Early hatching enhances survival despite beneficial phenotypic effects of late-season developmental environments

Abstract

Seasonal shifts in environmental conditions provide predictable cues to which organisms can respond in adaptive ways. For example, seasonal changes in temperature can induce phenotypes at different times of the year that have season-specific fitness benefits. Here, we tested the hypothesis that embryo responses to seasonal changes in thermal environments are adaptively matched to the timing of reproduction (environmental-matching hypothesis). We collected eggs of the brown anole lizard (Anolis sagrei) from early and late seasons, and exposed them to early and late thermal regimes that mimic nest temperatures. After measuring offspring morphology and performance, we quantified their survival in the field. Females had higher fecundity, but produced smaller eggs, early in the season compared with late in the season. Late-season eggs exposed to late thermal regimes had relatively high survival, but early-season eggs exposed to early thermal regimes had similar survival rates to those exposed to mismatched conditions. Late-season nest temperatures and late-season eggs produced offspring that were relatively large and fast runners. However, despite phenotypic benefits of late-season conditions, early-season hatchlings had greater survival in the field. Our results do not fully support the environmental-matching hypothesis but suggest that selection favours seasonal shifts in reproductive investment of mothers (high early-season fecundity) over plastic responses of embryos to seasonal environmental changes.

1. Introduction

Environmental change across seasons (e.g. temperature, photoperiod) is highly predictable and provides cues for organisms to initiate phenological activities (e.g. reproduction, migration [1–3]). Variation in environmental conditions also induces plastic phenotypes [4–6], resulting in phenotypic shifts across the season [7–9]. Theory predicts that natural selection should favour plasticity in temporally or spatially heterogeneous, but predictable, environments [10]. For example, larva of the moth Nemoria arizonaria exposed to chemicals in their diet develop a morphology that mimics oak flowers that are present early in the season. As the season progresses, the chemical composition of the diet changes, which induces caterpillar morphology to resemble twigs at a time when oak flowers are gone [11]. These environmentally induced seasonal changes in cryptic morphology have season-specific benefits because they conceal the caterpillars from predators. Similar adaptive matching of plastic phenotypes with seasonal environmental change has been described in diverse organisms [5,12,13].

Embryonic development is particularly important as environmentally induced phenotypes during this life stage have lasting consequences into adulthood [14,15]. In many reptiles, for example, environmental variation during egg incubation has significant effects on fitness-related phenotypes of offspring [16,17]. Moreover, nest temperatures shift in predictable ways across the reproductive season and may induce phenotypes that are suited to specific times of the year. Unfortunately, the adaptive value of environmentally induced phenotypes is poorly understood because fitness is difficult to measure, and effects of ecologically relevant conditions are rarely assessed [18–22]. Most studies that test the effects of developmental temperature use constant or arbitrary fluctuating temperatures in laboratory experiments, which are typically not experienced in nature. Rather, eggs experience temperatures that differentially fluctuate among days. This thermal complexity hinders studies that aim to quantify the effects of natural nest temperatures on offspring phenotypes. How embryos produced at different times of the year respond to season-specific temperatures is also poorly studied [22]. Research that evaluates the phenotypic effects of season-specific nest temperature will provide new insights into the adaptive significance of phenological plasticity.

To assess the effects of seasonal variation in developmental temperatures, we studied the brown anole lizard (Anolis sagrei), a species from Cuba and the Bahamas with non-native populations in the southeast USA [23]. This species has an extended reproductive season (from March to October) during which temperature increases predictably through spring and summer. Females produce about one egg every 5–10 days, which results in early-season eggs experiencing different thermal conditions than late-season eggs. Moreover, post-hatching competitive environments shift in predictable ways over the season; early-hatched lizards experience relatively low intraspecific competition, but as eggs continue to hatch and offspring grow through the season, these individuals become competitors with those that hatch late in the season. Accordingly, eggs incubated under late-season conditions produce larger offspring with faster running speed than those incubated under early-season conditions [22]. Females also produce relatively large eggs late in the season [24], resulting in seasonal differences in resources (i.e. yolk) for developing embryos. These plastic responses of embryos and shifts in maternal investment might prepare late-hatched offspring for a competitive environment. Given these predictable seasonal increases in nest temperature, maternal allocation and competition, natural selection could shape an adaptive match between early versus late eggs with their respective season-specific environments. Although seasonal variation in maternal allocation and temperature may have fitness consequences, whether seasonally driven plasticity is adaptive remains unknown.

In this study, early- and late-season eggs were exposed to early- and late-season temperatures to determine if the timing of oviposition is adaptively matched to the respective seasonal thermal environments (i.e. environmental-matching hypothesis [25]). By releasing laboratory-incubated offspring in the field, we assessed the interactive effects of timing of oviposition and season-specific nest temperature on survival. We addressed two predictions. First, we tested the prediction that early-season embryos that experience early-season temperature will have relatively high survival, and the hatchlings will have enhanced fitness-related phenotypes and survival compared with those that experience late-season temperatures (and vice versa for late-season embryos). By contrast, a mismatch between timing of oviposition and season-specific nest temperature would increase embryo mortality, produce low-quality phenotypes and reduce hatchling survival. Second, we tested the prediction that enhanced phenotypes late in the season (due to shifts in maternal allocation and positive effects of increased incubation temperatures) would compensate for negative consequences of hatching late. Support for these predictions will provide evidence that plastic responses of embryos or shifts in maternal investment are adaptively matched to season-specific environments.

2. Methodology

(a). Animal housing and egg incubation

We collected two breeding colonies of A. sagrei (80 females and 40 males each) from Palm Coast, FL, USA, near the start of reproductive season (9 February 2015) and late in the reproductive season (15 July 2015) to obtain early- and late-season eggs, respectively. Adult females were housed individually and males were rotated among cages weekly to encourage frequent mating. Lizards were maintained under standard conditions [26], and room lighting was adjusted to mimic seasonal changes in photoperiod.

Pots of soil in each cage were searched for eggs three times per week beginning 9 March for the first cohort and 15 July for the second. Eggs were weighed and randomly assigned to a treatment where early- and late-season eggs were incubated at temperatures that mimicked early- and late-season thermal regimes of nest sites in the field (see below). Eggs were incubated individually in glass jars (59 ml) with substrate collected from the field site (substrate water potential: −150 kPa) and covered with plastic wrap sealed with a rubber band.

(b). Experimental design

Thermal regimes were determined by placing iButtons into 24 potential nests at the field site in 2013. Because nest temperature data were not available for A. sagrei, we chose potential nest sites that were likely to encompass the environmental conditions that eggs experience in nature. Potential nests were approximately 2 cm deep and were located across a range of microhabitats (open to shaded areas and under natural cover objects similar to where A. sagrei eggs are found [27]). IButtons were programmed to record temperatures every 2.5 h from 1 April to 15 October.

Data from the 24 potential nests were averaged to establish two fluctuating thermal regimes (using programmable incubators) that mimicked temperatures during early and late periods of the season (figure 1). The thermal regimes carried out 45-day cycles beginning with temperatures recorded on 1 April or 15 July. Owing to temporal variation in egg production, not all eggs within a treatment experienced the exact same temperature regime. For example, an egg laid on 15 July experienced a thermal regime that mimicked conditions from 15 July to the date it hatched, whereas an egg laid 5 days later experienced thermal conditions from 20 July to the date it hatched. In addition, when thermal regimes reached the end of the 45-day period, the entire sequence of temperatures was looped back to day 1 (e.g. eggs placed in the incubator on day 40 spent their first 5 days at temperatures characteristic of the end of the cycle, and the remainder at the beginning of the 45-day cycle). Although not all eggs within a treatment experienced the exact same temperature regime, the early temperature regime was about 6°C cooler on average and exhibited 1.8 times as much variance compared to the late regime (figure 1). These temperature regimes maintained ecologically relevant conditions and captured the potential thermal effects experienced in the field.

Figure 1.

Average thermal regime of potential nest sites (n = 24) in the field (a). The segments of the thermal regime within the rectangles represent the two incubation temperature treatments programmed into the incubators (b). Experimental incubation regimes mimicked nest temperatures of 45-day periods early in the season (1 April–15 May 2014, blue arrows) and late in the season (15 July–28 August 2014, red arrows). The combination of early- and late-season collection of eggs and early versus late incubation temperatures created a 2 × 2 experimental design. Solid arrows represent treatments that were matched between seasonal cohort and season-specific nest temperature, and dashed arrows represent treatments that were mismatched. Sample sizes represent the number of eggs allocated to each treatment. The mean temperature for the early-season and late-season thermal regimes were 21.0°C and 26.9°C, respectively. Variances for the early-season and late-season thermal regimes were 6.68 (minimum = 14.7°C, maximum = 27.2°C) and 3.72 (minimum = 23.1, maximum = 32.6°C), respectively. (Online version in colour.)

(c). Hatchling phenotypes and survival

Hatchlings were measured in terms of snout–vent length (SVL), tail length (TL) and mass, and were sexed within 24 h of hatching. Each lizard was uniquely marked with a toeclip and briefly housed individually under standard conditions [26] before release.

Sprint speed was assessed 7–9 days after hatching. Sprint trials were performed by chasing hatchlings along a 1 m racetrack (placed at a 20° angle) with an artists' paintbrush. Each individual was raced five times with a 2 min rest between trials. Photocells at 25 cm intervals along the track triggered a stopwatch as the lizards ran past. As a measure of running behaviour, the number of times hatchlings stopped during the 1 m distance was recorded for each trial. Trials were performed at 28°C. The fastest sprint speed over 25 cm for each individual was used in the statistical analyses.

Hatchlings were eligible for release after they had been raced, and release dates varied between one and three weeks post-hatching (first cohort release dates, n = 233: 16 May, 31 May, 15 June, 10 July, 31 July 2015; second cohort, n = 331: 2 September, 10 October, 23 October, 18 November 2015). Hatchlings were released on an island (approx. 50 m diameter) in the Matanzas River located 19.6 km from the parents’ collection site. The island consisted of an open area with sparse mangroves and other small plants and a central forested area with cedar trees and little underbrush. The island eliminated dispersal, and thus facilitated recapture efforts and survival estimates.

A preliminary recapture effort was conducted on 10 October 2015. At this point, not all hatchlings had been released, as many late-season eggs incubated under the early-season temperature treatment had not yet hatched. Five people collected lizards for 8.33 h (306 lizards captured). Based on a later sampling effort in March 2016, 18 known survivors were not detected during this preliminary effort. Non-experimental resident A. sagrei were also collected, measured, marked and released to serve as a natural control group. All lizards were measured for SVL, TL and mass.

A second and more thorough recapture effort was performed to assess overwinter survival of experimental and resident individuals. Five people collected lizards from 14 to 17 March 2016 over 32.82 h (623 lizards captured). Based on surveys of a nearby island population, detection probability is 73% with an effort of 75 person hours (D.A.W. 2017, unpublished data). Because the capture effort was substantially greater in the current study (164.1 person hours), we are confident that most lizards were captured in this population during the March sampling effort. All lizards captured at this time were euthanized and deposited in the Auburn University Museum of Natural History.

(d). Statistical analysis

All analyses were performed using SAS software (SAS Institute, Inc., v. 9.4, 2016). Seasonal variation in reproductive investment was quantified with analysis of variance (ANOVA) using seasonal cohort (early versus late) as an independent variable. Separate analyses quantified cohort differences in latency to reproduce, egg mass and inter-egg interval. Latency to reproduce was the number of days between the first egg laid within a cohort and the first egg laid for each female. Inter-egg interval was the number of days between successive eggs averaged for each female. Egg mass was the mean egg mass per female. Fecundity (number of eggs per female) was compared between seasonal cohorts with analysis of covariance, using latency as a covariate.

Two-way mixed-model ANOVAs assessed the effects of seasonal cohort (early versus late), incubation temperature (early versus late regimes) and their interaction on incubation period and offspring phenotypes. Offspring sex had no major effect on any trait (electronic supplementary material, table S1), and was not included in the statistical models. Generalized linear mixed models assessed the effects of seasonal cohort, incubation temperature and their interactions on egg survival and hatchling survival in the laboratory. Tail length, sprint speed and average number of stops were log-transformed to normalize data. Egg mass was a covariate when testing the main effects on SVL and hatchling mass. SVL was a covariate for analyses of TL and hatchling mass (body condition). Mass at the time of the running trials and log-transformed number of stops were used as covariates for analysing sprint speed. Maternal identity was a random effect in all models.

Growth rate was analysed with a linear mixed model with body mass at hatching as a covariate. Growth rate in the field of recaptured individuals was the change in mass divided by the number of days between release and recapture. Hatchling survival (binomial distribution, logit-link function) in the field was analysed with generalized linear mixed models, using body mass at hatching and release date as covariates. Maternal identity was a random effect for analyses of growth and survival. Both dependent variables were based on recapture data collected in October 2015 and in March 2016. Because late-season eggs under early-season temperatures had not hatched by the October recapture effort, we could not perform full factorial analyses for these variables in October. Thus, we performed two separate analyses of growth and survival to October. First, the effects of incubation temperature on growth and survival were assessed only for the early-season cohort. For the second analysis, temperature treatments were combined to compare early versus late cohorts. Full factorial analyses were performed for growth and survival measured in March 2016 (i.e. independent variables were incubation regime, seasonal cohort, their interaction, hatchling mass and release date). Interactions between release date and the main effects (incubation regime and seasonal cohort) on survival were never significant, and these terms were removed from final model. For survival, individuals recaptured in March but not detected in October (n = 18) were scored as present in the October analyses.

Comparisons of survival between the laboratory-incubated and resident hatchlings were performed with χ² tests. For meaningful comparison, we only used resident lizards less than 22 mm SVL, which is the maximum length at hatching. In addition, comparisons of survival between laboratory and resident hatchlings were restricted to the time period from October to March. These two restrictions avoided comparisons across different times of the year and between lizards from different size or age classes.

3. Results

(a). Reproductive investment

Early-season eggs were smaller (n = 490; mean ± s.e.: 0.169 ± 0.001 g) than late-season eggs (n = 538; mean ± s.e.: 0.185 ± 0.001 g; F_1,872 = 43.1, p < 0.001). On average, early-cohort females took longer to begin nesting than those in the late cohort (F_1,152 = 73.9, p < 0.001; figure 2a), but early-cohort females had relatively short inter-egg intervals (F_1,151 = 12.6, p < 0.001; figure 2b). Owing to these cohort differences in latency and oviposition frequency, the average number of eggs produced per female did not differ across seasonal cohorts (F_1,153 = 1.2, p = 0.276). However, when fecundity was adjusted for latency, early-cohort females produced more eggs than those from the late cohort (F_1,151 = 71.6, p < 0.001; figure 2c).

Figure 2.

Comparisons of reproductive investment between female A. sagrei from early- and late-season cohorts. (a) Latency to begin oviposition since the production of the first egg within cohorts. (b) Average number of days between production of successive eggs (inter-egg interval) produced by females. Data are reported as means ± 1 s.e. (c) Number of eggs produced by females (fecundity) relative to latency to begin oviposition. Blue circles represent early-cohort females and red circles represent late-cohort females. (Online version in colour.)

(b). Incubation and hatchling phenotypes

Egg survival and incubation duration were affected by seasonal cohort, incubation temperature and their interaction (table 1; electronic supplementary material, table S2). Late-season eggs had higher survival than early-season eggs, particularly if they were exposed to corresponding late-season temperatures (figure 3a). In addition, eggs exposed to late-season temperatures had higher survival than those exposed to early-season temperatures. Eggs incubated under late-season temperatures hatched on average 37.2 days earlier than those from the early-season temperature regime (figure 3b), resulting in a 47.1% difference in incubation duration. Late-season eggs also had shorter incubation periods than early-season eggs irrespective of temperature, but the difference was only by 2.6 days (9.5% difference).

Table 1.

Effects of seasonal cohort (early versus late), season-specific temperature (early versus late) and their interaction on embryo development and hatchling phenotypes. Covariates for each analysis are described in the text. Italics values are statistically significant. Effect sizes are reported in electronic supplementary material, table S2.

trait	seasonal cohort	temperature	interaction
egg survival	F1,870 = 10.6; p = 0.001	F1,870 = 12.5; p < 0.001	F1,870 = 7.9; p = 0.005
incubation duration	F1,646 = 38.4; p < 0.001	F1,646 = 25 394.7; p < 0.001	F1,646 = 24.7; p < 0.001
snout-vent length	F1,646 = 80.6; p < 0.001	F1,646 = 0.6; p = 0.453	F1,646 = 0.6; p = 0.443
body mass	F1,646 = 52.7; p < 0.001	F1,646 = 10.9; p = 0.001	F1,646 = 0.5; p = 0.481
body condition	F1,646 = 21.7; p < 0.001	F1,646 = 7.1; p = 0.008	F1,646 = 3.2; p = 0.074
tail length	F1,642 = 19.8; p < 0.001	F1,642 = 114.6; p < 0.001	F1,642 = 8.0; p = 0.005
sprint speed 25 cm	F1,524 = 4.3; p = 0.039	F1,524 = 12.1; p < 0.001	F1,524 = 0.4; p = 0.541
average stops 1 m	F1,530 = 81.8; p < 0.001	F1,530 = 34.1; p < 0.001	F1,530 = 5.3; p = 0.021
survival in laboratory	F1,644 < 0.1; p = 0.903	F1,644 = 37.1; p < 0.001	F1,644 < 0.1; p = 0.971
growth to October	F1,16 = 4.3, p = 0.055	F1,16 = 1.1; p < 0.316	—
growth to March	F1,20 = 0.6, p = 0.441	F1,20 = 0.02; p = 0.890	F1,20 = 0.4; p = 0.513
survival to October	F1,297 = 8.8; p = 0.003	F1,296 = 2.1; p = 0.144	—
survival to March	F1,412 = 4.6; p = 0.032	F1,412 = 0.8; p = 0.373	F1,412 < 0.3; p = 0.616

Figure 3.

Effects of seasonal cohort, incubation temperature and their interaction on embryo development and hatchling phenotypes for A. sagrei. (a) Egg survival, (b) incubation duration, (c) hatchling SVL, (d) hatchling body mass, (e) hatchling body condition (residual score from mass/length regression), (f) hatchling TL, (g) running speed over 25 cm, (h) number of stops made by hatchlings when running over 1 m and (i) survival in the laboratory. Statistical results are reported in table 1. Error bars represent 1 s.e. (Online version in colour.)

In general, the late-season eggs and late-incubation regime produced larger offspring that ran faster than those produced from the early-season eggs and early-incubation regime (table 1). Late-season offspring had a greater SVL, TL, mass and body condition than early-season offspring (figure 3). Although incubation temperature did not affect SVL, the late-season temperature produced hatchlings with relatively long tails and weighed more than those produced under the early-season temperature. These patterns remained when body size was corrected for egg size and when body mass was adjusted for SVL (i.e. body condition). Interactions between seasonal cohort and incubation temperature for hatchling morphology were not significant after correcting for multiple comparisons.

Late-season hatchlings were faster runners than early-season hatchlings, and those from the late incubation treatment were faster than those incubated at the early-season temperatures (figure 3g). Late-season hatchlings made more stops than early-season hatchlings, but hatchlings incubated at late-season temperatures had fewer stops than those from early-season temperatures (figure 3h). Offspring that experienced late-season incubation temperatures had higher survival in the laboratory than offspring exposed to early-season temperatures (figure 3i).

(c). Growth and survival in the field

Most individuals (n = 383) were released prior to the October recapture effort, and an additional 181 individuals (all late cohort) were released over the following month, resulting in a total of 564 hatchlings released. Of these, 19 and 25 individuals were captured in October and March, respectively. Survivors that were not detected in October (n = 18) but were found in March raised our number of October survivors to 37 individuals. Thus, survival rates were 9.7% in October and 4.4% in March. Of the lizards released by October, significantly fewer individuals were recaptured from the late cohort compared with those from the early cohort (table 1 and figure 4a). For the early cohort, temperature treatment had no effect on survival to October (table 1). Survival to March was also affected by the seasonal cohort, but not by incubation temperature; again, individuals from the early cohort had relatively high survival (table 1 and figure 4b). Neither release date nor body mass had an effect on survival to October (release date: F_1,297 = 1.6, p = 0.205; body mass: F_1,297 = 0.3, p = 0.606) or March (release date: F_1,412 = 0.5, p = 0.485; body mass: F_1,412 = 1.0, p = 0.329). Growth in the field was not affected by incubation temperature, seasonal cohort or their interaction (table 1).

Figure 4.

Effects of seasonal cohort and incubation temperature on hatchling survival of A. sagrei in October 2015 (a) and March 2016 (b). Statistical results are reported in table 1. (Online version in colour.)

During the October recapture efforts, 49 resident individuals qualified as hatchlings and six of these were recaptured in March. The recapture rate of resident hatchlings (12.2%) was significantly higher than that of laboratory-incubated hatchlings (4.2%; χ² = 19.8, p < 0.001).

4. Discussion

Our experiment tested the hypothesis that embryos produced at different times of the year are adaptively matched to season-specific environmental conditions. This prediction was partially supported by analyses of egg survival, but the general lack of significant interactions between seasonal cohort and season-specific temperature on other phenotypes did not support this prediction at post-hatching stages. We also show partial support for our second prediction that late-season temperatures induce, and late-season offspring develop, phenotypes that are often associated with high fitness (e.g. large size, fast running speed); however, those phenotypes did not compensate for hatching late. Overall, the benefits of hatching early in the season outweighed the phenotypic benefits of late-season developmental conditions.

Late-season thermal regimes induced many developmental and phenotypic attributes that are typically associated with high fitness. Egg survival was increased under late-season temperatures, particularly for eggs produced late in the season (figure 3a), which is in the direction predicted by the environmental-matching hypothesis. This relatively high egg survival for matched conditions suggests that late-season embryos may be adapted or primed for late-season thermal regimes. In addition, developmental rate increased substantially (by nearly 50%) under the relatively warm late-season temperatures, which is consistent with other ectotherms [17,28]. An increased developmental rate could increase egg hatching success late in the season because it reduces the amount of time eggs would be exposed to predators or adverse ambient conditions.

Post-hatching phenotypes differed between early- and late-season offspring. Late-season eggs produced hatchlings that were relatively large and ran rapidly, and those from late-season incubation treatments produced hatchlings that were heavier, faster, had longer tails, made fewer stops when running and had higher survival in captivity than those from early-season incubation treatments. Previous work shows no morphological effects of constant incubation temperatures on hatchling A. sagrei [29], but the effects observed in the present study are consistent with our recent work that examines fluctuating incubation temperatures [22]. Together, these findings reinforce the importance of simulating natural nest temperatures because they induce significant phenotypic variation that might otherwise be masked by artificial conditions [30,31]. Whether the plastic responses observed here were shaped by natural selection or were merely a passive consequence of the physical impacts of temperature is unknown. Moreover, the impact of seasonal cohort and incubation regime on sprint speed could also be explained by seasonal differences in thermal performance curves. That is, because locomotor performance was tested at a single temperature (28°C, which is slightly lower than the optimum for Bahamian A. sagrei [32]), the greater running speeds of individuals from the late cohort may reflect the shape of the curve, rather than an effect of plasticity. Regardless of the mechanism, however, the late-season phenotypes observed here might compensate for negative effects of late hatching where offspring enter a more competitive environment than early-hatched individuals [22,24]. For example, relatively large size can benefit hatchlings during territorial disputes [33], and fast speed can aid in escaping agonistic interactions (frequently observed in young anoles [34]) or capturing prey in a competitive environment.

Recapture efforts, however, yielded no evidence that late-season phenotypes provide benefits over hatching early in the season; the late-season cohort experienced relatively low survival. These results reinforce the importance of early oviposition or hatching [35–39] and demonstrate that phenotypic consequences of the developmental environment are relatively minor compared with those of hatching early in the season. Indeed, the benefits of large size and fast running speed may be minor because late-hatched individuals will still be relatively small due to the longer growth period of early-hatched individuals. Moreover, enhanced growth and survival of early-hatched offspring is likely due to reduced competition with larger conspecifics early in the season [37]. Given the rapid growth of A. sagrei hatchlings [40], individuals that hatch early in the season can reach a sexually mature size by the end of that same season. Similar benefits of early birth or hatching have been documented in numerous taxa [37,41,42].

The survival benefit to early-season hatchlings suggests that natural selection should favour females with relatively high fecundity early in the season. Consistent with this expectation, many species produce relatively large clutches early in the season [43–45] (but see [46]). For A. sagrei, which produces one egg at a time, we show that the inter-egg interval is shorter early in the season compared to later (figure 2), a pattern consistent with other work on this species [47]. Although reproduction was delayed for early-cohort females, these individuals produced the same amount of eggs as the late-cohort females. Thus, when corrected for latency, early-cohort oviposition rate was relatively high (albeit egg size was relatively small). Owing to high survival of early-season offspring and their rapid growth to sexual maturity, relatively high fecundity by females early in the season should increase maternal fitness. Thus, selection should favour seasonal shifts in fecundity in a direction consistent with this pattern (more egg production early in the season), particularly in short-lived species (like A. sagrei) that rarely survive to a second reproductive season [48].

One caveat of survival estimates in the field involves detectability. The small island where lizards were released eliminated concern for dispersal from the study site, and our final sampling effort should have provided high detectability. However, survival was still relatively low compared with that reported in another study of A. sagrei on a nearby island [49]. In that study, juvenile survival was 15% but included individuals with larger body sizes (up to 39 mm SVL) than those in the present study; therefore, lower hatchling survival would be expected in our study. The survival rate of resident hatchlings, however, was similar to that reported previously [49] and was significantly higher than that of our laboratory-incubated hatchlings. One explanation is that selection had already operated on phenotypic variation in the resident population prior to capture efforts, and thus sampling was biased towards individuals that had already survived in the field. Similarly, field-caught Sceloporus undulatus have higher survival in the field than laboratory-reared individuals [50]. In addition, as field-caught lizards hatched into their natural environment, they may have an advantage by being familiar with their surroundings.

The impacts of the timing of oviposition and developmental environments on offspring phenotypes and survival have been demonstrated previously in a range of organisms [35,51], but the ecological implications are difficult to interpret because experiments often do not mimic ecologically relevant conditions [22,52]. Our novel approach provides an advancement in how researchers could design treatments to better reflect ecologically relevant incubation environments. Rather than constant conditions or continuously repeating thermal fluctuations each day [22,52–54], we simulated thermal cycles that are experienced in the field at different times of the year. Thus, incubation environments were controlled, yet temperatures mimicked the complex thermal environments found in nature. Future laboratory studies could use similar approaches to explore a variety of related questions, and we urge researchers to use and refine our approach to enhance the ecological relevance of their work.

Our study also has broad implications for discerning the roles of reproductive timing and plasticity in shaping variation in fitness in natural populations, particularly in changing environments. For example, changes in microclimate due to anthropogenic factors (e.g. climate change, habitat modification) can shift the timing of reproduction, so that environments are less predictable, resulting in inappropriate phenotypes being produced in a given season [5,55]. Such seasonal changes can negatively impact individuals and populations in diverse taxa [13,56–58]. We show that the timing of reproduction contributes most to offspring survival and potentially overshadows the contribution of plastic responses to variation in survival. Many organisms benefit from early hatching or birth [55–58], but the length of the reproductive season may influence the degree of this benefit; indeed, survival differed between our seasonal cohorts, but release date within cohorts did not explain variation in survival. Relatively long reproductive seasons (as in A. sagrei) could expose early versus late offspring to environments with very different selective pressures. If the level of plastic response is constrained from producing phenotypes extreme enough to deal with season-specific fitness consequences (e.g. due to physiological limits, plasticity may not induce a body size that is large enough to compete early-hatched individuals), adaptive phenological plasticity is unlikely to arise. Thus, researchers that examine adaptive matching across temporal scales in response to changing environments must consider the influence of seasonal length of reproduction and the physiological limits of plastic responses [59].

Overall, we tested whether the timing of oviposition is adaptively matched with corresponding seasonal incubation temperatures. Although our study system meets critical assumptions of the environmental-matching hypothesis (i.e. that changes in developmental environments are predictable and correlate with post-developmental environments [25,60]), we found limited support for this hypothesis during the embryonic stage and no support during post-hatching stages. Instead, by decoupling the effects of season and season-specific temperatures, we demonstrated significant fitness advantages of hatching early in the season despite positive phenotypic effects of late-season environments. The lack of support for the environmental-matching hypothesis could be driven by the overwhelming benefit of hatching early in the season, which could constrain the evolution of adaptive seasonal plasticity of embryos. We also showed that natural thermal variation induces phenotypic variation in wild populations. Given the overarching advantages of hatching early, selection may be more likely to operate on shifts in maternal investment (i.e. seasonal variation in reproductive investment) than on the phenotypic responses of embryos. Such patterns may be generalized to organisms that have prolonged and seasonal reproductive cycles, and should be considered in studies that investigate the consequences of environmental change. Controlled laboratory studies that simulate natural environments combined with field measures of survival, such as ours, are necessary for pinpointing the relative importance of environmentally induced seasonal variation in phenotypes and seasonal shifts in reproductive investment.

Ethics

All guidelines and procedures for the use of animals were approved by the Institutional Animal Care and Use Committee at the University of Alabama at Birmingham (Protocol 140710215), and the Guana Tolomato Matanzas National Estuarine Research Reserve.

Data accessibility

The dataset is available at the Dryad Digital Repository (http://dx.doi.org/10.5061/dryad.68711) [61].

Authors' contributions

Both authors conceived and designed the study, analysed the data, wrote the paper and approved the final version of the manuscript for publication. P.R.P. performed the laboratory work and data collection. Both authors agree to be held accountable for the content of this manuscript.

Competing interests

We declare we have no competing interests.

Funding

Funding was provided by University of Alabama at Birmingham Department of Biology, Auburn University Department of Biological Sciences, Alabama Academy of Sciences, The Society for Integrative and Comparative Biology and Sigma Xi.