Embryo Development and Sex Ratios in the Red-Footed Tortoise (Chelonoidis carbonarius) at Masculinizing Temperatures
Abstract
Incubation temperature is a factor that can affect several traits in turtles such as body size, growth, shape, and sex in species with temperature-dependent sex determination. A clear understanding of these effects is particularly important in threatened species such as the red-footed tortoise (Chelonoidis carbonarius), classified in Colombia as Vulnerable mainly due to the capture of wild individuals to be kept as pets. A previous study on the effect of incubation temperature on sex determination concluded that constant temperatures of 31°C and 33°C are lethal to C. carbonarius embryos and that a temperature of 29°C produces 100% females. Although this showed that C. carbonarius has a temperature-based sex determination mechanism, its full reaction norm is still unknown. To fill in the gaps, we incubated 160 eggs from 47 nests at constant temperatures of 24°C, 26°C, and 28°C. The hatching success rates were 55%, 53%, and 60%, respectively, with extremely long incubation periods of 213, 164, and 138 d. Of the 75 neonates obtained, 58 reached 8 mo of age, at which time 45% (n = 26) were sexed. The sexual proportions (% males) obtained were 88.9%, 60%, and 0% from the 24°C, 26°C, and 28°C treatments. From the calculated thermal reaction norm, we estimated that the pivotal temperature is 26.05°C (95% CI = 25.31–26.69) and the transitional temperature range is from 24.34°C to 27.77°C. Incubation temperature also affected incubation period and hatchling body size and growth. Additionally, we found a significant effect of female body size and mass on egg size and mass, supporting a fecundity selection hypothesis of body size. Finally, we inspected hatchlings for sexual shape dimorphism using geometric morphometrics. We found significant differences in anal notch and plastron shape at 7 d of age, but only differences for anal notch at 1 and 2 mo of age. However, the statistical signal was weak and the results varied with age; thus this sexing technique was unreliable. Given the low critical thermal maxima for embryo survivorship (31°C or less), together with a 100% feminizing temperature of 27.9°C, global climate change coupled with deforestation constitute imminent threats to this species due to demographic impacts such as low recruitment rates and skewed primary sex ratios.
Several lines of evidence sustain the opinion that global environmental temperature has been increasing since the preindustrial (1850–1900) period (Intergovernmental Panel on Climate Change [IPCC] 2014). Climate change can affect turtles in different ways, such as by altering their habitats, growth rates, fecundities, nesting phenologies, and predation rates (Butler 2019). Extreme environmental temperatures in particular can affect turtles with temperature-dependent sex determination (TSD), increasing the risk of skewing their primary sex ratios, which in turn causes a reduction in effective population size and reduces genetic variability (Valenzuela et al. 2019; Singh et al. 2020). Empirical studies, however, show that some species may have mechanisms to potentially mitigate the effects of extreme temperatures during the incubation period, such as heat-shock proteins (Bentley et al. 2017), maternally derived estrogens (Singh et al. 2020), modifying the timing of nesting (Janzen et al. 2018), selection of thermal microenvironments for nesting (Janzen et al. 2018), and delayed hatching (Constanzo et al. 2004; Rafferty and Reina 2012). The state of knowledge on these topics has grown in some taxa such as sea turtles (Booth et al. 2004; Booth 2006; Staines et al. 2019); however, the information is scarce in other groups such as terrestrial turtles (Butler 2019). This is the case for the terrestrial tortoise family Testudinidae, one of the most threatened clades of the order Testudines. Of the 65 modern species (living or extinct since CE 1500; Lovich et al. 2018) in this family, 10.8% (n = 7) are already extinct and 73.8% (n = 48) are considered Threatened (Critically Endangered, Endangered, and Vulnerable [CR+EN+VU]) (Rhodin et al. 2018). The lack of information on the sex determination mechanism for this family is of concern. Only 16.9% (n = 11) of the species are known to have a TSD-Ia mechanism (summarized in table 1 in Hernández-Montoya et al. 2017; Sancho et al. 2017), but the mechanism is unknown for the remaining 83.1% (n = 54).
Incubation temperature, as well as maternal effects, can also affect other hatchling traits such as hatching success rates, initial body sizes and masses, body shape, and growth rates (Booth et al. 2004; Páez et al. 2009; Ceballos and Valenzuela 2011; Ceballos et al. 2014; Butler 2019). Understanding how temperature and maternal effects influence these traits is important because they can affect hatchling survival and performance, and therefore fitness (Booth et al. 2004; Staines et al. 2019; Tezak et al. 2020). These effects are poorly known in Testudinidae and unknown in Chelonoidis spp. in particular.
The red-footed tortoise, Chelonoidis carbonarius, is an important species because, in spite of its wide distributional range from Nicaragua to Argentina (Uetz et al. 2019), it is considered a Vulnerable species in Colombia (Resolución 1912 de 2017, Ministry of Environment and Sustainable Development 2017) and is listed in Appendix II of the Convention on International Trade in Endangered Species (CITES) (Turtle Taxonomy Working Group [TTWG] 2017). This tortoise is threatened by a constant extraction of individuals from natural populations to be consumed or held as pets (Morales-Betancourt et al. 2015), as well as by fragmentation and destruction of their habitats (Vargas-Ramírez et al. 2010). This tortoise has a body size up to 44.9 cm straight-line carapace length (SCL) (Gallego-García et al. 2012). According to Aponte (2001) and Aponte et al. (2003), wild females reach maturity between 20.9 and 22.3 cm curved line carapace length (CCL) and between 9 and 17 yrs of age, depending on habitat quality. Compared with females, adult males are larger, have a more concave plastron, a larger supracaudal carapacial scute that extends beyond the plastron edge, a larger and thicker tail, and a wider and deeper anal notch (Pritchard and Trebbau 1984) Recently, the sex determination mechanism of C. carbonarius was evaluated (Hernández-Montoya et al. 2017), and it was found that constant incubation temperatures of 31°C and 33°C were lethal, while 29°C produced 100% females, although with only a 52% hatching success rate. Those results confirmed C. carbonarius TSD, but the full reaction norm remained unknown.
Our general objective in this study was to describe the effects in C. carbonarius of additional lower, constant incubation temperatures and of maternal effects on hatchling phenotypes: sex, body shape (sexual shape dimorphism), initial body size, and growth rates. Specifically, we predicted the following. 1) The transitional temperature range that produces both sexes, the pivotal temperature that produces a 50:50 male:female sex ratio, and the 100% male-producing temperature should be between 24°C and 28.2°C. 2) As for many other turtle species, we expected that incubation temperature would affect hatchling morphology and body growth rates. 3) We expected a significant and direct effect of the females' body size (SCL and mass) on clutch size and also an effect of egg size and mass on hatchling size, mass, and growth rates. 4) Finally, we expected to be able to find sexual differences in the plastron shape in hatchlings consistent with the sexual shape dimorphism exhibited in adults in this species (specifically, males with wider and deeper anal notches).
METHODS
Egg Collection and Incubation. — This study followed the egg collection and manipulation methods used by Hernández-Montoya et al. (2017) to obtain comparable results. Specifically, we collected 160 eggs from 47 nests of C. carbonarius between 0 and 6 d after oviposition. Eggs were collected between 5 October and 13 November 2017 from a C. carbonarius captive rearing facility located in Puerto Perales (Antioquia, Colombia). The females were weighed (Pesola® 80010 Spring Scale, 10 kg) and measured (Haglöf Mantax® caliper, 50 cm) for SCL and straight-line carapace width (SCW).
Eggs were individually marked and transported to the Herpetology Laboratory at the University of Antioquia following the protocols described elsewhere (Páez et al. 2009). Once in the laboratory, each egg was cleaned with a 1% hypochlorite solution, weighed on a digital scale (Lexus®, 0.1-g precision), and measured with a digital caliper (Brown & Sharpe®, 0.01-mm precision). Next, eggs were randomly assigned to 1 of the 3 incubation temperatures (Binder® incubators, 0°C precision, n = 54, 55, and 51 eggs in the 24°C, 26°C, and 28°C conditions, respectively) on a bed of humid commercial vermiculite prepared in a 2:1 weight ratio of distilled water:vermiculite (medium grain size). Each container was then weighed and placed within a Ziplock™ plastic bag. Substrate moisture variation was minimized by replenishing the weight lost from evaporation with distilled water each week (Páez et al. 2009). Also, containers were rotated among shelves each week to compensate for potential temperature gradients within the incubators. Embryo development was examined by candling the eggs every 2 wks. From the total of 160 eggs collected, only 83.7% (n = 134) exhibited blood circulation, confirming embryo development. The remaining eggs presented a bad odor or exhibited white fluff or reddish spots on the shell surface and were discarded after confirming whether the embryo had died or the egg was infertile. We tested for effects of incubation temperature on incubation period, defined as the time from the egg laying date until the egg pipped. We used a linear mixed effects model with temperature as a fixed effect and maternal origin as a random effect. For this purpose, we used the functions “lmer” and “ls_means” from lmerTest package (Kuznetsova et al. 2017) in the R environment (R Core Team and contributors worldwide, www.r-project.org).
Hatchling Rearing and Growth Analyses. — After egg pipping, neonates remained inside the incubator until the vitello was fully absorbed (1–6 d). Then, hatchlings were weighed, measured, and their carapaces and plastrons were photographed, and these data were collected again each month until the hatchlings had grown sufficiently (8 mo of age or 90 g of body weight; Hernández-Montoya et al. 2017) for sexing. Measurements taken were (mm) SCL, SCW, CCL, and straight-line plastron length (SPL). Photographs were taken using a digital camera (Nikon COOLPIX P510®) on a tripod placed vertically above the turtle (in a zenith angle) and with a piece of metric tape near the turtle as a size scale. The hatchlings were reared at the Santa Fe Zoological Park, Medellín, Colombia, and received water ad libitum and a diet based on local fruits (Carica papaya, Cucurbita maxima, Daucus carota, Manguifera indica), vegetables (Arachis pintoi, Phaseolus vulgaris, Beta vulgaris, Pennisetum purpureum, Colocasia esculenta), flowers (Hibiscus rosa, Taraxacum officinale), and cooked chicken dusted with calcium powder once a week.
To examine whether there were differences among the three incubators in terms of egg mass and initial hatchling body size (SCL) and body mass we conducted an analysis of variance (ANOVA) using linear models. Because egg size varied substantially (mean egg mass of clutches ranged from 36 to 66 g) and egg size is known to influence initial hatchling size and growth rates in other species (Páez et al. 2009; Ceballos and Valenzuela 2011; Ceballos et al. 2014), we examined the effects of incubation temperature on initial hatchling body size and juvenile body size at 5 mo of age using mixed effects models, with incubation temperature as a fixed effect and maternal origin (female identification) as a random effect.
Maternal Effects. — In addition, we also tested for potential maternal effects (female SCL, SCW, and mass, and body condition index [BCI]) on clutch size (number of eggs/nest), egg size (mass and major diameter), hatchling size (SCL, SCW, and mass), and growth, again using mixed effects models. Because SCL alone may not take into account the nutritional status of the females, we used BCI = mass/SCL3 (Bjorndal et al. 2000) for the analyses.
Sexing and Sexual Dimorphism. — From the total hatchlings initially obtained (n = 75) only 77% (n = 58) reached 8 mo of age; and from these 8-mo-old hatchlings a sample of 45% (n = 26) were sexed by laparoscopy or by direct observation of the gonads, with histological confirmation when needed (hatchlings with “known sex”; Fig. 1). The remaining hatchlings (sample sizes varied across ages) were sexed using geometric morphometrics (as explained below) using the body shape of hatchlings with known sex as a baseline.
Citation: Chelonian Conservation and Biology: Celebrating 25 Years as the World's Turtle and Tortoise Journal 20, 1; 10.2744/CCB-1441.1
Laparoscopies followed the same procedure as in Hernández-Montoya et al. (2017). Briefly, we used a rigid laparoscope (Storz Hopkins® 30°, 1.9-mm diameter) connected to a cold light and anesthetized the individual using ketamine with diazepam (60 mg/kg intravenously [IV] and 0.5 mg/kg intramuscularly [IM], respectively), plus applying lidocaine locally (2 mg/kg) (Carpenter and Marion 2017). When venous injection was not possible due to small body size, the injections were administered IM with an additional supplementation of ketamine and diazepam 10 min later (25 mg/kg and 0.5 mg/kg, respectively). After the sexing procedure was finished (Hernández-Montoya et al. 2017), we administered an anti-inflammatory drug and antibiotics (ketoprofen at 2.2 mg/kg IM and amoxicillin at 10 mg/kg IM) as a precaution. We attempted to sex the remaining individuals by quantifying their plastron shape using 2D geometric morphometrics, using the plastron shape data from those individuals with known sex as references. For this purpose, we took digital photographs of the plastron of all hatchlings each month until they were 8 mo old. With the photographs, we digitized 23 fixed landmarks around the entire plastron as well as 7 fixed and 9 sliding landmarks around the anal notch (Fig. 2). A Generalized Procrustes Analysis (GPA) was performed (Rohlf and Slice 1990) to obtain a set of shape and centroid size (a surrogate of size) variables for each individual, and those variables were used in subsequent statistical analyses. Digitalization of landmarks was conducted using the TPS software (TPS Util, TPS Digit) (Rohlf 2015), and the GPA procedure was conducted using the Geomorph package (Adams et al. 2019) in the R environment.
Citation: Chelonian Conservation and Biology: Celebrating 25 Years as the World's Turtle and Tortoise Journal 20, 1; 10.2744/CCB-1441.1
To estimate if there was sexual shape dimorphism in the hatchlings with known sex, we ran a linear model using shape as the dependent variable and sex as the independent variable. This model was performed with function “lm.rrpp” (linear model evaluation with randomized residuals in a permutation procedure) available in the RRPP package (Collyer and Adams 2018) in the R environment. When an effect of sex on shape was significant, then a classification function was performed on the shape data of the remaining hatchlings to estimate their sex. The function “classify” in the RRPP package (Adams et al. 2019) was used for this purpose.
Sex Determination Mechanism. — Using the average incubation temperatures experienced by the hatchlings and their sexes, we described the thermal reaction norm. While sex determination is influenced by mean temperature and daily fluctuations in temperature (Georges et al. 2004), the standard deviations of the temperatures recorded in this study (0.2°C–0.5°C) are not expected to produce a high embryo developmental rate that could bias the sex results, plus the 95% confidence interval (CI) of the estimated pivotal temperature was larger: 25.31–26.69 (see “Results” below).
The reaction norm was estimated using the function “tsd” (estimates the parameters that best describe temperature-dependent sex determination; Girondot 1999), available in the Embryogrowth package (Girondot and Kaska 2014), which uses the number of females and males as the dependent variable and the incubation temperatures as the independent variable. This function estimates the parameters that best describe the reaction norm for sex ratio: the pivotal temperature, the transitional range of temperatures (TRT; lower and higher limits), and the 100% male- and female-producing temperatures. Using the “tsd” function, 4 models were tested: logistic, Hill, A-logistic (asymmetric logistic), and Flexit (flexible logistic) (Abreu-Grobois et al. 2020), and the best model was selected using the Akaike Information Criterion (AIC). Once the best model was selected, the parameters were estimated with maximum likelihood, but their 95% CIs were estimated using the Metropolis Hastings Markov chain Monte Carlo procedure with “tsd_MHmcmc” and “P_TRT” functions in Embryogrowth (Abreu-Grobois et al. 2020).
RESULTS
Egg Incubation and Hatching. — The actual temperatures experienced by the eggs in the incubators were 24.3°C (SD = 0.5°C), 26.0°C (SD = 0.2°C), and 27.9°C (SD = 0.2°C). From the total of 160 eggs collected, only 83.7% (n = 134) exhibited blood circulation, confirming embryo development. Of the 134 apparently viable eggs, only 56% (n = 75) hatched successfully, specifically, 55.1% (n = 27), 53.3% (n = 24), and 60% (n = 24) from the 24.3°C, 26.0°C, and 27.9°C incubation conditions, respectively. This relatively low hatching success was not related to the egg collection time delay (between 1 and 6 d after oviposition) (F1,5 = 1.635, p > 0.257) but did vary significantly among females, irrespective of incubation temperature (χ2 = 86.053, df = 41, p < 0.001).
The incubation period was shorter with higher temperatures (F2,57 = 202.4, p < 0.001, while holding maternal origin as a random effect), and all 3 pairwise comparisons (24°C–26°C, 24°C–28°C, 26°C–28°C) were significant (p < 0.001). Specifically, it was 214 d (range = 178–251, SD = 41.2), 165 d (range = 150–202, SD = 13.6), and 138 d (range = 120–171, SD = 13.9) from the 24.3°C, 26.0°C, and 27.9°C incubation conditions, respectively.
Clutch, Incubation Temperature, and Maternal Effects on Size and Growth. — Average female mass (n = 47) was 4.7 kg (range = 3.3–6.5, SD = 0.671), average SCL was 31.2 cm (range = 27.2–36.7, SD = 2.0), and average BCI was 0.157 (range = 0.105–0.197). Likewise, average egg mass was 54.6 g (range = 32.1–70.0, SD = 7.9) and average egg diameter was 46.1 mm (range = 38.3–50.8, SD = 2.5). Average clutch size was 3.5 eggs (range = 1–8, SD = 1.3).
We found no significant relationship between female BCI, mass, or SCL with clutch size (BCI, F1,45 = 0.229; p > 0.63; mass, F1,45 = 1,095, p > 0.30; SCL, F1,45 = 1.187, p > 0.28). Likewise, female BCI was marginally significant with egg mass (F1,71 = 4.208, p = 0.044) and was not significant with egg length (F1,71 = 0.0524, p = 0.819). However, we found that heavier females were associated with heavier eggs (p < 0.0001, R2 = 0.364) and larger females (SCL) also had larger eggs (p < 0.0001, R2 = 0.305).
Likewise, while accounting for egg weight and using female identity as a random variable, we found no significant effect of female size (mass, SCL, and SCW) on initial hatchling sizes (mass, SCL, and SCW) (mass, F1,32 = 1.75, p = 0.19; SCL, F1,22.4 = 0.007, p = 0.93; SCW, F1,51.6 = 0.35, p = 0.35) or on 5 mo of age hatchling sizes (mass, SCL, and SCW) (mass, F1,64 = 0.50, p = 0.48; SCL, F1,64 = 0.156, p = 0.69; SCW, F1,26.2 = 1.81, p = 0.18). The effect of egg weight, however, had a significant effect on initial hatchling size and was still apparent 5 mo later (p < 0.05 all cases). Because eggs from each clutch were divided at random among the 3 incubators, there were no differences among incubators in egg mass (ANOVA, F2,57 = 0.438; p = 0.10). However, we found significant differences among clutches in egg mass (ANOVA, F28,31 = 6.581, p < 0.0001).
Because of this significant maternal effect, the effect of incubation temperature on hatchlings size (at hatching and 5 mo of age, n = 73) was tested with the female identity as a random effect using linear mixed effects models. We found a significant effect on initial hatchling carapace size (SCL, F2,49.4 = 6.035, p < 0.005; SCW, F2,50.3 = 4.389, p < 0.02), with hatchlings from the coolest incubation temperature (24°C) significantly larger than individuals from the other 2 temperatures (see post hoc pairwise comparisons in Fig. 3). Five months later, the effects of incubation temperature on hatchlings size changed to the opposite direction. That is, the effect was significant on body mass, SCL, CCL, and SPL (body mass, F2,51.2 = 4.31, p < 0.019; SCL, F2,47.2 = 4.155, p < 0.022; CCL, F2,49.1 = 4.344, p < 0.019; SPL, F2,49.4 = 3.579, p < 0.036), but this time with the 5-mo-old hatchlings from the coolest incubation temperature being smaller, i.e., exhibiting growth rates significantly slower than those of individuals from the highest incubation temperature (28°C) (Fig. 3).
Citation: Chelonian Conservation and Biology: Celebrating 25 Years as the World's Turtle and Tortoise Journal 20, 1; 10.2744/CCB-1441.1
Post-hatching Survival and Sex Ratios. — Of the 75 hatchlings obtained, 17 died (22.7%) for different reasons: congenital malformations (n = 4), a salmonellosis outbreak (confirmed by microbiological test, n = 8), and during the postoperative phase of the laparoscopies (n = 5). Of the 58 hatchlings that reached 8 mo of age, we were able to sex 26 individuals through laparoscopy, direct inspection of the gonads, or with histology.
The sex ratios obtained were 11.1% female (8 males and 1 female) at 24.3°C, 40% female (6 males and 4 females) at 26.0°C, and 100% female (0 males and 7 females) at 27.9°C. We expected to be able to sex the remaining 32 hatchlings using as a reference the average shape of those individuals with known sex using geometric morphometrics. However, sexual shape dimorphism (SShD) at earlier ages (from hatching until the fifth month of life) was weak and inconsistent. That is, SShD of the plastron was only significant at hatching (F1,14 = 3.360, p = 0.005), but not at older ages, and SShD in the anal notch was significant only at hatching and at 1 and 2 mo of age (F1,24 = 2,467, p < 0.042; F1,24 = 2,603, p < 0.035; and F1,24 = 2,603, p < 0.035 respectively), but not at older ages. In addition, when SShD of the plastron was used to sex the hatchlings, their predicted sexes changed over time, indicating that this was an unreliable sexing method.
Temperature-Dependent Sex Determination. — The thermal reaction norm was estimated using the incubation temperatures and sex ratios obtained in this study through laparoscopies or necropsies, plus those from Hernández-Montoya et al. (2017), in which 28.9°C produced 100% females (n = 20). The model that best fitted the data was the logistic (AIC = 10.46, Akaike weight = 0.33), although very close to the Hill model (AIC = 10.73, Akaike weight = 0.29), the A-logistic (AIC = 10.91, Akaike weight = 0.26), and the flexit model (AIC = 12.65, Akaike weight = 0.11). Using the logistic model the following parameters were obtained: pivotal temperature = 26.05°C (95% CI = 25.31–26.69), TRT = 3.44, TRT lower limit = 24.34°C (95% CI = 22.05–26.91), TRT higher limit = 27.77°C (95% CI = 25.17–29.69), and the 100% feminizing temperature = 27.9°C (Fig. 4). The 100% masculinizing temperature was not found within the range of evaluated incubation temperatures, but the closest was 88.9% at 24.3°C.
Citation: Chelonian Conservation and Biology: Celebrating 25 Years as the World's Turtle and Tortoise Journal 20, 1; 10.2744/CCB-1441.1
DISCUSSION
From data available (from this study and Hernández-Montoya et al. 2017), we confirm that the sex determination mechanism of C. carbonarius is the TSD-Ia pattern, in which low temperatures produce males and high temperatures produce females. The lowest temperature evaluated was 24°C, and we still did not obtain only males, indicating that the 100% masculinizing threshold lies below 24°C. It will be necessary to document the minimum viable incubation temperature for this species and the sex ratios obtained to inspect for a potential TSD-II pattern, in which females are produced at both low and high incubation temperatures and males at intermediate temperatures (Ewert et al. 1994). However, this possibility seems unlikely given that 11 other species of Testudinidae exhibit TSD-Ia (Hernández-Montoya et al. 2017, and references therein; Sancho et al. 2017), with 100% male-producing temperatures between 25°C and 29°C, all higher than the lowest temperature we evaluated (24°C). This low thermal reaction norm is expected given that C. carbonarius inhabits humid rain forest (Medem 1962; Gallego-García et al. 2012; Hernández-Montoya et al. 2017), while most of the Testudinidae species inhabit mostly dry desert or warm coastal environments (Ernst and Barbour 1989).
Female C. carbonarius in natural habitats lay their nests in shady areas by excavating a superficial hole with their posterior legs; the nests sometimes are left uncovered (Mondolfi 1955). Superficial nests are exposed to warmer and more variable temperatures than are deeper nests, even for turtle species that lay relatively shallow nests (Georges 1992). Historically, vegetation cover might have provided enough habitat heterogeneity to allow females to choose among differing microenvironments when nesting. However, habitat degradation in many areas of Colombia is extensive, particularly in the dry forest of the lowlands of the Caribbean and Magdalena regions. If vegetation cover is reduced or nonexistent, the ability of females to mitigate extreme incubation conditions behaviorally would be limited (Refsnider et al. 2013). This is why we suggest that, given the low critical thermal maxima for embryo survivorship (31°C or less) and a 100% feminizing temperature of 27.9°C, global climate change and deforestation constitute imminent threats to this species by affecting its demographics through low recruitment rates and skewed primary sex ratios.
The long duration of the incubation period appears to be a characteristic of this species and not a result of the particular range of temperatures used in this study. Mondolfi (1955), Medem et al. (1979), and Hernández (1997) reported incubation periods from 105 to 228 d under both natural and artificial incubation conditions. The apparently natural low rate of development of this species might represent another threat to it as global climate change also increases seasonal temperature fluctuations and stochasticity.
We found that female size was not related to clutch size, similar to other reports for this species (Castaño-Mora and Lugo-Rugeles 1981) but contrary to the findings of Hernández and Boede (2008), studying this species in captivity at the Llanos of Venezuela, who found that larger females laid both bigger eggs and larger clutches. However, we did find a positive relationship of female size and mass on egg size, but not on hatchling size and mass nor on growth. Nevertheless, egg size was positively related to initial hatchling size and mass, an effect that was maintained to the fifth month of age. This suggests that, in C. carbonarius, female fitness might increase with larger body sizes, via larger egg size, because their larger hatchlings would enjoy higher survivorships (Paitz et al. 2007; Janzen and Warner 2009). This species lays several clutches per season (3.2 clutches/yr; Vanzolini 1999), so larger females also might exhibit greater annual reproductive effort by laying more clutches each year rather than by producing larger clutches. In a recent pilot study on the relationship between maternal body size and total annual egg production in this species, we documented a significant female effect on egg mass and, after controlling for female effects, a significant decrease in egg mass but not clutch size in successive clutches (V. Rave, C.P.C., and V.P.P., unpubl. data). Among the maternal effects commonly demonstrated in turtle species are the influence of egg mass on multiple hatchling phenotypic characteristics such as body size and shape, growth, performance, and survival rates (Ashmore and Janzen 2003; Mitchell et al. 2015). In this study, when we included egg mass as a covariate, maternal influences on hatchlings size and growth were no longer significant, implying that even if there are additional differences among clutches related to the genetic constitution of the female, her mate or mates, or the composition of the egg yolks (nutrients, hormones, antibodies, or antioxidants), these were not detectable in our results. The maternal effect we demonstrated is important because hatchling body size, mass, and growth rate are all important fitness components to offspring in turtle species (Janzen et al. 2000, 2007).
In addition, the BCI index we calculated for C. carbonarius (0.2 ± 0.018) was well below that calculated for other turtle species (Chelonia mydas: 1.42 ± 0.015, Seminoff et al. 2003; Actinemys marmorata: 1.05 ± 0.04, Ashton et al. 2015). This low index may be due to different reasons associated with captive rearing, such as poor body condition, hydration state, urinary bladder/gut contents, reproductive state, or a mixture of these factors (Bonnet et al. 2001; Seminoff et al. 2003). In addition, we found that female BCI was a weak predictor of maternal effects. In our study, the largest females were underweight and still had a high reproductive output (i.e., larger eggs); we therefore recommend use of female mass or SCL instead of BCI in future studies.
We also found that the masculinizing incubation temperature (24°C) initially produced hatchlings of larger body sizes compared with those females obtained from the higher temperatures; however, this effect was reversed 5 mo later (the male juveniles were smaller). While a low incubation temperature may facilitate the conversion of yolk energy reserves into body tissues (Díaz-Paniagua et al. 1997) during embryo development, making male hatchlings initially larger than females, it is curious that we documented a reversal wherein faster female growth rates resulted in their being larger than males at 5 mo of age. According to Charnov and Bull (1977), in TSD species it is the sex that benefits most from large size as adults (males in C. carbonarius) that should exhibit the faster growth rates; however, in our results, 5-mo-old female juveniles (produced at 28°C, Fig. 4) were larger than males. But similar reversals of body sizes between sexes due to differences in posthatching growth rates also have been described in other species with a type Ia TSD pattern (Demuth 2001; Du et al. 2007). To better understand why such growth rate variation occurs in this or any species, it will be necessary to conduct longer-term studies of growth rates that consider multiple factors such as incubation temperature, the adult size differences between sexes, and even competition between individuals for resources.
Our geometric morphometric analyses of SShD in the plastron and anal notch were weakly significant but variable throughout ontogeny, which differs from results reported for other turtle species (mentioned above). We cannot discard the possibility that a small sample size (8 males and 12 females in this study) was one potential explanation. Alternatively, other parts of the body might be inspected, such as the relative sizes in the opening in the shell for accommodating the limbs and tail (Kaddour et al. 2008; Bonnet et al. 2010) or other traits related to courtship, as seen in Testudo species (Willemsen and Hailey 2003).
Hatching success rates in our study were low (55.1%), and our results suggest that it was most related to female (nest) effects. In fact, hatching success was also low (56%) in the previous study using C. carbonarius clutches obtained from the same captive rearing facility (Hernández-Montoya et al. 2017). We hypothesize a poor nutritional state of the females (Hernández 1997) or even genetic factors (i.e., this may be an endogamous turtle collection, given the number of hatchlings with malformations seen in this study and also reported by Hernández-Montoya et al. 2017). Future captive reproduction programs oriented toward the conservation of C. carbonarius should recognize the unusually low pivotal temperature of this species and the possibility that long incubation periods also may be due to embryo diapause, in attempts to increase hatching success rates and obtain a mix of both sexes (Rafferty and Reina 2012; Romito et al. 2015; Hernández-Montoya et al. 2017).
Photographs of (A) an ovary of a 6-mo-old Chelonoidis carbonarius hatchling, 94.6 g body weight, from the 28°C incubation condition; and of (B) a testis of a 5-mo-old hatchling, 99.2 g body weight, from the 26°C incubation condition. Photographs were taken during necropsy through a stereoscope (×4). Microphotographs of histological cuts of (C) an ovary and (D) a testis. Scale bar = 50 µm. Abbreviations: K = kidney, O = ovary, Ov = oviduct, T = testis.
Illustration of the fixed (black circles) and sliding (white circles) landmarks digitized in the entire plastron (left) and around the anal notch (right) in Chelonoidis carbonarius to estimate sex using geometric morphometrics.
Average body mass (g) and body linear measurements (mm) of Chelonoidis carbonarius at two ages: 7 d after hatching and 5 mo (n = 67), by incubation temperature. Different letters in pairwise comparisons (of the same age) indicate incubation temperatures have a different effect.
Reaction norm of the sex determination mechanism of Chelonoidis carbonarius estimated with the data from this study and a previous study (Hernández-Montoya et al., 2017). Black dots are sex ratios observed with their 95% confidence interval [CI], the plain curve shows the maximum likelihood model with its 95% CI as dashed lines, the vertical dash-dotted line shows the pivotal temperature (26.05°C), and the dotted lines indicate the lower (24.33°C) and higher (27.77°C) limit of the transitional range of temperatures.
Contributor Notes
Handling Editors: Jeffrey E. Lovich and Cristina A. Jones
