Cryptic Triploids and Leaky Premating Isolation in an Odontophrynus Hybrid Zone

: The diploid Odontophrynus cordobae and its autopolyploid counterpart O. americanus (4n) co-occur in a small-sized contact zone in Central Argentina, together with numerous specimens of a cryptic triploid taxon. Additionally, we monitor another ﬁve localities inhabited by taxa of this species complex within a radius of 76 km north and south of the contact zone to look for possible co-occurrence of taxa. In this study, we analyze the reproductive interactions in this breeding assemblage covering three levels of ploidy. We focus on the advertisement call structure as a powerful mechanism of premating isolation and on the mating preferences of males and females in the natural habitat. Advertisement calls of triploids were indistinguishable from those of tetraploids, and both differed signiﬁcantly in pulse rate from diploids. Analyses of 21 amplexi demonstrated that heterospeciﬁc matings dominated breeding: one between a diploid female and tetraploid male, and ten between triploids and tetraploids. At three localities with syntopic diploids and tetraploids, premating isolation was almost perfect, preventing the formation of triploids. Therefore, we question the putative origin of triploids from hybridization and discuss alternatives. This unique system, including three bisexual taxa of distinct ploidy, which interact reproductively, remains to be investigated in more detail to fully understand the mechanism stabilizing its persistence.


Introduction
In Amphibia, 56 out of the currently known 8426 species are polyploid [1,2]. The number of chromosome sets usually rises in even number, as exemplified in the anuran genus Xenopus with diploid, tetraploid, octoploid, and dodecaploid species [3]. Genome duplication may result from autopolyploidy, like in tetraploid Odontophrynus americanus, or from allopolyploidy, like in many tetraploid toads Bufotes spp. [4][5][6][7][8][9][10]. Polyploidy can lead to post-reproductive isolation from the diploid parental species because hybrids are often inviable or infertile [11]. Bufotes toads are the exception from the rule, being presently the only known amphibian species complex that includes diploid, triploid, and tetraploid bisexually reproducing taxa [8,9]. Triploids coexist sympatrically with diploid or tetraploid taxa in Kyrgyzstan, where all three ploidy levels interact reproductively [8]. The tetraploid taxon involved is the allopolyploid B. pewzowi (B. lastastii ♂× B. perrini ♀), the diploid one B. perrini (taxon names according to [2,12]). Triploid males-females are absent-are most probably hybrids from 2n mothers and 4n fathers, i.e., their chromosome sets combine two from B. lastastii and one from B. perrini, but if they are fertile, an origin through backcrosses with 2n females may be possible as well [8,12]. The advertisement calls of triploids were significantly distinct from the diploid parental species but indistinguishable from those of Prediction 1. Hybridization due to natural mating between the diploid O. cordobae and tetraploid O. americanus is rare to absent. Rationale: we expect that the significant difference in advertisement call structure allows the females to distinguish between conspecific and heterospecific males. Effective premating isolation based on female choice may be undermined by males that clasp coincidentally heterospecific females in the three localities with syntopic populations [18]. Character displacement in pulse rate may increase the interspecific differences in these localities [26]. Prediction 2. Advertisement call structure of triploids resembles more O. americanus than O. cordobae. Rationale: analogous to Bufotes and Hyla taxon groups of three ploidy levels, we expect polyploids irrespective 3n or 4n to call with lower pulse rates as diploids [8,9,16]. Prediction 3. Frequent heterospecific matings occur between O. americanus and triploid specimens, but not with O. cordobae. Rationale: we expect that premating isolation among specimens >2n ploidy levels is leaky due to the assumed similarity of advertisement calls in polyploid taxa.

Study Localities and Sampling Design
We studied ploidy, external morphology, and advertisement call features of Odontophrynus specimens (americanus-group) in six localities in the province of Cordoba, Argentina ( Figure 1, Table 1). The populations were situated along a 125 km transect beginning at the type locality of diploid Odontophrynus cordobae (Villa General Belgrano) in the north and ending at a population of tetraploid O. americanus in Rio Cuarto [18]. Between these localities marking the geographical extremes of the transect was the area of contact between O. cordobae and O. americanus, where recently triploid specimens were discovered [20]. Study sites were visited at least once per year during the breeding period of Odontophrynus spp (November to March), starting in 1998 at Villa General Belgrano and Rio Cuarto, in 2004 at Rio de los Sauces, El Cano, and Coronel Baigorria, and in 2014 at La Escondida. these localities marking the geographical extremes of the transect was the area of contact between O. cordobae and O. americanus, where recently triploid specimens were discovered [20]. Study sites were visited at least once per year during the breeding period of Odontophrynus spp (November to March), starting in 1998 at Villa General Belgrano and Rio Cuarto, in 2004 at Rio de los Sauces, El Cano, and Coronel Baigorria, and in 2014 at La Escondida.  The sampling protocol at each locality began with recording the advertisement calls of the males, followed by the collection of the recorded male for ploidy determination and morphometric measurements. Ploidy was determined by erythrometry based on a blood smear; individuals were released again in situ following measurements. In these surveys, we obtained the complete data set on reproductive males.
There were some complementary surveys, too, in which we recorded exclusively advertisement calls or captured Odontophrynus present at a breeding site independent of gender or calling activity, yielding either bioacoustic data or morphometric and ploidy data. The analyses of mating preferences at the La Escondida locality were based on two surveys, in which we encountered several pairs in amplexus: N = 13 on 6 November 2019, and N = 8 on 5 February 2020. Pairs were collected, placed separately in plastic containers, and finally processed by taking a blood sample and measuring morphometric features. All pairs were released again in situ and were observed to continue reproduction. Our field study was authorized by the Environmental Secretary of the Córdoba Government (A01/2013).

Recording and Analysis of Advertisement Calls
Advertisement call series were recorded in situ using a Walkman Digital AudioTape (DAT) SonyTM TCD-100 with stereo microphone ECM-MS907 SonyTM and a High Resolution Flash Recorder MarantzTM PMD671 with microphone SennheiserTM K6/ME66. Throughout each recording, the microphone was held at around 1 m distance from the focal calling male. Immediately after recording, water and soil temperatures (to the nearest 0.1 • C) were registered at or near the calling site. Please note that not all recorded males could be captured.
Stereo recordings were converted to mono at a sampling rate of 44.1 kHz and resolution of 16 bits, and subsequently, sonograms and oscillograms were prepared using Raven Pro 1.6 (Cornell University, Ithaca, NY, USA) and ADOBE Audition 1.0. Sonograms (Adobe Systems Inc., München, Germany), and frequency analyses were obtained by applying Blackman-Harris Fast Fourier transformation with an FFT size of 1024 Hz. Call structure was characterized by measuring six variables: (1) call duration (ms), (2) pulses per call, (3) pulse rate (pulses/s), (4) pulse duration (ms), (5) interpulse interval (ms), and (6) dominant frequency (Hz) [18,27]. We measured 3-6 single calls per advertisement call series emitted by the same frog. To avoid the introduction of intra-individual call variation, we used the average of all single measurements per call series as the representative of an individual male for further data analysis.

Erythrometric Ploidy Determination and Morphometry
Blood samples were collected from the vena angularis to prepare microscope slides with air-dried smears of fresh blood [28]. Photographs of blood cells were obtained by using an Axiophot Microscope (Carl Zeiss, Jena, Germany) at a magnification of 1000× with a Canon Powershot G6 Digital Camera (Canon, Krefeld, Germany) and ZoomBrowser EX (Canon, Krefeld, Germany)and saved in TIFF files. We measured the length (L) and width (W) of 40 randomly chosen erythrocytes and their respective nuclei for all individuals using ImageJ software (Version 1.0, Wayne Rasband, NIH, freeware) [29]. Cell and nuclear area were calculated assuming an ellipsoid shape (L × W × π/4). Ploidy levels were estimated based on the size of the nuclear area of the erythrocytes [29,30]. In 63 randomly chosen individuals, erythrometric ploidy determination was corroborated using chromosome set analyses. These individuals were transported to the laboratory and euthanized for karyological ploidy determination [31].

Statistical Analyses
All variables were first tested for normality. If data were skewed, they were normalized using log10-transformation. Descriptive statistics included arithmetic means, corresponding standard errors, and 95% confidence intervals.
Snout-vent length was tested for sexual dimorphism and taxon-specific difference in an analysis of variance (two-factor ANOVA, type III sums of squares) using sex and taxon as  Table S1. A principal component analysis (PCA) was used to detect potential differences in shape among taxa without a priori delimitation of groups. The extraction criterion for principal components was an eigenvalue >1. In morphometric data sets, PC1 represents size, and PC2 and PC3 represent size-independent aspects of shape. A discriminant analysis was used to test for morphometric differentiation among the taxa, which were a priori delimited. Variables were standardized by dividing the difference between value and arithmetic mean by the standard deviation. We considered a rate of correct classification of at least 80% as an indicator of significant morphological differentiation.
Advertisement call variation without a priori delimitation of groups was explored using a factor analysis on the standardized call and temperature variables. The complete bioacoustic data set is given in the supplementary Table S2. Extracted principal components were submitted to an orthogonal VARIMAX-rotation to yield factor loading of original variables close to 1 (strong association) or 0 (no association). A discriminant analysis was used to test for bioacoustic differentiation among the taxa, which were a priori delimited. Again, we considered the criterion of 80% correct classification as an indicator of significant call differentiation. The six call variables (see Section 2.2) recorded in allopatric and sympatric populations of each taxon were compared using an ANCOVA with population type as a fixed categorical factor and soil and water temperature as continuous co-variables to control for the impact of temperature on call structure.
Significance level was set at alpha = 0.05. All statistical procedures were performed using the program package Statgraphics Centurion version 18.1.01 (64-bit; Statgraphics Technologies, Inc., The Plains, VA, USA).

Odontophrynus Species Composition at the Study Localities
The taxon composition of Odontophrynus specimens collected or recorded during advertisement at the six study localities varied considerably along the transect ( Table 2). The northernmost breeding assemblage at Villa General Belgrano included exclusively O. cordobae, the two southernmost ones at Coronel Baigorria and Rio Cuarto exclusively O. americanus. Species identification was based on advertisement call features, ploidy assessment using erythrocyte features, and occasionally karyology. At the three central localities of the transect, the two species coexisted in varying proportions. Exclusively at the La Escondida locality, we detected additional triploid specimens. We did not attempt to quantitatively assess the abundance of taxa at mixed sites.
Principal component analysis of the standardized morphometric variables yielded three PCs with an eigenvalue >1 that explained 54.6% of total variance (Figure 2A). The size-related PC1 (eigenvalue = 4.46) accounted for 29.8% of total variance, the shape-related PC2 (eigenvalue = 2.70) and PC3 (eigenvalue = 1.02) for 18.0% and 6.8%, respectively. There was no detectable segregation of taxon groups with respect to PC2 and PC3.

Advertisement Call Features of Odontophrynus Taxa
Advertisement calls of diploid O. americanus and tetraploid O. cordobae consisted of a single pulse group, with pulses located at the beginning of the call having low intensity, followed by increasing intensity pulses, and a decrease in pulse intensity at the end of call ( Figure 3A,B). Pulse duration was shorter in O. cordobae and the number of pulses per call greater than in O. americanus, resulting in a higher pulse rate as a distinguishing call feature at any ambient temperature. The calls of triploid males resembled in all aspects the call of O. americanus ( Figure 3C). Discriminant analysis of the same data set yielded a single marginally significant discriminant function (eigenvalue = 0.17, canonical correlation = 0.3783, X 2 = 45.7, df = 30, p = 0.0329; Figure 2B). The second function did not contribute significantly to group separation function (eigenvalue = 0.05, canonical correlation = 0.2144, X 2 = 10.7, df = 14, p = 0.7112). The rate of correct classification was low in all taxa: 40.3% in O. cordobae, 41.6% in O. americanus, and 56.1% in the triploids. The misidentified specimens were equally distributed over the other two taxa. The three syntopic Odontophrynus taxa were indistinguishable using size-and shape-related morphometric characters.

Advertisement Call Features of Odontophrynus Taxa
Advertisement calls of diploid O. americanus and tetraploid O. cordobae consisted of a single pulse group, with pulses located at the beginning of the call having low intensity, followed by increasing intensity pulses, and a decrease in pulse intensity at the end of call ( Figure 3A,B). Pulse duration was shorter in O. cordobae and the number of pulses per call greater than in O. americanus, resulting in a higher pulse rate as a distinguishing call feature at any ambient temperature. The calls of triploid males resembled in all aspects the call of O. americanus ( Figure 3C). To explore the range of variation in the advertisement calls without a priori delimitation of taxon groups, we applied a factorial analysis to the data set, including six call parameters and two environmental temperature variables. There were three statistically independent factors with an eigenvalue >1, which explained 79.5% of total variance. Factor 1 (eigenvalue = 1.71, accounting for 21.4%) was significantly loaded by soil and water temperature, call duration, interpulse interval, and pulse rate, i.e., representing the temperature-sensitive structure of the call. Factor 2 (eigenvalue = 3.44, 43.0%) was significantly loaded by the number of pulses per call, pulse duration, and again pulse rate, i.e., representing the temperature-independent structure of the call. Factor 3 (eigenvalue = 1.21, 15.1%) was loaded exclusively by the dominant frequency. The factor scores of diploid O. cordobae individuals formed one homogeneous group, whereas those of the tetraploid O. americanus, together with those of the triploid individuals, formed an independent second one ( Figure 4A). To explore the range of variation in the advertisement calls without a priori delimitation of taxon groups, we applied a factorial analysis to the data set, including six call parameters and two environmental temperature variables. There were three statistically independent factors with an eigenvalue >1, which explained 79.5% of total variance. Factor 1 (eigenvalue = 1.71, accounting for 21.4%) was significantly loaded by soil and water temperature, call duration, interpulse interval, and pulse rate, i.e., representing the temperature-sensitive structure of the call. Factor 2 (eigenvalue = 3.44, 43.0%) was significantly loaded by the number of pulses per call, pulse duration, and again pulse rate, i.e., representing the temperature-independent structure of the call. Factor 3 (eigenvalue = 1.21, 15.1%) was loaded exclusively by the dominant frequency. The factor scores of diploid O. cordobae individuals formed one homogeneous group, whereas those of the tetraploid O. americanus, together with those of the triploid individuals, formed an independent second one ( Figure 4A).  Based on the data set including the studentized residuals of the six temperature-adjusted call parameters, a discriminant analysis was performed to maximize differences among the three a priori delimited Odontophrynus taxon groups. We obtained two significant discriminant functions (DF1: eigenvalue = 3.59, canonical correlation = 0.8846, Χ² = 377.8, df = 12, p < 0.0001; DF2: eigenvalue = 0.1, canonical correlation = 0.3186, Χ² = 24.8, df = 5, p = 0.0002). The rate of correct classification was 100% in O. cordobae, but considerably lower in O. americanus (64.0%) and in the triploids (73.7%; Figure 4B). The misidentified specimens were O. americanus confused with triploids and vice versa.  Based on the data set including the studentized residuals of the six temperatureadjusted call parameters, a discriminant analysis was performed to maximize differences among the three a priori delimited Odontophrynus taxon groups. We obtained two significant discriminant functions (DF1: eigenvalue = 3.59, canonical correlation = 0.8846, X 2 = 377.8, df = 12, p < 0.0001; DF2: eigenvalue = 0.1, canonical correlation = 0.3186, X 2 = 24.8, df = 5, p = 0.0002). The rate of correct classification was 100% in O. cordobae, but considerably lower in O. americanus (64.0%) and in the triploids (73.7%; Figure 4B). The misidentified specimens were O. americanus confused with triploids and vice versa. The temperature-adjusted call variables pulse rate and pulse duration were sensitive to the syntopic presence of other taxa ( Figure 5). In O. cordobae, syntopy with O. americanus alone or with additionally triploid increased significantly pulse duration compared to allopatric populations (ANCOVA with multiple average comparisons: p < 0.05). Pulse rate increased exclusively in the presence of triploids over the level of allopatric populations Diversity 2022, 14, 305 9 of 13 (ANCOVA with multiple average comparisons: p < 0.05). In O. americanus, the described response pattern of O. cordobae was reversed. The parameter level of triploids did not differ significantly from those of syntopic O. americanus but did so compared to allopatric populations of the tetraploid (ANCOVA with multiple average comparisons: p < 0.05). (ANCOVA with multiple average comparisons: p < 0.05). In O. americanus, the described response pattern of O. cordobae was reversed. The parameter level of triploids did not differ significantly from those of syntopic O. americanus but did so compared to allopatric populations of the tetraploid (ANCOVA with multiple average comparisons: p < 0.05).

Figure 5.
Sensitivity to the presence of one or two syntopic Odontophrynus taxa in temperature-adjusted call parameters pulse rate (A) and pulse duration (B). Each dot represents the least square mean and its corresponding 95% Bonferroni confidence interval, calculated using an ANCOVA on the log10-normalized response variables.
Local variations of the temperature-adjusted call parameters call duration and pulse interval did not show effects of sympatry with other Odontophrynus taxa.

Mating Preferences at the La Escondida Contact Zone
Homotypic pairing constituted 10 out 21 amplexi analyzed (Table 3). Yet, almost perfect premating isolation among taxa was limited to O. cordobae and O. americanus, and O. cordobae and triploids, respectively. We observed a single mismating of an O. americanus male with an O. cordobae female. In contrast, triploid specimens were more often found in amplexus with O. americanus than each taxon in homotypic pairings.

Discussion
We provide evidence that Odontophrynus toads of three ploidy levels interact reproductively in their natural habitat, the La Escondida locality in Central Argentina. This is the first example in an autopolyploid anuran species complex, and the second one for Figure 5. Sensitivity to the presence of one or two syntopic Odontophrynus taxa in temperatureadjusted call parameters pulse rate (A) and pulse duration (B). Each dot represents the least square mean and its corresponding 95% Bonferroni confidence interval, calculated using an ANCOVA on the log10-normalized response variables.
Local variations of the temperature-adjusted call parameters call duration and pulse interval did not show effects of sympatry with other Odontophrynus taxa.

Mating Preferences at the La Escondida Contact Zone
Homotypic pairing constituted 10 out 21 amplexi analyzed (Table 3). Yet, almost perfect premating isolation among taxa was limited to O. cordobae and O. americanus, and O. cordobae and triploids, respectively. We observed a single mismating of an O. americanus male with an O. cordobae female. In contrast, triploid specimens were more often found in amplexus with O. americanus than each taxon in homotypic pairings.

Discussion
We provide evidence that Odontophrynus toads of three ploidy levels interact reproductively in their natural habitat, the La Escondida locality in Central Argentina. This is the first example in an autopolyploid anuran species complex, and the second one for Amphibia [8]. While in the Bufotes pewzowi-breeding system exclusively triploid males are involved, the Odontophrynus-breeding system reported in this study is more complex, with triploid males and females mating with each other and with the tetraploid O. americanus. Heterospecific amplexi of triploids with the syntopic diploid O. cordobae were never observed. Consequently, our findings support the hypothesis that ploidy-dependent advertisement call structure shapes the reproductive interactions at the contact area of the three taxa. We highlight support in detail by contrasting bioacoustic analyses and field observations with the three predictions. As previously reported, the temporal structure of the advertisement calls differs mainly in the pulse rate because the pulse duration is significantly shorter in the diploids than in the tetraploids at a similar call duration [18,26,27]. The contrast between these features remained stable in syntopy, allowing for a highly effective discriminatory mate choice of phonotactically approaching females. Consequently, mismatings between diploids and tetraploids were very rare. In fact, the only heterospecific amplexus including a diploid and a tetraploid was between an O. cordobae ♀and an O. americanus ♂. As reproductive males counteract female choice by clasping actively any individual of appropriate size and shape that they notice in their vicinity, the heterospecific amplexus was most probably the result of rape rather than of female choice. As the Odontophrynus adults of all ploidy levels do not differ from each other in external morphology, the chance of erroneous mate choice of males increases. Mismatings between the morphologically similar diploid-tetraploid treefrogs Hyla chrysoscelis and Hyla versicolor support the case [34].

Prediction 2: Advertisement Call Structure of Triploids Resembles More O. americanus than O. cordobae
In agreement with call analyses in triploids of the Bufotes species group and of Hyla, triploid Odontophrynus produced advertisement calls that were indistinguishable from that of O. americanus [13,15,17,35,36]. A trend for larger cells in the midbrain auditory centers and larger spinal motor neurons of polyploids has been proposed to account at least partly for the different call structures of diploids, but current evidence is not unequivocal [14]. Whatever the proximate cause for the alteration of call structure from diploids to polyploids may be, the amazing similarity of advertisement calls of triploid and tetraploid Odontophrynus undermines effective premating isolation by female phonotaxis between these two ploidy levels. As the viability of triploids is like that of the parental species [25], unavoidable heterospecific matings may result in a significant genetic sink for the O. americanus population.

Prediction 3: Frequent Mismatings Occur between O. americanus and Triploid Specimens, but Not with O. cordobae
In fact, heterospecific amplexi between triploids and tetraploids constituted almost half of all amplexi analyzed, corroborating their incapacity to identify and choose a mate of the same ploidy. The high frequency of heterospecific amplexi and of homotypic triploid pairings raises the question of whether triploids are fertile. In the Bufotes species group, there are three triploid bisexual species (B. baturae, B. pseudoraddei and B. zugmayeri) that are fully fertile [6,12,37]. Presently, we do not have direct evidence for the fertility of the Odontophrynus triploids, but ongoing research focuses on this issue.

Putative Origin of Triploids and Their Local Long-Term Persistence
In this study, we identified three localities with syntopic occurrence of O. cordobae and O. americanus, but in only one we detected triploids. Hybridization of the paternal species occurs but is most probably too rare to account for the long-term persistence of numerous triploids at the La Escondida locality. Thus, hybridization is one possible explanation of the origin. Alternatively, triploidization may have been a spontaneous event, resulting from problems in the disjunction of chromosomes in gamete formation or in the first zygotic division [24]. A third possibility is pollution-induced effects on meiosis, which could result from the intensive local application of agrochemicals at the La Escondida area [38].
Whatever has caused the initial emergence of triploid Odontophrynus, the only plausible explanations for their continued presence in large numbers are unaffected viability and full fertility. Viability comparable to that of the paternal species has already been proven [25]. The second prerequisite for persistence remains to be investigated. If we postulate full fertility of triploids, gametogenesis is expected to produce haploid and diploid gametes. Offspring generated by backcrosses with O. americanus would yield partly the paternal species and partly triploids. Offspring generated from pure triploid pairing would additionally produce diploids, i.e., O. cordobae. In fact, the pulse rate of the advertisement calls given by the La Escondida diploids was significantly higher than that recorded at any other locality with O. cordobae. A possible cause could be character displacement, but alternatively may reflect differences from the gene pool of O. cordobae from the type locality, if specimens result from triploid mating rather than from dispersal of roaming individuals. Ongoing research will focus on this aspect as well.

Conclusions
The occurrence of numerous triploid specimens inhabiting an area of a few square kilometers together with diploid O. cordobae and tetraploid O. americanus is unique for an autopolyploid species complex. Reproductively interacting anurans of three ploidy levels were reported before exclusively in allopolyploid B. pewzowi-species complex [8]. While there is no doubt about the phenomenon itself, many open questions remain to be resolved before reaching a rudimentary understanding of interactions. Among the most urgent topics of future research are the process(es) leading to triploid formation, the assessment of fertility, and the mechanisms of gametogenesis. Another issue is the surprising geographic restriction of triploids to a single small-sized area, though the area of contact between O. cordobae and O. americanus is much larger. In conclusion, much remains to be learned from this peculiar species complex.