1In monogonont rotifers parthenogenetic reproduction allows population growth, and mictic (sexual) reproduction leads to the production of diapausing eggs. When amictic females are exposed to a mixis stimulus, they produce mictic daughters, whose eggs develop into males or, if fertilized, into diapausing eggs. Experiments showed that mictic offspring production is initiated by crowding in females of Brachionus angularis Gosse 1851, Epiphanes senta (O.F. Müller 1773) and Rhinoglena frontalis Ehrenberg 1853, just as it is in Brachionus calyciflorus Pallas 1766 and B. plicatilis Müller 1786.
2In B. calyciflorus, B. angularis, E. senta and R. frontalis, the propensity of amictic females to respond to crowding by producing mictic female offspring is low in the stem female hatching from a diapausing egg, but then increases after some generations. In many cases, only few mictic offspring are produced by crowded females of the second to the fifth generation, but the maximal response occurs only in later generations. Delayed sexual reproduction in early generations from the resting egg may be advantageous, because it first favours rapid population growth and later on maximizes resting egg production. However, it may be disadvantageous, if unpredictable environmental variation causes a population decline when sexual reproduction is still suppressed.
3The extent to which sexual reproduction is delayed varies among and within species. When strains from populations in temporary and permanent habitats were compared, sexual reproduction was significantly delayed in strains from temporary habitats in all species, whereas in B. calyciflorus and R. frontalis mixis was not significantly delayed in strains from permanent habitats. In E. senta mixis was significantly delayed in clones from both habitat types.
4Within all strains there was significant variation among clones in the propensity to produce mictic offspring, the extent to which sexual reproduction was delayed in the first generations after the stem female hatched, or both.
Diapause is a life-history trait that is found in the life cycle of many organisms. It has important implications both from a demographic and an evolutionary point of view. Individuals that enter diapause are usually protected against adverse environmental conditions. However, the deterioration of environmental conditions may often not be exactly predictable. Individuals entering diapause too early while favourable conditions continue to prevail, may experience a selective disadvantage, because they cease to reproduce and their gene frequencies in the population will decrease. On the other hand, individuals entering diapause too late will probably not survive.
Monogonont rotifers are ideal organisms to study the effects of diapause. Their life cycle is characterized by cyclic parthenogenesis and involves the production of fertilized diapausing eggs (Pourriot & Snell 1983; Gilbert 1992, 1993; Wallace & Snell 1991; Nogrady, Wallace & Snell 1993). Amictic females hatching from diapausing eggs reproduce parthenogenetically. They produce subitaneous eggs which develop immediately into diploid daughters. Parthenogenetic reproduction allows rapid growth of the population, which may consist of large numbers of different clones each derived from a separate diapausing egg (Gómez & Carvalho 2000). New diapausing eggs are produced by sexual reproduction (mixis) which is initiated by the production of mictic females. Mictic females are produced by amictic females in response to mixis stimuli which in some species as Brachionus calyciflorus or B. plicatilis are related to high population density or crowding (Gilbert 1992, 2002, 2003; Stelzer & Snell 2003). Mictic females parthenogenetically produce haploid male offspring if unfertilized. When fertilized by a male they produce diapausing eggs, which enter diapause.
The response to the mixis stimulus, i.e. the proportion of mictic offspring that is produced by amictic females, can be very variable. There are few reports of incidents where the proportion of mictic females in a population reaches almost 100% (Pourriot 1965; Aparici, Carmona & Serra 1996). In most cases a certain fraction of the offspring continues to be amictic. This is a bet-hedging strategy (Seger & Brockmann 1987; Philippi & Seger 1989), which ensures that diapausing eggs are produced to secure survival of the clone's genes in case of future habitat deterioration. At the same time it allows for continuing population growth in case environmental conditions remain favourable.
Recently it has been shown that the response to the mixis stimulus can also vary across different generations of a clone. Gilbert (2002) has found for clones of a strain of B. calyciflorus that the propensity to respond to the mixis stimulus and to produce mictic offspring is very low in females of the first generations after hatching from the diapausing egg. This propensity increases in later generations and reaches a maximum after about 12 generations. At that point about 50% of the crowded populations consist of mictic females. This transgenerational plasticity in the crowding response should have several important effects (Gilbert 2002, 2003). If the crowding response was not species-specific, as appears to be the case in B. plicatilis (Carmona, Serra & Miracle 1993), a generational delay in mixis would assure that sexual reproduction of a species is initiated at higher population densities of that species, when more mictic females and diapausing eggs can be produced. Perhaps more importantly, on the clonal level, a delay in sexual reproduction may prevent a clone which hatches into an environment already densely populated by conspecific clones from being induced to produce mictic offspring before it can increase its frequency in the population (Gilbert 2002, 2003).
Gilbert (2002) also showed that the mixis response varies significantly between clones in B. calyciflorus. Some clones consistently produce fewer mictic offspring than others. This may be important ecologically, because clones with a low propensity for sexual reproduction may risk local extinction before producing new diapausing eggs but may reach high population densities, whereas clones with a high propensity will produce at least some diapausing eggs at the cost of reduced population growth.
The trade-off between parthenogenetic population growth and sexual production of diapausing eggs is universal in monogonont rotifers (Serra & King 1999). Therefore, a delay of sexual reproduction may have evolved in many other species as well. In this study, we determine the extent of a mixis delay in early generations from the diapausing egg in Brachionus calyciflorus Pallas 1766, Brachionus angularis Gosse 1851, Epiphanes senta (O.F. Müller 1773) and Rhinoglena frontalis Ehrenberg 1853. Mixis stimuli for B. angularis, E. senta and R. frontalis have not been identified, but preliminary observations suggested that they are density dependent. Therefore, we first demonstrated that crowding induces the production of mictic offspring in these species. Then, we tested the hypothesis that mixis delay is a variable trait that may be present in some populations of a species and absent in others. We assumed that the trait would be beneficial by maximizing diapausing egg production in more predictable long-lived habitats such as permanent water bodies, and that it could be disadvantageous in unpredictably varying habitats such as temporary waters that might dry up while sexual reproduction is still suppressed. Accordingly in some species, we tested different strains that came from a variety of habitats, both permanent and temporary, and we compared the clones of each strain in their propensity to respond to crowding.
Materials and methods
Experiments were conducted with clones from strains of B. calyciflorus, B. angularis, E. senta and R. frontalis. The B. calyciflorus strain was kindly provided by T. W. Snell. It was originally collected from Piedmond Park Pond (Atlanta, Georgia), a permanent and highly eutrophic pond. For the experiment testing transgenerational plasticity in mixis response, distinct clonal lineages of a strain originated from females which hatched from diapausing eggs produced by clonal cultures of that strain. The B. angularis strain originated from Laguna los Juncos, a temporary pond on Estación Perito Moreno, Río Negro Province, Patagonia, Argentina, and was collected in April 2002. Females hatching from diapausing eggs in soil samples were used to establish different clonal lineages for the experiments. The strains of E. senta and R. frontalis were collected in April 2001 from permanent and temporary habitats at the Oder River near Schwedt (Brandenburg), Germany. One strain of each species came from the temporary floodplains of the Oder River, and another strain of each species was collected from a permanent pond near Bölkendorf, 17 km south-east of the floodplain location. Amictic females of these strains were collected in the field and used to start genetically distinct clone cultures (as shown by allozyme analyses, T. Schröder, unpublished data). Diapausing eggs from these different clone cultures were used to start the distinct clonal lineages for the experiments.
Cultures of B. calyciflorus, B. angularis and E. senta were maintained in modified MBL (Marine Biological Laboratory) medium (Stemberger 1981) on Cryptomonas erosa var. reflexa Marsson. Clones of R. frontalis were cultured in filtered (0·45 µm polyether sulphone membrane) lake water from Post Pond, Lyme, NH, on a diet of Rhodomonas minuta Skuja. Both algae were cultured on modified MBL medium. The algal cultures and the rotifer cultures of B. calyciflorus and B. angularis were kept at 20 °C and at a photoperiod of L : D 16 : 8 h. Experiments with these species were run under the same conditions. Cultures of E. senta and R. frontalis were maintained at 15 °C, which was necessary because both species had increased egg mortality at higher temperatures. All experiments with these two species were conducted at this temperature and the same light conditions as above.
tests of crowding as mixis stimulus
Mixis stimuli for B. angularis, E. senta and R. frontalis had not been identified so far. Known mixis stimuli (reviewed in Gilbert 1992) include crowding and changes in photoperiod, as well as the uptake of α-tocopherol which is limited to the genus Asplanchna. Therefore, it was tested whether females of E. senta would respond to crowding or differences in photoperiod as stimuli for mictic offspring production. Photoperiod proved to have no effect in E. senta (T. Schröder, unpublished data), and only effects of crowding were tested in the two other species. Newborn amictic females were cultured individually in a low-density and a high-density environment and their offspring were isolated and typed as mictic or amictic. Different volumes for the low-density and high-density treatments were used for each species, as the species differed considerably in size. It should be noted that these experiments were not conducted to determine threshold densities, therefore relatively high densities were chosen for the high-density treatments.
1Epiphanes senta. A high- and a low-density environment in combination with a short and a long photoperiod were tested in a 2 × 2 factorial experiment with females of a clone from the floodplain strain. Low-density treatments were Petri dishes (∅ 60 mm) filled with 15 ml medium for each female; high-density treatments were concavities of a 24-well tissue culture plate filled with 2 ml medium for each female. Each density treatment was combined with a long (L : D 16 : 8 h) and a short (L : D 7 : 17 h) photoperiod. Culture medium was MBL medium with C. erosa (3 × 104 cells ml−1).
2Brachionus angularis. Females were cultured at a low density by placing a single female in a Petri dish (∅ 35 mm) with 5 ml medium and at a high density by placing single females in concavities of a 96-well tissue culture plate filled with 400 µl medium. MBL medium with C. erosa (2 × 104 cells ml−1) was used in this experiment.
3Rhinoglena frontalis. Three different population densities were tested with a clone from the floodplain strain. Single females were cultured in Petri dishes (∅ 60 mm) filled with 10 ml medium for the low-density treatment, in concavities of a 24-well tissue culture plate filled with 800 µl medium for the intermediate density treatment, and in concavities of a 96-well tissue culture plate filled with 275 µl medium for the high density treatment. Culture medium was filtered lake water with Rhodomonas minuta (6·5 × 104 cells ml−1).
In all experiments, treatment conditions were kept constant by removing all offspring and transferring the parental females to fresh medium every day. Offspring were collected from Rhinoglena females until the mothers died. The experiments with B. angularis and E. senta were terminated earlier, and collection of offspring was stopped before the deaths of the mothers. Offspring were raised individually in concavities of 96-well tissue culture plates filled with 400 µl medium until they themselves produced eggs or offspring and could then be typed as mictic or amictic. Since determination of reproductive state in several oviparous species occurs well before the egg is deposited (Riggs & Gilbert 1972), there is no problem in using such a small volume to culture the offspring before they were typed. For each species, proportions of mictic offspring produced in the different treatments were compared by anova of arcsine-transformed values. The statistical analysis was done with JMP software (SAS Institute 2001).
transgenerational plasticity in the response to crowding
The propensity to produce mictic offspring in response to a crowding stimulus was tested in clones of B. calyciflorus, B. angularis, R. frontalis and E. senta in the first generation that hatched from the diapausing egg and in several of the following generations. Six clones of the B. calyciflorus strain were tested in the first, fourth and tenth generations. 12 clones of the B. angularis strain were tested in the first, sixth and tenth generations. In R. frontalis, five clones of the floodplain strain and six clones of the strain originating from the permanent pond were tested in the first, second, third, sixth and tenth generations. In E. senta, five clones of the strain collected from the permanent pond were tested in the first, sixth and thirteenth generations. In an additional experiment, another five clones of the strain from the permanent pond and five clones of the floodplains strain were tested in the first, fifth and seventh generations. All experiments were conducted in almost the same way as those described by Gilbert (2002) with another strain of B. calyciflorus. All clonal lineages were started with females hatching from diapausing eggs, these being the first generation. The first few offspring of these females, the second generation, were isolated and cultured individually at low population density in Petri dishes containing 15–20 ml medium. Third and subsequent generations were obtained by isolating the first offspring of the preceding generations and culturing them in Petri dishes.
To test a clonal response to crowding at a given generation, a single female of that generation was transferred to a Petri dish (∅ 35 mm) with 5 ml medium to start a population. Algal concentrations were 3 × 104 cells ml−1 in experiments with B. calyciflorus, B. angularis and E. senta, and 1 × 105 cells ml−1 in experiments with R. frontalis. In all experiments several populations were always started with siblings of the same generation to ensure that at least one of them would be amictic, and one of those was chosen for final analysis. The only exception would be the test with the first generation which was started just with the female that hatched from the diapausing egg. Each population started with a single female was allowed to grow so that the density in the Petri dish would gradually increase. The whole population was transferred to a new Petri dish with fresh medium every day in experiments with E. senta, and every other day in experiments with all other species, until the population size had reached about 30 females (40 females in populations of R. frontalis clones). The final population size was kept as constant as possible to standardize the mixis stimulus in all experimental populations of each species. The proportion of mictic females was determined in each final population. In experiments with B. calyciflorus and B. angularis, ovigerous females could be typed as amictic or mictic based on the size of the eggs they were carrying, and juvenile females were raised individually in concavities of 96-well tissue culture plates until they produced eggs and could be typed. Epiphanes senta deposits its eggs on the substrate, and R. frontalis females are viviparous. Therefore, all females in experiments with these species had to be transferred individually to concavities of tissue culture plates and raised until they produced eggs or offspring and could be typed.
Depending on species and experimental temperatures, it took about 4–10 days until the populations reached their final population size. At this time the populations consisted of daughters, grand-daughters and maybe great grand-daughters of the initial females. So the response to crowding is measured over three to four generations. For matters of simplicity we refer to these generations as the generation of the initial female that started the population.
Proportions of mictic females in the final populations were arcsine-transformed and analysed using anova and Tukey–Kramer honest significant difference tests for multiple comparisons. The different clones of a strain were used as replicates. Also, clone and generation effects were analysed for strains, using ancova. Statistical analyses were done with JMP statistical software (SAS Institute 2001).
mixis response to crowding in b. angularis, e. senta and r. frontalis
Amictic females of all three species produced a significantly higher proportion of mictic offspring in the high-density environment than in the low-density environment (Table 1). In B. angularis, there was almost no mictic offspring production at a density of 200 ind. l−1 (= 1 female 5 ml−1), whereas 31·4% of the offspring were mictic at the high density (1 female 400 µl−1). The difference was highly significant (anova; F1,34 = 33·74, P < 0·0001). In E. senta, females produced significantly more mictic offspring under high-density conditions, and there was no effect of photoperiod and no interaction of density and photoperiod (anova; F3,29 = 7·36, P = 0·0008; effect of density: F1 = 21·52, P < 0·0001; effect of photoperiod: F1 = 0·34, P = 0·566; interaction effect: F1,29 = 0·68, P = 0·418). In R. frontalis, mictic offspring production was also significantly increased by crowding (anova; F2,31 = 7·67, P = 0·002). Pairwise comparisons showed that females produced significantly more mictic offspring at the highest density (1 female 275 µl−1) than at the lowest density (1 female 10 ml−1). Mictic offspring production at the intermediate density (1 female 800 µl−1), although elevated, was not significantly different from that at the other two densities (Tukey–Kramer honest significant difference tests, P < 0·05).
Table 1. Mictic offspring production in Brachionus angularis, Epiphanes senta and Rhinoglena frontalis in low- and high-density treatments and, for E. senta, in long and short photoperiods. Offspring numbers of Rhinoglena females are given for the first 8 days of the experiments when all females were still alive
Mean % mictic offspring (1 SE)
Mean number offspring produced (1 SE)
No. of replicate females
E. senta (L : D 16 h : 8 h)
E. senta (L : D 7 h : 17 h)
High-density treatments did not negatively affect fecundity in any species (Table 1). Females in high-density treatments produced slightly more offspring than in low-density treatments. This difference was significant in R. frontalis (anova, F2,28 = 4·59, P = 0·019), but not in either B. angularis (Student's t-test, t34 = 1·71, P = 0·096) or E. senta (anova, F3,29 = 2·38, P = 0·090).
delayed mixis response to crowding
There were considerable differences among species, and also between strains within the same species, in the proportion of mictic females in the populations started with females of the first and several of the subsequent generations. In clones from the Georgia strain of B. calyciflorus the proportion of mictic females was already high in populations started with the stem female (Fig. 1a) hatching from a diapausing egg. In this strain, there was no effect of generation; proportions of mictic females in populations initiated with females from the first, fourth and eighth generation did not differ significantly (anova, F2,15 = 0·37, P = 0·698).
In B. angularis there was a significant increase in the proportion of mictic females when populations of generations 1, 6 and 10 were compared (Fig. 1b; anova, F2,32 = 3·39, P = 0·046). However, a considerable number of mictic females were already found in populations generated by females which hatched from the diapausing eggs.
Differences between strains from floodplain and pond habitats were observed in R. frontalis. In the pond strain proportions of mictic females were about the same in all generations tested. In the floodplain strain, the proportion of mictic females increased with generation (Fig. 2). A two-way anova with strain and generation as effects demonstrated that differences in the proportion of mictic females among generations and between floodplain and pond strain were significant (effect of generation: F1,49 = 6·20, P = 0·016; effect of strain: F1,49 = 9·71, P = 0·003). Also, there was a significant interaction effect (F1,49 = 4·07, P = 0·049), indicating that sexual reproduction in early generations of the floodplain strain was significantly lower than in early generations of the pond strain.
Mixis induction in both strains of E. senta was delayed for a long time after ex-diapausing females hatched from diapausing eggs (Fig. 3a). Populations initiated by females of the first and fifth generations were completely amictic. This was the case in both the floodplain and the pond strain. In the seventh generation there was a slight but significant increase in mictic females, but no significant difference between strains (two-way anova with generations and strains as effects; effect of generation: F2,24 = 3·44, P = 0·047; effect of strain: F1,24 = 0·26, P = 0·613). In the second experiment with clones from the pond strain (Fig. 3b), responses to the crowding stimulus in generations 1 and 6 were similar to those observed in generations 1–7 in the preceding experiment. However, in the populations started by females of the thirteenth generation, the proportion of mictic females rose to a mean of 23%. Differences in the proportion of mictic females were highly significant (anova, F2,12 = 26·3, P < 0·0001), with significant differences between the first and sixth, first and thirteenth, and sixth and thirteenth generations (Tukey–Kramer honest significance difference tests, P < 0·05).
correlation between habitat type and delay of mixis response
A comparison of the proportions of mictic females in the first and the last generations tested shows significant differences in most of the species and strains tested (Fig. 4). Only in the Georgia strain of B. calyciflorus and in the pond strain of R. frontalis were proportions of mictic females in populations initiated from first generation females not significantly lower than in subsequent generations. However, there was no clear pattern of mixis delay as a function of habitat type. The Georgia strain of B. calyciflorus and the pond strain of R. frontalis both came from permanent ponds and showed no mixis delay. In contrast, the E. senta strain, which also came from a permanent pond, showed no response to crowding in the first and fifth generations. Sexual reproduction was significantly delayed in all strains from temporary habitats.
clonal variation within strains
Clonal variation proved to be an important component in explaining variation in the mixis response to the crowding stimulus (Table 2). In all strains, clonal effects contributed significantly to differences in the mixis response. In B. angularis and the pond strain of E. senta, analyses of covariance revealed that both generation and clone effects significantly explained variation in the mixis response. In these strains, there were no significant interactions of clonal effects and generational effects, indicating that the propensity to produce mictic offspring increased with generation at a similar rate in all clones. In the Georgia strain of B. calyciflorus and the pond strain of R. frontalis, there was no effect of generation on the mixis response, but a significant clone effect indicated that the propensity of some clones to produce mictic offspring in response to crowding was consistently lower than in others. There was also a significant clone × generation interaction in the R. frontalis pond strain; there was a strong correlation between mixis response and generation in two clones (Pearson's correlation coefficient R > 0·99) and a weak or no correlation in the other four. Variation in the mixis response of the floodplain strain of R. frontalis was explained by an effect of generation and an interaction between clone and generation, because only one of the tested floodplain clones did not show an increase in the mixis response with generation.
Table 2. ancova results testing effects of clone and generation from the diapausing egg on the mixis response to crowding. Clone effects were tested with generation as a covariate. Significant P-values (P ≤ 0·05) are printed in bold
Species and strain origin
Brachionus angularis (temporary pond Argentina)
Generation × Clone
Epiphanes senta (permanent pond near Oder River)
Generation × Clone
Rhinoglena frontalis (permanent pond near Oder River)
Generation × Clone
Rhinoglena frontalis (floodplain of Oder River)
Generation × Clone
Brachionus calyciflorus (permanent pond Georgia)
Generation × Clone
So far, a transgenerational delay in the mixis response to crowding has only been described in detail for Brachionus calyciflorus (Gilbert 2002, 2003). The experiments of the present study clearly show that a mixis delay in the first generations after hatching from the diapausing egg also exists in several other species and that it may be a phenomenon that is widespread. Other studies indicate that such a delayed mixis response in the first generations also exists in Brachionus plicatilis (Hino & Hirano 1977) and Asplanchna brightwelli (Gilbert 1983). With the exception of A. brightwelli, all these species respond to density as a mixis stimulus, which thus far is the most common mixis-inducing stimulus among those known and reviewed by Gilbert (1992).
The extent to which the mixis response is delayed is not species-specific, but varies among strains. The B. calyciflorus strain from Georgia used in the present study showed no reduced response to crowding in early generations from the diapausing egg, whereas a strain of this species from Florida did (Gilbert 2002). Similarly, the strains of R. frontalis from the floodplain and the permanent pond near the Oder River showed significant differences in their mixis response to crowding. A significant suppression in this response was observed in early generations in the former strain but not in the latter. Although the two B. calyciflorus strains from Georgia and Florida originated from quite distant habitats, cross-mating tests and analyses of mitochondrial and rDNA sequences showed that they are genetically very similar and belong to the same biological species (Gilbert & Walsh, in press). Conversely, the R. frontalis strains originated from populations that are genetically differentiated with limited gene flow between them, even though their habitats are just 17 km apart. This has been shown by allozyme analyses of four polymorphic loci (T. Schröder, unpublished data).
The observed variation within strains suggests high variation within populations. It seems likely that mixis and mixis delay are traits that are under strong selection, so it can be expected that local adaptations in these traits evolve. Cyclic parthenogenesis in the life cycle of monogonont rotifers includes important trade-offs between parthenogenetic and sexual reproduction that are affected by a transgenerational delay of mixis. Population growth and increases in frequency of individual clones within the population depend on parthenogenetic reproduction of amictic females. Therefore allocating resources to the production of mictic offspring decreases a clone's intrinsic growth rate and thus its frequency within the population (Snell 1987; Serra & Carmona 1993; Serra & King 1999). Clones with a low propensity to produce mictic offspring have a selective advantage over those with a higher propensity as long as favourable environmental conditions permit the existence of parthenogenetically reproducing females (Fussmann, Ellner & Hairston 2003).
If environmental conditions deteriorate and only diapausing stages are able to survive, selection will no longer favour rapid parthenogenetic reproduction. Then those clones that have produced the most diapausing eggs up to that point should have a selective advantage. Obviously, a trait such as mixis delay should be selected against if deterioration of the environment is likely to occur after only a few generations from diapausing egg hatching.
While this study clearly demonstrated that there is intraspecific variation in mixis delay among strains, our results do not support a relationship between mixis delay and habitat permanence. The B. calyciflorus and R. frontalis strains that displayed no significant mixis delay in the early generations originated from populations in more predictable permanent habitats, whereas strains with a significant mixis delay were collected from temporary habitats, which were expected to be less predictable, such as the Oder River floodplains for R. frontalis and a temporary pond for B. angularis. Furthermore, the Florida strain of B. calyciflorus investigated by Gilbert (2002) came from a temporary pond. On the other hand, the proportion of mictic females was as high as 66·0 ± 2·3% (mean ± 1 SE of 10 clones) in experimental populations initiated by first generation females of an Australian strain of B. calyciflorus from a temporary floodplain of the Murray-Darling river basin (J. J. Gilbert, unpublished data). Mixis delay was most pronounced in the E. senta strains with no mictic female production in generations 1 and 5, irrespective of their origin from permanent or temporary habitats. It seems obvious that many generations of mixis delay should be selected against in very short-lived temporary habitats, likely to dry up before mictic reproduction can be induced. In these habitats a conservative bet-hedging strategy sensuSeger & Brockmann (1987) may be more appropriate, i.e. an early production of diapausing eggs as soon a possible at the expense of parthenogenetic population growth. Such a strategy may exist in populations of E. senta and also Hexarthra brandorffi in temporary ponds of the Chihuahuan desert in Texas, where females produce a high percentage of mictic offspring already in the first generations (E. J. Walsh, unpublished data). These ponds are very short-lived; they are filled with water by rain in the spring and may dry out again within a few days.
There may be several reasons why mixis delay may be absent in populations inhabiting more predictable permanent waters or present in populations from less predictable temporary habitats. The permanence of a temporary habitat has to be considered in relation to the generation times of the rotifers inhabiting it. For example, our results for E. senta indicate that sexual reproduction is suppressed for six to ten generations. Data from life table experiments (T. Schröder, unpublished data) suggest that these generations should develop within 4–5 weeks, assuming water temperatures between 5 and 15 °C. Continuously high water levels of the Oder River which lead to inundation of the floodplains for 6 weeks or longer occur with a frequency of about 70% over the years in March and April, when E. senta is most abundant (Schröder 2001). So, even with mixis delay, E. senta clones should be able to produce diapausing eggs and stock a diapausing egg bank in the floodplain sediments in most years so that local extinction can be prevented in years when flooding ends early.
Santer & Lampert (1995) have suggested that avoidance of food limitation may be a possible ultimate factor explaining summer diapause of certain cyclopoid copepod species. Likewise, periods of food limitation may be avoided by rotifers through diapause. In this case unpredictable variation in food resources – in permanent as well as in temporary habitats – may be another reason why a mixis delay could be disadvantageous. It has been argued that the production of diapausing eggs requires more food resources than the production of subitaneous eggs (Gilbert 1993), because they contain much more energy reserves than subitaneous eggs such as glycogen and lipids (Wurdak, Gilbert & Jagels 1978). Therefore, a conservative bet-hedging strategy with mictic offspring production as soon as population density permits a successful fertilization of mictic females and food resources are sufficient to produce energy-rich diapausing eggs, could be advantageous.
A striking result for all strains was the effect of clonal differences on the propensity to respond to the mixis stimulus. Previous studies with B. plicatilis and B. calyciflorus have shown significant variation among strains in the propensity to produce mictic offspring (Hino & Hirano 1977; Pourriot & Snell 1983; Carmona, Serra & Miracle 1994; Gilbert 2002), and in the timing of sexual reproduction (Aparici, Carmona & Serra 2001). Hino & Hirano (1977), Aparici et al. (2001) and Gilbert (2002) demonstrated variation among clones within the same strain. Genetic variation is necessary for the evolution of locally adapted traits affecting sexual reproduction, but on the other hand clonal diversity for propensity to produce mictic offspring, and for mixis delay, may also be maintained permanently in a population by the storage effect of the diapausing egg bank. Fluctuating selection may favour clones with mixis delay in some years and those without mixis delay in others, leading to varying success in sexual reproduction and varying recruitment of the different phenotypes to the diapausing egg bank in different years. The diapausing egg bank may then act as a buffer in years of poor recruitment. The storage effect has been associated with the coexistence of competing species and documented for field populations of cladoceran species (Cáceres 1997). In theory, it can also explain the maintenance of genetic variation within a species (Hairston, Ellner & Kearns 1996) and thus the coexistence of competing clones of a cyclic parthenogen in a temporally varying environment.
We thank Mr and Mrs Ernest Magnien for greatly appreciated research funds, Matthew P. Ayres for some advice with statistics, and two anonymous referees for comments that improved the manuscript. TS was supported by a postdoctoral fellowship from the Deutsche Forschungsgemeinschaft.