 1 Differences in AICc scores (ΔAICc) between each competing model and the best-fitting model (indicated by ΔAICc = 0) for six fish species of the family Poeciliidae. In all models, the response variable was embryo dry mass (transformed to natural logarithm). The addition and multiplication symbols represent additive and interactive effects of predictors, respectively
Full size table

To select the best model for each species we used the Akaike information criterion adjusted for small sample sizes (AICc; Burnham and Anderson 2002). The lowest AICc score was used to identify the model that provided the best fit to the data. However, in those cases where two or more models had strong support (i.e., models differing by less than two AICc units from the best-fitting model [ΔAICc < 2]), we selected the simpler model because the additional predictors did not substantially improve the model fit compared to the model with fewer parameters.

For each species, we report average values of brood size and RA (± one standard error). The only exception was RA of Po. viriosa because we could not obtain female dry masses and, consequently, we were unable to quantify this life-history trait for this species. To estimate offspring size at birth, we predicted embryo dry mass at stage 11 (mature embryo; Haynes 1995) from the particular model that included developmental stage as predictor with lowest AICc score (this particular model differed among species). If we detected an effect of female size on embryo mass, as indicated by female SL being included in the best-fitting model, we estimated offspring size at birth separately for small and large females. For this purpose, we classified females that measured less than the median SL as small females and those that measured more than the median SL as large females. Predictions of embryo mass at birth were obtained from models on the logarithmic scale, but we back-transformed them to the original scale (mg).

In addition, we compared among species the three life-history traits that we studied (embryo dry mass, brood size, and RA) accounting for the potential effect of female size on these traits (larger species may have larger embryos, broods, or RA; Promislow et al. 1992). For this purpose, we built five competing linear models that we fitted separately for each trait. These models included the following effects: (1) differences among species with no effect of female SL, (2) an effect of female SL without differences among species, (3) the additive effect of species and female SL, (4) the interaction between species and female SL, and (5) an intercept-only model in which the trait did not vary among species and was not influenced by female size. As in our previous models, embryo dry mass was transformed to natural logarithm. Given that RA is a proportion, we used an arcsine transformation (calculated as the arcsine of the square root of the proportion) to meet the assumptions of linear models (Zar 2010). In the case of brood size, which consists of counts of embryos, we implemented generalized linear models with log link function and Poisson distribution (Zuur et al. 2007). We also used AICc to select the best model for each life-history trait. Given that average female size differs among species (Table 2) and that life-history traits can be influenced by female size, we generated comparable estimates of embryo mass, brood size, and RA by calculating the expected value of these traits for a female of a common size for all species. For this purpose, we used the average female SL across all six species (32 mm). Therefore, based on the top model for each trait, we calculated the predicted values of embryo mass, brood size, and RA for a female of 32 mm SL of each species. These predicted size-adjusted values were obtained on the transformed scale and then back-transformed to their original scales (mg, number of embryos, and proportion of total female mass, respectively). All analyses were implemented in R version 4.1.3 (R Core Team 2022).

Table 2 Average values of life-history traits, including female size, for six fish species of the family Poeciliidae. In four species we detected an effect of female size on embryo mass and, hence, we report the predicted offspring mass at birth separately for small (S) and large (L) females. We were unable to quantify reproductive allotment for Poeciliopsis viriosa. Size-adjusted estimates account for differences among species in average female size and were calculated for a female of 32 mm standard length of each species
Full size table

Results

Size-dependent trade-off between number and size of offspring. Only in one species, Poeciliopsis latidens, we detected a stronger offspring size-number trade-off in small females. For this species, the top model included the interaction between female SL and brood size affecting individual embryo dry mass (Table 1). This model also included an additive effect of developmental stage. All other models had poor support (ΔAICc > 7 in all cases; Table 1). The statistical relationship between brood size and embryo mass was negative for both small and large females, but the negative slope was steeper for small females (β1 = -0.020) compared to the slope for large females (β1 = -0.013) (Fig. 1a). The relationship between developmental stage and embryo mass was negative, indicating that embryos of Po. latidens lose mass as development progresses (Fig. 1b).

Fig. 1
figure 1

Statistical effects of a brood size (number of offspring per brood) and b developmental stage on individual embryo dry mass (transformed to natural logarithm) of Poeciliopsis latidens. In (a), the dashed line and white circles represent small females, whereas the continuous line and black circles represent large females. In both (a) and (b), the lines represent the predicted relationships according to the model that we selected (see Table 1)

Full size image

We obtained the same results for Po. latidens when using data from all the broods present in each superfetating female as well as using data from a distinct brood. In both cases, the top model indicated an interaction between female SL and brood size affecting embryo mass as well as an additive effect of developmental stage [Electronic Supplementary Material (ESM) Table S1]. Also in both cases, the negative relationship between brood size and embryo mass was stronger for small females than for large females (ESM Fig. S1a, b) and embryos decrease in mass throughout development (ESM Fig. S1c, d).

Additional sources of variation for embryo size. In the other five species we did not detect a trade-off between number and size of offspring. Instead, we found striking differences among species in the sources of variation for individual embryo mass. In Gambusia sexradiata three models had strong support (ΔAICc < 2), all of which included the effect of developmental stage (Table 1). The top model only included this predictor variable, whereas the second and third models included female size and brood size, respectively, in addition to stage. Thus, adding these two other predictors did not substantially improve the model fit compared to the model that only included the effect of stage. In this species, embryo mass increases slightly as development progresses (Fig. 2).

Fig. 2
figure 2

Statistical effect of developmental stage on individual embryo dry mass (transformed to natural logarithm) of Gambusia sexradiata. The dashed line represents the predicted relationship according to the model that we selected (see Table 1)

Full size image

In Po. viriosa, two models had strong support and both included developmental stage and female size affecting embryo mass (Table 1). The top model indicated an additive effect of these predictors, whereas the second indicated an interactive effect. We selected the former model because it is simpler (i.e., adding the interaction term did not substantially improve the model fit). In this species, embryo mass decreases as development progresses and larger females produce larger embryos overall (Fig. 3). We obtained qualitatively similar results for Po. viriosa when analyzing data from all the broods present in each superfetating female as well as from a different brood per female. In both cases, the model that included an additive effect of stage and female SL provided the best fit to the data (ESM Table S1), larger females produced larger embryos, and embryo mass decreased throughout development (ESM Fig. S2a, b).

Fig. 3
figure 3

Statistical effect of developmental stage on individual embryo dry mass (transformed to natural logarithm) of Poeciliopsis viriosa. The dashed line and white circles represent small females, whereas the continuous line and black circles represent large females. Both lines represent the predicted relationships according to the model that we selected (see Table 1)

Full size image

In Priapella intermedia, a single model had strong support (Table 1). This top model indicated an additive effect of female size and brood size on individual embryo mass. Contrary to our expectation, the statistical effect of brood size on embryo mass was positive, indicating that females that produce numerous embryos are also capable of producing relatively large embryos (Fig. 4). Also in Pr. intermedia, larger females produce larger embryos overall. In Pseudoxiphophorus jonesii, four models had strong support, all of which included female size as predictor variable (Table 1). We selected the model that only included female size (ΔAICc = 1.38), because adding other predictors did not improve the model fit with respect to this simpler model. In Ps. jonesii, larger females produce noticeably larger embryos (Fig. 5). Finally, in Xiphophorus hellerii no predictor variable had an evident effect on embryo mass as indicated by the intercept-only model having strong support (ΔAICc = 0.63) (Table 1). Even though three other models also had strong support (ΔAICc < 2), none of these provided a better fit compared to the simplest intercept-only model. For illustrative purposes only, we show in Fig. 6 that embryo mass of X. hellerii does not change throughout development.

Fig. 4
figure 4

Statistical effect of brood size (number of offspring per brood) on individual embryo dry mass (transformed to natural logarithm) of Priapella intermedia. The dashed line and white circles represent small females, whereas the continuous line and black circles represent large females. Both lines represent the predicted relationships according to the model that we selected (see Table 1)

Full size image
Fig. 5
figure 5

Statistical effect of female size on individual embryo dry mass (transformed to natural logarithm) of Pseudoxiphophorus jonesii. The dashed line represents the predicted relationship according to the model that we selected (see Table 1)

Full size image
Fig. 6
figure 6

Individual embryo dry mass (transformed to natural logarithm) of Xiphophorus hellerii plotted against developmental stage. The dashed line represents the predicted effect of developmental stage according to the model that only included this predictor. Notice that embryo mass remains relatively constant throughout development

Full size image

Interspecific variation in life-history traits. We observed wide variation among species in the three life-history traits that we examined. The smallest offspring were observed in the two species of the genus Poeciliopsis (between 0.69 and 0.85 mg; Table 2). The largest offspring occurred in large females of Priapella intermedia (1.87 mg) and Pseudoxiphophorus jonesii (2.17 mg). In five of the six study species, mean brood sizes were relatively small (between 7.1 and 11.3 embryos), whereas in the sixth species (X. hellerii) females produced remarkably large broods (37.9 embryos on average; Table 2). Finally, three species (G. sexradiata, Ps. jonesii, and X. hellerii) had relatively similar RA (between 0.101 and 0.129; Table 2). The smallest RA was observed in Priapella intermedia (0.057) and the largest in Poeciliopsis latidens (0.172).

Differences in female body size among species (Table 2) only partially explain the observed interspecific variation in life-history traits. The model including the interaction between species and female SL unambiguously provided the best fit to all three traits (Table 3), indicating that the influence of female size on embryo mass, brood size, and RA differed among species (Fig. 7). Despite being medium-sized species, Priapella intermedia and Pseudoxiphophorus jonesii produce the largest embryos (Fig. 7a). In the case of number of offspring, the largest broods are indeed produced by the largest species, X. hellerii (Fig. 7b). In contrast, the greatest RA occurred in the smallest species, Poeciliopsis latidens (Fig. 7c). Size-adjusted estimates of these traits also confirmed that, for a female of the same size (32 mm SL), Priapella intermedia and Pseudoxiphophorus jonesii produce the largest embryos, X. hellerii produces the largest broods, and Poeciliopsis latidens has the greatest RA (Table 2). In addition, size-adjusted estimates also revealed that, despite their small sizes, females of both G. sexradiata and Po. latidens produce relatively large broods (Table 2).

Table 3 Differences in AICc scores (ΔAICc) between each competing model and the best-fitting model (indicated by ΔAICc = 0 and highlighted in bold) for three life-history traits of six fish species of the family Poeciliidae. Competing models represent that traits may vary among species and/or may be influenced by female size. The addition and multiplication symbols represent additive and interactive effects of predictors, respectively
Full size table
Fig. 7
figure 7

Statistical effects of female size on a individual embryo dry mass (transformed to natural logarithm), b brood size (also transformed to natural logarithm), and c reproductive allotment (transformed to arcsine square-root values) for six fish species of the family Poeciliidae. Fitted lines were derived from models that included the interaction between species and female size affecting each trait. The length of the lines represents the range of female sizes of each species

Full size image

Discussion

We have provided a series of quantitative information on brood size, offspring mass at birth, and reproductive allotment in Poeciliopsis latidens, Priapella intermedia, Pseudoxiphophorus jonesii, and Xiphophorus hellerii of the family Poeciliidae for which data on these life-history traits were still missing. In addition, we have provided evidence of a size-dependent trade-off between number and size of offspring in one of the six species that we studied. Females of Poeciliopsis latidens that produce more embryos do so at the cost of a reduction in the size of each embryo. This trade-off occurred in both small and large females of this species, but was stronger in the former. This pattern suggests that in this species, morphological constraints occur all throughout their life cycles, but are more severe in the early phases of adulthood. As females grow, such space limitations are relaxed slightly and the offspring size-number trade-off becomes less strong. However, we must notice here that we hypothesized that the trade-off should not exist or at least should be undetectable in large females, which is a pattern that did not occur in Po. latidens. Instead, the trade-off still persists as females grow, although to a lesser extent.

The body shape of species of the genus Poeciliopsis is relatively thin and fusiform (Zúñiga-Vega et al. 2007; Frías-Alvarez and Zúñiga-Vega 2016; Fleuren et al. 2018), compared to members of other poeciliid genera, which in general have deeper bodies (Gomes and Monteiro 2008; Zúñiga-Vega et al. 2011; Johnson et al. 2014). This particular morphology could explain why the offspring size-number trade-off still occurs in larger females, despite their larger body size and concomitant larger reproductive cavity. In other words, a fusiform body shape likely imposes space restrictions on all body sizes. However, this explanation is not consistent with the absence of this trade-off in the congeneric Po. viriosa, which also has a relatively thin morphology. An alternative explanation is a difference between species in food availability. The occurrence of trade-offs may depend on the external conditions rather than on (or in addition to) the intrinsic characteristics of the species (van Noordwijk and de Jong 1986). In particular, environments where food is abundant may allow females to acquire enough energetic reserves to produce numerous embryos without a detrimental effect on their individual size (Zera and Harshman 2001; Blanckenhorn and Heyland 2004). In contrast, low food availability entails energetic restrictions for females, which in turn impedes the allocation of nutrients to both number and size of embryos (Stahlschmidt and Adamo 2015; Mishra and Kumar 2019). Therefore, our collections of Po. viriosa may have come from a population where food is abundant and, in contrast, the females of Po. latidens that we examined may have come from a population where food is scarce. In addition, differences in female body size between Po. latidens and Po. viriosa can also explain the occurrence of this trade-off only in Po. latidens because females of Po. viriosa are substantially larger (37.9 mm SL on average) than females of Po. latidens (24.0 mm SL on average; Table 2). A larger body size provides a larger reproductive cavity and, hence, less space restrictions for offspring production (Saleh-Subaie et al. 2021).

Interestingly, we did not detect the offspring size-number trade-off in the other four species, suggesting a lack of morphological restrictions and/or abundant food sources. Three of these species (Priapella intermedia, Pseudoxiphophorus jonesii, and X. hellerii) have relatively large body sizes (see Table 2) and, therefore, larger reproductive tracts and looser constraints on offspring size and brood size. In fact, in Priapella intermedia, we detected a positive relationship between number and size of embryos, which is opposite to what we expected. This means that some females of this species are able to produce numerous large embryos. This ability may come from the relatively deep body of members of the genus Priapella (Riesch et al. 2012), which provides females with ample space within their reproductive cavities. In poeciliid fishes, morphology can indeed exert strong pressures on reproductive investment (Ghalambor et al. 2004; Quicazan-Rubio et al. 2019). Opposite to the apparent lack of morphological restrictions that we observed in Pr. intermedia, other species of this family have narrow bodies that impose limits on the number and size of young that females can produce. For instance, Alfaro cultratus has a narrow body and keeled ventral surface that apparently have impeded the evolution of increased reproductive allotment even in the presence of strong selective pressures that promote increased fecundity (e.g., in high-predation environments; Golden et al. 2021). Another interesting example of morphological constraints on reproduction is superfetation, which is a reproductive strategy that could have arisen in poeciliid fishes as an adaptive response to environments that favor streamlined morphologies (e.g., fast-flowing rivers; Zúñiga-Vega et al. 2010). Apparently, superfetation allows females to have high fecundities without substantial increases in the volume of their abdomens (Zúñiga-Vega et al. 2007, 2017; Fleuren et al. 2019).

Our findings also revealed that in four species (Poeciliopsis latidens, Po. viriosa, Priapella intermedia, and Pseudoxiphophorus jonesii) larger females produce larger embryos. A large size at birth can be advantageous under different ecological conditions. In environments with intense intraspecific competition, larger offspring of the least killifish (Heterandria formosa) and guppies (Poecilia reticulata) are better competitors and have higher fitness than smaller ones (Bashey 2008; Schrader and Travis 2012). In water bodies that contain toxic compounds such as hydrogen sulfide, a large size at birth confers a low surface/volume ratio and, consequently, less surface area per volume of body tissue is exposed to the toxins (Riesch et al. 2010). Large offspring likely have relatively low metabolic rates and, thus, low oxygen consumption, which could be beneficial in aquatic environments with low dissolved oxygen (Riesch et al. 2010). Moreover, intense cannibalism on juveniles also selects for larger offspring sizes, as documented in Poeciliopsis monacha (Thibault 1974; Weeks and Gaggiotti 1993). In numerous animal species, natural selection favors a large female size, which has been mostly explained as a result of the greater fecundity (i.e., larger broods) associated with a larger body (Cox et al. 2003; Lim et al. 2014; Pincheira-Donoso and Hunt 2017). Our findings indicate that, in some poeciliid species, a larger female size also conveys the selective advantage of larger newborns.

Larger females may also be capable of producing more embryos per brood (Zúñiga-Vega et al. 2011; Barneche et al. 2018) and our data confirms this positive relationship between female size and fecundity in five of our study species (with the exception of Pseudoxiphophorus jonesii; see Fig. 7b). At the interspecific level, we also observed this same pattern: the largest broods were produced by the largest species, X. hellerii. Directional selection for larger female size due to the fitness benefit of producing numerous offspring has been demonstrated in several taxa (Lefranc and Bundgaard 2000; Fox and Czesak 2006; Nali et al. 2014). Intriguingly, in all but one species reproductive allotment did not increase with female size (the only exception was Poeciliopsis latidens; see Fig. 7c), which indicates that even though females can increase the number and/or size of their offspring as they grow, the relative amount of energy that they invest in reproduction (relative with respect to their body size) remains constant throughout their lives. This unexpected finding suggests that a disproportionately greater reproductive investment by larger females may not result in a significant increase in the fitness of their young (i.e., may not increase the number of offspring surviving to sexual maturity). Instead, a disproportionately greater reproductive allotment may impose physiological costs. Interestingly, the only species in which reproductive allotment increased with female size was the same species in which we detected the trade-off between number and size of offspring, Po. latidens. Our data for this species indicate that large females of Po. latidens make a disproportionately greater reproductive investment through larger broods at the cost of a reduction in the size of each individual offspring.

Our study revealed three patterns of covariation among female size, number of offspring per brood, and embryo mass in poeciliid fishes. These distinct life-history strategies are likely shaped by ecological factors (both biotic and abiotic) and biological constraints. First, in three species (Po. latidens, Po. viriosa, and Priapella intermedia) both brood size and embryo mass increased with female size. Presumably, these species inhabit environments with high mortality rates that select for greater fecundity (Reznick et al. 1996), intense intraspecific competition that promotes large offspring size (Bashey 2008), and abundant food sources that allow females to simultaneously invest in both number and size of offspring (Taborsky 2006). Second, in two species (Gambusia sexradiata and X. hellerii) fecundity increases with female size whereas offspring mass remains invariant regardless of maternal size. Selection pressure for an optimal offspring size at birth may be strong in the habitats of these two species (Jørgensen et al. 2011). Another possibility is that smaller offspring are not viable and, hence, females are not able to further reduce offspring size in order to increase brood size (this is another plausible explanation for why the offspring size-number trade-off does not occur in these species). Alternatively, offspring size could be genetically constrained. Third, in one species (Pseudoxiphophorus jonesii) larger females produce larger embryos but brood size does not increase with female size. In this case, females may experience selection against a distended abdomen as a result of fast water currents or high predation risk, which are ecological conditions that impose strong pressures on swimming performance (Zúñiga-Vega et al. 2007; Wesner et al. 2011; Ingley et al. 2014). High fecundity entails a substantial increase in abdomen volume, which has a negative impact on swimming performance (Ghalambor et al. 2004; Quicazan-Rubio et al. 2019). Thus, a small brood size allows females of Ps. jonesii to maintain a less distended abdomen during pregnancy and, therefore, an efficient swimming performance. All these tentative explanations represent hypotheses that require empirical research.

In general, previous reports of the mode of maternal provisioning to developing embryos are not consistent with our findings. Specifically, Poeciliopsis viriosa has been considered as having incipient matrotrophy (Pollux et al. 2014; Furness et al. 2019, 2021), whereas the negative relationship between developmental stage and embryo mass that we observed in this species indicates that females rely entirely on yolk for embryo nutrition, without post-fertilization provisioning (i.e., strict lecithotrophy). In lecithotrophic species, embryo dry mass decreases throughout development due to metabolic costs (Reznick et al. 2002; Marsh-Matthews 2011). Furthermore, G. sexradiata, Priapella intermedia, Pseudoxiphophorus jonesii, and X. hellerii were previously considered as lecithotrophic (Furness et al. 2019, 2021), whereas in contrast our results suggest that these species are moderately matrotrophic (in the case of G. sexradiata) or at least capable of incipient matrotrophy (in the cases of Priapella intermedia, Pseudoxiphophorus jonesii, and X. hellerii). In G. sexradiata evidence of moderate matrotrophy comes from embryos increasing moderately in mass throughout development, which indicates that females actively transfer nutrients to embryos during gestation (Reznick et al. 2002; Marsh-Matthews 2011). In Priapella intermedia, Pseudoxiphophorus jonesii, and X. hellerii evidence of incipient matrotrophy comes from a lack of relationship between stage and embryo mass (depicted in Fig. 6 for X. hellerii), which indicates that embryos maintain a constant dry mass throughout development. In species with incipient matrotrophy, embryos feed mainly on yolk but obtain small amounts of nutrients directly from females during gestation, thereby offsetting the metabolic costs that otherwise would cause a reduction in embryo dry mass (Marsh-Matthews et al. 2001; Reznick et al. 2002; Pollux et al. 2009). The only exception was Poeciliopsis latidens because both our data and previous reports indicate that this species is lecithotrophic (Pollux et al. 2014; Furness et al. 2019, 2021).

Differences between studies can be explained by intraspecific variation in the mode of maternal provisioning to developing embryos, which in turn could arise from environmental effects on the relative amounts of pre- and post-fertilization sources of embryonic nourishment. There is now solid evidence that increased amounts of post-fertilization provisioning (i.e., higher degrees of matrotrophy) can be advantageous under particular conditions (e.g., where food is abundant or predation risk is high; Riesch et al. 2013; Gorini-Pacheco et al. 2018; Hagmayer et al. 2020) and that females of some poeciliid species can facultatively increase or decrease the nutrients that they actively transfer to embryos during gestation depending on the external conditions (specifically depending on food availability; Pires et al. 2007; Molina-Moctezuma et al. 2020).