Unique multiple paternity in the endangered big‐headed turtle (Platysternon megacephalum) in an ex situ population in South China

Abstract Understanding the mating system and reproductive strategies of an endangered species is critical to the success of captive breeding. The big‐headed turtle (Platysternon megacephalum) is one of the most threatened turtle species in the world. Captive breeding and reintroduction are necessary to re‐establish wild populations of P. megacephalum in some of its historical ranges in China, where the original populations have been extirpated. However, the captive breeding of P. megacephalum is very difficult and this may be due to its mysterious reproductive strategies and special behavior (e.g., aggressive temperament and territoriality). In this study, we achieved successful captive breeding of P. megacephalum by creating a habitat that mimics natural conditions and then investigated its mating system using microsatellite makers. A total of 16 clutches containing 79 eggs of P. megacephalum were collected, and 52 were hatched successfully over two breeding seasons. Of the 15 effective clutches, 6 clutches (40%) exhibited multiple paternity. There was no significant correlation between clutch size and multiple paternity, and no significant difference in hatching success between multiple‐sired and single‐sired clutches. However, there was significant correlation between male body size and the number of offspring, with higher‐ranked males contributing to more clutches. Our results provide the first evidence of multiple paternity and male hierarchy in P. megacephalum. These findings suggest that multiple paternity and male hierarchy should be considered in captive breeding programs for P. megacephalum, and creating a habitat that mimics natural conditions is an effctive way to achieve successful captive breeding and investigate the mating systems of this species.

investigate its mating systems using microsatellite DNA makers.
The main goals of this study were to reveal the characteristics of mating systems and provide a scientific basis for captive breeding of P. megacephalum. Detailed information for each animal is provided in Table 1. The turtles were seized from illegal animal markets and then raised in a turtle conservation center in Huizhou, Guangdong, South China. To encourage breeding, an mimicing natural habitat (resembling natural ponds) was created near a natural mountain stream. Each pond was rectangular, with an area of 20 square meters (4 m × 5 m). Half of the area was a water pool, and the other half was land. Flowing mountain spring water was introduced into the pool via pipes. The water temperature was 12-26°C during the year, the pH was approximately 6.4, and the water depth was 15-30 cm. Some stone caves were established in the pools as shelters for the turtles. The plants (e.g., Acorus tatarinowii, Blechnum orientale) from the natural habitats of the turtles were transplanted into the land section of the pools to create suitable nest microenvironments. Some living stream shrimps and small fish were introduced to the ponds to provide food for the turtles. In addition, fresh loaches, fish, river snails, and earthworms were also offered as supplementary food 2-3 times each week. The 20 turtles were divided into two groups and put into two ponds (pond I, pond II) separately (Table 1). In pond I, three males exhibiting significant differences in weight and body size were grouped with seven females, to help us to understand the effects of body size on social ranking. In pond II, five males and five females with only minor body size variations were grouped together.

| Sampling
From our observations, the laying season for P. megacephalum is from June to August, so we searched for eggs every day during this time. To prevent potential damage by snakes and mice, a cylindrical wire mesh cover was immediately placed over the nest. A total of 16 clutches of P. megacephalum were found in pond I (13 clutches) and pond II (3 clutches) during 2012-2013. Automatic temperature recorders (Taiwan Hengxin: AZ8829) were used to record the nest temperatures during incubation. After hatching, the newborn turtles were transported to housing for care. Tissue samples of the offspring were taken from the tail tips (0.5 cm). Oral swab samples, rather than tissue samples, were collected from the 20 parental turtles to reduce the potential sampling damage as much as possible.
Tissue samples and oral swab samples were stored at −20°C in 95% ethanol.

| Microsatellite genotyping
Genomic DNA was extracted from tail tissue and swab samples using the phenol-chloroform protocol and salting-out method, respectively (Sambrook & Russell, 2001). All individuals were genotyped at nine, previously developed microsatellite loci for P. megacephalum (Table 2). The PCRs were performed in volumes of 15 μl, which contained 1× Buffer, 1.5 mmol/L MgCl 2 , 200 μmol/L dNTPs, 0.2 μmol/L Primer (each), and 0.02 U/μl Taq DNA polymerase. The PCR conditions were 94°C for 3 min for denaturation; 34 cycles of 30 s denaturation at 94°C; 55°C for 30 s for annealing; 72°C for 30 s for extension; and final extension for 5 min at 72°C. The PCR products were separated with 10% polyacrylamide gel electrophoresis (PAGE) and were visualized using silver staining (Zhang, Wang, Chen, Lan, & Lei, 2007).

| Paternity analysis
The average allele numbers, mean expected heterozygosity, and polymorphic information content (PIC) of the microsatellite loci were calculated using CERVUS version 3.0 (Kalinowski, Taper, & Marshall, 2007). The parentage analysis module available in CERVUS was used to calculate the success of assignment of candidate parents to the offspring. CERVUS uses a log (base e) likelihood algorithm to calculate the likelihood ratio (LOD) of a candidate male being the true parent compared with an random male. LOD scores are calculated for all possible sires, and the difference between the two most likely candidates (delta) is calculated and provides an indication of the reliability of the assignment. The delta score is calculated at a confidence level of 95% and corresponds to an estimated frequency of false positives of 5% (Moon, McCoy, Mushinsky, & Karl, 2006;Roques et al., 2004). First, the candidate mothers of each clutch were verified.
Once the candidate mother and the offspring were matched, the paternal alleles were deduced from the comparison of both maternal and offspring genotypes. We assumed that clutches with more than two paternal alleles were multiple paternity. SPSS 17.0 Statistical Analysis Software and Pearson correlation coefficients were used to analyze correlations between variables (e.g., body size of female or male, clutch size, and the number of offspring).

| Social behavioral observations
To explore the influence of social hierarchies on the mating system, the social behaviors of big-headed turtles were observed every 2-3 days from April to November during 2012-2013. All turtles were marked for individual recognition with a white or red number code painted on the carapace. In order to reflect an individual's social rank, the territory size and the eating order of each turtle were observed.
Based on observation, females in pond I seem to have no territory (or territorial behavior) and they can move freely anywhere in the pool.
The males in pond I have their territories, and territorial behavior TA B L E 1 Body measurements of the 20 Platysternon megacephalum used in this study occurs only among males. The males in pond II seem to have no territory (or territorial behavior), and they can move freely anywhere in the pool. Except for the three males in pond I, the feeding order of other turtles in pond I and pond II seems random. So, in this study, we only analyzed the territory and eating order in males in pond I.
The territory of each male in pond I was approximate estimated by the irregular polygons generated by all occurrence site connections at the edge of active range. The feeding order of the males in pond I was analyzed by chi-square test based on the behavioral observation data.

| Egg hatching
A total of 16 clutches with 79 eggs (mean 5 eggs per nest) were collected, of which 13 clutches (63 eggs) were collected from pond I and 3 clutches (16 eggs) from pond II. The fluctuation of nest temperatures during the incubation period ranged from 11.5 to 32.5°C (mean 22.4°C), and the incubation period ranged from 78 to 97 days.
The fluctuation of relative humidity in the nest during the incubation period ranged from 63.2% to 100% (mean 97%). Nine of the eggs were destroyed by ants during the incubation period, and 18 eggs failed to develop. A total of 52 offspring successfully hatched from 15 clutches, and the total hatching rate was 74.3% (Table 3). Some of the nests, eggs, and hatchlings are shown in Figure 1.

| Paternity analysis
Nine microsatellite loci were successfully amplified in the 20 parents and 52 hatchlings. The average number of alleles per locus in all the samples was 3.89, the mean expected heterozygosity was 0.541, and the mean polymorphic information content (PIC) was 0.4071. Paternality was assigned to each of the 52 hatchlings at the 95% confidence interval (Table 3)

| Male social rank affects the number of offspring
All females in pond I and even the smallest female in pond II participated in successful reproduction, and 7 (78%) of the 9 females laid two clutches (Figure 2). There was no significant correlation between female body size and the number of offspring (r = .091, p > .05). For the males, however, there was a significant correlation between male body size and the number of offspring (r = .774, p < .05), with the larger males contributing to more clutches and thus producing more offspring. In pond I, the largest male (F1-1) fathered 6 clutches with 13 offspring, the second largest male fathered 9 clutches with 28 offspring, and the small male had no offspring. In pond II, the three larger males fathered 6 clutches with 11 offspring and the two smaller males had no offspring (Figure 3).
Behavioral observations showed that the three males in pond I had relatively fixed habitat sites (caves), but there are significant differences in their territory size. The bigger male (F1-1), the | 9873 GONG et al.
There were obvious eating orders among the male turtles in pond I (χ 2 = 1,011.105, df = 4, p < .01). Normally, male F1-1, with the biggest body size, climbed to the food site first and ate. When F1-1 left, F1-2 began to eat, followed by F1-3. The females had no obvious feeding order and usually took their food to secluded corners to eat. In pond II, most of the males and females ate at the same time, with no obvious eating order. However, the males and females always avoided eating together at the feeding site and they usually took their food to secluded corners to eat.

| D ISCUSS I ON
Multiple paternity has been found in some turtle species (Pearse & Avise, 2001), and our genetic data provide the first evidence of the occurrence of multiple paternity in P. megacephalum. Across the turtle species studied, high variability exists both in the frequency of multiple paternity and in the relative paternal contribution of males (Duran, Dunbar, Escobar, & Standish, 2015;Pearse & Avise, 2001;Roques et al., 2004). The overall incidence of multiple paternity in P. megacephalum was 40% (6 of 15 clutches; Table 3). Given the small sample size, this estimate for the incidence of multiple paternity may not represent the actual situation for wild populations. In this case, the incidence of multiple paternity in P. megacephalum is similar to that found in Chrysemys picta (over 30%) (Pearse, Janzen, & Avise, 2002), but lower than in Clemmys insculpta (50%) (Pearse & Avise, 2001), Chelydra serpentina (66%) (Galbraith, White, Brooks, & Boag, 1993), Graptemys geographica (71%) (Banger, Blouin-Demers, Bulté, & Lougheed, 2013), and Podocnemis expansa (100%) (Valenzuela, 2000). The variability in the prevalence of multiple paternity in turtle species may be influenced by a variety of environmental factors, population characteristics, species behavior, and reproductive strategies. In this study, given the relatively small sample sizes, and lack of comparative data, make it difficult to determine which of these factors influence the prevalence of multiple paternity in

P. megacephalum.
Several case studies in turtles indicated that clutches sired by multiple males contained significantly more eggs than those sired by only one male, for example, Testudo graeca (Roques et al., 2004) and Chrysemys picta (Pearse et al., 2002). These findings suggest that either a higher incidence of multiple paternity occurs in larger clutches or multiple paternity is helpful in increasing egg production.
However, we did not find significant correlations between clutch size and multiple paternity in P. megacephalum.
Direct benefits of multiple paternity include provisioning of resources to the female and paternal care of offspring (Moller, 2000;Richardson, Burke, & Komdeur, 2002). Indirect benefits may include improving the genetic quality of the offspring (Jennions & Petrie, 2000;Yasui, 1998), increasing genetic diversity among offspring against environmental variation (Loman, Madsen, & Hakansson, 1988;Reed & Frankham, 2003), and improving genetic compatibility between mating pairs (Petrie, Doums, & Moller, 1998;Rubenstein, 2007;Tregenza & Wedell, 2000;Zeh & Zeh, 1996). Genetic incompatibilities may block egg development (Kempenaers, Congdon, Boag, Robertson, & Boag, 1999). To a given female, the variation of genetic compatibility among males may affect the hatching rate of clutches (Pearse & Avise, 2001). In turtles, without paternal care, it seems no direct benefits from the males can be given to the female in a multiple paternity system. Therefore, indirect genetic benefits may be the most likely driver of multiple matings for female turtles. In this study, there was no significant difference in hatching success between multiple-and single-paternity clutches of P. megacephalum, and similar results were found in two other species (Chrysemys picta, Pearse et al., 2002;Graptemys geographica, Banger et al., 2013) with close evolutionary relationships (Crawford et al., 2015). The mean PIC reflects genetic diversity, and there was no significant difference in the mean PIC between multiple-and singlepaternity clutches of P. megacephalum. However, we suggest that a larger sample should be used to verify these results, especially using candidate fathers with significant genetic differences.
Male hierarchies also influence mating systems in turtles. In some turtle species, for example, Gopherus agassizii (Berry, 1986), Chelydra serpentina (Galbraith, Chandler, & Brooks, 1987), and Clemmys insculpta (Kaufmann, 1992), size-based male hierarchies affect reproductive success. Males at the top of a hierarchy often occupy larger and high quality territories, and can mate with multiple females (Galbraith et al., 1993). Based on our observations in this study, size-based male hierarchies exist in P. megacephalum and the males at the top of the hierarchy gained more resources (e.g., territory, nesting area, food resources) in pond I. Obviously, when females move throughout the whole pond, males with higher hierarchy have a higher mating opportunity than smaller males. However, our results indicated that the highest-ranking males did not fertilize the most clutches, although the lowest ranking males showed no evidence of paternity in any of the clutches. This may have been caused by a combination of factors, including both male hierarchies and sexual selection.
Female choice may have an important influence on paternity in clutches of P. megacephalum. For a female, mating with a higher ranking male may give her the chance to share more resources. In some turtles (e.g., Chrysemys picta), female fecundity is related to size; larger females are more attractive to males seeking to maximize genetic fitness per mating (Pearse et al., 2002). However, in our study, there were no significant correlations between female body size and clutch size in P. megacephalum. Wei et al. (2016) found that a male P. megacephalum always avoided mating with a female larger than itself. Conversely, a female of P. megacephalum may prefer to mate with larger males. Therefore, female choice may also contribute to males at the top of the hierarchy having more mating opportunities.
However, females also have a cryptic choice (sperm competition), among male sperm in their reproductive tracts, which may also influence the paternal contributions to clutches (Birkhead, 1998;Pearse & Avise, 2001). Sperm from the highest-ranking males may not necessarily be the most competitive. This may explain why the highest-ranking males in our study did not fertilize the most clutches; however, insufficient data are available to draw conclusions on this matter.
The prevalence of multiple paternity in turtles can also be influenced by a variety of environmental factors, the density of breeding individuals, the sex ratio of sexually mature individuals (Lambertucci, Carrete, Speziale, Hiraldo, & Donázar, 2013), and mate encounter rates within a reproductive season (Boomer et al., 2013;Rossi Lafferriere et al., 2016). In this case, the turtles were studied in an mimicing natural habitat (closed environment), not a real natural habitat (open environment). The density of breeding individuals of P. megacephalum in this study was 1 turtle per 2 m 2 , F I G U R E 4 Sketch of the male turtes� territory, habitata sites, and feeding site in pond I. AI, AII, and AIII represent three regions; the bigger male (F1-1) can move in AI, AII, and AIII; the medium-sized male (F1-2) can move in AII and AIII; and the smaller male (F1-3) can move only in AIII; HS I, the habitata site of F1-1; HS II, the habitata site of F1-2; HS III, the habitata site of F1-3