Gamete production patterns in the crossing experiment were largely in agreement with the rough gamete pattern from earlier studies. In addition, this study quantified the proportions of LR and R eggs laid by LR females and of rare gamete types. It also confirmed that the male-determining Y factor was confined to the L genome in the populations tested. The LRR males present in some of these populations could thus not have arisen from R genomes with Y factor. Nevertheless, the riddle of the LRR males was solved: 22% of the sperm from LR males from LRR-rich populations was LR sperm which, according to the model, was sufficient to explain the high proportion of LRR males in these proportions. In contrast, it remains unresolved whether LR females play a role in shaping genotype proportions in natural populations. Based on the gamete results, the model predicted adult equilibrium genotype compositions which are compared to empirical data below. The model also suggested that the proportion of tetraploid LLRR remains low unless their survival and/or reproductive output is substantially increased, and that inviability of non-hybrid genotypes is a precondition for the maintenance of the all-hybrid populations.
Gamete types
As expected from other studies, triploid frogs made almost exclusively haploid gametes with the genome they had in double dose. The only exception was LLR females that produced 11.1–11.5% LL gametes. Similar results were obtained with Swedish frogs by Jakob and Arioli [26]. They had an overall mean of 4.75% LL eggs from LLR females (n = 4), while their remaining triploids produced almost no unusual gamete types. Rare LL and RR ova are known from several areas and population types [32–35]. Reports on frequent LL gametes from Hungary [36], France [11] and possibly the Czech Republic [21] concern sperm – not eggs as in the present study. The Hungarian LL sperm was unreduced (made by apomixis) as opposed to most or all of the LL eggs, the RR egg and the RR sperm in the present study (made by automixis). Thus, LL gametes can apparently be formed by different cytological processes. Automixis also occurs in LR frogs: after the L genome is excluded, the R genome is not directly passed on into gametes, but first undergoes duplication and meiosis [37].
With 81.7% diploid LR and 18.1% haploid R eggs, R egg production in diploid females was also within the range of previous estimates (21.4%, n = 7, [24]; 25.0%, n = 4, [25]; 8.8%, n = 10, [26]). Together with the 18.1% (n = 19) in the present study, the overall, weighed average becomes 17.1% based on 40 frogs. Unfortunately for future studies, a count of large and small eggs gave an unreliable estimate of the ploidy in the viable offspring, because the small eggs often had very mixed genotypes and/or high mortality.
Sex determination
Sex determination in vertebrates can be either environmental or genetic, but only genetic sex determination has been found in amphibian populations studied [38, 39]. Both XY and WZ systems exist and shifts between them have been extraordinarily frequent in amphibian evolution [38]; XY and WZ systems can even coexist within the same species [40]. Amphibian sex chromosomes show little or no differentiation [38], which might explain the viability of polyploid amphibians: with no Y chromosome degeneration there is no X dosage compensation to be disrupted by polyploidy, as opposed to in for example birds and mammals [41]. However, the lack of sex chromosome differentiation complicates the study of sex determination, so that markers for DNA-sexing have been obtained for very few species [42].
In the present study, which was the first to investigate sex determination in the Scandinavian all-hybrid populations, sex was therefore determined by dissection. This gave slightly inconsistent results: 19–20 male and 2–3 female offspring had the wrong sex compared to the expectations from an L genome-confined Y factor. Low frequencies of the unexpected sex were also obtained in other studies and for unknown reasons ([43] and references therein). The reasons are unknown, but one possible explanation could be underdeveloped gonads, which are small and round and look like small testes (own observation in Swiss F1 hybrids). P. esculentus is known to have retarded ovary development, apparently because the special hybrid mode of gametogenesis creates complications [44]. Alternatively, unexpected offspring sex could result from spontaneous mixed genotypes. A missing Y factor could render unexpected females and a substitution or addition of a Y factor to an otherwise pure R sperm could render unexpected males.
LRR males
LRR males arose from LR sperm fertilizing R eggs in the crossing experiment, which suggests that LR sperm is responsible for their existence in some natural ponds. The 22% LR sperm from LR males found in the crossing experiment was sufficient to increase the LRR proportion to comprise almost half of the males and more than half of the females in the model population. The 22% LR sperm was mainly provided by only two out of seven LR males from LRR-rich populations; both from pond 089. Also in pond 089, Jakob and Arioli [26] found LR sperm in only one of three LR males, resulting in an overall mean of 4% LR sperm among them. In Alsønderup – the other LRR-rich population investigated – one out of four LR males produced LR sperm, i.e. none in the present, but one in a previous study [24]. With such large individual differences, large samples are required for reliable estimates.
As explained in the introduction, LRR males could also originate from R genomes with a Y factor and/or from RR eggs. R genomes with a Y factor are unlikely to occur, since in a sample of 15 LR and LRR males from LRR-rich populations no R genome with a Y factor was found, although the model predicted that, if present, they should spread and eventually replace L genomes with Y factors. It can still not be ruled out that RR eggs contribute to the formation of LRR males, as only four LRR females from LRR-rich ponds were investigated. Elevated proportions of RR eggs were, however, not observed in these four LRR females [26].
Gamete patterns might drive adult genotype proportions
LR sperm was apparently more common in LRR-rich populations than in normal populations in the present study, and for L sperm the trend was opposite. Similar striking differences in the sperm types between LR frogs from LLR-rich, LR-rich and LRR-rich populations were observed by Jacob and Arioli [26]. In contrast to in the present study, LR females in the study of Jacob and Arioli [26] also made more R eggs in LR-rich populations than in triploid-rich populations. Although in both studies, the sample sizes are too small for a statistic confirmation of these apparent population type-specific differences, the data suggests that gamete patterns may drive the genotype proportion in all-hybrid populations of P. esculentus. As extensive efforts to show relations between adult genotype proportions and ecological factors has been of rather limited success [26] this suggestion is a welcome alternative hypothesis that needs proper testing. If true, the number of evolutionary significant units relevant for conservation might be higher in P. esculentus than presently realized.
Modelled versus natural populations
The genotype proportions predicted for normal and LRR-rich all-hybrid populations matched available field data from a large sample of natural Swedish ponds and a subsample of LRR-rich ponds, respectively. The large sample consisted of 3000 frogs from 12–23 Swedish ponds with various genotype compositions sampled over 3 years [26]. Within males, the among-year range (compared to the model result in parentheses) was 33–60 (43)% LLR, 36–60 (57)% LR, 2–4 (0)% LRR; within females there were 15–28 (19)% LLR, 26–44 (42)% LR and 39–45 (39)% LRR. In the LRR-rich pond 089 there were within males (n = 103) 17 (14)% LLR, 58 (23)% LR, 5 (20)% LLRR, 20 (43)% LRR; within females (n = 216) 6 (9)% LLR, 29 (25)% LR, 0 (6) LLRR and 65 (59)% LRR [26]. The fit with Alsønderup was less good, but here the sample size was only 46 frogs.
The overall good fit between observed and modelled genotype proportions suggests that the model captured the essence of al least the normal all-hybrid populations, in spite of its simplicity. The simplifications included random mating, equal survival of LLR, LR and LRR and equal reproductive output for all genotypes, and were mainly motivated by insufficient empirical data. In the LE system, females prefer LL to LR males [45–47], but it is not known if females can and do distinguish between male genotypes in the all-hybrid populations. Concerning survival, LR probably survived better than triploids from eggs to 1-year-olds [25], but thereafter differences between genotypes disappeared. A capture-mark-recapture study on the Swedish ponds showed that local adult survival differed between sexes and genotypes, but overall the genotypes had similar annual survivals of around 30% (n = 329 [26]). In contrast, poor survival of LRR has been suggested by authors based in other areas [33, 48–50]. With respect to reproductive output, female fecundity depends on both female body size and genotype [26, 51], while fertilization success is apparently reduced in LR males producing several kinds of sperm [18, 26, 52]. In addition, reproductive output also depends on the genotype-specific proportion of aneuploid eggs and sperm that do not give rise to viable offspring. Such data is lacking; the present study only suggested that most eggs that died or were aneuploid came from LR females. Furthermore, male mating success is also important for the reproductive output of males. In the LE system, LL males with scramble competition behaviour have more mating success than territorial LR males [53], but no data are available on genotype-specific mating success of LLR, LR and LRR males.
Evolutionary potential of all-hybrid populations
According to the model, LLRR frogs needed a more than twofold advantage in either reproductive output, survival or a combination, to turn a normal or LRR-rich population into a pure LLRR population. In vertebrates, polyploidy tends to have little or no effect on body size [54], so no increased fecundity in females is expected. Increased reproductive output in LLRR is not unlikely because tetraploidy may result in more regular meiotic processes and, hence, a higher proportion of fertile gametes for both sexes. As this advantage should arise spontaneously in LLRR, and it has not yet helped LLRR increase in frequency, it is, however, unlikely to make LLRR increase further in the future. Concerning survival, field data from the Swedish study area do not suggest that LLRR have a selective advantage over the other genotypes. On the contrary: the proportion of LLRR decreased from 2.8% at the egg stage to zero at metamorphosis and among one-year old juveniles [25]. In ponds 089 and Alsønderup, the proportion of LLRR adults was also lower than expected from the gametogenetic pattern (see above). Although a broad variety of habitats have been investigated, it is, however, possible that the LLRR would have higher survival in a different habitat.
A recent study of the hybridogenetic Iberian minnow, Squalius alburnoides, provides strong evidence that di- and triploid hybrid populations can be an intermediate step on the way to a tetraploid species [9]. S. alburnoides (also called Leuciscus, Rutilus and Tropidophoxinellus) resembles P. esculentus most of the five other hybridogenetic complexes presently known. Most populations of this freshwater fish are composed of diploid and triploid hybrids, one parental species and sometimes backcrossed males of the other, now extinct, parental species [55]. Symmetrical tetraploids are common in low proportions, though not as low as in P. esculentus. In contrast, two newly discovered populations have 73% tetraploids with even sex ratios, normal meiosis and the capability to reproduce among themselves [9]. In addition, postzygotic isolation appears to have arisen between triploid and tetraploid forms. It was suggested that the success of the tetraploids is connected to the more upstream habitat of these populations, but this needs further investigation. These discoveries suggest that tetraploidization could also happen in P. esculentus if it be given sufficient habitat variation, space and time to evolve. Maybe it has even happened somewhere already and can be found if looked for.
Threats to all-hybrid populations
Normal all-hybrid populations will, according to the model, only persist when survival of LL and RR is zero. With moderate survival of parental genotypes, stable mixed populations of hybrids and non-hybrids will result; with LL survival above approximately 0.7 or RR survival above approximately 0.5, all or most hybrid genotypes go extinct. These results call attention to the possibility that introduction of water frogs is a potential threat to all-hybrid populations. It is not known why LL and RR genotypes have lower survival in natural all-hybrid populations. One possible explanation is that the parental species are at a selective disadvantage under Scandinavian environmental conditions. This explanation is, however, not very plausible, since P. ridibundus does occur on the very nearby Danish island of Bornholm, and P. lessonae lives both north, south and east of the all-hybrid populations. Another, not mutually exclusive, explanation is that the parental genotypes are homozygous for deleterious alleles that have become almost or entirely fixed in the genetically generally very depleted L and R genomes [25, 27].
If the reduced fitness of non-hybrids is due to homozygosity for deleterious mutations, an interesting question is what would happen if P. lessonae, P. ridibundus or P. esculentus with fewer or different deleterious mutations were introduced into the all-hybrid populations. Would they give rise to viable non-hybrids, spread and radically change the system; maybe even drive (most of) the hybrid genotypes extinct? Such a situation is currently observed in the Swiss LE system, where RR offspring used to die due to homozygosity of deleterious mutations [13], but now in many places survive and take over due to introductions of P. ridibundus [56]. Or would the viable alleles be scattered and swamped out by recombination in non-hybrids and triploids with the more numerous resident genomes? This is possibly the situation in northern Germany, where all-hybrid populations apparently persist without geographic isolation from P. lessonae and P. ridibundus populations further south. The answer might depend on the number of loci with deleterious alleles in high frequencies and on the number of introduced genomes without deleterious mutations at these loci. If introduced frogs include P. ridibundus males, however, the risk seems high that their R
y
will be invasive and replace L
y
in the recipient all-hybrid populations, with the previously mentioned consequences of modified equilibrium genotype proportions and slightly increased hybrid load.