Rapid evolution of sex frequency and dormancy as hydroperiod adaptations

Authors


Correspondence: Hilary A. Smith, Department of Biological Sciences, University of Notre Dame, Galvin Life Science Bldg, Rm 321, Notre Dame, IN 46556-0369, USA. Tel.: +1 303 549 4793;

e-mails: hsmith9@nd.edu; hasmith.biology@gmail.com

Abstract

Dormancy can serve as an adaptation to persist in variable habitats and often is coupled with sex. In cyclically parthenogenetic rotifers, an asexual phase enables rapid population growth, whereas sex results in diapausing embryos capable of tolerating desiccation. Few studies have experimentally tested whether sex–dormancy associations in temporary waters reflect evolution in response to the short hydroperiod selecting for diapause ability. Here, we demonstrate evolution of higher propensity for sex and dormancy in ephemeral rotifer cultures mimicking temporary ponds, and lower propensity in permanent cultures. Results are consistent with rapid evolution, with evolutionary changes occurring in a short timeframe (385 days, ≤ 84 generations). We also provide insight into mechanisms for rapid evolution in basal metazoans, discussing potential roles of new mutations, recombination and clonal selection.

Introduction

In ephemeral aquatic systems, taxa from macrophytes to zooplankton produce dormant stages, for example, resting eggs (diapausing embryos) or seeds that remain quiescent during dry spells and hatch when water returns (Brock et al., 2003). For many facultative asexuals and cyclical parthenogens that engage in asexual and sexual reproduction, dormant stages are the product of sex (Simon et al., 2002; Brock et al., 2003; Cáceres & Tessier, 2004; Serra & Snell, 2009). Sex–dormancy associations are found in organisms ranging from zooplankton to aphids (Simon et al., 2002; Serra & Snell, 2009); sex and dormant seed banks also play a role in persistence of aquatic plant communities (Lokker et al., 1997). Yet little is known of the mechanism and pace of evolution of dormancy and sex as adaptations to stressors such as desiccation.

Research on the loss of sex and dormancy suggests the potential for these traits to undergo rapid evolution, whereby contemporary evolutionary changes occur swiftly enough to affect ecological dynamics (Hairston et al., 2005; Ellner et al., 2011). Several studies have documented sex loss in the cyclically parthenogenetic rotifer (zooplankton) Brachionus calyciflorus in periods of months to years (Bennett & Boraas, 1989; Fussmann et al., 2003; Becks & Agrawal, 2010; Stelzer et al., 2010). Typically, asexual brachionid females produce subitaneous female eggs. The females also excrete a protein signal into the water that accumulates with increasing rotifer density (Snell et al., 2006). At high female densities, the signal reaches a threshold concentration that induces sex in a process analogous to microbial quorum sensing (Kubanek & Snell, 2008). A fraction of the asexual brachionid females start producing sexual daughters; the rest continue with asexual reproduction. The sexual daughters produce males (which are smaller, haploid, and do not feed), or if fertilized by a male, produce diapausing embryos that hatch after a period of obligate dormancy (Wallace & Smith, 2009; Wallace & Snell, 2010). Yet some rotifers lose ability for sex. It has been proposed that brachionid sex loss constitutes an eco-evolutionary feedback, with the ecological dynamic of density-dependent sex induction undergoing evolution, which in turn impacts population densities (Stelzer, 2012).

Sex–dormancy associations and their evolution provide an opportunity for investigating both the role and mechanism of rapid evolution. The significance of eco-evolutionary changes and rapid evolution has been questioned, but preliminary evidence suggests these forces do play important roles in community and ecosystem structure and function (Ellner et al., 2011; Becks et al., 2012). Rapid evolution of sex and dormancy may be critical for species' survival, particularly given the potential for climate change to alter hydroperiod and habitat permanence (Brooks, 2004). Field-based studies show and theoretical studies predict correlations of rotifer traits associated with sexual reproduction and dormancy to hydroperiod (Serra & King, 1999; Schröder et al., 2007; Serra & Snell, 2009; Gilbert & Diéguez, 2010; Campillo et al., 2011), with a tendency for higher levels of sex and dormancy in ephemeral habitats. Several studies have attributed rapid evolution and eco-evolutionary dynamics to clonal selection – including assays of rotifer sex loss (Fussmann et al., 2003), aphid growth rate (Turcotte et al., 2011) and cladoceran parasite resistance (Duffy & Sivars-Becker, 2007; Fussmann et al., 2007). However, the necessity of pre-existing variation and generality of clonal selection as the mechanism for rapid evolution of traits such as dormancy remain unknown.

Here, we describe an experimental evolution study designed to test the hypothesis that the requirement for dormancy in environments mimicking ephemeral or permanently filled aquatic habitats results in rapid evolution of diapause and sex, and discuss the role of clonal selection. Laboratory Brachionus plicatilis s.s. cultures were maintained for 385 days divided into six, 9-week growth seasons; ephemeral systems were reset (restarted from diapausing embryo hatchlings) at the end of each season, but permanent cultures were not reset. Hence, ephemeral cultures underwent five rounds of sex between seasons, whereas in permanent cultures, sexual embryos were removed and only asexual propagation contributed to population growth. We compared the cultures for evolutionary changes in sex-induction propensity, sex frequency, and the production and hatching frequency of diapausing embryos. We also assayed lifespan and fecundity for both asexual and sexual females to test for evolution of traits less directly associated with sex and dormancy, and monitored population density to assess a potential cost of sex and diapause to population growth. The results are consistent with rapid evolution of higher propensity for sex and dormancy in ephemeral as compared with permanent cultures.

Materials and methods

Cultures and hydroperiod treatments

Subcultures of 15 separate lineages of B. plicatilis s.s. from the pond Poza Sur, Spain, were provided by our colleague Dimas-Flores. The lineages differed in the density for sex induction (range, 2.2–149.7 females mL−1) as described in (Carmona et al., 2009). We confirmed all lineages were members of the species B. plicatilis s.s. with restriction fragment length polymorphism of cytochrome c oxidase subunit 1 and gel electrophoresis, following Berrieman et al. (2005). Ten asexually produced clonemates from each of the 15 lineages were used to inoculate each of six chemostat containers; chemostats were maintained for 385 days.

Cylindrical Cellift bioreactors (Ventrex; Portland, ME, USA), each holding ca. 570 mL of medium, were used as chemostat (continuous flow) culture containers, with a flow rate of ca. 150 mL day−1 controlled by a Manostat cassette pump. Inflow medium consisted of the green alga Tetraselmis suecica grown on F medium (Guillard & Ryther, 1962) and diluted with 15 ppt artificial seawater (ASW, Instant Ocean® sea salts) to 1 million cells mL−1. Chemostats were maintained in an Environmental Growth Chamber at 22 °C below a 40 W fluorescent light; location of the chemostats was randomized and altered every 9 weeks. Unless otherwise noted, all bioassays were conducted at these conditions (22 °C, constant light).

Three chemostats were randomly chosen to represent an ephemeral environment and three a permanent environment. At the end of every 9 weeks, all chemostats were cleaned to remove algal build-up on the walls. Debris settled at the base of the chemostat – including diapausing embryos, which sink – was collected and set aside. Permanent chemostats were re-inoculated with their adult animals and old medium after chemostat cleaning and removal of debris and diapausing embryos. Removal of diapausing embryos was conducted to reflect conditions of a deep permanent lake, where embryos may sink to layers too deep for exposure to hatching stimuli (e.g. light) (García-Roger et al., 2005). At the time of permanent chemostat cleanings, ephemeral chemostats were reset to mimic the desiccation and refilling of an ephemeral pond. For the reset, each ephemeral chemostat was emptied and then re-inoculated with fresh algae and hatchlings of diapausing embryos that had been harvested from that chemostat 6 weeks prior and kept in diapause (dark, 5 °C). Immediately before a reset, approximately 800 healthy-looking [categories I and II sensu (García-Roger et al., 2005)] diapausing embryos per ephemeral chemostat were incubated for ~ 74 h to hatch (22 °C, constant light from the 40 W lamp). All hatchlings were used to refill (restart) the chemostat (Fig. 1). Periods between these 9-week resets or cleanings are referred to as growth seasons. Conditions in permanent chemostats allowed for continuous asexual growth. A total of five resets between seasons in ephemeral chemostats, however, interrupted asexual growth and required re-colonization by hatchlings of sexually produced diapausing embryos.

Figure 1.

Schematic of timeline for chemostats and bioassays. Horizontal line with arrow represents the 385-days experiment. Vertical lines denote the initial inoculation and resets between 9-week growth seasons in ephemeral chemostats. Permanent chemostats were not reset. Symbols show approximate timing of events: c, collection of diapausing embryos for use at the next reset (ephemeral cultures only; end of 3rd week except postponed to day 44 in season 1); S, sex-induction density assay (begun 41–45 days into season); L, life history assay of lifespan and fecundity (begun 50–58 days into season); D/H, harvest of all diapausing embryos in each chemostat to assay diapausing embryo production and hatching (last day of the season). All hatching assays were performed 9.5 weeks after the chemostats were taken down on day 385. Sex frequency assays tested the frequency of sexual females (season 6 only). Population density data were used to assess changes in asexual female density (all seasons).

Bioassays

Sex-induction density

Bioassays testing relative propensity for sex induction were conducted at the first and final two growth seasons. Protocols were adapted from the study by Carmona et al., 2009; recording the female density at which males first appeared – signifying the transition to sex. Low densities suggest early induction or a high propensity for sex. To minimize maternal effects and allow induction densities to reflect intrinsic properties of females rather than chemostat conditions, we performed a pre-experimental step following (Carmona et al., 2009): that is, two asexual generations were maintained individually in Petri plates in low-density conditions (one female in 25 mL medium). After two generations, a group of females from each hydroperiod treatment was assayed for sex-induction density, with numbers of females tested as follows: season 1, N = 19 per hydroperiod; season 5, N = 22 ephemeral and 35 permanent; season 6, N = 30 ephemeral and 32 permanent. For the bioassay, females were allowed 10 days to reproduce. Adult females typically produced progeny at age 1–2 days and may continue reproducing for ~10 days. Owing to nonlinear dynamics in sex induction, culture volume can affect inducing densities (Carmona et al., 2011). Our use of a single, small volume (0.5 mL) precludes determination of absolute threshold densities for induction; we likely overestimate absolute thresholds. However, use of the same conditions and volume allows comparison of relative sex-induction propensity between hydroperiods by recording the female density when males first appeared.

Sex frequency

In the final season, we monitored the frequency of sexual relative to total females every 8 days. Asexual and sexual females are indistinguishable morphologically and can only be differentiated based on their progeny (asexual females produce females, sexual females produce males or diapausing embryos). On some dates, most adult females were nonovigerous; some may have been post-reproductive (sexual status indeterminate). Thus, to perform this assay on females whose sexual status we could determine, we analysed sexual female frequency of cohorts of subitaneous female eggs laid in the chemostat. Because brachionids' sexual status is determined in utero (Snell et al., 2006; Gilbert, 2007), the status of these eggs was determined while they formed in the chemostat. To begin the assay, we collected ovigerous asexual females from each chemostat. Females were maintained separately in 1 mL medium in 24-well plates until their eggs hatched; 1 hatchling per female was individually transferred to fresh medium and produced male or female progeny in ~2 days. Analyses compared mean frequencies of sexual females (N = 8 assays or days); each frequency was from testing 19–24 females per chemostat per assay.

Diapausing embryo production

At the end of each season, all diapausing embryos were collected by filtration of settled debris at the base of each chemostat (53 μm Nitex mesh), allowed to air dry and weighed. These end-of-season collections were used to assess diapausing embryo production (not for ephemeral resets). From each end-of-season collection from each chemostat, a random sample (~5 mg) of dried diapausing embryos was weighed, rehydrated in 4.75 mL 15 ppt ASW and sub-sampled (100 μL). All healthy-looking [categories I and II, sensu (García-Roger et al., 2005)] embryos in the sub-sample were counted (mean number of healthy-looking diapausing embryos per sub-sample ± 1 SE, 105 ± 13). These counts were used to determine the density of the dried embryos (embryos/gram), which was multiplied by the total mass of dried embryos from the chemostat to yield total (healthy-looking) diapausing embryo production per chemostat per season. From ephemeral cultures, at least 1000 embryos also had been harvested early in each season and put in diapause to be hatched for resets, so we added 1000 to estimates of ephemeral diapausing embryo production. Because we did not count total numbers of embryos collected for resets, diapausing embryo production reported for ephemeral cultures may slightly underestimate total values.

Diapausing embryo hatching

To test hatching ability, 100 of the healthy-looking diapausing embryos collected from each of the 6 chemostats at each season's end (see 'Diapausing embryo production', above) were incubated to hatch. Each sample of 100 embryos was placed together in 3 mL of 15 ppt ASW and incubated (22 °C, 40 W lamp) for 1 week; we recorded the hatching frequency. All embryos were incubated to hatch simultaneously; they had been maintained in diapause (dark, 5 °C) until 9.5 weeks after the end of the final (6th) season.

Population density

Density of total females, sexual females carrying diapausing embryos, and males was determined from manual counts every 4 days from pooling three 1-mL samples of the chemostat outflow (total of ~ 93 counts per chemostat in the 385 days). Most females were nonovigerous, so their sexual status could not be determined. To estimate asexual female density at each density count, we subtracted the density of unfertilized and fertilized sexual females from the total female density. Fertilized sexual female density was determined from direct counts of females bearing diapausing embryos. We estimated the number of unfertilized (male-producing) sexual females based on the male density count and mean individual fecundity of unfertilized sexual females that lay male eggs (from life history bioassays, below). In essence, we calculated the number of unfertilized sexual females required to produce the current male density. Our final calculation of asexual density was as follows: asexual density = total female density−(male density/daily fecundity of unfertilized sexual females)−fertilized sexual female density. This formula does not give an exact count of asexual density, as it does not account for factors such as the potential for more males to be produced during the day after the density count or for fertilized sexual females not currently carrying a diapausing embryo. However, application of the same formula allowed comparison of relative asexual density of ephemeral and permanent cultures in this study. Any estimated asexual density ≤ 0 owing to densities below detection was converted to 0.01 females mL−1 before statistical analysis and production of graphs.

Life history assay of individual lifespan and fecundity

During the first and the final two seasons, a life history bioassay was conducted to monitor total lifespan and fecundity of a cohort of asexual and unfertilized sexual females. We isolated ovigerous, asexual females from the outflow of each chemostat. Each ovigerous female was placed in 1 mL of medium (5 × 105 cells mL−1 T. suecica in ASW). The following day one neonate hatchling (F1 generation) per mother was transferred to fresh medium; the mothers were discarded. Every day until death, females of the F1 cohort were individually transferred to new medium, and their progeny (F2) were counted. For ephemeral cultures, the number of asexual females tested combining chemostats was 59, 66 and 71 F1 females in the 1st, 5th and 6th season, respectively (16 ≤ N ≤ 24 per chemostat per season); for permanent cultures, we assayed 53, 71 and 70 asexual females in the 1st, 5th and 6th season, respectively (17 ≤ N ≤ 24 per chemostat per season). For unfertilized sexual females, we tested 31, 42 and 65 F1 females from ephemeral cultures in the 1st, 5th and 6th season, respectively (8 ≤ N ≤ 24 per chemostat per season), and we tested 41, 29 and 22 females from permanent cultures in the 1st, 5th and 6th season, respectively (3 ≤ N ≤ 22 per chemostat per season). Lower numbers of sexual females reflected lower densities of sexual females in the chemostats.

Statistical analyses

Statistics were performed at α < 0.05 in SPSS v.18. Animals were randomized across wells and plates in all bioassays. We performed repeated measures linear mixed models analysed with restricted maximum likelihood (REML) on assays of sex-induction density, sex frequency, diapausing embryo production and hatching, and asexual female density. Transformations (log10 for counts; arcsine square root for frequencies) were made to improve normality. In all cases, chemostat was entered as the subject ID, with a random effect from the repeated measure of time. The intercept, hydroperiod, time and hydroperiod × time terms (Table 1) were input as fixed effects. Model selection was performed with Bayesian Information Criteria (BIC) and likelihood ratio tests (LRT), using an Unstructured covariance structure when multiple random effects were included, and otherwise using a Scaled Identity structure. In all cases, BIC and LRT showed the best model included a random effect for the intercept (subject = chemostat). Models were not significantly better (LRT with χ2), or appeared worse (higher BIC values), with inclusion of additional random effects (e.g. a random intercept plus a random hydroperiod treatment effect). Thus, models reported and used for significance testing only included random effects for the repeated measures term (time) and intercept (Table 1).

Table 1. Results for repeated measures linear mixed models (REML analysis) of the influence of hydroperiod treatment and time on sex-induction density, sex frequency, diapausing embryo production and hatching, and asexual female density. Time represented growth season in sex induction, diapausing embryo production and hatching, and asexual density analyses. Time represented day of the measurement in season 6 for the sex frequency assay. Degrees of freedom are given for the numerator (d.f.N) and denominator (d.f.D). Chemostat was entered as the subject ID, with random effects from the repeated measure of time, and a random intercept (subject = chemostat). The intercept, hydroperiod, time, and hydroperiod × time terms in the table were input as fixed effects
BioassaySeasonsInterceptHydroperiodTimeHydroperiod × Time
d.f.N/d.f.D F P d.f.N/d.f.D F P d.f.N/d.f.D F P d.f.N/d.f.D F P
  1. a

    Random factors are as follows. Sex-induction density: repeated measures variance (± 1 SE) = 0.261 (± 0.030), Wald Z = 8.627, < 0.001; intercept variance (± 1 SE) = 0.088 (± 0.074), Wald Z = 1.187, = 0.235. Sex frequency: repeated measures variance (± 1 SE) = 0.050 (± 0.011), Wald Z = 4.472, < 0.001; intercept variance (± 1 SE) = 0.003 (± 0.007), Wald Z = 0.450, = 0.653. Diapausing embryo production: repeated measures variance (± 1 SE) = 0.148 (± 0.040), Wald Z = 3.742, < 0.001; intercept variance (± 1 SE) = 0.026 (± 0.036), Wald Z = 0.707, = 0.480. Diapausing embryo hatching: repeated measures variance (± 1 SE) = 0.011 (± 0.003), Wald Z = 3.742, < 0.001; intercept variance (± 1 SE) = 0.009 (± 0.008), Wald Z = 1.168, = 0.243. Asexual female density, all counts: repeated measures variance (± 1 SE) = 1.480 (± 0.089), Wald Z = 16.628, < 0.001; intercept variance (± 1 SE) = 0.257 (± 0.193), Wald Z = 1.332, = 0.183. Asexual female density, last 10 counts: repeated measures variance (± 1 SE) = 1.070 (± 0.081), Wald Z = 13.266, < 0.001; intercept variance (± 1 SE) = 0.216 (± 0.165), Wald Z = 1.306, = 0.192.

Sex-induction propensity1, 5, 61/7.937111.080< 0.0011/7.9370.4860.5061/150.80820.740< 0.0011/150.8087.9260.006
Sex frequency61/29.836113.988< 0.0011/29.8364.6660.0391/405.3810.0261/400.1010.752
Diapausing embryo production1–61/23.3551038.893< 0.0011/23.3550.1830.6721/286.7690.0151/288.5360.007
Diapausing embryo hatching1–61/11.050168.717< 0.0011/11.0501.8030.2061/289.1330.0051/281.6640.208
Asexual female density, all counts1–61/6.053229.972< 0.0011/6.0530.0310.8651/552.9980.0420.8381/552.99814.706< 0.001
Asexual female density, last 10 counts1–61/6.972324.664< 0.0011/6.9720.0780.7881/3522.7090.1011/3525.9860.015

Chemostats were inoculated with identical populations initially and anticipated to evolve over time across seasons, but with the type of change affected by the hydroperiod regime. Thus, testing differences over seasons between ephemeral and permanent hydroperiods was performed by analysing the significance of the hydroperiod × time interaction. One exception was the sex frequency assay in which all measurements were taken in the last season; here, analysis focused on the main effect of hydroperiod.

Lifespan and fecundity did not appear to follow normal distributions and were analysed differently. Tests of total lifespan were performed with a log-rank Mantel-Cox test implemented in the Survival Kaplan-Meier analysis of SPSS with Strata = Season. Data were pooled from the last two seasons (5, 6) to compare final outcomes of hydroperiod treatment on lifespan. Generalized linear models with Poisson distribution, log link and Pearson chi-square scaling parameter were used to test for effects on fecundity with Wald's chi-square statistics. Analyses of fecundity tested for effects of time, hydroperiod treatment, the hydroperiod × time interaction term, and chemostat nested within hydroperiod (Table 2).

Table 2. Summary of log-rank (Mantel–Cox) tests of lifespan (data pooled from the last two seasons) and generalized linear models of fecundity (seasons 1, 5, 6)
BioassayRotifer typeChemostatHydroperiodTimeHydroperiod × Time
χ 2 d.f. P χ 2 d.f. P χ 2 d.f. P χ 2 d.f. P
LifespanAsexual0.92610.336
 Sexual1.82910.176
FecundityAsexual28.2214< 0.0010.96110.327134.7841< 0.0013.05510.081
 Sexual3.60140.4630.93210.3341.31710.2510.61110.434

Results

Sex induction, sex frequency and diapausing embryo production

Rotifers in ephemeral vs. permanent hydroperiod treatments evolved significantly different propensities for inducing sex across seasons (= 0.006) (Fig. 2a, Table 1). The relatively higher densities for male production by permanent cultures in later seasons reflected evolution of decreased propensity to induce sex. Before inoculation in the chemostat, induction densities ranged from 2 to 150 females mL−1 (Carmona et al., 2009). Our results from season 1 are similar (10–139 females mL−1). By the end of the experiment, inducing densities for some females in permanent populations had greatly increased (18–1610 females mL−1 and 18–1486 females mL−1 in seasons 5 and 6, respectively). Between seasons 1 and 6, mean sex-induction density for permanent cultures rose from 55 to 688 females mL−1, with an increase in the median from 48 to 697 females mL−1 (Fig. 2a). In seasons 5 and 6, a substantial number of females in permanent systems failed to produce males (up to 87% of females in one permanent culture in season 5); final densities observed on day 10 for females that did not produce males during the assay were recorded as their inducing density.

Figure 2.

Higher levels of sex in ephemeral chemostats. Error bars are ± 1 SE. (a) Mean female density for sex induction (first appearance of males) in the first and final two 9-week seasons averaged across the three ephemeral vs. the three permanent cultures, showing differential evolution between hydroperiods (= 0.006). Higher densities in permanent cultures for later seasons reflect evolution of lower sex-induction propensity. (b) Frequency of sexual relative to total females was significantly higher in ephemeral vs. permanent cultures in the 6th growth season (= 0.039). Circles represent mean sex frequency from assays every 8 days. Interpolating lines between data points are included to aid visualization.

Frequency of sex in the final season, quantified as the proportion of sexual relative to total females, was higher in ephemeral vs. permanent chemostats (= 0.039) (Fig. 2b). However, within this final season, there was no differential evolution between ephemeral and permanent cultures (the hydroperiod × time interaction term was not significant, = 0.752) (Table 1).

Over time, total diapausing embryo production per season (Fig. 3) evolved to higher levels for ephemeral than permanent populations (= 0.007) (Table 1). Hatching frequency of the diapausing embryos did not show differential evolution between hydroperiods across seasons (= 0.208) (Table 1), but overall declined in later seasons (= 0.005).

Figure 3.

Total diapausing embryo production during each season in ephemeral and permanent cultures. Bars represent mean numbers of embryos (± 1 SE). Differential evolution in ephemeral and permanent hydroperiods was significant, with higher production by ephemeral cultures in later seasons and lower production by permanent cultures (= 0.007).

Population density and life history

Evolution led to higher mean asexual female densities for permanent compared with ephemeral cultures in later seasons (< 0.001) (Fig. 4, Table 1), with percentage differences between hydroperiods ≥ 76% in seasons 3–6. Although the decrease in ephemeral density at a reset (i.e. re-inoculation with hatchlings of the 800 diapausing embryos) may have contributed to the results, rapid population growth makes it unlikely that resetting alone explains the lower size of ephemeral cultures. We re-analysed the asexual female density, excluding data from the first 4–7 population density counts (first 16–28 days of each season) in all chemostats to account for the time when ephemeral chemostat densities may be less owing to the population bottleneck at ephemeral resets. In this re-analysis, we compared asexual density between hydrope-riods based on only the final 10 density counts (last ~ 40 days of each season) and still found significant divergence across seasons towards higher asexual female density in permanent cultures (= 0.015) (Fig. 4, Table 1).

Figure 4.

Mean asexual female density per season (± 1 SE). Interpolating lines between data points are included to aid visualization. Asexual female density was higher for permanent than ephemeral cultures in later seasons, as seen analysing all density counts (< 0.001), or focusing on the last 10 counts and excluding the first 4–7 counts (16–28 days) of each season (= 0.015).

Lifespan and fecundity did not evolve in response to hydroperiod. Data pooled from the last two seasons revealed that the difference in lifespan between ephemeral and permanent cultures was not significant for asexual or unfertilized sexual females ( 0.176) (Table 2). Mean lifespan in days (± 1 SE) for asexual females was 11.0 (± 0.3) in ephemeral and 10.5 (± 0.2) in permanent chemostats. For unfertilized sexual females, lifespans were 8.5 (± 0.3) in ephemeral and 8.3 (± 0.3) in permanent chemostats. Generalized linear model tests of Wald's chi-square statistic of total lifetime fecundity showed no evolutionary response to hydroperiod treatment across seasons for asexual or unfertilized sexual females ( 0.081) (Table 2). Pooling data from the final two seasons, for asexual rotifers the mean number of progeny per female (± 1 SE) was 18.3 (± 0.3) in ephemeral and 17.1 (± 0.4) in permanent cultures. For sexual females in the final two seasons, mean progeny per female was 13.0 (± 0.4) in ephemeral and 13.2 (± 0.4) in permanent cultures.

Discussion

Here, we provide experimental evidence that ephemeral habitats requiring dormancy for survival, contrasted to permanent environments, impose differential selection pressure leading to rapid evolution (≤ 84 generations, or 385 days) of sex propensity and dormancy. Our experimental evolution approach was designed to mirror the pressures imposed by hydroperiod in nature by requiring or eliminating the need for dormancy, itself the outcome of sex in brachionid rotifers. We observed multiple responses at the individual (e.g. sex-induction density) and population level (e.g. total diapausing embryo production). In nature, aquatic habitat permanence correlates with gradients in community composition and adaptive life history traits (Wellborn et al., 1996; Simon et al., 2002; Brock et al., 2003; Jocque et al., 2010). Our laboratory experiment demonstrates that rapid evolution may allow for hydroperiod adaptation in situ, suggesting a causal mechanism for these correlations. The requirement of diapausing embryo production for survival in ephemeral cultures was sufficient for the evolution of higher sex frequency and dormancy relative to permanent populations. Results of this study also suggest the potential for rapid evolution of dormancy to facilitate evolutionary rescue (Kinnison & Hairston, 2007) from stressors such as altered hydroperiod regimes, an anticipated effect of climate change (Brooks, 2004).

Unlike some studies that have demonstrated complete loss of sex in continuous cultures of B. calyciflorus rotifers in periods of months to years (Bennett & Boraas, 1989; Fussmann et al., 2003; Stelzer, 2007; Becks & Agrawal, 2010; Stelzer et al., 2010; Fussmann, 2011), we demonstrate a decline in sex frequency in permanent B. plicatilis cultures. Our emphasis on the evolution of sex frequency is congruent with studies reporting variation in sex propensity among clones for both rotifers and cladocerans (Schröder & Gilbert, 2004; Tessier & Cáceres, 2004; Gilbert & Schröder, 2007; Carmona et al., 2009). Ephemeral chemostats maintained higher frequencies of sexual females in the final season. Lower responsiveness to sex-induction stimuli immediately after the reset may explain the increase in sex frequency between days 7 and 15 (Fig. 2b) in ephemeral cultures. Brachionid diapausing embryo hatchlings and the first few generations of descendants often show reduced propensity for sex (Gilbert, 2002). The overall difference in sex frequency between hydroperiods may be due to the evolution of propensity for sex induction. In permanent systems at later seasons, we observed initiation of male production by females at densities higher than any observed initially (> 900 females mL−1 for some females in seasons 5 and 6), consistent with evolution of lower propensity for sex, but not complete loss of induction ability. Over longer timeframes, such gradual declines in sex propensity could culminate in loss of the capacity for sex. Slight increases in inducing densities for ephemeral chemostats across seasons could reflect mild inbreeding depression because of the low (N = 15) number of lineages in each original population (Tortajada et al., 2009). Alternatively, this may reflect adaptation to laboratory conditions or genetic slippage from sex (Lynch & Deng, 1994).

In addition to evolution of sex induction, diapausing embryo production increased in ephemeral and decreased in permanent cultures. This may reflect the higher frequency of sexual females in ephemeral cultures, and/or evolution of fecundity for sexual, fertilized females. The results support the finding of a negative correlation between hydroperiod of ponds in nature, and B. plicatilis diapausing embryo production (Campillo et al., 2011). The initial large increase in diapausing embryo production by ephemeral cultures between the first and second season may be due to the very strong selection pressure imposed on ephemeral populations by only using diapausing embryo hatchlings for re-colonization after resets. Decreased production by permanent cultures was not evident until later seasons. Delaying sex induction until a higher population density is reached can increase total diapausing embryo production (Serra et al., 2005); larger populations can produce more diapausing embryos because more females are available to reproduce. Yet despite their lower female density, ephemeral populations still produced more diapausing embryos than permanent cultures in later seasons. We found no differential evolution of diapausing embryo hatching success between hydroperiod treatments.

Our results are consistent with the theme of eco-evolutionary dynamics, with evolution of rotifer density-dependent sex induction affecting the ecological dynamic of population density (Stelzer, 2012). This experimental evolution study reinforces theoretical predictions that higher proportions of sexual offspring increases the cost of sex (Stelzer, 2011) and empirical evidence for lower population growth in brachionid clones that engage in sex vs. obligate asexual clones (Stelzer, 2012). In the final season, asexual female density was significantly lower in ephemeral compared with permanent populations, supporting the historical idea of a cost of males and sex to population growth (Maynard Smith, 1978). Life history assays of cohorts showed no significant difference in fecundity or lifespan between hydroperiods for asexual or unfertilized sexual females. We cannot here rule out the possibility of maternal effects influencing the observed lifespan and fecundity, and future analyses would benefit from a pre-experimental step as performed in the sex-induction assay, but any maternal effects influencing these traits in our bioassay would be anticipated to occur in the chemostats as well. Thus, we suggest that differences in asexual population size are not owing to the evolution of asexual fecundity. The lower population density of ephemeral chemostats partially may be attributed to the bottleneck occurring at resets. Future studies designed to keep similar population sizes between hydroperiods at resets would be helpful to verify that our results do reveal eco-evolutionary dynamics and a cost of sex (and dormancy) to population growth. Part of the cost of sex may be indirect in brachionids and incurred because of the fact that sexually produced embryos undergo diapause, which itself may slow population growth. For additional consideration of dormancy and the cost of sex in rotifers, we refer readers to Serra & Snell, 2009.

This study also demonstrates the need to further consider the role of recombination and mutation in rapid evolution of metazoans. Previous studies have proposed rapid evolution occurred through clonal selection, changes in the frequency of initial clones (i.e. genotypes) (Fussmann et al., 2003, 2007; Turcotte et al., 2011). Yet we report that females in permanent chemostats in later seasons induced sex at densities higher than we observed in the first season and higher than that reported by Carmona et al. (2009) for any lineage in our founding populations. The new, evolved sex-induction densities we observed could have occurred through recombination or segregation leading to novel arrangements of the genetic material. These meiotic processes seem unlikely in permanent cultures because of our removal of the sexually produced diapausing embryos. However, we cannot rule out the potential that some small fraction of diapausing embryos hatched in the permanent cultures before removal at each season's end, allowing some meiotic recombination. Ameiotic recombination represents another possible source of genetic variation (Omilian et al., 2006), as do new mutations. It is known that inheritance of the op locus in B. calyciflorus results in inability to respond to the cue for sex induction via Mendelian inheritance; loss of sex is recessive (Stelzer et al., 2010; Scheuerl et al., 2011). However, op heterozygotes do not differ in sex-induction density from homozygous dominant rotifers (Scheuerl et al., 2011). Thus, recombination at this locus cannot fully explain the evolution of decreased sex-induction propensity in permanent cultures.

Future research on the mechanisms of the rapid evolution reported here also should address effects of population dynamics (e.g. numbers of generations, population size). The portion of ephemeral populations that survived a reset existed as a diapausing embryo for the final 6 weeks for seasons 1–5; thus, permanent cultures experienced ~84 generations, whereas ephemeral cultures would have experienced ~37 generations with the remaining time in dormancy. The greater number of generations in permanent cultures, their larger population size, and absence of the population bottleneck that occurred in ephemeral cultures at resets could have fostered the ability of natural selection to act in permanent populations. Perhaps this promoted fixation of mutations, for example, for lower sex propensity. Conversely, higher frequencies of sex in ephemeral cultures could promote natural selection, leading to higher fitness and promoting evolution and adaptation to the ambient hydroperiod regime (Burt, 2000; Becks & Agrawal, 2012). Although it could be argued that one should compare equivalent numbers of generations for studies of evolution, our experiment was designed to test effects of the environmental driver of hydroperiod by mimicking temporary or permanent aquatic habitats. In nature, one might expect ephemeral pond populations to spend greater time in dormancy. Additional research is needed to study how dormancy itself may affect rapid evolution by altering numbers of active generations.

Acknowledgments

We thank A.M. Fields for laboratory assistance and N. Dimas-Flores and E.M. García-Roger for rotifer subcultures. Comments from R.L. Wallace, D.B. Mark Welch, K.E. Gribble, B. Hecox-Lea, M. Serra, and anonymous reviewers improved this study and manuscript. This research was supported by National Science Foundation Grant BE/GenEn MCB-0412674 to T.W.S.

Ancillary