We analyzed the ovaries of 69 green turtles (Chelonia mydas) preyed upon by jaguars (Panthera onca) during the nesting season at Tortuguero, Costa Rica. Our findings revealed a bimodal distribution of the diameter of vitellogenic follicles. In addition to pre-vitellogenic follicles, ovaries contained larger "dominant" follicles that would be ovulated in the current nesting season and smaller "non-dominant" follicles that were destined for resorption. Small dominant follicles increased in diameter by up to 66% during the nesting season, indicating ongoing yolk deposition. Analysis of yolk composition showed that small dominant follicles had higher percent water content than large dominant follicles. Thus, dry matter deposition rather than hydration is responsible for the pre-ovulatory increase in diameter of green turtle dominant follicles during the nesting season. Furthermore, percentages of lipid, nitrogen (N), and phosphorus (P) were constant across green turtle vitellogenic follicles, which underscores that the increase in follicle size results from provisioning with yolk containing similar proportions of these nutrients. Atretic follicles had higher water and lower P percentages than dominant follicles, indicating an accelerated resorption of phosphorus over lipids and N, which could be due to the importance of this nutrient for eggshell production. Finally, the yolk available for resorption in the green turtle ovary during the nesting season was not enough to supply the energy needed for 83% of the studied females to complete vitellogenesis, which means they must either mobilize it from body stores or non-dominant follicles or acquire more energy through foraging. These insights into follicular dynamics and nutrient provisioning highlight the ongoing reproductive investments made by female green turtles at Tortuguero. Understanding these dynamics is crucial for devising effective conservation strategies aimed at supporting the reproductive health and resilience of green turtle populations in their critical nesting habitats.

Study Site

At Tortuguero, on the Caribbean coast of Costa Rica (Figure 2), approximately 100,000 green turtle (Chelonia mydas) nests are recorded (Bruno et al. 2020; Restrepo et al. 2023) and over 300 nesting green turtles are preyed upon by jaguars (Panthera onca) annually (Troëng 2000; Arroyo-Arce and Salom-Pérez 2016). Jaguars usually attack a sea turtle through its neck and pull the contents out of the body cavity with their paws, often leaving ovaries and oviducts undisturbed.

Assisted by National Park rangers and members of the local community, we conducted necropsies of green turtles preyed upon by jaguars at Tortuguero less than 12 hours after the predation event. All activities were conducted under research permits (SINAC-ACTo-DIR-RES-057-2021, 013-2022, 025-2023).

Jaguars eating sea turtles in Tortuguero may return to the kill for several days afterwards to feed (Arroyo-Arce and Salom-Pérez 2016). Therefore, we first identified a predation event from afar by the number of black vultures (Coragyps atratus) in the area. Upon finding a depredated green turtle, we monitored the site for 10 minutes using binoculars to see if the jaguar was still present. If the jaguar was sighted (Figure 3), we left the area, if not we proceeded to conduct a necropsy of the turtle. We collected either one or two ovaries from each necropsied turtle and returned to the Tortuguero National Park’s “Cuatro Esquinas” headquarters where we proceeded with analyses.

Size distribution of ovarian follicles

To determine if all follicles were at maximum size when females arrived at the nesting beach, from 2021 to 2023 we used a caliper to measure the maximum diameter of all vitellogenic follicles (more yellow and larger than pre-vitellogenic follicles). Additionally, we measured all recently formed ovulatory scars, which were evident on top of fluid-filled vesicles (see Figure 1). We also weighed each ovary to the nearest 500 grams and used a graduated cylinder to measure the volume of each ovary by water displacement to the nearest 50 mL.

In 2021, we divided one ovary of four females into four segments of similar volume according to the position in the body axis (anterior, mid-anterior, mid-posterior, and posterior) and counted follicles and scars separately in each segment (Figure S1). Finally, when possible, we counted the number of eggs in each oviduct.

Using R package diptest (Maechler 2024), we used a Hartigan’s dip test to assess if the distribution of the diameters of green turtle vitellogenic follicles was bimodal. We also used two methods to determine the optimal number of clusters for k-means clustering: the elbow method (Syakur et al. 2018) and the silhouette method (Kodinariya and Makwana 2013). We then performed k-mean clustering to further analyze the distribution of green turtle ovarian follicles.

Based on results of the k-mean clustering, we classified follicles larger than 18 mm as "dominant". Follicles smaller than 18mm in diameter were classified as “non-dominant” (see Results). We assumed dominant follicles would have been ovulated if the female green turtle had lived, and we assumed non-dominant follicles would have been resorbed by the end of the nesting season (Hodgen 1982; Etches and Petitte 1990; Fortune et al. 1991, 2001; Nielsen et al. 2016; Mayor et al. 2023). As follicles may undergo atresia at any stage of development (Price et al. 2016), some dominant follicles were probably resorbed by the end of the nesting season.

The number of clutches remaining in a female that would have been laid in that season (clutches left) was calculated by counting the number of dominant follicles, whereas the number of clutches a female had already laid in the season (clutches laid) was calculated by counting the number of recently formed ovulatory scars. These numbers were then divided by 110, which is the average clutch size for green turtles in this population (Bjorndal and Carr 1989; Bruno 2019). Further, if we had only measured and counted follicles in one ovary, we assumed that both ovaries contained the same number of scars and follicles and multiplied the result by two to obtain the number of clutches laid and clutches left per female. We confirmed this assumption (see Results section).

We classified female green turtles as being in the first or second half of their individual nesting season based on the proportions of clutches laid and clutches left. Females in the first half of their nesting season had fewer clutches laid than clutches left, while those in the second half had an equal or greater number of clutches laid compared to clutches left. Clutch frequency, defined as the number of clutches deposited by a turtle within a nesting season, was estimated by summing the number of clutches laid and clutches left.

Nutrient analyses

To understand what was being added to dominant follicles during the green turtle nesting season, in 2022 and 2023 we collected approximately 10 large and 10 small dominant follicles and a random sample of atretic follicles per female. We dried collected follicles at 60°C in a food dehydrator until a stable mass was achieved.

We exported samples from Costa Rica under CITES permits (2021-CR5515/SJ, 2022-CR5969, 2023-CR6500) and imported into the United States under CITES permits (20US72454/9, 21US72454/9, 23US72454/9). We processed samples in the laboratory of the Archie Carr Center for Sea Turtle Research at the University of Florida.

We used a random number generator to select five large (LF) and five small (SF) dominant follicles per female. We also used all the dehydrated atretic follicles (AF). We ground each of the selected dried follicles to a uniform particle size using a mortar and pestle. We subsampled ground follicles and quantified lipid content by extraction with hexane in an Ankom system according to manufacturer instructions. Another subsample of the same follicles was used to quantify nitrogen and phosphorus content via a modified Kjeldahl method (Bouchard and Bjorndal 2000). Finally, using the remainder of the sample, we quantified organic and mineral content of the yolk by combusting the dried samples at 600°C for 8 hours. Throughout this study, we express water composition as a percentage of wet mass, while all other nutrient compositions are expressed as percentages of dry mass.

Follicles from 12 females we used for water analyses had to be excluded from the nutrient analyses because samples were damaged during transport. Further, some of the females did not have atretic follicles, and sample mass of some follicles was not sufficient for all nutrient analyses. Therefore, number of females and follicles used per analysis varied; sample size and results of each analysis are in Tables 1 and 2.

Nutrient investment and resorption

We measured the initial diameter of each vitellogenic follicle in the ovaries we studied, and we measured dry weight only for the follicles we used for nutrient analyses. Therefore, to understand how much yolk and energy had been invested and would be further invested into green turtle ovarian follicles during the nesting season, we first used a linear regression (Figure S2) to predict dry weight (g) of each follicle in the studied ovaries:

 The dry weight of each dominant follicle (predicted or measured) was taken as the yolk investment already made by the female. To estimate the additional yolk (in grams) that would be invested in each follicle during the nesting season, we subtracted the dry weight of each dominant follicle from the weight of the largest dominant follicle in the same ovary. The dry weight of all non-dominant follicles was considered as the yolk available for resorption by the female.

We then calculated the total grams of yolk each female had deposited and would have deposited and resorbed throughout the nesting season. If follicles were measured and counted in both ovaries, we summed the yolk amounts from each ovary. If only one ovary was measured, we multiplied the yolk sum by two to estimate the total yolk investment. Finally, using the average values of lipid, nitrogen, and phosphorus per gram of dry yolk specific to each female, we determined the total grams of each nutrient that would be deposited into and resorbed from green turtle ovarian follicles during the nesting season at Tortuguero.

Statistical analyses

We used R program version 4.4.0 (R Core Team 2024) for all statistical analyses, and we generated plots with package ggplot2 (Wickham 2016). We checked if data were parametric by using a Shapiro-Wilks test in base R to assess normal distribution and a Levene’s test or Bartlett’s test in package car (Weisberg 2019) to assess data homoscedasticity.

We used a Welch’s T-test to investigate whether the number of dominant follicles varied significantly according to the stage of the nesting cycle each female green turtle was in (first or second half). We used a Wilcox test in package rstatix (Kassambra 2023) to test if the number of non-dominant follicles differed significantly between the two stages of the nesting cycle.

We investigated whether nutrient content of yolk was different among follicle types (fixed effect) while blocking for the effect of female id (random effect) in a complete block design. Because of the much smaller sample sizes of AF, we first ran these models using all follicle types (AF, SF, and LF) and then again using only dominant follicles (SF and LF). For these analyses, when the data did not fit the assumptions for parametric statistics, we used a Durbin test in package PMCMRplus (Pohlert 2023). In case of a significant result (p < 0.05) in the analyses among the three groups, we used a DurbinAllPairs test to understand the variation between groups. When the data fit the assumptions for parametric statistics, we used a general linear mixed effects model (LMM) in the lme4 package (Bates et al. 2015). In case of a significant result (p < 0.05) in the analyses among the three groups, we used a TukeyHSD test with a Bonferroni correction to understand the variation between groups.