The interplay between natural and sexual selection in the evolution of sexual size dimorphism in Sceloporus lizards (Squamata: Phrynosomatidae)

Abstract Sexual size dimorphism (SSD) evolves because body size is usually related to reproductive success through different pathways in females and males. Female body size is strongly correlated with fecundity, while in males, body size is correlated with mating success. In many lizard species, males are larger than females, whereas in others, females are the larger sex, suggesting that selection on fecundity has been stronger than sexual selection on males. As placental development or egg retention requires more space within the abdominal cavity, it has been suggested that females of viviparous lizards have larger abdomens or body size than their oviparous relatives. Thus, it would be expected that females of viviparous species attain larger sizes than their oviparous relatives, generating more biased patterns of SSD. We test these predictions using lizards of the genus Sceloporus. After controlling for phylogenetic effects, our results confirm a strong relationship between female body size and fecundity, suggesting that selection for higher fecundity has had a main role in the evolution of female body size. However, oviparous and viviparous females exhibit similar sizes and allometric relationships. Even though there is a strong effect of body size on female fecundity, once phylogenetic effects are considered, we find that the slope of male on female body size is significantly larger than one, providing evidence of greater evolutionary divergence of male body size. These results suggest that the relative impact of sexual selection acting on males has been stronger than fecundity selection acting on females within Sceloporus lizards.

In many vertebrate and invertebrate taxa, the magnitude of SSD changes systematically with mean body size, either increasing or decreasing as body size increases (Fairbairn et al., 2007;Webb & Freckleton, 2007). The former pattern is common in species where males are larger than females, while the latter occurs commonly in species in which females are the larger sex. Both patterns are explained by greater evolutionary divergence in male size, compared with female size; a pattern known as Rensch's rule (Fairbairn, 1997;Rensch, 1950). This allometric trend is usually attributed to sexual selection acting on male body size (Fairbairn et al., 2007;Stillwell et al., 2010). The converse trend, where female size varies more than male size, is less common, but seems to be the result of strong fecundity selection acting on females (Fairbairn et al., 2007;Foellmer & Moya-Laraño, 2007;Webb & Freckleton, 2007). Lizards exhibit a broad range of SSD. However, in the majority of species, males are larger than females (Cox, Butler, & John-Alder, 2007;Cox, Skelly, & John-Alder, 2003), mainly because body size often determines success in agonistic encounters, and it is correlated with dominance and territoriality (Carpenter, 1995;McMann, 1993;Molina-Borja, Padron-Fumero, & Alfonso-Martin, 1998;Perry et al., 2004). Nonetheless, in some species, females are larger than males, suggesting that fecundity selection may have favored the evolution of large female body size because it may allow females to (1) accommodate more offspring (Cox et al., 2003;Stuart-Fox, 2009;Zamudio, 1998) and (2) increase the capacity for storing energy to be invested in reproduction (Calder, 1984;Pincheira-Donoso & Tregenza, 2011).
In some viviparous species, the embryos develop in a placenta with little or no shell forming, whereas in other species, the female retains the eggs within the uterus until development is complete. In any case, because placental gestation or extended egg retention requires more space within the abdominal cavity associated with an increased gestation period (Pincheira-Donoso & Tregenza, 2011;Qualls & Shine, 1995), it has been suggested that the females of viviparous lizards possess larger body size or greater abdomens than their oviparous relatives (Braña, 1996;Scharf & Meiri, 2013;Yan-Yan et al., 2012).
The lizard genus Sceloporus serves as an excellent example of SSD in lizards. This is a widely distributed genus (from southwestern Canada to northern Panama), which can be found in several environments and along broad altitudinal ranges (0 to >4,000 m; Sites et al., 1992;Smith, 1939). There are both oviparous and viviparous species in the genus (Méndez-de la Cruz, Villagrán-Santa Cruz & Andrews, 1998). In the majority of species, males are the larger sex and exhibit a conspicuous coloration formed by belly and gular patches. However, these characteristics are also present in the females of some species within the group (Calisi & Hews, 2007;Carpenter, 1978;Fitch, 1978;Jiménez-Cruz et al., 2005;Köhler & Heimes, 2002;Ramírez-Bautista & Pavón, 2009;Ramírez-Bautista et al., 2008;Weiss, 2006). In addition, conspicuous coloration is also present on the dorsum, including the head, tail, and limbs (e.g., Sceloporus minor, S. aureolus, S. horridus : Köhler & Heimes, 2002;. The sexual coloration in males, principally the belly and gular patches, is related to species recognition, territory defense, agonistic interactions, and courtship (Carpenter, 1978;Martins, 1994;Sites et al., 1992;Wiens, Reeder, & Nieto Montes de Oca, 1999), which suggests that sexual selection has generated much of the divergence among males and females in Sceloporus lizards. However, in other species, females are larger than males (Fitch, 1978), suggesting that in these species, selection on female fecundity has been stronger than sexual selection on males.
In this study, we explore the relationship between female body size, fecundity and reproductive modes, and the potential impact of these relationships on body size divergence between females and males of Sceloporus lizards. In addition, we tested Rensch's rule in order to evaluate the relative impact of sexual selection on the evolution of SSD, and we performed an ancestral character reconstruction to infer the evolutionary trends of SSD in these lizards. We expected differences in body size between oviparous and viviparous females and that these differences affect the body size relationships between the sexes. Nonetheless, if sexual selection has been the main force driving the evolution of SSD in Sceloporus, we predict that the regression of male size on female size will have a slope steeper than 1, following the Rensch's rule.

| Data collection
Our study comprised data collected for 56 Sceloporus species, four Urosaurus species and Petrosaurus thalassinus for a total of 61 evolutionary units (Urosaurus and P. thalassinus were used as outgroup taxa). The Sceloporus species sampled included all major species groups of the genus (Leaché, 2010;Wiens et al., 2010); 41 species were oviparous and 20 were viviparous (Table 1). We performed a literature search for data on snout-vent length (SVL; a standard measure used as a proxy for lizard size; Cox et al., 2003;Losos, 1990) for both females and males and clutch/litter sizes (number of eggs or embryos) for the species studied. We collected information from the literature by executing searches on Google Scholar using the terms "snout-vent length," "clutch size," "litter size," "number of eggs/ embryos," "sexual size dimorphism," or "reproductive cycle" for a list of species of the  (2013), Fitch (1978Fitch ( , 1985, Vitt (1977) (Continues) Scholar was used as the search engine instead of other engines because it cataloged full-text versions of published papers. Moreover, terms that were included in our search like "clutch size," "litter size," and "snout-vent length" were not the principal focus of the papers, and the phrases were usually referred to only briefly. Thus, we were less likely to locate the pertinent information using literature databases that contain only keywords, titles, and abstracts (see Dornhaus, Powell, & Bengston, 2012). We excluded data in which the number of vitellogenics follicles were reported as part of clutch size, because the follicular atresia may occur in any stage of the ovogenesis, including previtellogenic and vitellogenic follicles, and thus does not represent an accurate estimation of clutch/litter size (Méndez-de la Cruz et al., 2013). For species with data on more than one clutch per reproductive season, we used the average of all clutches reported in the literature.

| Phylogenetic reconstruction
We inferred the phylogenetic relationships between the 56 stud-  Wiens et al. (2010) recognized these taxa as putative species, which was also supported by more recent evidence (Grummer et al., 2015). Our inclusion of these taxa as distinct evolutionary lineages was not an endorsement of their recognition as different species, but we did not want to ignore important previous taxonomic work on these groups (see Pérez-Ramos & Saldaña de La Riva, 2008;Wiens & Reeder, 1997;Wiens et al., 2013).
We used MUSCLE algorithm (Edgar, 2004) to align each gene data set using the default parameters in the software MEGA (version 7; Kumar, Stecher, & Tamura, 2016). We then used the software MESQUITE (Maddison & Maddison, 2015) to combine the sequences of each gene, and to make the final concatenated matrix for all genes (see below). We provide the GenBank accession numbers of the sequences used in Appendix S1. Our concatenated alignment consisted of genetic information from 61 terminals (56 Sceloporus species, five outgroups taxa) and 11,113 characters. We estimated the best partition scheme and nucleotide substitution models for the data using the greedy algorithm of PARTITIONFINDER (version 1. Our analysis only considered the phylogeny that resulted from a concatenated matrix of both mitochondrial and nuclear loci, utilizing a total evidence approach for Sceloporus species and outgroup taxa.
Although this approach may be controversial because nuclear and mitochondrial genes may have incongruent histories due to incomplete lineage sorting and exhibit different substitutions rates (see Maddison, 1997

| Comparative analyses
We converted the molecular branch lengths from the Bayesian analysis to units of time using a penalized likelihood method (Sanderson, 2002). For branch length conversion, we used the R (version 3.1.3; R Core Team 2015) package "ape" (Paradis, Claude, & Strimmer, 2004) and performed all the comparative analysis on the resulting ultrametric phylogeny. For more details, see Appendix S1.

| Reproductive modes, female body size, fecundity, and SDI
We used the phylogenetic generalized least squares (PGLS) model to test for an association between fecundity, body size, and reproductive mode. The PGLS approach incorporates phylogenetic information into linear models to account for the statistical nonindependence of residuals using a variance-covariance matrix (see Martins & Hansen, 1997) specified by the phylogeny. For all models, the maximum likelihood value of the weighting parameter λ was estimated simultaneously with the models (Gonzalez-Voyer & Kolm, 2010; Revell, 2010).
The λ parameter indicates whether trait evolution is independent of the phylogeny (λ = 0) or evolving according to Brownian motion (λ = 1). Intermediate values of λ suggest a process in which the effect of the given phylogeny is weaker than expected by Brownian motion evolution (Pagel, 1999). The models were fitted as implemented in the R package "caper" (Orme et al., 2012). The first model included fecundity (dependent variable), log 10 SVL female (independent variable), and reproductive mode (categorical independent variable) as well as the interaction between SVL and reproductive mode. In order to evaluate the impact of fecundity on SDI, we first saved the residuals of the previous model and then constructed a model with SDI as the dependent variable, reproductive mode as a categorical independent variable, and the fecundity residuals as a covariate. We used the residuals to eliminate potential confounding effects associated with female body size.

| Rensch's rule and ancestral reconstruction of SDI
Rensch's rule predicts that the slope of a regression of male body size on female body size will be steeper than 1. To test this prediction in the studied species, we used the phylogenetic independent contrasts method (PIC method; Felsenstein, 1985), as implemented by the R package "caper" (Orme et al., 2012) to control for the phylogenetic nonindependence of species (Harvey & Pagel, 1991). We examined the studentized residuals for outliers > |±3|, but found none in our data set. Also, in order to verify whether the standardized contrasts are independent from their estimated nodal values (see Felsenstein, 1985), we plotted the standardized contrasts against their estimated nodal values using the "plot" function provided by "caper". Ultimately, we tested the allometric relationship between independent contrasts of log 10 SVL male (dependent variable) and log 10 SVL female (independent variable) by fitting major axis regression using the R package "smatr" (Warton et al., 2012). Major axis regression offers an accurate approach to test the null hypothesis of isometry (h 0 : β = 1), because both variables were measured on the same scale and residual variance is minimized in both x and y dimensions, rather than the y dimension only (Cox et al., 2007;Pincheira-Donoso & Tregenza, 2011;Warton et al., 2006). Given that the mean value of contrasts is expected to be zero (Sanabria-Urbán et al., 2015), we forced the major axis regression through the origin. We used the Wald statistic (r w ) and confidence intervals (95%) of the slope to test the null hypothesis (see Warton et al., 2006). In addition, in order to explore the evolutionary trends in body size and SDI, we performed an ancestral character reconstruction following Revell (2013). This method estimates the maximum likelihood value for internal nodes and then interpolates the states along the branches of the tree (see Revell, 2013Revell, , 2014 for details).
For the reconstruction and visualization of ancestral state reconstruction of SDI (see Figure 3), we used the R package "phytools" (Revell, 2012).

| Reproductive modes, body size, and fecundity
After controlling for phylogenetic nonindependence among of the species studied, the results of the PGLS analysis were highly significant (r 2 = 0.3, F 3,57 = 8.025, p = .0001). We found a strong and significantly positive relationship between body size and fecundity

| DISCUSSION
Once we control for phylogenetic effects, our results confirm a strong relationship between female body size and fecundity, suggesting that in Sceloporus lizards selection on fecundity has had a main role on the evolution of female body size. However, regardless of the reproductive mode (oviparous or viviparous), the size of females of Sceloporus is similar and has evolved in a similar fashion. We must point out that the similar response in the relationship of body size with increase in the clutch/litter size between both reproductive modes does not imply that the overall reproductive output (i.e., reproductive fitness of the female's life) is similar. The potential impact of fecundity selection on the different reproductive modes may be underestimated (Niewiarowski et al., 2004;Pincheira-Donoso & Hunt, 2015;Shine, 2005 The differences in the gestation period between reproductive modes do not have any impact on the evolution of SSD, but according to the Renchs' rule, the slope of the regression of males on females is significantly steeper, providing evidence of greater evolutionary divergence in male size than in female size. Fitch (1978) noted that the high variation of SSD in Sceloporus lizards, and the implications of sexual and natural selection in order to explain the differences in body size between females and males. For lizard species in which body size often determines male mating success, males are typically larger than females (Cox et al., 2007). Body size often determines success in agonistic encounters, and it is correlated with dominance and territoriality (Carpenter, 1995;Martins, 1994;McMann, 1993;Molina-Borja et al., 1998;Perry et al., 2004). However, in other species, females are larger than males, suggesting that fecundity selection may have favored the evolution of larger-than-average female body size (Cox et al., 2003;Zamudio, 1998). Furthermore, as Sceloporus lizards follow Rensch's rule, it can be argued that this allometric trend is the result of sexual selection favoring large male body size and that the relative impact of sexual selection on males has been stronger than fecundity selection on female body size (Fairbairn, 1997;Fairbairn et al., 2007;Pincheira-Donoso & Tregenza, 2011).
The reconstruction of the evolution of SSD in Sceloporus lizards suggests that the ancestor and most of the extant species show a pattern of male-biased SSD. This could indicate that directional sexual selection acting on males has been greater than the selection acting on female fecundity. Territoriality and aggressive behaviors are common in Sceloporus: These are mainly associated with defense of mates in males (Martins, 1994), and resources (e.g., food, water, perches) in both sexes (Cooper & Wilson, 2007;Martins, 1994;Vinegar, 1975c;Woodley & Moore, 1999). In general, larger individuals have an advantage when defending territories in agonistic encounters (Martins, 1994;Swierk, 2014). However, female-biased SSD has evolved independently at least six times ( Figure 3). Perhaps in these taxa, selection on fecundity has been stronger than sexual selection. Nonetheless, it is possible that in these species, sexual selection has also favored small male body size (see Cox et al., 2007;Olsson et al., 2002) and within species. However, the information available for female preference and agonistic interactions between males are, in the majority of species, severely scarce or absent (see Martins, 1994;Swierk, 2014 Sites et al., 1992). The similarity between our results and those reported in Liolaemus suggests that fecundity selection may have driven the divergence in female body size but that the diversifying effects of sexual selection may often exceed fecundity selection on females in both genera.
The genus Sceloporus includes more than 90 species and has been proposed as a group with an accelerated diversification rate (Bell, Smith, & Chiszar, 2003;Leaché, 2010;Wiens et al., 2010). Sceloporus F I G U R E 3 Maximum likelihood ancestral reconstruction of SDI for 56 species of Sceloporus and five outgroup taxa performed in R package "phytools" (Revell, 2012). For the analysis, we used the ultrametric phylogeny and the values of SDI estimated for each species. The values in the color ramp represent the range of SDI registered for the study species. Negative values indicate male-biased SSD (blue to paleyellow) and positive values female-biased SSD (palepurple to red). Open and filled circles indicate, respectively, oviparous and viviparous lizard species We thank four anonymous reviewers who also made invaluable suggestions to the manuscript.