Individual Variability in Reproductive Success Determines Winners and Losers under Ocean Acidification: A Case Study with Sea Urchins

Background Climate change will lead to intense selection on many organisms, particularly during susceptible early life stages. To date, most studies on the likely biotic effects of climate change have focused on the mean responses of pooled groups of animals. Consequently, the extent to which inter-individual variation mediates different selection responses has not been tested. Investigating this variation is important, since some individuals may be preadapted to future climate scenarios. Methodology/Principal Findings We examined the effect of CO2-induced pH changes (“ocean acidification”) in sperm swimming behaviour on the fertilization success of the Australasian sea urchin Heliocidaris erythrogramma, focusing on the responses of separate individuals and pairs. Acidification significantly decreased the proportion of motile sperm but had no effect on sperm swimming speed. Subsequent fertilization experiments showed strong inter-individual variation in responses to ocean acidification, ranging from a 44% decrease to a 14% increase in fertilization success. This was partly explained by the significant relationship between decreases in percent sperm motility and fertilization success at ΔpH = 0.3, but not at ΔpH = 0.5. Conclusions and Significance The effects of ocean acidification on reproductive success varied markedly between individuals. Our results suggest that some individuals will exhibit enhanced fertilization success in acidified oceans, supporting the concept of ‘winners’ and ‘losers’ of climate change at an individual level. If these differences are heritable it is likely that ocean acidification will lead to selection against susceptible phenotypes as well as to rapid fixation of alleles that allow reproduction under more acidic conditions. This selection may ameliorate the biotic effects of climate change if taxa have sufficient extant genetic variation upon which selection can act.


Introduction
Environmental factors directly affect populations by selecting resilient individuals. Selection at the gametic level, or during early life, has strong and immediate effects at the population level, carrying over into subsequent life stages. Heritability of this resilience leads to cascading adaptive effects in subsequent generations. For example, in free-spawning marine organisms, sperm selection during fertilization plays a key role by determining the nature and diversity of genotypes in the subsequent generation [1,2] and thus their resilience to environmental change.
Rising atmospheric carbon dioxide levels are a key driver of environmental change, and will likely lead to rapid ocean acidification [3,4]. With gametes possessing no, or only limited, buffering capacities against CO 2 -mediated pH changes in seawater, the dynamics of fertilization and subsequent development are likely to be affected in all free-spawning marine organisms, with potentially severe implications [5,6]. Yet we know little about the relative fitness of individuals within species under the predicted acidification of the ocean.
The sensitivity of reproductive processes to ocean acidification has thus far been assessed from mean responses of mixtures of gametes and/or larvae obtained from multiple individuals [7][8][9][10] (but see [11]). However, the key determinant of reproductive success in a future ocean is not the average response, but the proportion of successful offspring contributed by each individual under the changed environmental conditions. Individual-level responses to ocean acidification have been examined to some extent in larval development processes [12,13], but not closely in fertilization processes. In this context, the importance of naturally high variability that is observed in fertilization success of individual pairwise crosses [14,15] becomes apparent: not all matings are equal. Consequently, acidification-mediated impacts on reproductive success and subsequent development might result in flow-on consequences for genetic diversity and population demographics [12].
The early life history stages of echinoderms are particularly useful for studies of fertilization success, as these species are experimentally tractable and ecologically important, often acting as ecosystem engineers [16,17]. Here, we investigate the effects of CO 2 -induced ocean acidification on the early life history stages in the Australasian sea urchin Heliocidaris erythrogramma, focussing on intra-specific variation in responses, which can be highly variable for this species [18]. Following the A1FI-scenario from the IPCC's 4 th assessment report [3], we compared the effects of present day conditions for southeast Australia with the end-of-century scenario (year 2100; pCO 2 = 970 matm.0.3 pH unit reduction) and a high-CO 2 scenario (year 2300; pCO 2 = 1600 matm.0.5 pH unit reduction). Observed effects on sperm swimming behaviour were applied within an established fertilization kinetics modelling framework [19,20] to predict fertilization outcomes of single urchin pairs at each pCO 2 level. These were then compared to observed results from fertilization experiments conducted in the laboratory.

CO2 Treatment
Experimental CO 2 treatments were achieved by bubbling a mixture of air and CO 2 through filtered seawater (FSW; 0.22 mm filtered). pH (NBS scale) was maintained using microprocessorcontrolled CO 2 micro-injection systems in separate FSW-tanks. Systems were set to maintain a pH change (DpH) of 20.5, 20.3 or 0 pH units (control, no addition of CO 2 ), resulting in treatment pH values of 7.6, 7.8 and 8.1. Dissolved oxygen levels were maintained by slow bubbling of filtered air. pH levels were checked prior to experiments. Total alkalinity was determined once by titration [21] (all seawater used in experiments was taken from a closed recirculating system in which alkalinity was controlled to be constant). Parameters of the CO 2 system (Table 1) were calculated with CO2SYS [22] using the dissociation constants of Dickson & Millero [23].

Experimental Animals
Heliocidaris erythrogramma (test diameter = 50.361.3 mm, mean 6 S.E.) were collected during their spawning season between February and March 2011, from shallow subtidal areas at Long Bay (33u57.59S, 151u15.29E) and Bare Island (33u59.29S, 151u13.59E), Sydney, Australia. Animals were immediately transported to Macquarie University and held in tanks with flowing seawater for up to one week maximum before being used for experiments. Individuals were collected fresh each week, alternat-ing between sites and those from different sites were never mixed. Three collections were done at Long Bay (males/pairs A-D on 7 Feb 2011, J-M on 21 Feb 2011, R and S on 14 March 2011; Table 2) and two at Bare Island (males/pairs E-I on 14 Feb 2011 and N-Q on 1 March 2011; Table 2).

Collection of Gametes
Gametes were obtained by injecting urchins with 0.8-1 ml of 0.5 M KCl through the peristomal membrane, followed by gentle shaking. This concentration of KCl reliably induced gamete release without being lethal. Individuals were used once (for convenience, the term ''individual'' will be used here to refer both to individual males [for sperm speed and motility measures] and to individual pairs [for fertilization success measures]). Eggs were collected in FSW, diluted to a concentration of 50 eggs ml 21 and incubated in seawater at one of the three pH conditions for 10 min prior to use in experiments. Sperm were collected ''dry'' and held on ice until use to extend their lifespan. Experiments were conducted immediately after spawning. All experiments were done in a constant temperature room at a standard temperature of 20.561uC (mean 6 range).

Sperm Motility and Speed
Data were obtained for each of 19 male urchins. The motility assay followed that of Havenhand & Schlegel [24]. Briefly, 0.5-1 ml of freshly collected sperm from an individual male was diluted into 1 ml of seawater of each pH immediately prior to use (10 replicate sperm suspensions for each CO 2 treatment and male). Sperm concentrations across assays ranged consistently from 1-2*10 4 sperm ml 21 . A 60-70 ml drop of sperm suspension was placed between an albumin-coated microscope slide and coverslip, separated by a 0.75 mm thick O-ring. Sperm movement was recorded for 2s at the midpoint of the drop, at 25 frames s 21 , using a digital camera (Sumix SMX-160) mounted on a compound microscope (Olympus BX51). Pilot experiments showed illumination by the microscope lamp has no impact on the temperature inside the drop during videotaping (time of slide on microscope: approx. 10s). All recordings (one per sperm suspension) were done within 30s after creation of each sperm suspension. Videos were post-processed and analysed using CellTrak 1.3 (Motion Analysis Corporation) to determine sperm speed and percentage of motile sperm (i.e., sperm moving faster than 15 mm s 21 on average). Only sperm classified as motile were used for sperm speed analyses.

Fertilization Success
Data were obtained for 18 pairwise crossings using the same males used in the sperm motility experiments (above). Each male was crossed with one female (no female was spawned for the first male, hence n = 18). Each individual was used only once (18 males and 18 females in total). For each of the three CO 2 treatments, nine four-fold serially diluted sperm-concentrations and one control (FSW only) were prepared in 6-well plates holding 6 ml of seawater. Sperm concentrations and CO 2 treatments were distributed across multiple 6-well plates to preclude plate effects. Approximately 200 eggs were added to each well in filter dishes (25 mm diameter 6 20 mm height, 80 mm mesh floor). Eggs and sperm were mixed and left for 10 min to fertilize. Eggs were then rinsed twice (with water of the relevant pH) to remove sperm and left for 2 h to develop, (typically this was to the 4 cell stage). Fertilization success rates (FSR) were determined by counting the proportion of cleaved eggs in vivo. Sperm concentrations in the stock solutions were checked post-hoc using a haemocytometer (Neubauer improved). All fertilizations were conducted within 15 min of obtaining the gametes.

Modelling of Fertilization Kinetics
We used modelling [19,25] to combine the results from motility and fertilization experiments to investigate the importance of pHinduced changes in sperm motility in the fertilization process. Fertilization kinetics curves [20,26] were fitted to the measured fertilization data for each cross and CO 2 treatment. Fertilization efficiency (F e ) and polyspermy block efficiency (B e ) were estimated using least squares [27]. Sperm speeds and percentage of motile sperm for each male were taken from video analyses. Sperm concentration for modelling was defined as the product of the observed sperm concentration in the serial dilution and observed percent motility of that male in that treatment. For the control assay (pH 8.1) we identified the sperm concentration that yielded maximum fertilization success (F max Control ; Figure 1), the sperm concentration that yielded 50% of F max (F 50 Control [28]) and the respective fertilization success rates at each of these sperm concentrations (''FSR max Control '' and ''FSR 50 Control ''). For each of the CO 2 treatments fertilization success at F 50 Control was observed (''FSR obs '') for each cross. These values were then compared to the fertilization success at F 50 Control that was obtained from the model using the sperm speed and percent motility values observed for each treatment (''FSR mod '').

Data Analyses
All percentage data were arcsine transformed prior to statistical analyses and checked for normality [29]. Levene's test was used to assess for homogeneity of variances among individuals and among treatments. Two-factor Analysis of Variance (ANOVA) was used to assess pH effects on sperm speed and percent motility across all males (pH fixed, male random). Tukey's test was used to compare post-hoc differences among means. The magnitude of responses of percent motility and fertilization success to pH treatments was assessed using logarithmic responses ratios (LnRR; natural log of treatment response divided by control response [30]). Mean LnRRs and 95% confidence intervals of fertilization success were determined by bootstrapping in R (100,000 iterations; [31]). Regression analysis was used to assess relationships between observed and modelled fertilization outcomes. All statistical tests were carried out using SPSS TM .
In contrast, average sperm speed was not significantly affected by acidification (p = 0.710; Fig. 2B, Table 2&3B), although again there were significant differences in responses between individuals (P,0.001; Table 3B). Upper and lower bound 95% CIs for individual response ratios (LnRRs) of sperm speed were equivalent to +2.4% to 21.1% (DpH = 0.3), and +2.9% to 21.4% (DpH = 0.5; Fig. 2C). There were no significant differences in sperm parameters between males from different sites at any pH level ( Table 2).

Fertilization Success
Ocean acidification substantially increased the variance of observed (FSR obs in Fig. 3A) and modelled fertilization success (FSR mod in Fig. 3A).
The overall effect of DpH on observed fertilization success (Table 4) was not statistically significant (P = 0.9, Table 5A), however bootstrapping showed that responses were highly variable between pairwise crosses (Fig. 3B, Table 4). The 95% CIs around the mean log response ratios (LnRR) varied from moderately positive (11% and 14% increases in fertilization success at DpH = 0.3 and DpH = 0.5, respectively) to strongly negative (#44% and #79% decreases at DpH = 0.3 and 0.5, respectively; Fig. 3B). There was no significant difference in maximum fertilization success between pairs from different sites (P = 0.202).
Modelling the effects of the observed changes in sperm percent motility and sperm swimming speeds on fertilization success yielded predictions that broadly mirrored the patterns seen in observed measurements (FSR mod in Fig. 3B, Table 5B): for DpH = 0.3 the modelled LnRR 95% CI ranged from an equivalent of 0 to 234.8%, and for DpH = 0.5 from 23.4% to 236.1%. Regression analyses (Fig. 4) revealed that 34% of the observed change in fertilization success could be attributed to changes in sperm motility at DpH = 0.3, but only 4% at DpH = 0.5.

Discussion
Our finding that the effects of acidification on sperm swimming behaviour (Fig. 2C, Table 2), and fertilization success (Fig. 3, Table 4), vary significantly between individuals is biologically important. Differences in individual responses are the raw material for effective selection, and especially so at the critical life stage of fertilization [2]. However, inter-individual variation has been overlooked in most previous, group-mean based investigations.  The shortcomings of analysing group means are demonstrated by comparing the overall non-significant effect of ocean acidification on mean fertilization (Table 5) with the substantial interindividual variation in response we observed (FSR obs in Fig. 3B). The majority of individual pairs had reduced fertilization success under acidified conditions, however, some pairs showed increased fertilization success (Fig. 3B, Table 4). This illustrates the importance of examining individual responses: it is individuals that contribute differentially to the next generation -not group means. Consequently, approaches that assess the group mean response ignore evolutionarily important effects of rare individuals that may contribute disproportionately to the next generation. This example also emphasizes the need for adequate sample sizes in order to capture the variety of individual responses to ocean acidification, particularly in species with high inter-individual variation.
The effects of acidification on fertilization success have often been tested using inappropriately high sperm concentrations. Using a sperm concentration that yields maximum fertilization success in controls (F max in Fig. 1) can generate misleading or uninformative results, because assays may be saturated and therefore unresponsive to subtle, but biologically important, differences in fertilization [32], and because such assays cannot detect possible positive effects of the treatment. Had such high concentrations been used in our experiments, the observed increases in fertilization success seen in some pairs would not have been detected (Fig. 3). Choosing 50% of the maximum fertilization success as the response variable (F 50 in Fig. 1; [28]) allowed us to detect both negative and positive pH impacts on fertilization success, while ensuring maximum sensitivity in our assays.
Previous investigations of the effects of ocean acidification on fertilization success in H. erythrogramma have found contradictory results [11,33,34]. These differences may partly be explained by the use of saturated assays in some studies. Previous studies also used smaller sample sizes than used here (5 M 6 F pairs [11], 3 replicates [33,34], vs 18 M 6 F pairs in this study). The use of gametes mixed from multiple individuals [33,34] also precludes observation of intra-specific variation. Apparently conflicting results may also be explained by the intra-specific variation demonstrated in our experiments. Fertilization success in some pairs was negatively impacted by acidification (confirming [11]) whereas other pairs showed no (or little) net response (confirming [33,34]). Thus, we suggest that much of the controversy around the response of fertilization success in H. erythrogramma to ocean acidification can be attributed to a combination of the factors discussed above.
Our observation that lowered seawater pH did not affect the speed of motile sperm (Fig. 2B) but rather decreased the proportion of motile sperm ( Fig. 2A) contrasts partly with earlier reports [11]. Sea urchin sperm are stored immotile in an acidic environment inside the testis after development, which inhibits respiration and metabolic processes prior to release [35][36][37][38][39]. Upon spawning, the difference between intra-cellular pH in the testis and extracellular pH in the seawater triggers mitochondrial activity and thus motility. Ocean acidification may reduce the pH gradient upon spawning to a point where it is insufficient to activate the sperm mitochondrion. Since we observed a decrease in the proportion of motile sperm under acidified conditions ( Fig. 2A), but the swimming speed of those motile sperm did not decrease (Fig. 2B), this strongly suggests an effect on activation of the sperm mitochondrion but not on mitochondrial function once activated. As sperm have no actively transcribing nuclear genes or biochemistry, the most parsimonious explanation for our results is that genetic variation in mitochondrial membrane protein genes explains some of the observed inter-individual variation in sperm swimming behaviour, and hence fertilization success.
Variation in fertilization success may also have been influenced by parental environmental history, although our results suggest this was not likely to have been a significant driving factor. Individuals were collected from small areas of uniform habitat, and variance within single populations was as large as differences between populations (the converse would be expected if environmental history effects were more influential than genetic diversity). Early cleavage stages (2 h post-fertilization) are also largely independent of transcription of paternal DNA [40]. Consequently we suggest that variation in sperm swimming behaviour and sperm-egg binding compatibility [41,42] are the most likely explanations for the observed variance in responses to acidification. More broadly, for free spawners such as H. erythrogramma the chance of successful fertilization depends on gamete concentra-tions, gamete life span and (hence) on the distance between spawning individuals [14,43]. The density and abundance of H. erythrogramma varies substantially in situ -from ,1 to .50 individuals.m 22 [44,45], occurring in a range of even to patchy distributions. Low fertilization success due to sperm limitation may be common between widespread clusters of urchins because sperm longevity is short [46][47][48]. If many individuals within populations display decreased sperm motility due to ocean acidification, sperm limitation downstream of a spawning site may be exacerbated [49,50], further decreasing reproductive success and ultimately reducing the number of individuals that contribute to future generations [15,51].
Our results support the concept of 'winners' and 'losers' of climate change; a concept often proposed over the last decade [52][53][54][55], and applied to many organisational levels such as alleles within genes, individuals within populations, and species within ecosystems. Here we apply the 'winners' and 'losers' concept at the individual level. The substantial variation in sperm motility and fertilization success we observed in response to predicted ocean acidification (Figs 2, 3) shows that some individuals (here, males or male x female pairs, see above) are better equipped than others to cope with acidification. Future ocean acidification will likely reduce the proportion of fertilizations by acidification-sensitive gametes ('losers'), and increase the proportion of fertilizations by acidification-resistant gametes ('winners'). Whether these 'winners' are selected due to genetic traits (such as sperm swimming ability ( Fig. 2A) and sperm-egg binding compatibilities [41,42]) or due to non-genetic maternal traits (such as egg condition [56]) will be  immaterial to the selection of the 'winners' per se. It also remains unclear whether 'winners' of climate change-induced selection for fertilization success will actually remain 'winners' across the entire life-cycle. Nonetheless, it should be remembered that the heritability of the selected traits will of course strongly influence the future adaptation potential of the progeny and hence the longterm adaptive benefits of 'winner' status. If this observed variability is heritable, it will have important implications for urchin populations in a near-future acidified ocean. Loss of less competitive gamete genotypes will reduce overall genetic diversity, at least initially. This loss may be counteracted by genetic drift and new mutations, although in the short term these are likely to be trivial compared to the effects of selection. The long-term fitness consequences, however, will depend on the fitness benefits of traits that covary with acidification-resistance in gametes, and the extent to which recombination in second and subsequent generations gather advantageous alleles in some individuals. Any selection that reduces available genetic diversity leaves future populations less capable of tolerating further perturbations [57,58], and this underscores the need for analyses of the fitness of the descendants resulting from experiments such as those we have described here.
In summary, despite an increasing number of studies focusing on the effects of ocean acidification on the early life history of marine organisms [59,60], very few studies have investigated individual-level responses to changing oceanic conditions. This is perplexing in light of the growing evidence for a degree of interindividual variation that exceeds the noise of baseline variability [24]. Conducting adequately replicated studies to investigate interindividual variability in response to marine climate change is imperative if we are to understand the capacity for selection and adaptation of marine species.