Anuran swingers: misdirected mating attempts occurred early during anuran diversification

Abstract

Promiscuity, explosive breeding and male-biased operational sex-ratios can affect the strength of male selectivity and may play major roles in the expression of misdirected matings (with conspecific males, other species, corpses or objects) in anuran amphibians. Yet, misdirected amplexus occur in multiple species diverging from these reproductive traits, suggesting that the expression of such behaviour is widespread, and may have occurred early during anuran diversification. Using two methods of ancestral-trait reconstruction which predict unknown states, we found a very high probability that most anuran genera express misdirected amplexus, suggesting that this by-product of anuran reproductive strategies is likely to have occurred early during anuran diversification. The persistence of misdirected amplexus suggests that its infrequent occurrence may outweigh individual costs of breeding attempts with unfit mates. We found a recent exponential increase in reported observations of misdirected amplexus, which, along with increased research effort and publication rate, may reveal an effect of environmental perturbations known to promote the occurrence of these behaviours.

amphibians, ancestral state reconstruction, generalized hidden Markov model, heterospecific matings, hidden state prediction, reproduction

INTRODUCTION

In sexually reproducing organisms, a universal aspect of reproduction involves the ability to recognize conspecifics of the opposite sex to produce offspring and thus improve the chances of reproductive success and overall fitness (Emlen & Oring, 1977; Mayr, 2017). Although hybridization is possible between some species and can promote evolutionary diversification (Dowling & Secor, 1997), incompatibility between heterospecifics usually leads to reproductive failure (Hettyey & Pearman, 2003; Kudo & Kasagi, 2005). This generates strong selection to evolve traits that promote mechanisms of reproductive isolation for matings to occur exclusively between conspecifics (Winter & Rollenhagen, 1990; Rodríguez et al., 2004). Yet, heterospecific matings are often observed across diverse phylogenetic groups, and have various consequences, including species displacement, spatial or temporal segregation, changes in life history parameters, and reproductive character displacement (Gröning & Hochkirch, 2008). Studying these situations has offered opportunities for investigations into mate selection (Pfennig, 2007), mating signals (Wyman et al., 2011), genetic incompatibility (Kyogoku & Wheatcroft, 2020), reproductive isolation (Vega-Sánchez et al., 2022), reproductive interference (Kyogoku, 2015), hybridization (Hartke & Rosengaus, 2011) and the role of geographical distribution (i.e. allopatry vs. sympatry) in these evolutionary processes (Verrell, 1994; Ruokolainen & Hanski, 2016; Kyogoku & Kokko, 2020).

Heterospecific (misdirected) mating attempts have been commonly observed among anuran amphibians (Serrano et al., 2022). This is presumably due to the specific reproductive biology of this taxonomic group. Reproduction in anurans involves amplexus, the mating position in which a male grasps a female prior to and until the completion of oviposition and fertilization (Wells, 2007). Amplexus can last for a significant duration (up to several days, Wells, 1977, 2007) and are relatively conspicuous as compared to secretive or short matings that are difficult to observe in many other taxa. The conspicuousness of amplectant pairs in anurans has led to numerous observations of misdirected amplexus in this highly diverse amphibian lineage (Serrano et al., 2022). These misdirected amplexus can involve conspecific males (Mollov et al., 2010; Lago Londero et al., 2018) and anuran heterospecifics (both male and female individuals), but can also include numerous observations of amplexus with phylogenetically disparate species (e.g. caudates, fish, turtles, Simović et al., 2014; Zimić et al., 2018), dead individuals (i.e. necrophilia, Pintanel et al., 2021) and inanimate objects (e.g. Mollov et al., 2010; Sodré et al., 2018).

Both the geographical and phylogenetic uneven distributions of misdirected amplexus among anurans (Serrano et al., 2022) – although presumably linked to taxonomic bias in research intensity – have led to several hypotheses to explain the occurrence of such atypical behaviour (Serrano et al., 2022). These hypotheses involve the role of breeding strategies (explosive vs. prolonged breeding, Fouquet et al., 2021), the role of scramble competition (Hettyey & Pearman, 2003; Schmeller et al., 2005; Hettyey et al., 2014), the lack of or alteration of male–female or species discrimination (Schmeller et al., 2005; Cheong et al., 2008; Simmons & Narins 2018), the influence of other traits such as longevity or morphology (Schmeller et al., 2005), or, potentially, a combination thereof (Serrano et al., 2022). In such conditions, it would be advantageous for males to arbitrarily be attracted to – and thus clasp – any female-looking object in order to increase mating probabilities (Serrano et al., 2022). This is supported by the geographical distribution of species in which misdirected amplexus were more frequently recorded (Neartic and Palaeartic, Serrano et al., 2022). In these areas, suitable reproductive conditions are restricted in time, favouring explosive breeding and thereby promoting higher occurrence of indiscriminate amplexing behaviour (Pearl et al., 2005; Canestrelli et al., 2017). In turn, this may help to explain the phylogenetic distribution of misdirected amplexus, which occur more frequently in three anuran families (Bufonids, Ranids and Hylids, Serrano et al., 2022) in which explosive breeding and high levels of male–male competition are widespread (Pearl et al., 2007; Gaitonde et al., 2016; Bell et al., 2017).

Communal, explosive breeding strategies are likely to be a primary cause of misdirected amplexus. However, misdirected amplexus have been observed worldwide in species with contrasting breeding strategies, including those with prolonged breeding (e.g. Höbel, 2005; Ferreira et al., 2019) and asynchronized mating periods (Mollov et al., 2010), in which mating occurs individually on land or on trees rather than communally in water (Lu et al., 2016; Ferreira et al., 2019; Vásquez-Cruz et al., 2019), with amplexus initiated by female signals (e.g. Wells, 1977), and in which mate selection occurs through acoustic cues (Gerhardt, 1994; Nityananda & Bee, 2011). These exceptions have led to the intriguing possibility that the expression of misdirected amplexus is ubiquitous across anurans. In turn, this suggests that the expression of atypical misdirected amplexus – and thus conditions promoting this behaviour – may have occurred early during anuran diversification. Importantly, this idea does not invalidate previously highlighted hypotheses but rather roots these hypotheses more deeply at the basis of anuran phylogeny, prior to diversification processes that induced variations from the putative ancestral reproductive characteristics of this taxonomic group (Gomez-Mestre et al., 2012; Carvajal-Castro et al., 2020).

In this study, we tested whether misdirected amplexus occurred early during anuran diversification based on a comprehensive dataset of published observations in anurans using state-of-the-art ancestral state reconstruction approaches which allow unknown states to be predicted, because these data are relatively scarce across the whole anuran diversity. We predicted that, as observations of misdirected amplexus are distributed across all continents (see Serrano et al., 2022) and are observed in species with contrasting life history traits, this by-product of anuran reproductive strategies is likely to have occurred early during anuran diversification.

MATERIAL AND METHODS

Data collection

We used the dataset assembled by Serrano et al. (2022) along with missing references and additional data published since then (Supporting Information, Appendices A and B). We followed the same methodology as described by Serrano et al. (2022). Briefly, we searched articles mentioning misdirected amplexus using key words such as ‘frog’, ‘toad’, ‘amphibian’, ‘anuran’, ‘anurans’, ‘Anura’ and ‘Amphibia’ combined with ‘amplexus’, ‘abnormal’, ‘unusual’, ‘interspecific’ and ‘heterospecific’. We did not check for photographic records shared via social media, but did keep those assembled by Serrano et al. (2022). We deliberately excluded from our analyses (see below) records of necrophilia (Davian behaviour) and amplexus with inanimate objects as these events may represent extreme and atypical situations. From this list of species, we updated taxonomic identity using Jetz & Pyron (2018) and Frost (2021).

In total, our database included 280 records of misdirected (interspecific) amplexus, representing 159 species across 70 genera (Supporting Information, Appendices A and B).

Statistical analyses

The structure of our dataset led to several analytical constraints for ancestral state reconstruction. Indeed, the absence of published records of misdirected amplexus in numerous species does not necessarily mean that such atypical behaviour does not occur in these species. As a consequence, we could not univocally assess the absence of this behaviour in species in which misdirected amplexus has not been observed or published to date. This leads to the existence of unknown states in a large number of species. To take into account such complexity, we used up-to-date analytical approaches to compute ancestral states: the hidden state prediction (HSP) method using the package castor (v.1.7.3) (Louca & Doebeli, 2017), and the generalized hidden Markov model (HMM) method using the package corHMM (v.2.8) (Beaulieu et al., 2022), with an ER model [equivalent rates, corrected Akaike information criterion (AICc) = 15.277, compared to AICc = 16.475 for the all rates different (ARD) model], and no fixed nodes probability. Based on a collection of known characters in a limited number of species, these methods allow for unknown character states to be predicted at the tips of a phylogenetic tree as well as for intermediate nodes. For the same reasons (impossibility of obtaining univocal data demonstrating that a given species does not express misdirected amplexus), we set a fixed probability of 0 for ‘no misdirected amplexus possible’ in the few species (genera Adelphobates, Andinobates, Dendrobates and Oophaga) that are known to have lost amplexus for reproduction (Carvajal-Castro et al., 2020). Our rational was that if these species do not display amplexus during reproduction, they are arguably unlikely to display misdirected amplexus. Using this approach, a single conflict emerged for the genus Nanorana, which has been classified as ‘no amplexus’ in Carvajal-Castro et al. (2020) but for which we found a published observation of an amplexus (Lu et al., 2016). In this case, we kept species of this genus in the category ‘performing amplexus’.

We used TimeTree of Life (Kumar et al., 2022) to produce the phylogenetic tree used for our analyses. The anuran phylogenetic tree available in TimeTree of Life (accessed 15 April 2022) contained 3505 species (Kumar et al., 2022), from which 159 species have been observed completing misdirected amplexus. Thus, 20 species were classified as ‘no misdirected amplexus possible’ and 3324 species with unknown states. Consequently, we set the genus – rather than the species – as the taxonomic unit to perform our analyses. This approach allowed us to reduce the frequency of unknown states to 82.3% (N = 67 genera) rather than 94.8% using species as the taxonomic unit.

Using the tip states estimated from both methods, we illustrated the accumulation through time of lineages (lineages-through-time plots) in each of the mapped states, allowing us to determinate the time of diversification (Helmsestter et al., 2022), using the package phytools (v.1.1.7) (Revell 2012). We computed ER models, through selection via AICc (AICc = 20.783 compared to AICc = 22.767 for the ARD models), with 100 iterations.

All data analyses were performed using R 3.6.3 (R Core Team, 2020) and Rstudio v.1.1.419.

RESULTS

Ancestral trait reconstruction

Both statistical approaches (HSP and HMM) yielded similar results, but with small differences.

Using HMM, the probability that misdirected amplexus occurred at the ancestral node of the current anuran phylogeny was very high (0.998, Fig. 1). Overall, across all nodes in the phylogenetic tree, the mean probability that misidirect.ed amplexus occurred was also very high (0.962 ± 0.163). The few nodes for which this probability was relatively low were highly clustered (Fig.1) and corresponded to instances when species lacked amplexus for breeding (information that we used to set ‘no misdirected amplexus possible’, see Methods). These nodes represented 4.25% (N = 17) of all nodes, with probabilities that misidirected amplexus occurred ranging from 5.556E⁻⁷ to 0.419 (0.204 ± 0.168). All other nodes had probabilities > 0.973.

Figure 1.

Phylogenetic tree displaying the probability that misdirected amplexus occurred at each node and tip calculated using the HMM method. Genera of the species that have been observed in heterospecific matings appear in blue font. Genera of the species that are unlikely to be involved in misdirected amplexus appear in red font. Genera for which this information was unknown appear in black font.

Using HSP, the probability that misdirected amplexus occurred at the ancestral node of the current anuran phylogeny was very high (0.976, Supporting Information, Appendix C). Overall, across all nodes in the phylogenetic tree, the mean probability that misdirected amplexus occurred was also very high (0.944 ± 0.145). In contrast to the HMM method, the distribution of probabilities was multimodal. Very few nodes had a probability equal to 0 (N = 6, 1.5% of all nodes) and, similarly to the HMM method, these nodes were highly clustered (Supporting Information, Appendix C) and corresponded to instances when species lacked amplexus for breeding. There was a second group of nodes with intermediate probabilities ranging from 0.648 to 0.878 (0.770 ± 0.059, N = 63, 15.7% of all nodes, Supporting Information, Appendix C). All other nodes had probabilities > 0.946, among which 290 had probabilities equal to 1 (Supporting Information, Appendix C).

Lineage through time

This additional set of analyses was convergent with ancestral trait reconstruction. Using tip states estimated from the HMM approach, this analysis dated the appearance of misdirected amplexus at the oldest node of the phylogenetic tree we used (220 Mya, Fig. 2), while the loss of this character (i.e. the emergence of species that do not perform misdirected amplexus) was much more recent (60 Mya, Fig. 2). The results from the tip states estimated from the HSP method yielded similar results, except for the date of misdirected amplexus loss, which was estimated to have emerged more recently (30 Mya, Supporting Information Appendix D).

Figure 2.

Accumulation through time of lineages in each of the mapped states (misdirected amplexus or not) using tip states estimated from the HMM method. This analysis dated the appearance of misdirected amplexus at the oldest node of the phylogenetic tree we used (220 Mya), while the loss of this character (i.e. the emergence of species that do not perform amplexus) was much more recent (60 Mya). The vertical dashed line represents the expected transition to the no misdirected amplexus state.

DISCUSSION

Ancestral state reconstructions indicated that there is a high probability that most anuran species exhibit misdirected amplexus. More importantly, these analyses also indicated that these atypical mating attempts may have first occurred early during anuran diversification, suggesting that the conditions promoting this behaviour (i.e. explosive breeding and scramble male–male competition because of highly biased operational sex-ratios) may also have occurred in anurans prior to the remarkable diversification of reproductive modes in these taxa (Gomez-Mestre et al., 2012; Crump, 2015; Carvajal-Castro et al., 2020). In support of this finding, our analyses dated the emergence of this behaviour at the earliest node in our dataset (~220 Mya). Although it is plausible that the emergence of this behaviour may be older or younger given the uncertainties in our dataset, it is noteworthy that this period is situated after the split between anurans and caudates, but before the separation of Gondwana (San Mauro, 2010), which dovetails relatively well with the fact that misdirected amplexus have been observed worldwide (see Serrano et al., 2022). Taken together, these elements reinforce the notion that these atypical behaviours may have occurred early during anuran diversification, rather than being linked to multiple independent origins of a reproductive dead-end. Similarly, our results (HMM) suggest that the loss of misdirected amplexus occurred ~60 Mya, a period during which a simultaneous diversification of anuran clades occurred (Feng et al., 2017).

Previous studies have investigated ancestral reproduction modes of anurans. Gomez-Mestre et al. (2012) have demonstrated that aquatic breeding, egg-laying and larval development are all ancestral traits in anurans despite the high diversity of reproductive modes exhibited currently (e.g. terrestrial reproduction, direct development). More recently, Carvajal-Castro et al. (2020) have shown that amplexus has been conserved across the evolutionary history of anurans despite several evolutionary transitions in amplexus type. In combination with current hypotheses about the conditions that cause atypical misdirected amplexus (Schmeller et al., 2005; Serrano et al., 2022), our results provide further insights into reproductive modes of ancestral anurans. Indeed, if misdirected amplexus is promoted by explosive breeding with scramble male–male competition (Serrano et al., 2022), then our analyses suggest that such conditions were presumably basal to the anuran diversification. Unfortunately, it was not possible to test for covariation patterns between the occurrence of misdirected amplexus and other reproductive traits that may maintain such behaviour (i.e. explosive breeding, rival density, biased operational sex-ratio, aquatic breeding) because we could not robustly assess the univocal presence or absence of misdirected amplexus in groups for which this behaviour has not been observed to date. We believe that increasing natural history observations will be crucial to circumvent such a caveat and to formally assess the reproductive traits that promote heterospecific matings in anurans. Our dataset showed that the frequency of observations of misdirected amplexus in anurans has exponentially increased since the early 2000s (Supporting Information, Appendix E) which is promising to test for covariation patterns between the occurrences of misdirected amplexus and other reproductive traits. Photographic posts on social media or naturalist applications (such as iNaturalist) can play a key role in observations of unusual behaviours and help scientists to build more comprehensive datasets (Qi et al., 2019; Edwards et al., 2021).

Although hybridization can promote evolutionary diversification (Dowling & Secor, 1997) – a process that has been found in anurans (Kraus, 2015; Pabijan et al., 2020) – incompatibility between heterospecifics usually leads to reproductive failure (Hettyey & Pearman, 2003; Kudo & Kasagi, 2005). Why then is misdirected amplexus so widespread across anurans? We believe that the answer may lie in the relative individual vs. populational (or indeed species) costs of the reproductive failure induced by misdirected amplexus. Indeed, under the conditions that may promote misdirected matings (explosive breeding with scramble male–male competition), it would be advantageous for males to arbitrarily be attracted to – and then clasp – any female-looking object in order to increase mating probabilities (Serrano et al., 2022). In such circumstances, increased mating probabilities with conspecifics are likely to outweigh the individual costs of atypical indiscriminate breeding attempts with unfit mates (including other anuran species, phylogenetically disparate species, dead bodies or inanimate objects, Mollov et al., 2010; Simović et al., 2014; Sodré et al., 2018; Shin et al., 2020; Pintanel et al., 2021). For example, amplexus attempts with conspecifics of the same sex are often observed across anurans (Lago Londero et al., 2018; Mollov et al., 2010), a behaviour that we did not include in our analysis, and specific strategies (e.g. ‘release call’, Schmidt 1972) have evolved to avoid reproductively useless homosexual amplexus (Toledo et al., 2015). Overall, fierce competition in small water bodies during short breeding seasons is likely to promote relatively weak mate discrimination but, in turn, promiscuity induced by explosive breeding may decrease the costs of infrequent misdirected amplexus – and thus reproductive failure – at the population level. In addition, in areas where species promiscuity is high and misdirected amplexus are frequent, females have developed strategies to increase the likelihood of being released by a heterospecific male (but with partial loss of fitness, Hettyey et al., 2009) and/or to adjust mate preference to reduce hybridization risk (Bell & Zamudio, 2012; Calabrese & Pfennig, 2022). Theoretical modelling approaches could be useful to test for this hypothesis. Such an approach would allow us to test the relative contribution of mate and rival density (i.e. operational sex-ratio) but also the number of heterospecific species breeding simultaneously, and the duration of the breeding season (i.e. explosive vs. prolonged breeding).

There are some caveats to our investigations, notably linked to the fact that we based our analyses on sparse information. That is, we could not univocally assess the absence of misdirected amplexus in species in which such behaviour has not been observed to date. Thus, we had to rely on indirect information to identify species in which misdirected amplexus were unlikely to occur (i.e. species in which amplexus have been lost during diversification, Carvajal-Castro et al., 2020). In addition, we performed our analyses with a relatively low number of species in which misdirected amplexus have been observed relative to the whole anuran diversity [N = 159 over about 6700 species (Jetz & Pyron 2018), representing 2.4% of the anuran current diversity]. As a consequence, we used two different statistical approaches developed to infer unknown states at the tips and nodes of phylogenetic trees (HSP and HMM, see Methods). Despite the usefulness of these analytical tools to decipher evolutionary pathways when available data are scarce, we emphasize that the low number of taxa in which misdirected amplexus have been actually observed may have hampered our ability to detect putative independent transitions across the anuran phylogeny as those found for amplexus types (Carvajal-Castro et al., 2020). Despite these caveats, we think that the phylogenetic distribution of species in which misdirected amplexus have been observed across anurans (i.e. these observations were not clustered within restricted families or genus, Fig. 1; Supporting Information, Appendices A and C) as well as the geographical distribution of these observations (that have been found worldwide, Serrano et al., 2022) strengthen our results.

Recent studies have highlighted that current environmental perturbations linked to anthropogenic activities and associated global change (e.g. droughts, habitat modifications, anthropogenic noise, invasive species) may promote the occurrence of misdirected amplexus (Cory & Manion, 1955; Pearl et al., 2005; D’Amore et al., 2009; Canestrelli et al., 2017; Candolin, 2019; Simmons & Narins 2018; see Serrano et al., 2022). Thus, it is plausible that the recent exponential increase in the frequency of observations of misdirected amplexus in anurans (Supporting Information, Appendix E) is linked to such processes in combination with increased research effort and increased publication rate. If such an hypothesis holds true, anthropogenic perturbations of mate discrimination in reproducing anurans may ultimately have stronger impacts on the persistence of these endangered taxa.

To conclude, our results suggest that misdirected amplexus, which could lead to reproductive failure, are widespread across the anuran phylogeny and may have arisen prior to anuran diversification. Although we suggest that the persistence of such atypical behaviour during evolutionary time may have been possible because infrequent occurrence within populations may outweigh individual costs of indiscriminate breeding attempts with unfit mates, the current exponential increase in this behaviour may reveal a worrisome effect of environmental perturbations linked to global change. Increased investigations and observations of these behaviours (either in the scientific literature or through citizen science) could allow these different hypotheses to be tested.

SUPPORTING INFORMATION

Additional supporting information may be found in the online version of this article on the publisher’s website.

Appendix A. List of species that have been observed in misdirected amplexus and corresponding literature sources. Only species that mistakenly amplected other species are listed and used in our analyses.

Appendix B. References used for Appendix A.

Appendix C. Phylogenetic tree displaying the probability that misdirected amplexus occurred at each node and tip calculated using the HSP method. Genera of the species that have been observed in heterospecific matings appear in blue font. Genera of the species that are unlikely to be involved in misdirected amplexus appear in red font. Genera for which this information was unknown appear in black font.

Appendix D. Accumulation through time of lineages in each of the mapped states (misdirected amplexus or not) using tip states estimated from the HSP method. The vertical dashed line represents the expected transition to the no misdirected amplexus state.

Appendix E. Number of studies reporting heterospecific misdirected amplexus in anuran amphibians relative to the year of publication. Statistical analysis (quasi-Poisson GLM) indicates a recent exponential increase of misdirected amplexus observations (Estimate = 0.0597, SE = 0.0085, t  = 7.027, P < 0.001).

Acknowledgments

We thank the CNRS for funding. LLS was supported by a PhD grant from La Rochelle Université. We warmly thank Jean-Pierre Vacher for discussions on the hypothesis tested in this paper, as well as for sharing his expertise on ancestral state reconstruction and on anuran natural history. We thank Filipe C. Serrano, Juan C. Díaz-Ricaurte and Marcio Martins for making their dataset available, which was very helpful to complement our initial collection, and two anonymous referees for providing insightful comments.

Data Availability

The data are available as supporting information (appendices A and B).