Morphological convergence in a Mexican garter snake associated with the ingestion of a novel prey

Abstract Morphological convergence is expected when organisms which differ in phenotype experience similar functional demands, which lead to similar associations between resource utilization and performance. To consume prey with hard exoskeletons, snakes require either specialized head morphology, or to deal with them when they are vulnerable, for example, during molting. Such attributes may in turn reduce the efficiency with which they prey on soft‐bodied, slippery animals such as fish. Snakes which consume a range of prey may present intermediate morphology, such as that of Thamnophiine (Natricinae), which may be classified morphometrically across the soft–hard prey dietary boundary. In this study, we compared the dentition and head structure of populations of Thamnophis melanogaster that have entered the arthropod–crustacean (crayfish)‐eating niche and those that have not, and tested for convergence between the former and two distantly related crayfish specialists of the genus Regina (R. septemvittata and R. grahamii). As a control, we included the congener T. eques. Multivariate analysis of jaw length, head length, head width, and number of maxillary teeth yielded three significant canonical variables that together explained 98.8% of the variance in the size‐corrected morphological data. The first canonical variable significantly discriminated between the three species. The results show that head dimensions and number of teeth of the two Regina species are more similar to those of crayfish‐eating T. melanogaster than to non‐crayfish‐eating snakes or of T. eques. It is unclear how particular head proportions or teeth number facilitates capture of crayfish, but our results and the rarity of soft crayfish ingestion by T. melanogaster may reflect the novelty of this niche expansion, and are consistent with the hypothesis that some populations of T. melanogaster have converged in their head morphology with the two soft crayfish‐eating Regina species, although we cannot rule out the possibility of a morphological pre‐adaptation to ingest crayfish.


| INTRODUCTION
Some of the most compelling test cases for adaptive evolution involve morphological convergence (Schluter, 2000), which is predicted to evolve when organisms experience similar functional demands on their phenotype (Schluter, 2000;Vincent, Brandley, Herrel, & Alfaro, 2009).
Snakes are very good subjects for studying feeding morphology because their head is directly involved in feeding (Cundall & Rossman, 1984;Dwyer & Kaiser, 1997). They consume their prey whole; therefore, some morphological attributes of their typical prey should be associated with trophic morphology (Hampton, 2011(Hampton, , 2013Mori & Vincent, 2008;Vincent et al., 2009). To consume prey with external body features such as the hard exoskeletons of arthropods, snakes require specialized morphology (e.g., piercing teeth) or behavior such as targeting arthropods when they are vulnerable, such as when molting, as the hard exoskeleton is both slippery to grasp and hard to pierce.
The tribe Thamnophiine (family Natricinae) comprising North American semi-aquatic snakes, includes Thamnophis melanogaster (Mexican black-bellied garter snake), an aquatic dietary specialist that is sympatric with a freshwater crustacean, the crayfish Cambarellus montezumae, but eats crayfish only in 3.0% of the area of sympatry (Manjarrez, Macías Garcia, & Drummond, 2013; Figure 1). The 35% of prey consumed by T. melanogaster were crayfish eaten only when recently molted, so with the exoskeleton as yet unhardened (Manjarrez et al., 2013). Extensive dietary studies of Thamnophis species have failed to reveal crayfish ingestion, except in a rare record for T. proximus (0.8% of individuals with crayfish in stomachs; Hampton & Ford, 2007). Therefore, the rarity of crayfish ingestion in the focal cluster of populations of T. melanogaster (Alfaro & Arnold, 2001;de Queiroz, Lawson, & Lemos-Espinal, 2002) suggests crayfish eating represents a niche invasion that has not yet expanded to more populations.
We explored possible morphological differences in dentition and head structure within T. melanogaster by comparing individuals from crayfish-eating versus non-crayfish-eating populations and included in this comparison both soft crayfish-eating Regina species and the aquatic generalist Thamnophis eques (Mexican garter snake).
Thamnophis eques is sympatric with T. melanogaster over most of its range (Rossman, Ford, & Siegel, 1996) and represents a control for geographic determinants of head morphology. We predicted a morphological convergence between crayfish-eating T. melanogaster and Regina species that specialize in eating soft crayfish.
Consuming soft crayfish may not require specialized teeth, but because of their vulnerability during the molt, crayfish seek refuge and must be sought in burrows and crevices, which would impose different demands on the head morphology of a snake that often preys in the open and guides its strikes visually (Drummond, 1983;Macías Garcia & Drummond, 1995). Dwyer and Kaiser (1997)

Evidence of crayfish ingestion by Thamnophis melanogaster
Yes No morphology of garter snakes (Thamnophis) is associated with the ingestion of soft prey (Britt, Clark, & Bennett, 2009;Savitzky, 1983).
Thamnophis melanogaster is a snake that specializes in underwater foraging and feeds mainly on soft-bodied aquatic prey such as fish (ca. 50%), tadpoles and leeches. It has a narrow head (Rossman et al., 1996), similar to that of other species that feed on aquatic soft prey (Dwyer & Kaiser, 1997;Hibbitts & Fitzgerald, 2005), and curved, pointed, and backward-directed maxillary teeth, suitable for piercing through soft skin (Rossman et al., 1996). This species is located within the mono- Drummond & Macías Garcia, 1989;Rossman et al., 1996).

| MATERIALS AND METHODS
We measured 80 crayfish-eating T. melanogaster individuals from 10 populations (Manjarrez et al., 2013) and 88 non-crayfish-eating individuals from 29 populations adjacent to the crayfish-eating populations (Table 1). All snakes were captured in the wild. In addition, we examined 19 specimens of R. grahamii and 81 of R. septemvittata at the Florida Museum of Natural History, University of Florida (Table 1).
We also included 42 T. eques (Table 1). Snakes were mostly adults or of a size close to that of the adults (Table 1; Appendix 1).
Four variables were used to characterize head structure: (1) jaw length (distance from the posterior edge of the posterior-most supralabial scale to the anterior tip of the rostrum; King, 2002), (2) head length (distance from the snout tip to the posterior-most portion of the parietal bone), (3) head width (widest part measured while applying pressure on the posterior portion of the head to spread the quadrates and mandibles laterally; Miller & Mushinsky, 1990), and (4) number of maxillary teeth.
Although often used in similar studies (King, 2002;Miller & Mushinsky, 1990), we did not use gape index in our analyses because this is a composite of several of the above measures. An exploratory analysis showed that gape index (computed as the area of an ellipse with major and minor axes equal to jaw length and head width; Miller & Mushinsky, 1990) is highly correlated with the three head measures in T. eques and in the two feeding morphs of T. melanogaster (Appendix 2). Accordingly, this index does not add information to the analysis beyond that provided by jaw length, head length, and head width (King, 2002). We also measured snout-vent length (SVL, Table 1) and recorded the snake gender.

| Statistical analyses
We ascertained whether head measurements differed between sexes.
As the head variables are influenced by snake size, sexes were compared using one ANCOVA for each species (n = 5) and head measurement (n = 4), entering SVL as a covariate (n = 20 ANCOVAs; see  Garcia and Drummond (1988), Drummond and Macías Garcia (1989), Manjarrez (1998) Regina and Thamnophis are two genera of semi-aquatic North American snakes (Natricinae Thamnophiine). Thamnophis snakes were collected at ponds and rivers in two watersheds in Central Mexico, while Regina were museum specimens (see Materials and Methods). Kaysville, Utah, USA).

| RESULTS
The residuals of number of teeth and jaw length, head length, and head width on SVL contributed significantly to the discriminant function (all p < .000001), which correctly classified 64.8% of the 310 snakes on the basis of four significant canonical variables which explained 100% of the variance in the data ( Values of CV1 obtained from the discriminant function analysis of morphological variation among means of snake species (which T A B L E 2 Number of snakes classified as Regina grahamii, R. septemvittata, Thamnophis eques, crayfish-eating and non-crayfish-eating T. melanogaster (Natricinae Thamnophiine) by a discriminant function analysis performed using the residuals from linear regressions of number of teeth and three log-transformed head shape variables, on SVL T A B L E 3 Canonical coefficients from a discriminant analysis to assort individual snakes belonging to Regina grahamii, R. septemvittata, Thamnophis eques, and T. melanogaster (Natricinae Thamnophiine) from two dietary morphs; crayfish eating and noncrayfish eating (see Table 2) explained 59% of the variance) increased with jaw and head length, and decreased with head width and number of maxillary teeth, according to the coefficients shown in Table 3, and they differed significantly between genera and between Thamnophis species, but not between Regina species, nor between T. melanogaster dietary morphs ( Figure 2a; Table 4). Values of CV2, which explained about one-third (32.6%) of the variance, increased with head length and decreased with jaw length and head width (  After correcting for body size, we found no difference in the number of maxillary teeth between the two Regina species, but we found difference between the two morphs of T. melanogaster. Crayfisheating T. melanogaster had 2.28 more teeth than non-crayfish-eating conspecifics (Student-t 187 = 2.92, p = .001), which themselves had 6.0 more teeth than R. grahamii and 6.3 more than R. septemvittata F I G U R E 2 Principal canonical variates obtained from a discriminant function analysis of morphological variation among snake species Regina septemvittata, Regina grahamii, Thamnophis eques, and Thamnophis melanogaster (Natricinae: Thamnophiine) with two dietary morphs, crayfish eating and noncrayfish eating.
Thamnophis snakes were captured in the wild at two Mexican drainages, while Regina were museum specimens (see Materials and Methods). (a) Principal canonical variables (CV)1. Equal letters represent statistical similarity when we compared the canonical variates among groups (one-way ANOVA). (b) Plotting the Principal canonical variables 2 versus 3 reveal morphological proximity between the crayfish-eating morph of Thamnophis melanogaster, and the two species in genus Regina which also prey on newly molted crayfish (F 1,190 = 227.6, p = .0001). We found no difference in the number of maxillary teeth between T. eques and either morphs of T. melanogaster.

| DISCUSSION
We found a large overlap in head morphology and number of teeth between the several species/morphs examined, yet we also found evidence consistent with the hypothesis that the head morphology of soft crayfish-eating T. melanogaster should more closely resemble that of the two soft crayfish-eating species of Regina than that of noncrayfish-eating conspecifics.
Crayfish ingestion in only some locations can be explained by subtle environmental differences between localities (Arnold, 1981), for example, spatiotemporal availability of crayfish or differences in use of microhabitats by T. melanogaster. However, a sampling suggests that, if anything, crayfish are more abundant in ponds where snakes do not eat them that in ponds where they do (Appendix 4).
Although significant, the magnitude of the apparent morphological convergence between crayfish-eating T. melanogaster and the two Regina species is small. This may be because invasion of this dietary niche is recent, thus even if challenging, crayfish consumption has not had time to shape head and tooth morphology. Alternatively, the selective pressures from soft crayfish predation on head/tooth morphology could be weak, for instance because crayfish-consuming populations mostly feed on other prey such as fish, tadpoles, and leeches (cf., Forsman & Shine, 1997;Manjarrez et al., 2013). Additionally, other adaptive demands on head morphology may be more important (Rossman & Myer, 1990), while optimal capture and handling of crayfish may require only minor morphological modification (both in T. melanogaster and R. septemvittata and R. grahamii). Indeed, both Regina species have been described as having head and tooth morphologies similar to those of generalist Thamnophiinae snakes (Dwyer & Kaiser, 1997), suggesting that specializing on crayfish does not induce major morphological adaptation.
Snakes preying on soft crayfish may occasionally attack slightly harder ones as these occupy the same refuges and their surface chemicals are capable of eliciting a predatory response (Manjarrez, 2003). If occasionally successful, these attacks could select for morphological adjustments to profit from such encounters. Weak selective pressure of this kind may be operating in both soft crayfish-eating Regina species and in soft crayfish-eating T. melanogaster, slowly yielding minor convergence.
The small effect size of our evidence for convergence may reflect the novelty of this niche expansion by T. melanogaster (Arnold, 1981).
No phylogeographic analysis has been made, but the restricted geographic expansion of crayfish ingestion (only 3% of the total area of sympatry of crayfish and T. melanogaster; Manjarrez et al., 2013) and its location close to the southern limit of the snake's distribution (the Natricinae originated further north) suggests that crayfish ingestion by T. melanogaster is a recent development (Lozoya, 1988).
It has been proposed that dental morphology in snakes is associated with dietary preferences (e.g., Britt et al., 2009). Thamnophis melanogaster has maxillary teeth that are curved, pointed, and oriented to pierce soft prey such as vulnerable molting crayfish. Only a few snake species ingest hard preys, and they have specialized teeth. For example, R. alleni and R. rigida have maxillary teeth with rounded tips for handling hard crayfish (Dwyer & Kaiser, 1997), whereas Storeria has long maxillary teeth that allow the extraction of land snails from their shells (Rossman & Myer, 1990). The higher number of maxillary teeth in crayfish-eating T. melanogaster (34.1 ± 3.9 teeth) compared with their congeners (32.2 ± 4.9 teeth) and soft crayfish Regina is unlikely to be an adaptation to ingest soft crayfish per se, as this runs against the trend of fewer maxillary teeth. We should, however, not dismiss too readily the possibility that having more teeth is adaptive when preying on soft crayfish, because different combinations of teeth number, head, and jaw morphology may represent equivalent mechanical solutions to the same problem (see also Arnold, 1993).
The limited scope of morphological microevolution associated with adopting a crayfish diet could also be interpreted as evidence for T. melanogaster being morphologically pre-adapted to ingest crayfish.
Our multivariate analysis supports this hypothesis because in relation to CV2, which explained a third of the variance in the original variables, Regina species and T. melanogaster cluster together and away from T. eques (Figure 2). Thamnophis is a monophyletic group that originated in the Mexican highlands ~5-6 million years ago (Mao & Dessauer, 1971;de Queiroz et al., 2002), whereas Regina is a polyphyletic group first found in North America 4-5 million years ago (Guo et al., 2012), making it more recently evolved than Thamnophis. Consequently, crayfish consumption by T. melanogaster could represent recent dietary convergence (analogy) with Regina rather than homology resulting from the common ancestor of Regina and T. melanogaster and more primitively shared with T. eques. The rarity of soft crayfish ingestion within populations T. melanogaster supports the hypothesis of analogous behavior, and it is more likely a phenomenon of invasion of a new feeding niche in an aquatic diurnal species (Hibbitts & Fitzgerald, 2005).
In conclusion, our analyses suggest that T. melanogaster shows morphological convergence in head and tooth parameters with two Regina species, potentially associated with the ingestion of a novel prey, newly molted crayfish, by the genus Thamnophis.
T A B L E 4 Canonical means used in the classification of individual Regina grahamii, R. septemvittata, Thamnophis eques, crayfish-eating and non-crayfish-eating T. melanogaster (Natricinae: Thamnophiine) by a discriminant function analysis based on jaw length, head length, head width, and number of maxillary teeth (see Tables 2 and 3