Use of field‐portable ultrasonography reveals differences in developmental phenology and maternal egg provisioning in two sympatric viviparous snakes

Abstract A thorough understanding of the life cycles underlying the demography of wild species is limited by the difficulty of observing hidden life‐history traits, such as embryonic development. Major aspects of embryonic development, such as the rate and timing of development, and maternal–fetal interactions can be critical features of early‐life fitness and may impact population trends via effects on individual survival. While information on development in wild snakes and lizards is particularly limited, the repeated evolution of viviparity and diversity of reproductive mode in this clade make it a valuable subject of study. We used field‐portable ultrasonography to investigate embryonic development in two sympatric garter snake species, Thamnophis sirtalis and Thamnophis elegans in the Sierra Nevada mountains of California. This approach allowed us to examine previously hidden reproductive traits including the timing and annual variation in development and differences in parental investment in young. Both species are viviparous, occupy similar ecological niches, and experience the same annual environmental conditions. We found that T. sirtalis embryos were more developmentally advanced than T. elegans embryos during June of three consecutive years. We also found that eggs increased in volume more substantially across developmental stages in T. elegans than in T. sirtalis, indicating differences in maternal provisioning of embryos via placental transfer of water. These findings shed light on interspecific differences in parental investment and timing of development within the same environmental context and demonstrate the value of field ultrasonography for pursuing questions relating to the evolution of reproductive modes, and the ecology of development.


| INTRODUC TI ON
The ability to accurately assess population-level outcomes of demographic changes requires measuring responses across the full life cycle of demographic events that make up a species' life history. This applies both to evolutionary-ecological applications (e.g., understanding how life-history differences evolve; Cole, 1954) and measuring fitness in wild populations (McGraw & Caswell, 1996) as well as conservation applications, such as identifying sources and sinks in spatially structured populations (Runge, Runge, & Nichols, 2006) and assessing limiting factors across an animal's life-cycle (Marra, Cohen, Loss, Rutter, & Tonra, 2015). In many cases, observing complete life cycles of plants and animals in the wild can be difficult if not impossible. Examples of hidden or unobservable demographic variables that have plagued researchers' ability to make strong inferences include partitioning mortality among seasons in migratory species (Marra et al., 2015), quantifying and following long-distance dispersal events (Cain, Milligan, & Strand, 2000), and observing life-history stages that occur underground (Muñoz, Miller, Sutherland, & Grant, 2016).
One of the major challenges in studying development in the wild is that while gravidity or pregnancy can be fairly straightforward to diagnose in many cases, it is difficult to noninvasively obtain more fine-grained information about embryonic development in the field. Often the timing of fertilization, duration of gestation, and location of oviposition/birth are unknown, particularly for more secretive species. Furthermore, while impressive progress has been made studying the dynamics of development in the laboratory in a range of species, many basic features, such as plastic differences in developmental rate and environmentally-mediated variation in maternal-fetal interactions, have remained largely unstudied in wild populations. Ultrasonography has been used to great effect in monitoring development in humans, laboratory, domestic and zoo animals (Hildebrandt, Brown, Hermes, & Goritz, 2003;Hildebrandt, Göritz, & Hermes, 2006), but use in the field has been more limited.
However, field-portable ultrasound systems can now offer highresolution images of development in a range of species, large and small, and are well-posed to help fill in gaps in our knowledge on development in the wild and provide valuable information for demographic and fitness analyses (Gilman & Wolf, 2007;Hildebrandt, Hermes, Jewgenow, & Göritz, 2000).
Squamate reptiles-snakes and lizards-are a particularly important and interesting group to improve our understanding of development in the wild. First, there is remarkable diversity of reproductive mode within this group, as viviparity has independently evolved from oviparity more than 100 times (Shine, 1983;Blackburn, 2006). Furthermore, while even viviparous snakes and lizards remain primarily lecithotrophic (i.e., nutrition is derived primarily from yolk), there is extensive variation in the degree of maternal provisioning via placenta (Blackburn, 2006). Second, while some important work has been done linking environmental factors to lizard developmental phenology in the wild, only very limited data are available for snakes (Urban et al., 2014). As snakes and lizards are ectotherms and may be particularly vulnerable to changes in environmental temperatures, an understanding of how a changing environment affects development in these species is a critical component of understanding demographic and fitness consequences of climate change. Finally, ultrasonography has already been shown to be a highly effective tool in reptiles, providing valuable information on clutch/litter size, egg mass, and developmental stage (e.g., Gilman & Wolf, 2007;Stahlschmidt, Brashears, & Denardo, 2011;Lorioux et al., 2013;Lourdais et al., 2015).
Water is required for the conversion of yolk to embryonic tissue in both oviparous and viviparous species (Thompson & Speake, 2002;Vleck, 1991), and water balance has critical ramifications for embryonic survival (e.g., Aubret, Bonnet, Shine, & Maumelat, 2003;Cagle, Packard, Miller, & Packard, 1993;Vleck, 1991;Warner & Andrews, 2002). Recent work in the viviparous aspic viper (Vipera aspis) has made ground-breaking use of ultrasonography to monitor changes in egg volume over gestation in captivity. They found that eggs can increase in size by over 200%, primarily due to maternal water transfer . Other research on V. aspis has also shown that there is a trade-off between maternal and fetal hydration, such that when mothers are water-deprived, they become dehydrated, but their eggs derive water via placental transfer to the same degree as eggs in nondehydrated mothers . In general, these findings call for further study of maternal water provisioning in other viviparous species, so that we can begin to understand how fluctuations in water availability might affect developing embryos in the wild.
We use two species of garter snake inhabiting discrete montane meadow habitats to begin an investigation of the relative effects of environmental and phylogenetic factors on embryonic development in the wild. These two species, Thamnophis sirtalis (common garter snake) and T. elegans (western terrestrial garter snake) (Figure 1), are estimated to have diverged around 6.9 million years ago; T. sirtalis is part of a small clade that is an outgroup to all other garter snakes species, including T. elegans (Manier & Arnold, 2005;de Queiroz, Lawson, & Lemos-Espinal, 2002). In the Sierra Nevada mountains in Lassen County, California, these two species share hibernation and retreat sites, bask in the same grasses throughout the day, and show an almost complete overlap in diet within meadows (Kephart, 1982).
Furthermore, these sympatric species experience the same annual fluctuations in photoperiod, temperature, and precipitation, as well as prey availability. Both species are capital breeders, with frequency of reproduction highly contingent on availability of anuran prey the previous year (Bronikowski & Arnold, 1999;Miller, Clark, Arnold, & Bronikowski, 2011). As such, this system forms a kind of natural experiment in which to explore development in two closely related species in the context of the same general environmental conditions. Both T. sirtalis and T. elegans, like all garter snakes, are viviparous.
While garter snake embryos are primarily lecithotrophic, with the yolk being the major source of nutrients, functional studies suggest that there is placental transfer of respiratory gases, nutrients, and water from the mother to the developing egg (Blackburn & Lorenz, 2003a,b;Blackburn, Stewart, Baxter, & Hoffman, 2002;Hoffman, 1970;Stewart, Blackburn, Baxter, & Hoffman, 1990). Furthermore, morphological studies of the placenta in T. sirtalis, T. ordinoides, and T. radix (the latter both closely related to T. elegans-de Queiroz et al., 2002) indicate that maternal and fetal components of the placenta become more vascularized during the last half of gestation; furthermore, the shell membrane (used for shell deposition in oviparous species, but still present in viviparous species) shows a reduction in thickness as gestation progresses, bringing maternal and fetal tissues into closer proximity (Blackburn & Lorenz 2003b;Blackburn & Flemming, 2009;). Thus, prior evidence points to the likelihood that garter snakes, like many other viviparous reptiles, will show evidence of maternal provisioning during development.
Our study employed field-portable ultrasonography to test whether sympatric T. elegans and T. sirtalis showed similar developmental phenology and evidence of maternal provisioning of embryos.
We compared distributions of females with embryos at different developmental stages between species over a given time period, and tested for differences in egg volume across developmental stages.
We hypothesized that similarities in both phenology and maternal provisioning could result from living under the same general environmental conditions. Alternatively, differences in phenology and provisioning could indicate interspecific differences in genes or gene by environment interactions underlying developmental phenotypes.
In pursuing these questions, we provide not only basic information regarding dynamics of development in this system, but also demonstrate the feasibility and value of a method for determining otherwise hidden patterns of developmental life history in the wild within and across species and environments.

| Study system and ultrasonography
All procedures in this study were conducted with approval from Eggs/follicles, which are positioned linearly caudal to the gall bladder in garter snakes, were counted and classified into the following general developmental stages: preovulatory, early development, middle development, and late development (Figure 2). Vitellogenesis is the process of yolk deposition into the oocyte prior to ovulation from the ovary. Preovulatory vitellogenic follicles were identified F I G U R E 1 Gravid female western terrestrial garter snake (Thamnophis elegans). Photograph taken by A.M. Sparkman as less echogenic (i.e., bright-appearing) than ovulated eggs, more variable in size, irregularly shaped, and often overlapping and interspersed with anechoic (i.e., dark-appearing) prevoulatory follicles (Gilman & Wolf, 2007;Matayoshi, de Souza, Ferreira, Prestes, & dos Santos, 2012;Prades, Lastica, & Acorda, 2013;Schumacher & Toal, 2001). Early, middle, and late developmental stages corresponded roughly to Zehr's 1-10, 11-24, and 25-32 developmental stages for Thamnophis sirtalis, respectively (Zehr, 1962), and represent broader groupings of Zehr's more fine-grained categories. Postovulatory eggs with no embryo visible were classified as early development.
Eggs classified as middle development had an embryonic sac clearly visible (though still small relative to the yolk sac), with indistinct echogenic skeletal elements. Eggs classified as late development had prominent, enlarged embryonic sacs with the cranium and vertebral coil evident. Eggs within females appeared to be in the same stage of development, allowing us to classify all embryos for a given female as a single developmental stage (although note that in some reptiles, differences in developmental stage can be evident among embryos within the same female-see Gilman & Wolf, 2007). While this study was cross-sectional in design, we recaptured and subsequently ultrasounded two T. elegans females and one T. sirtalis females again at later dates within the same year. We provide information on changes in developmental stage and egg volume within these individuals, for comparison with our findings based on the cross-sectional analysis.

| Developmental phenology
All statistical analyses were performed using SAS 9.3 (SAS Institute Inc). We tested for differences in developmental stage between the late development). We tested for main effects of species, year, and source population. We also included ultrasound date in Julian days as a covariate, to control for potential temporal differences in capture rate between species over the study period, as well as interactions with species and each of the other three variables.

| Maternal provisioning
We tested for evidence of maternal provisioning by testing for dif- trade-off can exist between the number and size of offspring (Seigel & Ford, 1987), we also included maternal snout-vent length (SVL) and litter size (i.e., number of eggs counted during the ultrasound) as covariates. We tested for two-way interactions between species and all other main effects.

| Developmental phenology
Species, source population, and ultrasound date were all highly significantly associated with developmental stage (Table 1). T. sirtalis embryos were at more advanced developmental stages over the study period than T. elegans embryos, with the majority of T. sirtalis females having embryos at middle or late stages of development, but the majority of T. elegans females having embryos at the early stage of development (Figure 4). Populations also varied in proportion of individuals in different developmental stages. Furthermore, more advanced developmental stages were more prevalent at later dates.
We dropped all interactions between species and other variables, as well as year from the final model, as these were nonsignificant.
It is worth noting, however, that there were 13 T. elegans individuals in 2017 that had preovulatory vitellogenic follicles, a stage not observed in any other year.
Note that for some populations, one or the other species was not sampled each year. As we detected a source population difference in developmental stages, we also ran a similar analysis restricted only to the two populations, MAH and PAP, from which we consistently had representatives of both species, to confirm that our species differences could not simply be explained by population differences. Our findings remained the same, with differences in distribution of developmental stages for each species being very similar in the restricted compared to the full analysis; thus, we present only the full analysis including all populations here.

| Maternal provisioning
The final model for mean egg volume included species, developmental stage, source population, body size, litter size, and the interactions between species and developmental stage, and species and body size ( T. elegans eggs were significantly larger than T. sirtalis eggs by late development. The three recaptured females showed species-specific patterns of change in egg volume over time that were generally consonant with our cross-sectional data-although of course this is a low sam- T. sirtalis (slope = 0.016 ± 0.003).

| D ISCUSS I ON
Our study indicates that in spite of sharing the same microhabitats and being dependent upon the same primary prey, two closely related species of snake exhibit marked differences in developmental life history (Table 3). First, we found that pregnant T. sirtalis females in this region had embryos at more advanced developmental stages than T. elegans females over the study period for all 3 years ( Figure 2). Second, while both species showed a difference in mean egg volume between early and late developmental stages, the difference in volume was significantly less pronounced in T. sirtalis than in T. elegans (27% vs. 68% increase, respectively; Figure 3). Our primary method involved sampling a population cross-section of females with embryos in different stages; however, repeated measures from a few female snakes showed a similar pattern, with T. sirtalis embryos showing only a modest change in volume over roughly a week's time in contrast to a more substantial change in T. elegans (2% in 8 days vs. 27% in 7 days, respectively). We interpret each of these findings in more detail below and explore their ramifications for future study of development in the wild.

| Developmental phenology
A potential proximate cause of species-specific differences in developmental phenology could be variation in responsiveness to environmental cues for Spring emergence. Temperature is thought to be one of the major environmental cues stimulating emergence from hibernacula and commencement of spring reproductive activities in reptiles (Krohmer & Lutterschmidt, 2011). In our study system, T. sirtalis adults may emerge at cooler springtime temperatures, allowing eggs to be fertilized and ovulated at an earlier date than those of T. elegans. The fact that T. sirtalis has a more extensive northerly geographic range than any other reptile (Logier & Toner, 1961)

| Maternal provisioning
We found strong evidence of maternal provisioning by both species of garter snake. We hypothesize that the change in mean egg volume across developmental stages we observed is primarily due to placental water uptake, as proposed in other viviparous reptiles (e.g., Dauphin-Villemant & Xavier, 1986;Lourdais et al., 2015). Why T. elegans embryos should experience a greater transfer of water than T. sirtalis embryos is unclear. Interestingly, study of placentation in garter snakes has not uncovered notable differences in morphology between T. sirtalis and T. ordinoides or T. radix (the latter both closely related to T. elegans) that would predict such a difference (Blackburn & Lorenz, 2003a,b;Blackburn et al., 2002)-although the underlying morphological differences may be nuanced, and potentially even variable among populations. At some point in its evolution since the last common ancestor with T. sirtalis, T. elegans may have experienced environmental conditions that selected for greater maternal transfer of water during development, fine-tuning maternal-fetal interactions in the already-viviparous state common to all garter snakes. Alternatively, this feature may have been derived more recently in T. elegans populations residing within this particular geographic region.
We did find that while both species showed evidence of a similar trade-off between egg size and number (i.e., where females with larger litters tended to have smaller eggs), there was a stronger positive relationship between body size and egg number in T. sirtalis than in T. elegans (Figure 7). This suggests that while there may be energetic constraints underlying the trade-off between egg size and number, physical space is an additional constraint on egg size that is more pronounced in T. sirtalis. If selection (past or present) has favored a life-history strategy in T. sirtalis that involves maximization of number of offspring in the body cavity, it may be that a trade-off with offspring size is expressed via maternal water allocation. One study in the viviparous aspic vipers (Vipera aspis) reported that larger litters resulted in less embryonic sac fluid per offspring (Bonnet et al., 2017). Thamnophis sirtalis females did have significantly larger litters on average than T. elegans females in this sample, controlling for body size (Table 3). Furthermore, anecdotal data involving recapture of one T. elegans female after 21 days suggests that the increase in egg volume may result in increasingly more tightly packed eggs ( Figure 6). If T. sirtalis eggs are already more tightly packed to begin with, this may disallow extensive water acquisition by embryos during gestation.
Conversely, if T. elegans females are unable to maximally utilize their body cavity for reproduction-perhaps due to energetic constraints (note that different populations in this study had significantly different mean egg volumes, which may indicate energetic constraints)-the trade-off between water allocation and offspring size may have been relaxed, allowing selection to favor more advanced placentotrophy (i.e., a greater degree of maternal provisioning via placenta). Future study of a greater number and diversity of garter snakes and other closely related viviparous clades would be useful in testing hypotheses regarding the interacting effects of phylogeny, body size, and litter size on changes in egg volume over development, to document the extent of variation within the viviparous reproductive mode, and the degree to which these relationships are canalized or plastic.

| CON CLUS ION
In this study, we describe a type of natural experimental system where two sympatric, closely related species experience many of the same environmental fluctuations, and yet differ in key aspects of developmental life history. Furthermore, we document intraspecific population differences in developmental phenology evident even on this small geographic scale. In general, this study provides a foundation for future work using ultrasonography in free-living animals to document patterns of maternal provisioning in the evolution of viviparity, and potential impacts of fluctuations in water availability on maternal/fetal interactions. Our research also demonstrates the usefulness of ultrasonography for investigating the dynamics of developmental phenology in the wild, in a manner that may be suitable for a wide range species, providing valuable information for demographic and evolutionary analyses, and effects of environmental change that would otherwise be difficult to discern.

Support for this field research was provided by a National Science
Foundation grant (IOS-1558071 to AMB, AMS, & DAWM). We are also grateful to Karen Martin for her continued support and hospitality at Eagle Lake.

CO N FLI C T O F I NTE R E S T
None declared.

AUTH O R CO NTR I B UTI O N S
AMS conceived and designed the study. All authors participated in the field work. AMS and KJC compiled and analyzed the data. AMS wrote the manuscript with contributions from DAWM; other authors provided editorial advice.

DATA ACCE SS I B I LIT Y
Data are archived in Mendeley (Sparkman, 2017).