Study Site

This study was conducted in the Claro Cocorná Sur River, bordering the town Estación Cocorná in the Municipality of Puerto Triunfo, Department of Antioquia, Colombia. The site is located at an altitude of 150 m above sea level, with average temperatures ranging between 26° and 28°C and annual mean rainfall surpassing 2,000 mm (Caballero-Acosta et al. 2001).

Egg Collection

During the December 2012 nesting season, we monitored 14 river beaches (Ceballos et al. 2014b) and collected 227 freshly laid eggs from 10 nests. Daily boat trips were made every morning from approximately 0500 to 0800 hrs to collect eggs laid the night before. Eggs were taken to our field station located by the river in the town of Estación Cocorná. Once in the lab, eggs were individually marked, measured (length and width) with a caliper to the nearest 0.1 mm, and weighed with an electronic balance with 0.1 g precision.

Experimental Design

Eggs were selected randomly from the clutches as they were being collected in the field and allocated systematically to the incubation treatments until all eggs from the same clutch were used. We first filled the 3 constant-temperature treatments (following Páez et al. 2009) and then the 6 shift-twice treatments. Temperature in the constant treatments varied slightly such that the temperatures that eggs experienced were 29.2°C ± 0.8°C (Control 29°C), 31.2°C ± 0.4°C (Control 31°C), and 34.6°C ± 1.2°C (Control 34.7°C).

To estimate the onset and duration of the TSP, we used a shift-twice design: eggs were exposed to a masculinizing temperature at the beginning of incubation and groups of eggs were later switched to a feminizing temperature starting at different times during the incubation period, then returned to the original masculinizing temperature until hatching (Table 2). Two alternatives were tested by using 2 masculinizing temperatures: 29°C (which promoted a slower embryonic development) and 31°C (which promoted a relatively faster embryonic development). For both masculinizing temperatures, the feminizing temperature pulse was applied during 1 of 3 10-d periods: days 21–30, 31–40, or 41–50 of incubation, each of which falls within the middle third of the incubation period (Bull 1980; Wibbels et al. 1994; Valenzuela 2001).

Table 2. Experimental design indicating actual incubation temperatures (°C), incubation period, hatching success, and sex ratio (% females and 9% sexing error). The 2 lighter shadings indicate the 2 masculinizing temperatures while the darkest shading indicates the feminizing temperature. Numbers in bold indicate the time of the TSP (when the confidence interval of % female hatchlings obtained does not encompass zero). ^* = sex ration values not statistically different from zero given the error rate of the sexing method.

View Fullscreen

We incubated eggs in plastic containers using the original sand where eggs were laid in the nesting beach as incubation substrate (Ceballos et al. 2014b). Although the moisture content of the sand was not quantified, the initial moisture level was maintained by weekly replenishing lost water as determined by the weight of the container. To account for any potential temperature clines within the incubators, we rotated the containers weekly, from front to back and in clockwise fashion. We monitored the temperature within the incubators hourly throughout the incubation period using 3–4 HOBO JR5 Temp dataloggers (Onset Computer Corporation; accuracy 0.1°C). Hatchlings were removed from the incubators once their remnant yolk was internalized (between 5 and 7 d after pipping), weighed and measured (straight-line length and width of the carapace and plastron), and photographed individually. Hatchlings were tagged by attaching 3 colored beads to the 10th marginal scute of the carapace (Galbraith and Brooks 1984), raised in captivity for 3–4 mo, and released back in their natal river.

Development of a Geometric–Morphometric Sexing Method

The sex of 20 individuals incubated at constant temperatures was assessed by gonadal histology 3.5 mo after hatching, and results matched the expectations (Sánchez-Ospina et al. 2014): all 10 individuals from the feminizing temperature (34.7°C) were females, and all 10 from the 2 masculinizing temperatures (5 from 29°C and 5 from 31°C) were males. These individuals were then used to estimate the sex of the remaining hatchlings whose sex was unknown. For this, we quantified the shape of the carapace, plastron, and anal notch of the 20 individuals of known sex (prior to dissection) using geometric morphometrics and calculated discriminant functions based on their shape to distinguish males from females (Valenzuela et al. 2004; Ceballos et al. 2014a).

To quantify the shape of the carapace, plastron, and anal notch, all turtles were photographed at 7 d (hereafter referred to as hatchlings) and 3.5 mo of age (hereafter referred to as juveniles) with an Olympus SP-500 UZ digital camera. The number of hatchlings (n = 164) differed from the number of juveniles (n = 147) because some individuals escaped from the enclosure or were depredated by birds. We used these photographs to digitize 29 fixed landmarks at the intersections of the scutes of the carapace, 21 fixed landmarks in the plastron, and 7 fixed and 12 sliding landmarks (Bookstein 1996) in the contour of the anal notch (Fig. 1). Five individuals with deformities or supernumerary scutes were excluded from analysis. We then performed a Generalized Procrustes Analysis (Rohlf and Slice 1990) to obtain centroid size (a surrogate for carapace and plastron sizes; Bookstein 1996) and calculated shape variables for each of the 3 anatomical body parts: 54 variables from the carapace, 38 from the plastron, and 34 from the anal notch (Fig. 2). Morphometric analyses were performed using TpsDig, TpsRelw, TpsUtil, and TpsSplin software (Rohlf 2001, 2003, 2004).

View Figure

• Download as Powerpoint
• Download Figure

View Figure

• Download as Powerpoint
• Download Figure

To estimate the sex of turtles of unknown sex, we calculated discriminant functions using the shape of the carapace, plastron, and anal notch at both ages for the 20 individuals sexed by histology and estimated the percentage of correct classification (Valenzuela et al. 2004). Cross-validation of these functions was performed by using 75% of these individuals of known sex to calculate a discriminant function that was then used to predict the sex of the remaining 25%, which was then compared with their true sex. Cross-validation analyses were done with the function “Predict” from R package MASS (Venables and Ripley 2002).

We used the morphological trait whose function had the highest cross-validation rate to estimate the sex of the remaining individuals. To assess the significance of the classification, we used a Procrustes ANOVA, which is a linear model that uses the linear distances among individuals instead of using the covariance matrix and is identical to running a permutational MANOVA (Adams and Otárola-Castillo 2013). Procrustes ANOVA is advised when sample size is low (20 individuals whose true sex is known) relative to the number of shape variables. These linear models were built using sex as main factor; carapace, plastron, or anal notch shape as response variables; and clutch of origin as covariate to account for maternal effects (Páez et al. 2009). Analyses were conducted using the function procD.lm in the statistical package Geomorph (Adams and Otárola-Castillo 2013) in software R version 3.0.1 (R Development Core Team 2013).

Ontogeny of Sexual Dimorphism

We characterized sexual size and shape dimorphism in hatchlings (n = 164) and juveniles (n = 147) as reported for its congener P. expansa (Ceballos et al. 2014a) to identify subtle differences that could be used as diagnostics for the nonlethal sexing of hatchlings.

Embryonic Development

As a proxy for the embryonic stage that P. lewyana may have reached in each of the 10-d periods of the experimental treatments in this study, we used a statistical model developed for the congener P. expansa (Valenzuela 2001) that accounts for the cumulative effect of temperature on development. Daily rates at which cumulative temperature units (CTUs) accrue is a strong predictor of the days it takes for the embryos to reach particular developmental stages in P. expansa, and the thermosensitive period in this species occurs approximately between developmental stages 17 and 24 (Valenzuela 2001). Thus, assuming that the developmental biology of P. lewyana resembles that of P. expansa, we calculated the daily CTUs accumulated by P. lewyana in each of the 10-d periods of the experimental treatments in this study and predicted whether the embryos had reached or surpassed stages 17 and 24 of development using the equations for P. expansa (Valenzuela 2001, fig. 8). Mean daily CTUs were calculated by subtracting 28°C (the developmental zero temperature for P. expansa) from each hourly temperature recorded, summing the daily CTUs, and dividing by the number of days of the period (20, 30, 40, and 50 d).

Nonintrusive Geometric–Morphometric Sexing and Sexual Dimorphism

Results from our sexing procedure indicated that the plastron shape of hatchlings was the best sex diagnostic because it had the highest cross-validation rate (91%), followed by plastron shape of juveniles (85%; Fig. 2). Thus, we used the plastron shape of hatchlings to classify individuals of unknown sex as male or female and estimated the sex ratio (% females) obtained from our 9 incubation treatments (Table 2).

Sexual size and shape dimorphism changed over time and differentially for distinct body parts. Namely, the carapace and plastron size were monomorphic at hatching (p = 0.335 and 0.192, respectively) but became sexually dimorphic by the third month of age (p < 0.001; Table 3) when females were larger than males by 4.5% for the carapace and 6.4% for the plastron. Contrastingly, the carapace, plastron, and anal notch were sexually dimorphic in shape at hatching, and differences became more pronounced by the third month of age (p < 0.05 and p < 0.001, respectively; Table 4). Specifically, males exhibited a wider carapace and plastron and a more acute angle of the anal notch than females, differences which became accentuated with age (Fig. 2).

Table 3. ANOVA test of the effect of sex and clutch of origin on the size of Podocnemis lewyana at 7 d and 3.5 mo of age.

View Fullscreen

Table 4. MANOVA tests of the effect of sex and nest of origin in the body shape of Podocnemis lewyana at 7 d and 3.5 mo of age.

View Fullscreen

Effect of Constant Incubation Temperatures on Sex Ratios

The sex ratio obtained in our constant temperature treatments varied slightly from those previously reported (28°–33°C ± 0.4°C = 100% males and 34.7°C ± 0.4°C = 100% females; Páez et al. 2009). The sex ratio from our Control 29°C treatment was 7.7% ± 9% (0%–17%) females; thus, it did not differ from the expected 0% females. However, sex ratios from our Control 31°C and Control 34°C were 18% ± 9% females (9%–27%) and 86% ± 9% females (77%–95%); thus, they differed from the expected 100% males and 100% females, respectively. These results indicate that our observed temperatures (31.2°C ± 0.4°C and 34.6°C ± 1.2°C) fall within the transitional range (values producing > 0% and < 100% females) for our population (Table 2).

Effect of Shift-Twice Temperatures on Sex and the Onset and Duration of the TSP

In general, sex ratios from the slow-development shift experiments (29°–34°–29°C) were more male-biased than those from the fast-development shift experiments (31°–34°–31°C; Table 2). The sex ratios observed in our slow-shift experiment indicate that the TSP included days 20–40 of the incubation period because some females were produced in these treatments when none were expected if the TSP had been inactive (Table 2). This TSP encompassed 28% of the incubation period (71 d) and fell roughly within the middle third. On the other hand, the sex ratios observed in the fast-shift experiment indicate that the TSP included days 20–50 of the incubation period, again because some females were produced in all 3 treatments (> 18% females produced in the Control 31°C treatment). This TSP encompassed 52% of the incubation period (58 d) and fell within the middle half.

The difference between the average baseline temperatures of the slow (29.3) and fast (31.3) treatments suggests that 2°C accelerated embryo development 13.3 d on average. Additionally, within each of these 2 treatment types, there was a trend for embryonic development to take longer with later onset of exposure to the feminizing temperature pulse (Table 2). Interestingly, the timing of exposure to the feminizing temperature had a different effect on sex ratios in the slow and fast treatments. Namely, more males were produced at delayed feminizing pulses in the 29°–34°–29°C treatments, whereas more males were produced at the intermediate pulse time in the 31°–34°–31°C treatments (Table 2).

Embryonic Development

The approximate embryonic stages presumably attained by P. lewyana in this study, calculated using the model of P. expansa, were somewhat concordant with our sex-ratio observations obtained in the slow- and fast-shift treatments, with differences mostly about the end of the TSP (Table 5). According to our approximation: 1) P. lewyana embryos in the slow treatments had not reached stage 17 by day 20 and would have been between stages 17 and 24 before day 50. This result supports our finding that the TSP was active during days 20–40 because embryos are estimated to have reached stage 17. However our data shows that the TSP in P. lewyana had ended by day 40, which would be earlier than stage 24 (the end of the TSP in P. expansa, which would have been reached by day 50 under the rate of CTU accumulation experienced by P. lewyana in this treatments). 2) Podocnemis lewyana embryos in the fast treatments would have reached around stage 17 by day 20, were estimated to be between stages 17 and 24 until day 40, and likely surpassed stage 24 by day 50. Therefore, given the estimated developmental stages and the TSP of P. expansa, the TSP of P. lewyana should have been active between days 20 and 40. Yet our sex-ratio data show that the TSP of P. lewyana was active for a longer period, between days 20 and 50 (past the estimated stage 24 of P. lewyana). Potential explanations for these discrepancies are that developmental rates for P. lewyana and P. expansa are different, such that estimating embryonic stages in P. lewyana from P. expansa is not entirely accurate or that the TSP differs between both species.

Table 5. Approximate embryonic stages attained by Podocnemis lewyana embryos based on models from the congeneric species Podocnemis expansa (Valenzuela 2001).^a

View Fullscreen

Nonintrusive Estimation of Individual Sex by Geometric–Morphometric Techniques

Diagnosing the sex of individuals is important to study the ecology and evolution of species, and nonintrusive methods are particularly crucial to monitor the sex ratios of endangered species. Molecular methods have been developed for some taxa (reviewed in Literman et al. 2014), including Podocnemis (Lance et al. 1992), but they can be cumbersome and expensive and may require highly specialized equipment. Geometric–morphometric techniques are extremely powerful to compare the shape of individuals based on photographic records alone and can distinguish very subtle differences that are undetectable to the naked eye. Here, we developed an effective geometric–morphometric sexing method for P. lewyana and validated its accuracy at estimating individual sex with a subset of hatchlings whose sex was identified by gonadal histology. We found that sexing via geometric–morphometric comparison of the plastron shape of hatchlings was highly accurate (91% rate of cross-validation), even more so than in previous studies in P. expansa, in which the correct cross-validation values ranged between 75% and 90% (Valenzuela et al. 2004; Ceballos et al. 2014a). It should be noted that a margin of error still exists (albeit small) when estimating sex ratios using this approach. Geometric–morphometric techniques have been successfully used previously to sex young hatchlings of the congener P. expansa obtained under 1) constant temperatures, 2) simple artificially fluctuating profiles, and 3) naturally fluctuating temperatures in the wild (Valenzuela 2001; Ceballos et al. 2014a). As such, the approach we used to sex P. lewyana hatchlings from our experiments should apply equally well to sexing hatchlings from natural nests.

Podocnemis lewyana Populations Differ in Their Sex Ratio Response to Temperature

Despite the thermal variability experienced in the Control 29°C (28.4°–30°C) and Control 31°C (30.8°–31.6°C) treatments, these temperatures were within the range of values reported as 100% masculinizing for a northern Magdalena River population (27.6°–33.4°C in Páez et al. 2009). However, sex ratios from our study population from the Claro Cocorná Sur River, a southern tributary of the Magdalena River, were more feminized (i.e., needed less accumulated heat to produce females) than those from the north. These discrepancies are not attributable to differences in mortality rates, which were comparable between studies, and instead suggest that these populations, which are approximately 475 km of river distance apart, may differ in their thermal sensitivity.

Geographic differences in the TSD reaction norm have been reported in many turtles and may have a genetic basis (Bull et al. 1982; Ewert et al. 2005; Burke and Calichio 2014; Rhen et al. 2015). The population differences observed in the present study may reflect local adaptation, particularly considering that the northern population inhabits a region with warmer climate (mean ambient temperature 31°C, elevation 33 m a.s.l.), compared with our southern population (mean ambient temperature 28°C, elevation 150 m a.s.l.). However, previous studies have detected low genetic variability and weak population differentiation in this species, which are most likely attributable to genetic drift (Vargas-Ramírez et al. 2007, 2012), such that other causes should also be considered. For instance, differences between populations may also be attributable to variation in maternal allocation of yolk hormones, as has been reported for P. lewyana (Páez et al. 2015) and other turtles (Janzen et al. 1998; Bowden et al. 2000; Ramsey and Crews 2007), or even trans-generational effects of incubation temperature reported in other reptile species (i.e., inheritance of the parents' thermal sensitivity at which their sex was determined during embryonic development; Warner et al. 2013). Further research is necessary to test these hypotheses.

Initial Incubation Temperatures Affect the Duration but Not the Onset of the TSP

Duration of the TSP was influenced by the baseline temperature used in the shift experiments, such that the higher masculinizing temperature (31°C) induced a longer-lasting TSP than using 29°C as the base temperature. Similarly, in E. orbicularis, shifting from a lower baseline to higher temperature induced a longer TSP, and shifting from a higher baseline to a lower temperature induced a shorter TSP (Pieau and Dorizzi 1981). Furthermore, in Malaclemys terrapin, the TSP occurred in the middle third of incubation in eggs maintained at 25°C, but it occurred during the last third of incubation in eggs kept at 31°C (Burke and Calichio 2014).

On the other hand, the finding that the fast and slow treatments with feminizing temperature pulses at days 21–30 produced females suggests that temperatures experienced early in development affected the duration of the TSP, but we found no evidence of an effect of early temperatures on the onset of the TSP. However, it should be noted that our experimental design does not permit an absolute identification of the onset of the TSP because all of our pulse experiments produced some females. In other words, our data indicate that the TSP was active some time during days 21–30, but our data do not rule out the TSP having already been active by day 21. Thus, future research using earlier pulses and shift-one experiments will be needed to narrow down the exact onset and the end of the TSP.

Our findings have important implications for conservation because understanding sex determination is necessary to evaluate management programs of endangered reptiles with TSD to avoid or mitigate the effects of climate change. Given that the onset and duration of the TSP may vary depending on the temperatures experienced early in development as observed here, we recommend that conservation programs monitor the incubation temperature throughout the entire incubation period to better predict sex-ratio production. Our results also underscore the need for further research to understand the effect of fluctuating temperatures on sex determination in this and other endangered TSD taxa facing climate change (Neuwald and Valenzuela 2011). The geometric–morphometric approach developed here to sex P. lewyana should aid sex-ratio monitoring in the lab and in the wild and, thus, constitutes an important contribution to the toolkit for the conservation of this charismatic turtle.