Study species, sampling sites, and taxon assignment

All European water frogs were collected in the upper Oder River drainage (Central Europe; Czech Republic; Fig. 1, Table 1). In total, 249 individuals from all three Pelophylax taxa were caught with a hand net at 16 locations during years 2002 and 2008. Pelophylax kurtmuelleri from Greece, a sister species of European P. ridibundus, was used as an outgroup in a phylogenetic tree construction. Males were distinguished from females by the presence of vocal sacs and nuptial pads. Taxon affiliation was determined on the basis of external morphological characters [22, 32], and identification was later verified genetically with allozyme markers (see the “Protein electrophoresis” section). Tissue samples obtained from the finger tips or muscles were stored at − 20 °C for allozyme analyses and in 96% ethanol for DNA analyses.

No.*	Pop. type	Sampling site	Latitude Longitude	Genotype	No. of f/m/juv	MLG type	No. of MLGs	Hemiclone ID
1	R-E male	Albrechtičky	49.708056 18.098889	RR	8/9/-	RR	17
				RL	-/12/-	R	12
						L	1	OderL1
2		Bílovec	49.769722 18.032222	RR	1/-/-	RR	1
				RL	-/2/-	R	2
						L	1	OderL1
3		Darkovice	49.757778 18.213056	RR	3/5/-	RR	8
				RL	-/4/-	R	4
						L	1	OderL1
4		Dolní Benešov	49.912222 18.120000	RR	4/2/-	RR	6
				RL	-/4/-	R	4
						L	1	OderL1
5		Ostrava	49.858889 18.265278	RR	5/4/-	RR	9
				RL	-/5/-	R	5
						L	1	OderL1
6	L-E	Bravantice	49.766944 18.084444	LL	1/-/-	LL	1
				RL	7/-/-	R	NA
						L	7
7		Břidličná	49.916389 17.359444	LL	-/4/-	LL	4
8		Český Těšín	49.764444 18.591667	LL	3/7/-	LL	10
				RL	5/-/-	R	NA
						L	5
9		Dobrá	49.676667 18.392222	LL	7/22/2	LL	31
10		Důl Staříč	49.667222 18.258611	LL	-/2/-	LL	2
				RL	13/-/-	R	NA
						L	13
11		Horní Bludovice	49.744444 18.457500	LL	6/15/-	LL	21
				RL	2/2/-	R	NA
						L	4
12		Horní Domaslavice	49.694722 18.470000	LL	3/4/-	LL	7
				RL	7/1/-	R	NA
						L	8
13		Karviná-Doly	49.825000 18.483056	LL	-/2/-	LL	2
				RL	10/3/-	R	NA
						L	13
14		Louky	49.817222 18.576944	LL	1/-/-	LL	1
				RL	10/4/-	R	NA
						L	14
15		Prstná	49.915000 18.560556	LL	3/5/-	LL	8
				RL	3/-/-	R	NA
						L	3
16		Trnávka	49.683333 18.181389	LL	8/10/3	LL	21
				RR	-/-/1	RR	1
				RL	4/-/-	R	NA
						L	4
OG	KK	Greece	39.873166 22.732813	KK	1/-/-	KK	1

R-E male P. ridibundus-P. esculentus male populations, L-E P. lessonae-P. esculentus populations, OG used as an outgroup, RR P. ridibundus, RL P. esculentus, LL P. lessonae, KK P. kurtmuelleri, NA not analyzed

*The localities are numbered according to Fig. 1

Protein electrophoresis

For genotype determination, six allozyme loci previously identified to be diagnostic for Pelophylax taxa were used [20, 33]. Approximately 1 g of skeletal muscle (or gonads) was homogenized on crushed ice for 20 s in an equal volume of Tris NaCl extraction buffer (pH 8.5; [34]) using an Ultra-Turrax homogenizator (IKA-WERK). The homogenate was then centrifuged at 13,500g at 4 °C for 20 min. Enzymes obtained from these tissues, namely aspartate aminotransferase (Aat; EC 2.6.1.1), glucose-6-phosphate isomerase (Gpi; EC 5.3.1.9), glycerol-3-phosphate dehydrogenase (G3pdh; EC 1.1.1.8), l-lactate dehydrogenase (Ldh-1; EC 1.1.1.27), phosphoglucomutase (Pgm-2; EC 5.4.2.2), and phosphogluconate dehydrogenase (6-Pgd; EC 1.1.1.44), were analyzed by horizontal potato starch gel electrophoresis [20, 34]. Subsequently, gels were cut into three 2-mm-thick slices and stained with appropriate allozyme chromogenic detection methods [35–37]. Stained gel slices were photographed and the agar layers were transferred to filter paper, then dried and stored as part of the protocol. The visualized allele products were designated from “a” fastest to “e” slowest according to their mobility. Samples which revealed unclear patterns were reprocessed.

DNA extraction, microsatellite genotyping

Genomic DNA was extracted using a NucleoSpin commercial kit (Macherey-Nagel GmbH and Co.) with the epMotion 5075 automated pipetting system (Eppendorf). We amplified 17 microsatellite loci: Ga1a19, Re2caga3, Re1Caga10, RICA1b6 [38], RlCA1b5, RlCA5, RlCA18 [39], RlCA2a34, GA1a23, Rrid169A, Rrid059A, RlCA1a27, Rrid135A [40], Res16, Res20, Res22 [41], and Rrid013A [39, 42], using the redesigned primer sets described in Hermaniuk et al. [43]. We followed Pruvost et al. [31] for species-specific characterization of the aforementioned markers. Polymerase chain reaction (PCR) was performed according to Christiansen and Reyer [40]. Fragment lengths were determined using an ABI 3730 Avant capillary sequencer (Applied Biosystems, Zug, Switzerland) and an internal size standard (GeneScan-500 LIZ); alleles were scored with GeneMapper v. 3.7 (Applied Biosystems, Zug, Switzerland).

Preparing datasets used for all analyses

Due to the presence of null alleles, raw microsatellite genotypes of P. lessonae and P. ridibundus were checked for potential genotyping errors with Micro-Checker version 2.2.3 [44]. This method estimates frequencies of null alleles with the Brookfield 2 null allele estimator, which treats nonamplifications as data and regards them as null homozygotes when calculating null allele frequencies [45]. In each population, Micro-Checker tested every locus for departure from the Hardy-Weinberg (HW) genotypic equilibrium. We found one locus (Res20) in P. lessonae and four loci (Res16, Rrid069A, RICA5, Re1Caga10) in P. ridibundus with a potential presence of null alleles, and therefore applied a correction for null alleles at these loci following Wagner et al. [46]. The after-correction analysis in Micro-Checker did not detect any locus with a presence of null alleles. The software MSA v. 4.05 [47] was used to determine summary statistics like mean microsatellite allele numbers (AN), observed heterozygosities (H_OBS), and expected heterozygosities (H_EXP) within populations.

Based on the definition of hybridogenetic reproduction, diploid hybrid water frogs from Central Europe transmit one haploid chromosome set clonally to gametes (therefore termed a “hemiclone” [19]), whereas the other set is discarded and regained for the next generation by mating with an individual of the sexual parental species. As our aim was to test the origin of a particular hemiclone in this vertebrate species, we analyzed sexual and clonal genomes of hybrid individuals separately. First, we sorted lessonae and ridibundus genomes according to the allele species specificity known from the literature [31]. Then, the correctness of allele separation was tested visually assuming that one allele per locus was received from a sexual mate, and therefore such allele has to be present (and always was in our study) in the gene pool of a sympatric sexual population. Our approach of separating sexually and clonally inherited alleles in hybrid genomes was 100% successful.

While for hybrids from R-E system populations both ridibundus and lessonae genomes were included in our analyses, we used only lessonae genomes of hybrids from the L-E system populations because in this system we were not interested in the origin of ridibundus genomes. We determined a hemiclone by a multilocus genotype (MLG), defined by the identical combination of alleles found in our microsatellite analyses. A minimum of three samples exhibiting the same allele composition was a clear indication that the genome was inherited clonally (P_SEX value mentioned below) and did not originate from a sexual donor [48]. As most statistical programs dealing with microsatellite data—including GenAlEx and Populations used in this study—are not designed to compare haploid and diploid data, we transformed the haploid genotypes to diploid forms adding a second allele as missing data. This was done for lessonae and ridibundus genomes of P. esculentus (MLG’s) separately. To diploidize these haploid genotypes for GeneClone, we simply doubled lessonae and ridibundus genomes. Detailed information about the programs used are given below.

Flow cytometry

In order to determine ploidy levels, all individuals were analyzed by flow cytometry using blood samples following the standardized methodology [49]. A drop of blood was added into 70% ethanol and immediately shaken to prevent clotting. Chicken blood was used as a reference standard for cell size measurement. Relative nuclear DNA content was measured with DAPI fluorochrome using the Cystain two Step High Resolution DNA Staining commercial kit (Partec GmbH, Münster, Germany). Fluorescence intensity of 5000 stained nuclei was measured with a Partec PAII flow cytometer at a speed of 0.5 μl/s. Flow cytometric histograms were evaluated using FloMax 2.52 (Partec GmbH, Münster, Germany).

Statistical analysis of microsatellite data

In order to determine the origin of the all-male P. esculentus lineages, we analyzed genetic relationships between hybrids and their parental species on the basis of allele frequencies of up to 17 microsatellite loci using Structure v. 2.3.1 [50], GenAlEx v. 6.5 [51], Populations v. 1.2.32 [52], and GeneClone v.2.0 [53].

Taxon determination

In total, we identified 108 individuals of P. lessonae, 42 individuals of P. ridibundus, and 98 individuals of P. esculentus using six diagnostic allozyme loci (Additional file 3: Table S3) and 17 microsatellite loci (Additional file 2: Table S2). In P. lessonae, there were four monomorphic (Aat, Gpi, Pgm, G3pdh) and two polymorphic allozyme loci (Ldh-1, 6-Pgd); in P. ridibundus, two monomorphic (Aat, G3pdh) and four polymorphic loci (Ldh-1, Gpi, 6-Pgd, Pgm) were found. In P. esculentus, all six loci were polymorphic. The allelic products of three loci (G3pdh, Pgm, Gpi) displayed an atypical expression in 12 individuals. Here, four P. ridibundus possessed alleles characteristic of the lessonae genome in G3pdh and Pgm, one P. lessonae possessed alleles characteristic of the ridibundus genome in Pgm, and seven P. esculentus possessed either two lessonae or two ridibundus specific alleles in both Gpi and Pgm (Additional file 3: Table S3). Whether the shared alleles represent introgressions caused by crosses between individuals of the parental species and hybrids or ancestral polymorphisms is not clear.

Genetic diversity

Overall, all 17 microsatellite loci were polymorphic among 249 Pelophylax individuals (Additional file 4: Table S4). The loci Re2caga3, Res22, and Rrid169A amplified only in the P. ridibundus genome while the loci Res20, RICA1a27, and RICA18 amplified only in P. lessonae genome. Two loci (RICA5, Ga1a23) that are assumed to be specific for P. lessonae [39, 40], and one locus (Rrid135A), suggested to be specific for P. ridibundus [40], amplified in both P. lessonae and P. ridibundus samples.

A total of 187 alleles were detected, with 6 - 17 alleles per polymorphic locus (Additional file 2: Table S2, Additional file 4: Table S4 and Additional file 5: Table S5). The allele called 83 in RICA1b6 was marked as non-specific because it was amplified in the genomes of both P. lessonae (12 individuals) and P. ridibundus (four individuals). The presence of one or two alleles per locus (Additional file 2: Table S2) supported the flow cytometric data that all individuals were diploid. In addition, allele-dosage effects as indications of polyploidy were not observed.

Analyses of population structure using microsatellite data

Likelihood values provided by Structure converged during the runs, and results did not notably change between replicates. Evanno et al.’s [54] method implemented in Structure Harvester indicated K = 2 as the most probable number of genetic clusters for the dataset. Three types of diploid individuals were found, i.e., those with a q value ranging between 0.004–0.009 and 0.991–0.996, respectively, representing sexual P. lessonae or P. ridibundus, and those with intermediate q values (0.440–0.560) representing hybrid P. esculentus, which is in accordance with results from the determination based on diagnostic markers (Fig. 2a).

The PCA identified three groups of microsatellite MLGs, the first two principle components accounted for 62.15% (axis 1) and 11.47% (axis 3) of genetic variation, respectively (Fig. 2b). Cluster 1 (green and dark blue symbols) grouped P. lessonae and lessonae-specific MLGs of P. esculentus sampled in L-E system populations. Cluster 2 (yellow symbols) included lessonae-specific MLGs obtained from P. esculentus in R-E populations while cluster 3 (red and blue symbols) grouped P. ridibundus and ridibundus-specific MLGs of P. esculentus from R-E populations.

Genealogical analyses made with the program Populations revealed two distinct clusters of lessonae-specific MLGs. One is characteristic of lessonae-specific MLGs found in P. esculentus from R-E populations; the second comprises genotypes of P. lessonae individuals (Fig. 2c and Additional file 6: Figure S1). Contrary to the previous pattern, ridibundus-specific MLGs found in P. esculentus from R-E populations did not represent a separate lineage but instead clustered with P. ridibundus.

Identification of MLGs and hemiclones

In the dataset which contained missing data for some loci, GenAlEx estimated a total of 185 MLGs among all Pelophylax individuals investigated. Among 150 P. lessonae and P. ridibundus individuals, the program distinguished 150 MLGs. Considering the 27 P. esculentus from R-E male populations, ridibundus genomes were represented by 27 MLGs, whereas the lessonae genomes exhibited only eight MLGs (Fig. 3, Additional file 7: Table S6). One lessonae-specific MLG was present in 14 lessonae genomes of P. esculentus, while the remaining seven MLGs were discovered in one to three lessonae genomes. In contrast to MLGs obtained from ridibundus genomes, all lessonae-specific MLGs shared the same alleles; their differences are only caused by missing data for some loci (Additional file 7: Table S6). Therefore, we consider the eight lessonae-specific MLGs detected in the hybrid males of the R-E populations as a single MLG. This MLG is further suggested to represent a clone (or a hemiclone from an individual level) designated as ODERL1.

Considering only MLGs with complete allelic data, GenClone generated 122 MLGs among 143 genotypes. Twenty-one lessonae genomes, all originating from P. esculentus of R-E male populations, were represented by only a single MLG. The P_SEX value of this MLG ranged from 0.15 to 1.50E−41, indicating clonal inheritance of ODERL1.

Detection of all-male P. esculentus lineage

The upper Oder River valley is mainly inhabited by P. ridibundus of both sexes and only males of P. esculentus (R-E system; populations 1-5, Table 1), whereas populations of the L-E system (typically represented by P. lessonae and P. esculentus of both sexes) live outside the valley. Since 2001, no hybrid female was detected among 377 hybrid individuals in localities of the R-E system (Table 1 and unpublished data), which makes this region a newly discovered area for the coexistence of bi-sexual P. ridibundus with diploid all-male P. esculentus hybrids. Together with the occurrence of similar populations downstream the Oder River drainage basin [8, 9, 28, 60], this points to Central Europe as the area of origin of R-E male populations.

P. esculentus males are active in maintaining their own all-male hybrid lineage

Hybrid males are usually infertile [61] and therefore considered as a by-product of hybridizing sexual species or a sexual male species with a unisexual female species [7, 62]. In the L-E system of Western Europe, P. esculentus males usually result only from crosses between P. esculentus females and heterogametic P. lessonae males. Backcrosses between such hybrid males and P. lessonae females result exclusively in P. esculentus females, due to the presence of clonal ridibundus genomes in the sperm that obviously contain female-determining factors [21, 22, 63].

Two lines of evidence support our finding that P. esculentus males from the upper Oder River valley are fertile and maintain the all-male hybrid lineage, by crossing with syntopic P. ridibundus females, i.e., they do not originate from primary hybridizations between their parental sexual species, or between sexual males and hybrid females.

First, our extensive sampling did not reveal the presence of P. lessonae or even a single hybrid female, necessary for the origin of diploid hybrid males [21]. The second line of evidence comes from independent statistical tests of allelic variation of the nuclear loci. The individually specific MLGs and the presence of both sexes indicate that both investigated taxa, P. ridibundus and P. lessonae, came from randomly mating sexual populations. Ridibundus alleles of P. esculentus males were grouped together with alleles of P. ridibundus males and females in two different clustering approaches (Fig. 2a, b). Moreover, 27 ridibundus genomes from P. esculentus males represent 27 MLGs, i.e., again all combinations are unique. It is therefore reasonable to assume that P. esculentus received its haploid ridibundus genomes from sympatric P. ridibundus females. On the other hand, we suggest that the lessonae genomes found in P. esculentus males from the Oder River are transmitted clonally only by these hybrids over the generations because the lessonae genome of P. esculentus males differed not only from the genome of sympatric P. ridibundus individuals but also from the genome of P. lessonae included in this study (Fig. 2a, b). The statistical test for the P_sex value of these 27 lessonae genomes (1.50E⁻⁴¹) supports its clonal inheritance.

All-male hybrids represent a single hemiclone

Hemiclonal hybrids are genetically identical for half of the diploid parental genome [64] which they clonally pass on to the next generation. Analyses of ridibundus and lessonae allelic richness in all-male P. esculentus revealed that 13 loci within the ridibundus genome were polymorphic with the observed allelic frequencies (0.036–0.842). Together with the abovementioned observation that single ridibundus MLGs were not shared among individuals at all, it is reasonable to assume that these ridibundus genomes came from recombinant eggs of sexually reproducing P. ridibundus. In the lessonae genome, however, we observed only monomorphic loci with allele frequencies of 1.00. Therefore, it is proven that the lessonae genome is clonally inherited and represents a single clone (Figs. 2 and 3, Additional file 2: Table S2).

The characteristic of the lessonae genome ODERL1 only producing male hybrids is in accordance with the XX/XY type of sex determination hypothesized for the R-E system much earlier by Uzzell et al. [9]. However, our finding that all hybrid males possess an identical lessonae (ODERL1) is new finding. Most animal systems with female unisexuality comprise multiple clonal MLGs, as spined loach fishes of the genera Cobitis [65] and Poeciliopsis [66] and the Phoxinus eos-neogaeus complex [67]. Even where hybrids form a monophyletic group, e.g., Poecilia formosa, populations show a fairly high level of clonal diversity [68]. In Pelophylax populations of the L-E system, where both hybrid males and females coexist, the diversity of transmitted ridibundus clones is high [48, 69].

Origin of the hemiclonal line

Apparently, the existence and maintenance of all-male unisexuality among animals in general, and in Pelophylax water frogs in particular, hinge on the formation of a specific genome from a particular sexual species. In European water frogs, the genome of P. lessonae represented by the MLG ODERL1 seems to play a key role in the formation of all-male P. esculentus lineages. Supporting evidence comes from rare Central- and West-European mixed populations, where all-male P. esculentus are triploid and originate from clonal sperm with a diploid lessonae genome and haploid ridibundus eggs [26].

Despite the evidence for a single origin of the ODERL1 clone and the fact that it has evolved from the P. lessonae genepool, this clone does not cluster with other lessonae-specific MLGs (Fig. 2a, c) indicating clear differences between the ODERL1 clone and other P. lessonae genomes. Due to high distance between the two datasets of lessonae alleles, we hypothesize that present P. lessonae populations inhabiting areas close to the Oder River valley are not the most recent donors of the lessonae genome found in the all-male hybrid lineage of the R-E system populations investigated. It also seems unlikely that the lessonae clone recently originated from local P. lessonae. Because microsatellite markers have high mutation rates [70], an old hemiclone would comprise several MLGs that would merge into one multilocus lineage [65]. In contrast to this, we found a single lessonae-specific MLGs in all hybrids males, indicating a quite recent hybridization event. On the other hand, we cannot exclude a potential selective sweep in the lineage that might be caused by positive natural selection for the ODERL1 hemiclone [65]. Although age estimates of the ODERL1 clone and comparison with more distant P. lessonae populations are yet to be obtained, we hypothesize either an older in situ origin or a possible ex situ origin of the ODERL1 clone, i.e., some distance away from the local P. lessonae genepool.