Morphometry of two cryptic tree frog species at their hybrid zone reveals neither intermediate nor transgressive morphotypes

Abstract Under incomplete reproductive isolation, secondary contact of diverged allopatric lineages may lead to the formation of hybrid zones that allow to study recombinants over several generations as excellent systems of genomic interactions resulting from the evolutionary forces acting on certain genes and phenotypes. Hybrid phenotypes may either exhibit intermediacy or, alternatively, transgressive traits, which exceed the extremes of their parents due to epistasis and segregation of complementary alleles. While transgressive morphotypes have been examined in fish, reptiles, birds, and mammals, studies in amphibians are rare. Here, we associate microsatellite‐based genotypes with morphometrics‐based morphotypes of two tree frog species of the Hyla arborea group, sampled across a hybrid zone in Poland, to understand whether the genetically differentiated parental species also differ in morphology between each other and their hybrids and whether secondary contact leads to the evolution of intermediate or transgressive morphotypes. Using univariate approaches, explorative multivariate methods (principal component analyses) as well as techniques with prior grouping (discriminant function analyses), we find that morphotypes of both parental species and hybrids differ from each other. Importantly, hybrid morphotypes are neither intermediate nor transgressive but found to be more similar to H. orientalis than to H. arborea.

however, both lineages may rather form hybrid zones (Harrison, 1990;Maroja et al., 2015;Payseur & Rieseberg, 2016). In the latter ones, various recombinants over several generations present excellent systems to study natural organismal genomic interactions as a result of the evolutionary forces acting on certain genes and phenotypes (Abbott et al., 2013) and reflecting the dispersal of the animals (Barton & Hewitt, 1985). Hybrids may often exhibit intermediate phenotypes, compared to their parental lineages (Kierzkowski et al., 2011(Kierzkowski et al., , 2013Szymura, 1993), including poor adaptations to their ancestral ecological niches (Svedin et al., 2008;Vamosi et al., 2000).
In vertebrates, transgressive morphotypes have been examined in fish (Stelkens et al., 2009), reptiles (Robbins et al., 2014), birds (Campagna et al., 2018), and mammals (Boel et al., 2019;Larsen et al., 2010). Morphometric studies in amphibian hybrid zones have long been focusing on the variation in transects or rarely associating morphotypes and genotypes, but then, to our knowledge, mostly without investigating the occurrence of transgressive morphotypes (Babik & Rafiński, 2004;Fijarczyk et al., 2011;Gollmann et al., 1988;Kuchta, 2007). Here, we study an anuran hybrid zone by associating microsatellite-based genotypes and morphometrics-based morphotypes in tree frogs of the Hyla arborea group (sensu Faivovich et al., 2005) to understand whether secondary contact leads to the evolution of transgressive morphotypes.
At contact, both species, whose homologous X and Y sex chromosomes are undifferentiated (Stöck et al., 2011, exhibit restricted introgression at sex-linked compared to autosomal markers (Dufresnes, Majtyka, et al., 2016). Recent data from the Ukraine also suggest that both species also differ regarding their bioacoustics (Smirnov, 2013).
At first glance, H. arborea and H. orientalis appear morphologically virtually identical. They reach a snout-vent-length of about 5 cm, often have a lettuce-green backside and a whitish to yellowish belly. Likely by similar physiological mechanisms as related tree frogs of the Hyla japonica group (Kang et al., 2016), both species are able to change their color and pattern against visually heterogeneous backgrounds. Between dorsal and ventral parts, usually from the nostrils to the after, runs a dark brown to black lateral stripe that is often bordered by a lighter upper margin and forms a characteristic sinus in the hip area ( Figure 1).

revealed two parapatric cryptic species in
Poland, in the present paper, using a unique dataset of genotypically  Dufresnes, Majtyka, et al., 2016). Because some morphological characters might be sexually dimorphic and since the usually much more cautiously behaving tree frog females are much harder to catch, morphological analyses were here restricted to males. Animals were caught at night with a dip net or manually in breeding ponds, using a flashlight.
For genetic analyses, DNA samples (buccal swabs) were taken using cotton swabs (Broquet et al., 2007). Each frog was measured and photographed in a standardized manner (details below) at the collection site and then released immediately.  (Falush et al., 2003(Falush et al., , 2007Pritchard et al., 2000) served to detect patterns of genetic population structure. For the previously determined number of clusters (K = 2; Dufresnes, Majtyka, et al., 2016) and using default settings of the program (Pritchard et al., 2010), we ran ten simulations (algorithm Markov chain Monte Carlo, MCMC), each of which consisted of preliminary analysis (burn-in) with 10,000 steps and 100,000 steps for the main analysis. Tree frogs were considered as "confirmed" nuclear hybrids between H. arborea and H. orientalis only, if 90% credible intervals (CIs) of their ancestry coefficient neither reached 0 nor 1. This conservative approach allows confidently assigned individuals to be distinguished from those with uninformative genotypes (Dufresnes, Majtyka, et al., 2016). As a result of repeated backcrossing and thereby introgression of mitochondrial DNA, some hybrid tree frogs exhibited a cyto-nuclear discordance (equally termed in our paper "cyto-nuclear hybrids"). These possessed mtDNA belonging to one species, in our case H. arborea but a nuclear assignment of Q ≥ 0.900 to H. orientalis (Figure 2; Stöck et al., 2021), and based on the available microsatellites could not necessarily be considered as nuclear hybrids. Therefore, these "cyto-nuclear hybrids" may likewise not exhibit a hybrid morphological situation and thus were morphometrically also analyzed as a separate group, in several settings (see below and Stöck et al., 2021).

| Morphometric measurements
Eleven (six direct and five image-based) morphometric measurements were taken (for abbreviations and definitions: Table 1)

| Statistical analyses
Initially, all data were tested for homogeneity of variances and normal distribution. For each morphometric parameter, we calculated mean, maximum and minimum, and standard deviation and performed a one-way ANOVA and Tukey's multiple comparison tests (HSD). To test if groups could be separated without prior hypotheses on group membership, we conducted principal component analyses (PCA). PCAs were run on the correlation matrices. Individual F I G U R E 1 Measurements based on standardized photographs of tree frogs. (a) dorsal view; LC -width of the head, Dop-distance between the eyes, Do-Dn -distance between the eye and the nostril, Dn -distance between the nostrils, Lo -eye diameter; (b) right side of the body; Lo -diameter of the eye, Ltym -diameter of the tympanum, Do-Dtym -distance between the eye and the tympanum. See also

| Animals and groupings of parental species and hybrids
From the initially 199 male tree frogs caught, 8 exhibited uninformative genotypes (in 2 males mtDNA could not be amplified; in 6 potential nuclear hybrids 90% credible intervals of their ancestry coefficients reached 0 or 1 and thus were not confirmed; cf. Dufresnes, Majtyka, et al., 2016). Thus, the full dataset comprised 191 male To better understand how hybrid status as well as body size translate into potential morphometric differences, we analyzed the data in three different groupings: (i), including pure H. arborea, pure H. orientalis, and pooled hybrids (i.e., both those based on nuclear microsatellites and those only exhibiting mtDNA introgression); (ii), as above, but all individuals < 39 mm were left out, because they present immature juveniles, in which the adult morphotype might not be completely established (see also Discussion on allometry); and (iii), as in (ii) but further subdivided the hybrids into two sub-groups: those with a clear signature of nuclear hybridization (nuclear hybrids as in Figure 2c and d) and those only detected based on mitochondrial DNA introgression (cyto-nuclear hybrids as in Figure 2e).

| Descriptive statistics and one-way ANOVAs
Comparisons between H. arborea, H. orientalis and their hybrids revealed significant differences between the means of many of the parameters as calculated for each of the three groupings (i-iii; Table   S1a-c). Specifically, all types of hybrids (grouping (i)) differed from H. arborea for 6 characters: width of the head, length of femur, diameter of the tympanum, distances between the nostrils and the eye, and distance between the nostrils (LC, F, LTym, Do-Dn, and Dn), but only for three from H. orientalis: tibia length, foot length and distance between nostrils (T, LP, Dn) and this bias increased for adult hybrids (> 39 mm; grouping (ii)). Finally, for grouping (iii) with a subdivision of hybrids, nuclear hybrids differed from H. arborea in 5 characters: femur length, foot length, distance between eye and  Table 1) and 2nd axes (strongly loading: Ltym, Do-Dtym), and less pronounced differences between each of the two species and the group of pooled hybrids along the 2nd and 3rd axes (strongly loading: Do-Dn).
A PCA for specimens > 39 mm (grouping ii) yielded three components with eigenvalues > 1 (Figure 3d Although for grouping (i), H. orientalis were not distinguishable from pooled hybrids, specimens > 39 mm (grouping ii) could be well discriminated.
Likewise, subdivided hybrids (grouping iii) were re-classified considerably into H. orientalis (17%-18% for nuclear and cyto-nuclear hybrids) than into H. arborea (9%-4%), respectively. DFAs calculated based on parental species only likewise led to the reclassification of a majority of hybrids into H. orientalis (Tables S6a-c).

| DISCUSS ION
In the contact and hybrid zone of two Eastern-European hylids, we have morphometrically examined genotyped male tree frogs and their natural hybrids. Our analyses showed the morphotypes of both parental species (H. arborea, H. orientalis) and their various (pooled or subdivided) hybrids to differ, with the hybrid morphology tending to be more similar to H. orientalis than to H. arborea. Despite some potential influence of known allometric ontogenetic changes in anurans as well as in Hyla (Shrimpton et al., 2021), our tests involving

Measurement Definition
Dn distance between the nostrils Do-Dn distance between the eye and the nostril straight distance from the front edges of the eye to the nostril Do-Dtym distance between the eye and the tympanum closest distance from the rear edge of the eye to the front edge of the tympanum Dop the distance between the eyes distance between the nearest edges of the eyes

| Expectations and preconditions for evolution of transgressive phenotypes in the study system
Transgressive segregation requires quantitative trait loci with antagonistic effects (opposite to the direction of mean phenotypic variances) in the parental populations (Albertson & Kocher, 2005;Rieseberg et al., 2003). Stabilizing selection, genetic drift, or varying selective regimes in the evolutionary history may promote the evolution of transgressive loci, while consistent selective regimes may favor the accumulation of alleles with steady effects (Albertson & Kocher, 2005;Orr, 1998

| Remarks on the distinguishability of the cryptic species H. orientalis and H. arborea in the field
Morphological criteria alone may be, obviously, often be misleading reading the hybrid status of particular individuals (e.g., Babik et al., 2003;Lamb & Avise, 1987). Our data set allowed comparisons be- orientalis," the latter soon being synonymized with the former (Boulenger, 1898).
However, Bedriaga (1890), describing the two forms as the "varieties" orientalis and arborea, likewise found nearly no morphometric differences. He stated that foot length (LP) is roughly equal to the length of tibia (T); also tibia (T) and femur (F) were of similar lengths.
The ratio of foot to tibia length (F/T) by Bedriaga (1890)  Perrin for help with microsatellite genotyping in his laboratory.

CO N FLI C T O F I NTE R E S T
None.

DATA AVA I L A B I L I T Y S TAT E M E N T
Morphometric data for all parameters (as shown in Table 1)