Discussion
We identified three loci carrying sex-linked SNPs in agile frog populations in Hungary. Based on a genome sequence assembly of a closely related species, the common frog, we assume that the sex markers reported here cover a suitably large region of the sex chromosomes (Rds1 and Rds3 being at 112 million nucleotides from each other). Furthermore, genetic sex based on Rds3 corresponded to the sexual phenotype in 95% of all laboratory-raised individuals, and all discrepancies were found to be likely cases of sex reversal (as discussed in more detail below). Therefore, we conclude that parallel usage of the best performing markers Rds1 and Rds3 is suitable for molecular sexing in the North-Central Hungarian populations, yielding at least 95% confidence for individual sexing (allowing for the possibility that the 5% mismatch between Rds3 genotype and phenotypic sex had been caused by recombination; Ezaz et al., 2006; Perrin, 2009; Stöck et al., 2013) and good statistical power for comparing populations or experimental groups. Because amphibian sex determination can vary even within species (Miura, 2017; Rodrigues, Merilä, Patrelle, & Perrin, 2014), the reliability of our sex markers should be tested before applying them in other, especially distant populations (Lambert et al., 2016; Rodrigues et al., 2014). However, genetic diversity of the agile frog is in general very low across Europe (Vences et al., 2013), suggesting that our markers may be sex-linked in other agile frog populations as well. Thus, our genetic sexing method enables further studies on environment-induced sex reversal in this declining species, potentially throughout its distribution range.
According to our markers, 6 out of 125 laboratory-raised froglets were genetically females (XX) with male phenotype (testes), despite being raised under controlled conditions with presumably no sex-reversing effects. There are several potential explanations to consider for these mismatches. First, phenotypic sex might have been erroneously categorized; however, we can exclude this possibility because the phenotype based on gonad morphology was corroborated by histology in the mismatching individuals. Second, the presence of sex races could result in false assumption of sex reversal; for example, in the common frog, some individuals develop ovaries first that turn to testes later Rodrigues et al. (2015). This would cause overestimation of the proportion of XY females, which we did not find in our study at all. Third, the mismatches may have been due to recombination (Ezaz et al., 2006; Perrin, 2009; Stöck et al., 2013); however, 4 out of the 6 concerned froglets had XY siblings in our sample, suggesting that both Rds3 and Rds1 genotypes of chromosome Y were normal in their families. Furthermore, all mismatching individuals showed some signs of poor condition, and we are not aware of any reason why recombination would be associated with the deficits we detected. The fourth interpretation is that the mismatching individuals were indeed sex-reversed, which we consider most likely. Recent studies suggest that sex reversal may be a natural phenomenon in ectothermic vertebrates (Holleley et al., 2016; Lambert, 2015; Lambert et al., 2019), due to dosage-dependent sex determination where stochastic variation in gene expression levels may lead to sex reversal (Perrin, 2016). Alternatively, but not mutually exclusively, sex reversals may result not only from random variation but also from stressful stimuli, as experiments with fishes showed that various forms of physiological stress can induce sex reversal, and ”stress hormones” (activated by the hypothalamus-pituitary-interrenal glands axis) mediate this process (Castañeda Cortés et al., 2019; Fernandino, Hattori, Moreno Acosta, Strüssmann, & Somoza, 2013). Therefore, we suspect that a few of our lab-raised animals experienced relatively high levels of physiological stress despite the generally favourable lab conditions, and this led to sex reversal. Their developmental abnormalities may have been either the cause or the consequence of the stress that ultimately caused their sex reversal; in either case, our findings suggest that sex reversal can be associated with reduced health and poor fitness prospects. For example, enlarged spleen may indicate infections (Hadidi, Glenney, Welch, Silverstein, & Wiens, 2008), and small body mass predicts low chances of surviving the winter hibernation (Üveges et al., 2016) and low future reproductive success (Reading & Clarke, 1995; Vági & Hettyey, 2016).
Despite the above findings suggesting that sex-reversed individuals might have poor viability in nature, we found a relatively high number of sex-reversed adults in free-living agile frog populations. Genetically XX phenotypic males made up ca. 20% of phenotypic males, and ca. 35% of genetic females, although the latter rate of female-to-male sex reversal is probably overestimated because we had relatively low capture success (small sample size) for females. These numbers are relatively high compared to those reported for natural populations of two other frog species (Alho et al., 2010; Lambert et al., 2019). Interestingly, we found no difference in body mass between sex-reversed and normal adult males, despite the fact that some of the sex-reversed juveniles in the lab had seriously reduced body mass. This suggests that those sex-reversed individuals that survive to adulthood in nature may be able to mate, because male body size influences success in competition for mates (Vági & Hettyey, 2016). Their reproduction might still fail, however, if sex reversal reduces fertility, as reported in fish (Senior et al., 2012) and indicated by some of our findings with the lab-raised froglets, i.e. three sex-reversed juveniles had small testes and two of them had testicular oogonia (intersex). However, other findings of our study suggest that at least some of the sex-reversed individuals may be fertile. First, four out of six sex-reversed froglets showed normal testicular histology, and three of them had relatively large testes. Second, we found one family that was likely to be sired by an XX male: 12 laboratory-raised animals that were randomly chosen as eggs from a single clutch were all XX individuals, which would have a very low chance of happening merely by accidental sampling if the clutch had the theoretically expected 1:1 sex ratio (ca. 0.0002 probability). Sex-reversed individuals were found to be fertile in some ectothermic vertebrates (Devlin & Nagahama, 2002; Edmunds, McCarthy, & Ramsdell, 2000; Holleley et al., 2015), and in common frogs XX males appear to be fertile and as successful in mating as XY males (Alho et al., 2010; Veltsos et al., 2019). If sex-reversed individuals do reproduce in nature, the biased sex ratios of their progeny may lead to changes in the population sex ratio, sex-chromosome frequencies, and ultimately the sex-determination system (Bókony et al., 2017; Quinn et al., 2011; Wedekind, 2017). Furthermore, the offspring of sex-reversed individuals may themselves be more susceptible to sex reversal, as suggested by empirical results from lab experiments (Holleley et al., 2015; Shao et al., 2014).
We found higher female-to-male sex-reversal frequency in breeding populations exposed to anthropogenic land use. However, due to the availability of agile frog populations, our capture sites with different levels of anthropogenic land use were unequally distributed such that most sites West/South of the river Danube had little anthropogenic influence whereas most sites East/North of the Danube were highly anthropogenic (Figure S5). Also, the lab-raised animals that we used for validating the markers originated from three western populations (Table 1). Therefore we cannot exclude the possibility that the differences we observed in genotype-phenotype mismatches among the free-living populations were due to phylogenetic correlation, i.e. an inherited tendency for more frequent sex reversal in populations East/North of the Danube, or different patterns of linkage disequilibrium between our markers and the master sex-determination gene in these populations (e.g. higher recombination rate in the eastern populations). For example, in common frogs, sex-chromosome differentiation in Switzerland is mainly explained by a major alpine ridge separating the populations (Phillips, Rodrigues, Jansen van Rensburg, & Perrin, 2020). Alternatively, a mutation on chromosome Y at Rds3 (e.g. at the primer binding site, resulting in null allele) or a more complicated sex-determination system present on the East side could cause spatial genetic population structure (Oike et al., 2017; Rodrigues et al., 2014). These alternative explanations could be ruled out using phylogeographic information based on neutral autosomal loci in our populations, or by sexing lab-raised froglets from the eastern populations as well, but unfortunately such data are not available. However, in the region we studied, there are no high mountains or other likely geographical barriers to gene flow between these populations, because rivers like the Danube are not expected to be significant barriers for migration in species like the agile frog (Decout, Manel, Miaud, & Luque, 2012). Further, our study sites lie relatively close to each other, mostly within ca. 40 km; genetic structure at such small spatial scale in agile frogs is more likely the outcome of habitat fragmentation than isolation by distance (Lesbarrères et al. 2006, Sarasola-Puente et al. 2012). Highways pose migration barriers for agile frogs (Lesbarrères et al. 2006, Sarasola-Puente et al. 2012); however, the distribution of main roads and highways between our study sites is more likely to reduce migration along a North-South cline than separating East from West Figure S5). Taken together, we have little reason to expect an East-West population differentiation in our study. Our only population East/North from the Danube that had low anthropogenic land cover (pond ”B” in Figure S5) had high frequency of XX males; however, this pond was created from a closed quarry and was subjected to reconstruction works about a decade ago. In our study of chemical pollutants in anuran habitats in 2017, we found the highest concentration of phthalates in this latter pond (Bókony et al., 2018). Therefore, we believe that our results reflect a genuine effect of anthropogenic environmental change on sex-reversal frequencies.
Our results suggest that both urbanization and agriculture may contribute to the observed relationship between sex-reversal frequency and anthropogenic land use. Both kinds of anthropogenic habitats are polluted by various chemicals, many of which have demonstrated sex-reversing effects (Eggert, 2004; Hayes et al., 2002; Kloas, Lutz, & Einspanier, 1999; Nakamura, 2013; Reeder et al., 1998; Tamschick et al., 2016). Our result that sex reversal occurred even in the least anthropogenic habitats concurs with our earlier finding that those habitats are not devoid of chemical pollutants either (Bókony et al., 2018; see also Figure S5). Furthermore, the increased female-to-male sex-reversal rate that we found in urban agile frog populations may as well be due to the urban heat island effect which makes urban ponds warmer than rural ponds (Brans, Engelen, Souffreau, & De Meester, 2018), given that high temperature during larval development is a known inducer of sex reversal (Bókony et al., 2017; Chardard, Penrad-Mobayed, Chesnel, Pieau, & Dournon, 2004; Lambert et al., 2018). This variety of chemical, thermal, and potentially other stressors might complicate the relationship between sex-reversal rate and anthropogenic land use. In line with this, no correlation was found between sex-reversal frequency and urbanization along a forest-suburban gradient in green frogs (Rana clamitans ; Lambert et al., 2019), although the frequency of testicular oocytes was found to increase with urban land cover (Skelly, Bolden, & Dion, 2010). Similarly, several but not all studies found a positive association between agricultural land use and amphibian intersex (Orton & Tyler, 2015), laryngeal demasculinization (Zlotnik, Gridi-Papp, & Bernal, 2019) and reduced spermatogenesis (McCoy, Amato, Guillette, & St. Mary, 2017). These reports together with our results emphasise the need for further studies on sex-reversal frequency and its causes in wild populations of vertebrates with environmentally susceptible sex determination. Adult sex ratio has shifted towards males over the last decades in some amphibian species (Bókony et al., 2017), and this skew might be a consequence of sex reversals becoming more common due to anthropogenic environmental changes of land use and climate. The high frequency of female-to-male sex-reversal we found in this study suggests male-biased sex ratios and consequently reduced effective population sizes that might especially affect populations living in anthropogenic environments. Thus, we urgently need data on the survival and reproduction of sex-reversed individuals and their demographic effects on natural populations; developing novel sex markers for non-model species will be a key step in this endeavour.