Christoph Vorburger, Institute of Zoology, University of Zürich, Winterthurerstrasse 190, CH-8057 Zürich, Switzerland. Tel.: +41 1 635 49 84; fax: +41 1 635 68 21; e-mail: email@example.com
The European waterfrog Rana esculenta (RL-genotype) is a natural hybrid between R. ridibunda (RR) and R. lessonae (LL) and reproduces by hybridogenesis, i.e. it eliminates the L-genome from the germline and produces gametes only containing the clonally transmitted R-genome. Because of the lack of recombination, R-genomes are prone to accumulate spontaneous deleterious mutations. The homozygous effects of such mutations become evident in matings between hybrids: their offspring possess two clonal R-genomes and are generally inviable. However, the evolutionary fate of R. esculenta mainly depends on the heterozygous effects of mutations on the R-genome. These effects may be hidden in the hybrid R. esculenta because it has been shown to benefit from spontaneous heterosis. To uncouple clonal inheritance from hybridity, I crossed R. esculenta with R. ridibunda to produce nonhybrid offspring with one clonal and one sexual R-genome, and compared their survival and larval performance with normal, sexually produced R. ridibunda tadpoles. Because environmental stress can enhance the negative effects of mutation accumulation, I measured the performance at high and low food levels. There was no indication that tadpoles with a clonal genome performed worse at either food level, suggesting that at least in the larval stage, R. esculenta benefits from heterosis without incurring any costs because of heterozygous effects of deleterious mutations on the clonally transmitted R-genome.
Sex is the predominant mode of reproduction in higher organisms, but many taxa have retained the potential to give rise to asexual forms. The taxonomic distribution of these forms is sporadic and asexuals usually have close relatives that are sexual, suggesting that on an evolutionary timescale, asexuals are short-lived (Maynard Smith, 1978). However, some ‘ancient asexual scandals’ challenge the generality of this pattern (Judson & Normark, 1996). The ubiquity of sex, despite its two-fold cost, still implies the operation of selective forces that bestow an advantage to sexual reproduction (Maynard Smith, 1978). Two general kinds of explanation are commonly invoked for this advantage: (1) sex creates new gene combinations and thus accelerates adaptation to changing environments, and (2) sex eliminates recurrent deleterious mutations (Kondrashov, 1988; Maynard Smith, 1988; West et al., 1999). Mutation-based models apply to both stable and variable environments, and are therefore appealing because of their generality. But presently, there is no common agreement on whether the rate of occurrence and the magnitude of effects of spontaneous deleterious mutations are sufficient to disfavour asexuals and thus provide a general explanation for the maintenance of sexual reproduction (e.g. Peck & Eyre-Walker, 1997; Kondrashov, 1998, 2001; Keightley & Eyre-Walker, 2000, 2001; Nachman & Crowell, 2000). It has also been argued that the synergistic interaction of both environmental and mutation-based mechanisms may be required to balance the two-fold cost of sex (West et al., 1999). Obviously, a better understanding of the characteristics of deleterious mutations is urgently needed (Peck & Eyre-Walker, 1997).
A useful study system should allow to compare the fitness of clonally and sexually transmitted genomes within the same species. This is possible in hybridogenetic taxa and their parental species. Hybridogens are natural hybrids which exclude the genome of one parental species from the germline and transmit the genome of the other parental species to the gametes without recombination (Schultz, 1969). A well-studied hybridogenetic taxon is the European waterfrog Rana esculenta L. (RL-genotype), a diploid, bisexual hybrid between the two parental species Rana ridibunda Pallas (RR) and Rana lessonae Camerano (LL) (Berger, 1967, 1968). Over large parts of central Europe, R. esculenta coexists with only one of its parental species, R. lessonae, forming the so called L-E-system (Uzzell & Berger, 1975). In this system, hybrids exclude the L-genome from their germline prior to meiosis and form haploid gametes containing the clonally transmitted R-genome (reviewed by Graf & Polls Pelaz, 1989). Rana esculenta acts as a sexual parasite of R. lessonae because it can only persist in the population by backcrossing with R. lessonae to produce R. esculenta offspring which again exclude the L-genome from their germline. Matings between hybrids also occur, producing nonhybrid offspring with two clonal R-genomes. However, these offspring are typically inviable and die at an early larval stage (Blankenhorn et al., 1971; Heusser & Blankenhorn, 1973; Berger & Uzzell, 1977; Binkert et al., 1982; Semlitsch & Reyer, 1992). Inviability can be attributed to deleterious mutations that were fixed in the clonal R-genomes through Muller’s ratchet and become homozygous in these tadpoles, providing evidence that clonal genomes are prone to accumulate harmful mutations (Graf & Müller, 1979; C. Vorburger, unpublished manuscript). However, the evolutionary fate of R. esculenta depends on the heterozygous rather than the homozygous effects of these mutations, because the clonal R-genome is always paired with a sexual, genetically variable L-genome that is unlikely to carry the same mutations.
Deleterious mutations are generally not completely recessive (Simmons & Crow, 1977; García-Dorado & Caballero, 2000), implying that in R. esculenta, mutation accumulation on the clonal R-genome may lead to a decline in fitness, making hemiclonal hybrids a ‘dead end road’ for these genomes (Milinski, 1994). But several studies on tadpoles have demonstrated that R. esculenta is fully viable and highly competitive (Semlitsch & Reyer, 1992; Semlitsch, 1993a; Rist et al., 1997). However, this does not necessarily indicate that heterozygous effects of deleterious mutations on clonal R-genomes are weak or absent, because R. esculenta was found to strongly benefit from spontaneous heterosis, i.e. F1 hybrids performed better than both parental species (Hotz et al., 1999). Heterosis may possibly compensate for detrimental effects of mutation accumulation.
For a valid test of heterozygous fitness effects due to deleterious mutations in clonal R-genomes, clonal inheritance must be uncoupled from hybridity. Therefore, I crossed R. ridibunda females both with R. ridibunda and R. esculenta males to produce nonhybrid R. ridibunda tadpoles, possessing either two sexual or one sexual and one clonal genome, and compared their survival and larval performance. Because it is known that environmental stress enhances the negative effects of deleterious mutation accumulation (Kondrashov & Houle, 1994; Shabalina et al., 1997; Szafraniec et al., 2001), I raised the tadpoles under benign and stressful conditions, i.e. under high and low food levels. I found no indication that tadpoles with a clonal genome were inferior at either food level, suggesting that at least presently, R. esculenta enjoys the full benefits of heterosis without suffering from detrimental effects of mutation accumulation.
Materials and methods
I obtained female R. ridibunda from a single population near Bavois (Ba), about 20 km north of Lausanne, Switzerland. These females provided the common source of sexual R-genomes with which all clonal and sexual R-genomes to be tested were paired. Rana ridibunda is not native to Switzerland but was introduced into this area from eastern Europe and Anatolia about 50 years ago (Grossenbacher, 1988). It has replaced the native waterfrogs and now forms pure populations in large areas of western Switzerland, including the collection site.
Males of R. ridibunda and R. esculenta were collected from three sites each. Using replicate sources of sexual and clonal R-genomes was important because local adaptation to breeding ponds may also affect tadpole performance. Taking the fathers and not the mothers from both taxa and different sources avoided confounding these effects with maternal effects and/or cytoplasmic inheritance. Male R. ridibunda came from three sites within western Switzerland: Arnex (Ar), Chavornay (Ch), and Penthaz (Pe). Male R. esculenta came from Alpnach (Al), 10 km south of Luzern, central Switzerland, from Hellberg (Hel), 20 km east of Zürich, northern Switzerland, and St. Margrethen (StM), 20 km east of St. Gallen, eastern Switzerland. These three sites contain native L-E-systems. It was previously found that only one R. esculenta hemiclone HEL 1 occurs in Hel (Semlitsch et al., 1996), i.e. the clonal R-genomes of all R. esculenta in the population share the same haploid multilocus genotype as defined by allozyme markers.
Hemiclone composition of populations Al and StM was determined by cellulose acetate electrophoresis of allozymes following standard procedures (Hebert & Beaton, 1993). Toe clips of 71 individual R. esculenta from Al and 44 R. esculenta from StM were analysed. Only one hemiclone could be identified in either population, which I designated as AL 1 and STM 1. These hemiclones shared the same R. ridibunda alleles at all six loci assayed and were also identical to HEL 1 [sAAT: e, GPI: a, LDH-B: c, MPI: c, PGM-2: d, PGDH: d; allele designations follow Hotz (1983) and Beerli (1994)]. Hence, based on the limited resolution provided by six polymorphic loci, R. esculenta from all three sources may possess R-genomes belonging to the same clonal lineage.
Clonal R-genomes from all three sources possess recessive deleterious alleles corresponding to at least one lethal equivalent. This is already suggested by the fact that R. ridibunda does not occur in these populations, and was confirmed by performing R. esculenta ×R. esculenta crossings. I crossed 5 R. esculenta pairs from Al in 1998 (unpublished manuscript), and 5 pairs from StM in 1999. Also in 1999, I obtained a total of 17 egg masses deposited by any of 15 male and 21 female R. esculenta from Hel during a behavioural experiment performed by H.-U. Reyer and T. Garner (unpublished work). Twenty tadpoles per cross or egg mass were raised in outdoor tanks under conditions that allow for >90% survival to metamorphosis in normal, sexually produced R. ridibunda tadpoles. However, all tadpoles died during early developmental stages.
Using artificial fertilizations (Berger et al., 1994), I crossed each of four female R. ridibunda from Ba with one male from all six source populations, producing four groups of six maternal half-sib families of tadpoles (Fig. 1). All tadpoles produced by these crossings were R. ridibunda, but those with R. ridibunda fathers possessed two sexual R-genomes whereas those with R. esculenta fathers possessed one sexual and one clonal R-genome. Rana ridibunda with a clonal and a sexual R-genome do not occur in natural L-E-systems. This ‘artificial’ genotype was generated in this experiment in order to assess the heterozygous fitness effects of deleterious mutations on clonal R-genomes in a homospecific genetic background, i.e. without the compensating effects resulting from heterosis in the hybrid R. esculenta. Crossing the same female with males of both taxa avoids confounding maternal effects with the comparison between clonally and sexually transmitted genomes. On the other hand, this crossing design confounds paternal taxon with offspring sex ratio because male R. esculenta only produce daughters. This results from the fact that males are the heterogametic sex (XY sex determination: Berger et al., 1988), and that for behavioural reasons, primary hybridizations occur between R. ridibunda females and R. lessonae males (Tunner, 1974). As a consequence, the clonally transmitted R-genome in the hybrid R. esculenta always contains an X-chromosome. Thus, by comparing tadpoles from R. ridibunda fathers with tadpoles from R. esculenta fathers, I compare tadpoles of both sexes with all-female tadpoles. This should not compromise the interpretation of the experiment, however, because in a previous study I found no significant difference in survival and larval performance between male and female R. ridibunda tadpoles (Vorburger, 2001, in press).
All frogs were captured at night by hand between 1 and 23 May 1999 and stored in a cool room at 10 °C until used for crossings. On 23 May I injected the females with the hormone LH-RH (H-7525, Bachem, Inc.) to induce ovulation. One day later, when ovulation began, I killed the males with 3-aminobenzoic acid ethyl ester (MS-222; A-5040, Sigma, Inc.) and removed their testes. I prepared sperm suspensions of a single male from each of the six sources by crushing both testes in Petri dishes containing 15–20 mL of aged tap water. Then I stripped the eggs of one female into the six sperm suspensions, cycling between males to avoid confounding potential effects of ovulation order with paternal effects. Subsequently, I prepared six new sperm suspensions for the next round of fertilizations following the same protocol. A total of four rounds were performed to complete the crossing design. On 25 May, 1 day after fertilization, I transferred the eggs from Petri dishes into larger containers filled with 1 L of aged tap water. These containers were stored at 20 °C in a climatized room where tadpoles hatched after approximately one week. On 8 June, when tadpoles were free swimming and had resorbed their yolk sacs (stage 25; Gosner, 1960), I randomly selected eight tadpoles per family and housed them individually in containers filled with 1 L of aged tap water. Four containers were assigned to the low and high food treatment, respectively. Containers were arranged on shelves into four randomized complete blocks along a temperature gradient related to shelf height. After the first 6 days, I changed the water in each container every 3 days. The room in which tadpoles were raised had no windows and was illuminated by fluorescent light at a 15:9 h light–dark cycle. The temperature during the experiment ranged between 19 and 22.5 °C and tended to increase gradually by 0.5–1 °C from the lowest to the highest shelf in the room.
The two food levels used to provide an environmental gradient along which to measure tadpole performance were chosen to reflect benign and stressful conditions. Tadpoles were fed with commercially available tadpole food (Dorswal, Roswal-Produkte AG, Zurich, Switzerland) after water changes. The two food rations were initially set at 5 and 15% of mean tadpole mass per day, but changed to 7.5 and 15% after 12 days because 5% appeared to be insufficient to support at least minimal growth. Mean tadpole mass was estimated separately for each treatment on every feeding occasion by randomly selecting eight tadpoles per treatment (2 from each block) and weighing them to the nearest mg. Because food rations were set at a fixed fraction of mean tadpole mass determined within treatment, food rations diverged as they increased during the course of the experiment as a result of slower growth in the low food treatment. After day 60 I changed the water and fed the tadpoles every 2 days to avoid pollution of the water because of the increasing amounts of food added.
On days 30 and 60 of the experiment, I weighed all surviving tadpoles to the nearest mg and determined their developmental stage after Gosner (1960). Larval growth and development are reliable indicators of fitness in several anurans (e.g. Berven & Gill, 1983; Smith, 1987; Berven, 1990), including waterfrogs (R. Altwegg, personal communication), because a short larval period and a large size at metamorphosis positively affect survival at the terrestrial life stage and the timing of maturity. When metamorphosis started (defined as the emergence of at least one forelimb; stage 42), I checked the containers daily for metamorphs, and recorded their larval period (number of days from the start of the experiment to forelimb emergence) and their size at metamorphosis (mass to the nearest mg at stage 42). The experiment was terminated after 120 days on 6 October, when only 11 tadpoles had not yet metamorphosed.
Tadpole mass at days 30 and 60, larval period, and mass at metamorphosis were analysed using mixed-model nested analyses of variance (ANOVAs), testing for the effects of male taxon, food level, the interaction of male taxon and food level, male source nested within taxon and the interaction of male source and food level. Male sources represent a random sample from a large number of known R. ridibunda populations and L–E-systems in Switzerland and were thus considered random effects. The main effects of block and individual female were also included in the model to account for environmental variation amongst shelves and for maternal effects, respectively. Mass at day 30 was log-transformed before analysis to obtain a normal distribution of residuals, the other responses fulfilled the assumptions of ANOVA without transformation. Developmental stage is a noncontinuous response variable on an ordinal scale and does not have an expected distribution. However, inspection of the data revealed that stages ranged over 10 levels at day 30 and 15 levels at day 60, and that their frequency distribution did not significantly deviate from normal. Hence I decided to analyse also developmental stage at day 30 and 60 with ANOVA. Because of mortality, the experimental design was not completely balanced. Thus, I performed all analyses using type III sums of squares, which account for unequal cell size and produce orthogonal tests of hypotheses (SAS Institute, 1989). Satterthwaite approximations were used to produce error estimates for F-tests.
Survival to metamorphosis was analysed with maximum-likelihood logistic regression using the logit link function with binomial errors in PROC GENMOD (SAS Institute, 1996). The effect of male source could not be included in this model because the resulting cell sizes became too small. Tadpoles that did not reach metamorphosis until the end of the experiment were excluded from this analysis as well as seven tadpoles that died because they jumped out of their rearing containers.
Overall, 82% of the tadpoles survived until metamorphosis. Survival differed significantly amongst blocks with the highest mortality occurring on the uppermost shelves (χ23=10.25, P=0.017). There was also significant variation amongst individual females in the survival probability of their offspring (χ23=11.08, P=0.011). Male taxon (χ21=1.46, P=0.227), food level (χ21=0.11, P=0.735), and the taxon × food interaction (χ21=0.73, P=0.392) did not significantly affect survival to metamorphosis.
Food level had a strong effect on tadpole growth and development. On days 30 and 60 of the experiment, tadpoles in the high food treatment were significantly heavier and further developed than tadpoles in the low food treatment (Table 1, Fig. 2). They also reached metamorphosis on average 16 days earlier and at a significantly higher weight (Table 2, Fig. 3).
Table 1. Mixed-model nested ANOVAS for mass and developmental stage of R. ridibunda tadpoles on days 30 and 60 of the experiment. Mass on day 30 was log-transformed before the analysis. Block, female, and male source (taxon) × food were tested against the residual, male taxon was tested against male source (taxon), and food level, male taxon × food, and male source (taxon) were tested against male source (taxon) × food. Because of slight imbalances in the design (mortality), Satterthwaite approximations were used to generate error estimates for F-tests (SAS Institute, 1989).
Table 2. Mixed-model nested ANOVAS for larval period (days from the start of the experiment to forelimb emergence), and mass at metamorphosis. F-ratios were calculated as in Table 1.
Offspring from the four females used for crossings significantly differed at all response variables I measured (Tables 1 and 2, indicating strong maternal effects on larval growth and development in R. ridibunda. Male taxon only affected tadpole mass at day 60 and the mass at metamorphosis: offspring of R. esculenta fathers were significantly heavier than offspring of R. ridibunda fathers (Tables 1 and 2, Figs 2 and 3). Variation amongst blocks was not significant at early larval stages (day 30), but became significant on day 60 for tadpole mass, and marginally so for developmental stage. There was also significant variation amongst blocks for mass at metamorphosis and larval period, with tadpoles from the higher shelves reaching metamorphosis earlier.
The interaction between male taxon and food level was only significant on day 30 of the experiment, when tadpoles with R. esculenta fathers exhibited a steeper response to changing food levels than tadpoles with R. ridibunda fathers for both mass and developmental stage (Table 1, Fig. 2). There was also significant variation amongst male sources nested within taxa for both responses on day 30, mainly because of differences amongst the three R. esculenta sources (Table 1, Fig. 2). However, all of these effects were no longer evident at later stages of larval development (Tables 1 and 2. The interaction of male source nested within taxon and food level was not significant for any response variable (Tables 1 and 2.
The purpose of this study was two-fold: first, to test whether clonal R-genomes with a known deleterious gene load of at least one lethal equivalent in the homozygous state also reduce the fitness of R. ridibunda tadpoles in the heterozygous state; and second to test whether heterozygous effects of clonal genomes on fitness depend on environmental conditions. There was no indication that R. ridibunda tadpoles with one clonal and one sexual R-genome (R. ridibunda × R. esculenta crosses) are inferior to normal tadpoles with two sexually transmitted genomes (R. ridibunda × R. ridibunda crosses). If anything, tadpoles with R. esculenta fathers performed better: the only significant difference between paternal taxa was that R. esculenta males produced tadpoles that were larger on day 60 of the experiment and also at metamorphosis. In anurans, larger size at metamorphosis generally enhances fitness by translating into higher survival at the terrestrial life stage and into earlier maturity (Smith, 1987; Berven, 1990). Whether this finding can be generalized beyond the populations I used remains to be investigated, because based on allozyme markers, I cannot exclude that that all R. esculenta used in this experiment belonged to the same clonal lineage.
There was an effect of the environment on relative performance in traits measured early in development (day 30): tadpoles from R. esculenta fathers showed a stronger response to a decrease in food availability. This would be expected if the effects of deleterious mutations on clonal genomes are aggravated through environmental stress. However, performance of tadpoles with a clonal and a sexual genome did not fall below that of tadpoles with two sexual genomes at the low food level, and they actually performed somewhat better at the high food level (Fig. 2a,b). At later stages of tadpole development, this genotype-by–environment interaction was no longer evident. Hence overall, there is no indication that clonally transmitted R-genomes in the heterozygous state have detrimental effects on performance of R. ridibunda tadpoles.
Negative results of this kind are generally difficult to interpret. It could be that (1) the low food treatment did not produce sufficient environmental stress to expose deleterious mutations to selection, or that (2) clonal R-genomes do not possess more deleterious mutations than sexually transmitted genomes, or else that (3) clonal R-genomes are, as expected, more loaded with deleterious mutations, but these mutations are nearly completely recessive.
The highly significant effects of food level on growth and development suggest that the low food treatment did succeed in producing environmental stress. But clearly, it did not push tadpoles to their tolerance limit, as survival did not differ between food treatments. Also, it cannot be excluded that a different type of environmental stress could reveal a disadvantage for tadpoles possessing a clonally transmitted genome, although food limitation is certainly a biologically reasonable type of stress for tadpoles which commonly encounter severe competition (Smith, 1983; Alford, 1999).
The inviability of offspring from R. esculenta ×R. esculenta matings proves that there are deleterious mutations on clonal R-genomes, but not necessarily that their load is higher than that of sexual genomes, because the fixation of a single recessive lethal mutation within a clonal lineage is sufficient to cause the observed inviability. So how could clonal R-genomes not have a higher deleterious gene load than an average haploid genome in a sexual R. ridibunda population? First, there may not have been enough time for deleterious mutation accumulation to proceed to a detectable level. Unfortunately, the evolutionary age of clonal R-genomes is not exactly known. Rana esculenta probably immigrated into central Europe from areas of sympatry between R. ridibunda and R. lessonae in eastern Europe after the last glacial period (Würm age), about 10 000 years ago. Assuming a mean generation time of 2 years, R-genomes in hybridogenetic R. esculenta may have been transmitted clonally for roughly 5000 generations, but this estimate is uncertain. Some hemiclones may be younger, because immigration from eastern Europe was permanently possible after the last glaciation, or even older, if ancient clonal lineages were already present in refugial areas at the end of the Wuerm age (H. Hotz, personal communication). But in any case, their age appears sufficient for the accumulation of substantial amounts of deleterious alleles even at large population sizes (Lynch & Gabriel, 1990; Gabriel et al., 1993; Lynch et al., 1993). A second possible explanation was suggested by C. Som (personal communication): clonal R-genomes are passed on through females more often than through males, because male R. esculenta only sire daughters (see Methods). As mutation rates are generally higher in males than in females (Redfield, 1994; Hurst & Ellegren, 1998), autosomes in the clonal R-genome of R. esculenta may experience a lower input of deleterious mutations than autosomes in a sexual species which spend on average equal amounts of time in males and in females (C. Som, personal communication). Nevertheless, clonal genomes tend to accumulate mutations because of the lack of recombination. The net effect of these counteracting factors is currently being explored with mathematical models by C. Som.
Leaving this intriguing possibility aside, it appears likely that clonal R-genomes have indeed accumulated deleterious mutations, but that their coefficients of dominance are low, at least too low to be detected in this experiment. Clearly, spontaneous deleterious mutations vary in their degree of dominance (Simmons & Crow, 1977; García-Dorado & Caballero, 2000), but interclonal selection may have eliminated all but highly recessive mutations.
Whatever the appropriate explanation, the outcome of this experiment suggests that, at least in terms of larval performance, the hemiclonal hybrids from my source populations enjoy the full benefits of heterosis without suffering from detrimental effects of mutation accumulation. This may in part explain the high ecological success of R. esculenta, which is common and widely distributed in Europe (Graf & Polls Pelaz, 1989), and has been shown to perform equally or better than both parental species under a variety of environmental conditions (Semlitsch & Reyer, 1992; Semlitsch, 1993a, b; Ristet al., 1997; Hotz et al., 1999). Thus at present, there is no indication that the hybrid R. esculenta is an evolutionary ‘dead end road’ for clonal R-genomes. That clonally reproducing vertebrates can reach a substantial evolutionary age has convincingly been demonstrated in the hemiclonal fish Poeciliopsis monacha-occidentalis, where the oldest known clonal lineage has an estimated age of at least 100 000 generations (Quattro et al., 1992), and in gynogenetic salamanders of the genus Ambystoma (Spolsky et al., 1992). Considering also invertebrates, plants, and fungi (reviewed by Judson & Normark, 1996), there is substantial evidence for the evolutionary antiquity of several clonal lineages, suggesting that mutation accumulation does not inevitably lead to their rapid extinction. Additional factors such as low origination rates as a result of developmental constraints (e.g. Corley et al., 2001), and reduced adaptability to changing environments may indeed be required to explain the rarity of asexual reproduction.
I wish to thank S. Röthlisberger and M. Stauber for help with the experiment, and numerous people who helped to collect frogs. R. Altwegg, D. J. Hosken, H.-U. Reyer, C. Som, J. Van Buskirk, and two anonymous reviewers provided helpful comments on the manuscript. This study was conducted with approval and under the ethical guidelines of the ‘Veterinäramt Kanton Zürich’ (license number 88/99), and supported by the Swiss National Fund (31–40688.94 to H.-U. Reyer).