Effects of Ocean Acidification and Warming on Sperm Activity and Early Life Stages of the Mediterranean Mussel ( Mytilus galloprovincialis )

Larval stages are among those most vulnerable to ocean acidification (OA). Projected atmospheric CO2 levels for the end of this century may lead to negative impacts on communities dominated by calcifying taxa with planktonic life stages. We exposed Mediterranean mussel (Mytilus galloprovincialis) sperm and early life stages to pHT levels of 8.0 (current pH) and 7.6 (2100 level) by manipulating pCO2 level (380 and 1000 ppm). Sperm activity was examined at ambient temperatures (16–17 ◦C) using individual males as replicates. We also assessed the effects of temperature (ambient and ≈20 ◦C) and pH on larval size, survival, respiration and calcification of late trochophore/early D-veliger stages using a cross-factorial design. Increased pCO2 had a negative effect on the percentage of motile sperm (mean response ratio R̄ = 71%) and sperm swimming speed (R̄ = 74%),


Introduction
Ocean acidification (OA) results from the uptake of carbon dioxide in seawater due to anthropogenic emissions [1].Together with global warming, it is predicted to cause significant changes in the marine environment over the coming century [2,3].Such changes are already occurring: the average ocean surface pH has dropped by 0.1 units since the beginning of the industrial revolution [1], and the global average surface temperature has increased by 0.7 • C during the last century [4].Atmospheric CO 2 concentration is predicted to reach 540-1020 ppm by the end of this century, which will reduce the open ocean surface water pH by an additional 0.12-0.36units depending on the latitude and ocean dynamics [5,6].
Increased pCO 2 affects a range of ocean chemistry parameters, including pH, dissolved organic carbon (DIC), and calcium carbonate (CaCO 3 ) saturation state [6].Although it is still uncertain how these changes will impact the biological systems, their effects will likely vary among both taxa and life stages [7].Larval stages, especially in calcifying species, are thought to represent the life-cycle bottleneck most vulnerable to OA [7][8][9][10][11].They may be more sensitive to perturbations due to low energy storage levels, potential disruptions in development of vital structures, and possible delays in metamorphosis, which may increase vulnerability to predators or affect transport mechanisms, removing them from suitable habitats [12].Moreover, many species already live close to their thermal tolerances [13,14], and increases in temperature can therefore further exacerbate the negative effects of climate change [15].Thus, synergistic effects of increased temperature and acidification on calcifying larval-stages may have significant consequences for marine populations and community structure [12,16].
Marine taxa exhibit variable responses to OA.In general, non-calcifying autotrophs might benefit from higher CO 2 availability, while calcifying organisms, apart from crustaceans, are negatively impacted, with bivalves and echinoderms being among the most vulnerable groups [7,8,10,11].However, ocean pCO 2 itself varies considerably with space, depth and time [17].While open ocean pH is relatively stable at a given location, near shore habitats experience large diel, tidal, seasonal and stochastic variability [18], which can be larger than the predicted decrease in open ocean pH during this century [19].Therefore, shallow-water benthic calcifying organisms might be expected to be adapted to a variable pH regime, and perhaps are not the most vulnerable taxa to OA [20].
Mytilus is a widespread genus of intertidal and shallow-water mussels consisting of several species worldwide [21,22].The Mediterranean mussel (Mytilus galloprovincialis) is an important commercial species and is native to the Mediterranean with a range northwards along the Atlantic Coast of West Europe [23].Due to its commercial importance and dispersal by shipping in ballast water, the mussel can be found in many regions worldwide (see a list of distribution records at [24]).Adult Mytilus shells consist of both calcite and aragonite, which are CaCO 3 minerals [25], whereas larval shells consist mostly of aragonite [26].The outermost CaCO 3 shell layer of adult Mytilus consists of calcite, which is covered by an organic periostracum [25].This organic layer may protect the CaCO 3 layers from dissolution, and calcite is less soluble than aragonite in low pH seawater [27].This may, therefore, result in adult and juvenile Mytilus being relatively robust to near future OA [20,[28][29][30][31].
Effects of OA on Mytilus seem to be most often identified in the first aragonitic calcification stages.Embryogenesis of Mytilus appears to be unaffected by elevated pCO 2 until the trochophore stage, after which growth, shell-strength, and normal development of calcified larvae become negatively affected [32][33][34].To our knowledge, effects of OA on fertilization and sperm quality of Mytilus have not been reported, but effects can vary considerably among taxa and even between male-female pair within a population [35][36][37][38][39].However, no study has followed OA effects on Mytilus from sperm activity to the early D-veliger stage, covering sperm kinetics, larval growth, and vital rates.This information is critical to evaluate hypotheses of high susceptibility of early life-stages, and to assess OA impacts on this key habitat-forming taxa with high commercial value.The aim of this study was to evaluate the effect of ocean acidification and warming on sperm activity, development, calcification and metabolism of early life-stages of M. galloprovincialis using a cross-factorial design and a high-end IPCC scenario predicted to occur by 2100 (SRES A2 [5]).In addition, we aimed to increase understanding in responses of early life-history stages of Mytilus genus to the climate change by comparing our results to the literature using meta-analysis.

Material and Methods
Experiments were conducted in temperature regulated climate rooms at the Mediterranean Institute for Advanced Studies (Mallorca, Spain).Artificial seawater adjusted to a salinity of 35 was used in the experiments by blending a commercial salt mixture (Instant Ocean Sea Salt) with distilled water [40].The water was mixed to ensure a complete dissolution of salt into the water and recirculated through a UV lamp for at least 24 h before use to reduce bacterial growth.The pH of the seawater was manipulated by aerating water with pCO 2 -adjusted air to each culture cylinder or tank separately (Figure A1).Ambient air was collected via aquarium pumps and passed through soda-lime columns.Precise volumes of CO 2 -stripped air and CO 2 gas were administrated using mass-flow controllers (Aalborg GFC17) and mixed in a container filled with marbles to achieve pCO 2 concentrations of 380 ppm (control, CSW) and 1000 ppm (treatment, TSW).These values correspond to the annual average atmospheric pCO 2 level of 2005 [41] and to the high-end projected level for 2100 [5], respectively.

Study Organism and Collection of Gametes
Adult Mediterranean mussels (Mytilus galloprovincialis) of ≈7 cm in length were obtained from a commercial mussel farm (Manuel Cabrera e Hijo S.C.P) located in Mahòn, on the coast of Menorca, Spain on 11th and 14th of April 2011.Temperature and salinity, measured nearby, were 15.9 and 37.5 • C respectively.Mussels were placed into an aerated storage tank in a cold room set to 14 • C. Spawning was induced the next day for mussels that appeared to be in a good condition.
Gametes were collected from individual mussels after inducing spawning with a combination of temperature shock (at 22 • C) and fluoxetine as described in Honkoop et al. [42].Typically, the mussels started to spawn 30 min after adding fluoxetine, which was then washed away by changing the water to freshly aerated seawater.The mussels were left to spawn for about 30 min.Sperm suspensions from individual males were pipetted from the highest concentrations close to the bottom of the beakers, and transferred to separate Eppendorf tubes for sperm activity experiments.Tubes containing sperm stock were stored in wet ice and used as soon after collection as possible (1 to 4.5 h after spawning).Eggs were collected from the beakers with a cut-tisp pipette to avoid damaging them.

Sperm Activity
A two-level experimental design (380 ppm = CSW, control; 1000 ppm = TSW, treatment) was used to assess pCO 2 effects on sperm swimming speed and percentage of motile sperm ("sperm motility" hereafter).Water for the experiment was obtained from tanks treated similarly to the experimental cylinders (Figure A1).The experiment was conducted in a temperature controlled room approximately at 16 • C following the methodology in Havenhand and Schlegel [43].A known volume (14-110 µL) of concentrated sperm stock (215-1777 × 10 3 cells µL −1 , counted using haemacytometer and 8 replicate subsamples) from an individual male was mixed with 1500 µL of treatment (TSW) and control seawater (CSW) in separate Eppendorf tubes.Temperature of TSW and CSW was maintained at 16 • C using a water bath.The final concentration of the suspension (≈16 × 10 3 cells µL −1 ) was adjusted to provide an optimal number of sperm cells for filming (20-40 frame −1 ).
Immediately after mixing, a drop of sperm suspension was placed in one of the chambers of a Leja 4-chamber microscope slide (20 µm depth, SC-20-01-04-B), which was then set on the stage of a compound microscope (Leica Leitz Labor-Lux S, 20× objective) equipped with an infrared filter and connected to a Panasonic 1.2 megapixel video camera via CCD-Video camera module.The camera system was connected to a computer and calibrated for length and width of the field of view.A film clip was recorded, the clips were edited to constant length (1 s, 25 frames) and analyzed using CellTrak 1.5 software (Motion Analysis Corporation, Santa Rosa, CA, USA).Sperm swimming speed was determined as average curvilinear velocity of motile sperm cells in each film clip [44].Sperm velocity threshold of less than 10 µm s −1 was used to calculate percentage of motile sperm.The threshold value was determined from histograms of speed data from multiple males.
This procedure was repeated 10 times for each sperm sample, in both treatment and control conditions.The mean of 10 replicate measurements was used in subsequent analyses as described in Havenhand et al. [35].Thirteen males were used as statistical replicates to evaluate the variation within the population.

Larval Experiment
The larval development experiment used a two-variable factorial design, consisting of two temperature treatments (low ≈ 17 • C and high ≈ 20 • C; see Table 1) and two pCO 2 treatments (380 and 1000 ppm).Each treatment combination was replicated three times, thus requiring twelve 6 L meta-acrylate cylinders (Figure A1).Cylinders had a funnel-shaped bottom and were aerated through a glass pipe from the bottom to equilibrate the water to the desired pCO 2 and to prevent sedimentation of larvae.Cylinder lids were fitted with a sampling port consisting of a silicon tube reaching mid-way down the cylinder.A 35 µm mesh filter was attached to the sampling port to extract water samples while leaving the larvae in suspension.
Table 1.Mean (±1σ, n = 3) pH, carbon chemistry and calcite saturation state parameters of experimental seawater for pCO 2 control (CSW, 380 ppm), treatment (TSW, 1000 ppm), temperature control (L, low) and treatment (H, high) cylinders 0.5 h after fertilization [(a.f., 2 h before addition of embryos (a.e.)] and 47.5 a.f.(44.5 h a.e.).S refers to salinity; T to temperature, pH T to pH in total scale; DIC to dissolved organic carbon; A T to total alkalinity; and Ω Ar to aragonite saturation state.Measured parameters were obtained at 25 • C and calculated for experimental temperatures using CO2SYS.Experimental temperature treatments were maintained within ±0.1 • C (PolyScience 9600 series), and monitored with temperature sensors connected to a data logger.In addition, pH sensors were deployed in two cylinders of each treatment combination.Two additional 10 L tanks of sea water continuously aerated to the desired pCO 2 concentrations were maintained at each temperature treatment to provide pretreated water at the relevant pCO 2 and temperature for refilling the cylinders after sampling and for the sperm motility assessment (Section 2.2).
Gametes were obtained by the methods outlined above (Section 2.1).Eggs were gently rinsed over a 20 µm sieve and pooled in a 1 L container.Sperm was passed through a 63 µm sieve to remove debris.Eggs from five individual females were combined and fertilized with a diluted sperm sample combined from 16 males, and left for 2 h in aerated water.Embryos were then rinsed over a 20 µm sieve, and sub-sampled to determine fertilization success and egg concentration.A total of ≈98,000 embryos (16 embryos mL −1 ) were added to each cylinder.Statistical population for the larval study consisted therefore of five females and 16 males, which were in turn a random subsample of the population in the Menorcan mussel farm.

Calcification
Net larval calcification was determined 48-55 h after fertilization using a modified version of the micro-diffusion technique for measuring coccolithophore calcification [45].Three replicates of unfiltered water containing larvae (40 mL each) were sampled from each cylinder, transferred into acid washed 60 mL polycarbonate bottles, and mixed with 147 kBq of 14 C-bicarbonate.One of the three replicates served as a blank, and was poisoned with concentrated borax buffered formalin.Radioactivity added to each bottle was determined by preserving 250 µL of each sample in 1 mL of ethonolomine in a separate scintillation vial.All replicates were then placed in a temperature bath for 4-6 h.Samples were filtered, processed, and acidified according to Paasche and Brubak [45], and finally frozen until later analysis.Radioactivity was measured from each filter in a scintillation counter (Tri-Carb 2810TR, Perkin Elmer) more than 8 h after the addition of 5 mL of scintillation cocktail.Radioactivity in blank samples was variable.Thus, the minimum blank radioactivity measured across the experiment was used to subtract from radioactivity measured on the live sample filters.Net calcification rates (µg C L −1 d −1 ) were calculated using measured DIC concentrations with the standard formula from Parsons et al. [46] and standardized to the estimated number of larvae in the subsample at the end of the experiment (see Section 2.3.3.).

Metabolism
Larval respiration rates were measured at the end of the experiment (51-57 h after fertilization) by determining the change in oxygen content from three replicate samples taken at the beginning and end of a 4.5 h incubation period.Incubations were performed after sealing the experimental cylinders.A deflated O 2 impermeable intravenous drip bag (medical supply PVC) was used to aid water sampling: Water was added to the drip bag through a silicon port which displaced the sample water inside the cylinders by pushing it out through the sampling port into 50 mL Winkler bottles, keeping the total volume of the cylinders stable and minimizing gas exchange.Oxygen content of the water samples was determined by the Winkler method [47] and determined by spectrophotometry at a wavelength of 466 nm [48].Mean oxygen consumption values for each cylinder were used as statistical replicates after standardizing to the estimated number of larvae in a cylinder at the end of the experiment (see Section 2.3.3.).

Larval Size and Survival
The experiment ended 58 h after fertilization, when most of the larvae the were at the late trochophore stage and some had reached the early D-veliger stage (Figure 1).Three subsamples of larvae from each experimental cylinder were fixed in borax buffered formalin.The mean number of larvae remaining in each cylinder was divided by the initial number of embryos added (98,000) and used as the estimate of survival rate.Sizes of 50 larvae from each cylinder were measured along the longest axis.Mean values from each cylinder were used as statistical replicates.

Carbonate System
Total hydrogen ion concentrations (pH) and dissolved inorganic carbon (DIC) were measured at the beginning [0.5 h after fertilization (a.f)] and near the end of the experiment (47.5 h a.f.) to determine if targeted CO 2 concentrations were achieved.Seawater pH was measured following standard operating procedure (SOP 6b from DOE [49]) at 25 • C. Triplicate samples were collected with a silicon tube and directly siphoned into glass cuvettes previously rinsed with sample water.Cuvettes were submerged in a 25 • C temperature bath for at least 30 min prior to measurement in a Jasgo 7800 spectrophotometer.The temperature bath was connected to a continuous flow water jacket cuvette support inside the spectrophotometer to ensure temperature stability throughout the measurement.Samples for DIC were collected in 10 mL amber glass ampoules, poisoned with 0.2 µL HgCl 2 , flame sealed and processed at the National Oceanography Centre in Southampton (UK).Measurements were calibrated using Certified Reference Material from the Dickson lab at the Scripps Institute for Oceanography.Water pH, pCO 2 , total alkalinity (A T ), and aragonite saturation state (Ω Ar ) were calculated for experimental temperatures with CO2SYS program [50] from pH and DIC as the carbon dioxide system parameters.Sensors were used to constantly monitor pH values during the experiment.Values measured with pH sensors are given in National Bureau of Standards scale (pH NBS ), whereas pH values measured with spectrophotometer are in total pH scale (pH T ).Therefore, pH NBS values are approximately 0.15 units higher than pH T values [51].

Statistical Analyses
The natural logarithm of the response ratio [R = XTreatment / XControl ; ln(R) = L = ln( XTreatment ) −ln( XControl )] was calculated from mean values obtained from replicate cylinders (n = 3) and males (sperm activity data, n = 13) [52].Means and the 95% confidence intervals (CIs) for L were calculated using Equations 1 and 2 from Hedges et al. [53], with the exception that t-instead of z-distribution was used to obtain more reliable CIs for small samples.The method led to slightly wider CIs than directly following Hedges et al. [53].Percentage values (survival and percentage of motile sperm) were arcsine-square root transformed before calculating L [54].Interactions between pH and temperature in the larval experiment were assessed using two-factor analysis of variance (2-ANOVA).The effect of time lag between sperm activation and activity measurements was examined with analyses of covariance and regression analyses using R [55].
Literature of OA effects on Mytilus larvae was examined by searching with Google Scholar using keywords "Mytilus", "ocean acidification", and "larvae" (5 relevant studies found on 6 November 2013).As larval size was the only parameter measured by every study, it was further evaluated by extracting pH, mean response and standard deviation for pCO 2 treatments and controls from the articles.L and 95% CIs were calculated assuming a normal distribution following Hedges et al. [53] directly.L was back-calculated to percentage response ratios (R) for graphical presentation only.Figures were made with ggplot2 package [56] for R.

Carbonate System
Average pH values in control cylinders remained between 8.0-8.1 pH T over the larval experiment, whereas pH in CO 2 treatment cylinders decreased gradually (Figure 2).The difference in pH values between treatment and control cylinders was approximately 0.2 units 2 h after addition of embryos (a.e.) (4 h a.f.), 0.3 units 8 h a.e.(10 h a.f.), and stabilized to approximately 0.35 units at the middle of the experiment.The mean difference measured with spectrophotometer 44.5 h a.e.(47.5 h a.f.) was 0.37 and 0.38 pH units (Table 1), and the average difference measured with sensors over the entire experiment was 0.38 and 0.40 pH units for low and high temperature treatments, respectively.pH T values measured from tanks for sperm motility treatment water corresponded to averages measured from cylinders.Aragonite saturation state (Ω Ar ) stayed well over the under-saturation limit (1) during the entire experiment.Average total alkalinity (A T ) calculated from pH and DIC varied between 2969 and 3226 µeq kg −1 .Average DIC varied between 2834 and 2964 µ kg −1 .

Sperm Activity
An increased pCO 2 had a substantial and statistically significant negative effect on percentage of motile sperm (95% CIs for R = 57%-88%, R = 71%) and swimming speed (95% CIs for R = 61%-89%, R = 74%) (Figure 3a).Average percentage of motile sperm was 48.1 ± 12.1 (σ) % and 28.2 ± 14.6 (σ) % for control and treatment, respectively.Average sperm swimming speed was 32.2 ± 7.8 (σ) µm s −1 for control and 23.7 ± 5.6 (σ) µm s −1 for treatment (Figure A2).Time from spawning had a significant negative effect on sperm swimming speed (Table 2, Speed), but the rate of decrease was not significantly different between control (CSW) and treatment (TSW) [Table 2, Ln( RSpeed )].Percentage of motile sperm was not significantly affected by time from spawning (Table 2, Motility).The pCO 2 treatment effect was larger than the sperm activity decline over time for both sperm swimming speed and percentage of motile sperm.

Larval Experiment
There were no significant interactions between pCO 2 and temperature treatments in the larval experiment (Table A1).None of the studied parameters were significantly affected by the elevated pCO 2 treatment (Figure 3b), whereas increased temperature had a large significant effect on larval metabolism increasing the respiration rate by a factor of four (Figure 3c).Mean responses of larval survival and size in elevated pCO 2 were almost identical to those in controls (survival RLow = 96% and RHigh = 100%; size RLow = 96% and RHigh = 102%) and 95% CIs were small (survival R Low = 78%-118% and R High = 59%-170%; size R Low = 88%-105% and R High = 92%-113%).On the other hand, 95% CIs of effect sizes for respiration (R Low = 31%-605% and R High = 40%-293%) and calcification rate (R Low = 18%-1401% and R High = 17%-548%) had a wide range indicating large variation in responses for these variables.Temperature significantly increased larval metabolism in both pCO 2 treatments ( RCSW = 481% and RTSW = 378%), although the 95% CIs around the mean response ratios were broad ( RCSW = 122%-1900% and RTSW = 121%-1181%).Temperature had a positive mean effect on calcification (348%), but the effect was not as strong in the high pCO 2 treatment as in the low pCO 2 treatment ( RCSW = 435% and RTSW = 260%).Variation in calcification rate with temperature was high as indicated by very large 95% CIs (R CSW = 128%-1480% and R TSW = 20%-3334%).Despite this high variability, there was a significant effect of temperature at control levels of pCO 2 , while the effect of temperature on calcification rate at high pCO 2 was non-significant.Larval survival declined at higher temperatures ( RCSW = 68% and RTSW = 71%), although the response ratios were marginally non-significant (R CSW = 46%-100% and R TSW = 46%-108%).The relatively large effect size and consistent differences between treatment and control cylinders for larval survival (see Figure A2) suggest that the statistical non-significance was due to the conservative method for calculating 95% CIs (i.e., if the method from [53] had been used directly, 95% CIs would have been R CSW = 50%-91% and R TSW = 51%-97%).In contrast, the effect sizes of temperature on larval size were small ( RCSW = 88% and RTSW = 93%), but statistically significant for control pCO 2 due to narrow 95% CIs (R CSW = 81%-95% and R TSW = 83%-104%).

Carbonate System
Open ocean, near-surface, pH values in the Western Mediterranean varied between 8.05 and 8.23 on the NBS scale in 1970s [57], and average surface pCO 2 vary between 395 and 442 µatm [17].Our experimental values (Figure 2) for control cylinders are close to these historical values, and treatment pH values are slightly lower than open ocean scenarios for the end of 2100 [5], yet within a reasonable range considering the variability of pH values in coastal systems [17,19,58].Ocean pH can be highly variable in near-shore locations: Hofmann et al. [18] reports daily variations of 0.2-0.3pH units for estuarine/near-shore locations and stochastic short term variations of more than 0.6 units.Therefore, it is possible that M. galloprovincialis populations may already experience treatment pH values for short periods [19].The gradually decreasing trend of pH values in the high pCO 2 treatment (Figure 2) may have reduced the potential pH shock experienced by the embryos and led to a gradual acclimation as recommended by Riebesell et al. [59].
The use of a commercial salt mixture (Instant Ocean) led to approximately 500-750 µ kg −1 higher total alkalinity (A T ) and dissolved organic carbon (DIC) values in control cylinders than has been measured for the Mediterranean Sea [57,60,61] (Table 1).Furthermore, DIC and A T remained at higher levels in both control and treatment cylinders compared to the predictions for the end of the 21st century [62].Despite these high values, aragonite saturation state (Ω Ar ) remained within an expected range in control [57,60] and treatment [62] cylinders.Whereas pH values were within an expected range, pCO 2 values were higher both in control [57] and treatment [62] than expected, indicating that a higher amount of CO 2 was required to acidify water in this experiment [51], hence explaining the high DIC values.Although we are not aware of any mechanisms by which A T directly affects fertilization or early development of bivalve larvae, it is possible that our Ω Ar for pCO 2 treatments would have been lower had we not used artificial seawater.Consequently, high A T could partly explain the lack of significant pCO 2 effects on calcification and shell growth (see Sections 4.3.1.and 4.3.3. ) as discussed by Range et al. [63].On the contradictory, higher concentration of CO 2 could have caused an increased stress to the larvae [58,64], which could potentially increase the probability of detecting a pCO 2 effect on larval growth and calcification.Nevertheless, high A T must be considered as an explanation contributing to our lack of significant effects.

Acidification Effects on Sperm Activity
We found a significant 25% reduction in percentage of motile sperm and sperm swimming speed as a result of elevated pCO 2 (Figure 3a).Reports for sea-urchins suggest sperm motility is negatively impacted by acidification [35,65].Negative effects of OA on sperm motility have also been reported for corals and sea cucumbers [66,67], while positive effects have been reported in one species of sea urchin [39].Reduced sperm swimming performance in marine broadcast spawners is likely to decrease fertilization in sperm limited low density populations [68][69][70][71][72][73].However, fertilization in broadcast spawners is a complex process affected by a range of parameters, such as egg size [68], gamete age [74], distance between individuals [75], chemical signals [76], gamete compatibility [77], and water movements [78,79].Predicting how reduced sperm swimming performance will impact fertilization success in a population, let alone the entire geographical range of a species, therefore, requires knowledge of the particular habitat the population is occupying.Nonetheless, the large changes in sperm swimming performance observed in our experiments are likely to have flow-on effects for fertilization success in free-spawning species such as M. galloprovincialis as the ocean pH decreases.Although temperature was not included in our sperm activity assessment, temperatures outside of a species' optimal range are likely to lead to a reduced sperm fertilization capability.Higher temperatures might also increase the swimming speed and decrease the longevity of sperm [72,80].
A potential bias in this study was caused by us not using 'dry-sperm', but introducing spawning with a temperature shock.This resulted in sperm being activated only at control pH and temperature, and lead to a variable time lag (1-4.5 h) between recording of sperm activity and activation of sperm.Although similar method has been used previously [67,73] in sperm studies, we have no way knowing to what degree, if any the pH shock affected sperm activity.Further studies are needed to judge whether this could cause a significant bias.Although the time lag had a significant effect on sperm swimming speed (Table 2), the rate at which the sperm swimming speed declined did not significantly differ between treatment and control.This makes us relatively confident that the observed effect was not caused by the time lag between sperm activity measurements and sperm activation.

Overall Effects in the Larval Experiment
We did not observe significant interactions between pH and temperature treatments (Table A1).Overall, warming had a more prominent effect on larval stages than pCO 2 , reducing performance (growth and survival) and increasing energy demand (respiration and calcification) (Figure 3b,c).Although the temperature effect is dependent on a species optimal temperature range, this finding is generally in line with observations on calcifying larvae [12,16].We did not observe significant effects of pCO 2 on larval survival, size, respiration or calcification (Figure 3b).

Acidification Effects on Larval Survival and Size
Relatively narrow 95% CIs and small effect sizes of pCO 2 for larval survival and size indicate that the lack of a significant response is not due to high variability in measured values, and that pCO 2 is likely to have a negligible effect for these variables (Figure 3b).The lack of a pCO 2 effect on larval survival is supported by Gazeau et al. [33], who found no effect for Mytilus edulis for the near-future pH range evaluated in this study.In contrast, increased pCO 2 over a range predicted to occur by 2100 has been reported to reduce larval growth of Mytilus (Figure 4).These effects [33,34,81] fall within the 95% CIs of our study indicating that our study design may not have been able to detect such small effects.Three studies with comparable ∆pH [33,34,82] had all lower Ω Ar values for pCO 2 treatments than we measured, highlighting the possibility that the usage of artificial seasalt might have attenuated the detrimental effect of pCO 2 treatment on larval calcification.Also, the short development time (55 h, until late trochophore-early D-veliger) could explain the lack of a significant effect in this study, although Gazeau et al. [33] found a significant effect already after 48 h of development in M. edulis.Bechmann et al. [82] reported a significant reduction in larval growth of M. edulis over a greater ∆pH than used in in this study.Reanalysis of their data using mean and standard deviation values for number of chambers (n = 7) as replicates (avoiding pseudoreplication [83,84]) gives a considerable negative effect size, but with large 95% CIs that indicate a non-significant effect of pCO 2 (Figure 4).From the few studies conducted across the Mytilus genus, it appears that larval growth is reduced by approximately 5% for a ∼0.3 unit decrease in pH, which is within the range of fossil fuel intensive ICPP scenarios for 2100 [5].It is not until OA levels beyond those estimated for 2100 that growth is reduced by more than 10%.[5,85] and projected IS92a maximum pH reduction following the burning of the remaining fossil-fuel resources (Max) [1].Red points [82] have been recalculated to reflect proper levels of replication (see Section 4.3.1.).This does present a paradox as near-shore Mytilus habitats are naturally exposed to such daily to seasonal pH differences due to metabolic activity of organisms in the euphotic zone, and fresh-water discharge [18,19,58,86,87].Pelagic larval stages may be a strategy to avoid such pH fluctuations during development, resulting in their continued sensitivity to experimental OA.Consequently, the pH variability might contribute to inter-population and site-specific variation in Mytilus reproductive success.It should be noted however, that the meta-analysis presented in Figure 4 does not consider long-term effects [88], genetic variability [81], trans-generational effects [89], nor effects of adult exposure during reproductive conditioning [90].It is possible that the effect of OA in the figure is over-stated, and that the Mediterranean population of M. galloprovincialis is more adapted to variable pH regime than populations that experience less pH variation.However, if Mytilus populations were not able to adapt to decreasing pH, ocean acidification may subject Mytilus larvae to even lower pH values, and larger fluctuations, than currently experienced, with potential negative consequences.

Temperature Effects on Metabolism, Larval Survival and Size
Positive effects of temperature on respiration rates are consistent with known effects of temperature on ectotherm metabolism [91][92][93], and demonstrate that our experimental design was capable of detecting differences, even when variation among replicates was high for some response variables (Figure 3c).We found modest negative mean effects of warming on larval survival and size (=growth) between 17 • C (T Low ) and 20 • C (T High ), although the effect size of temperature on larval size was only around 10% and the effect of larval survival was barely non-significant.These negative effects are in contrast to those reported by Sánchez-Lazo and Martínez-Pita [94], who found increased larval performance (increased growth and survival, and faster development) for M. galloprovincialis larvae grown at warmer temperatures (20 or 24 • C) instead of 17 • C. Sánchez-Lazo and Martínez-Pita [94] tested larvae from mussels growing in Southern Andalucia, approximately 900 km southwest of the site in Menorca where our mussels were collected.Optimal growth temperatures of 20-25 • C for M. galloprovincialis larvae have been reported by several studies [95][96][97], however, the performance of M. galloprovincialis larvae is known to vary depending on the population even within a small geographical range [98,99].Both adult and larvae of M. galloprovincialis seem to be sensitive to temperatures higher than 24-26 • C, after which the growth decreases and mortality increases [14,95,100].Adaptation of M. galloprovincialis populations to temperatures of their respective habitats could explain the observed difference between previous studies [94][95][96][97] and our results.

Calcification
The response of calcification to increased pCO 2 was non-significant (Figure 3b), while the effect of temperature was accompanied by large effect sizes, which were significant under control conditions and non-significant in the high pCO 2 treatment (Figure 3c).Aragonite saturation (Ω Ar ) values for the high pCO 2 treatment cylinders varied between 1.6 and 2.0.These values are above saturation (1), but within a range expected for 2100 (see Section 4.1).Studies have reported generally negative relationships between calcification rate and saturation state the effects starting well before under-saturation [31,[101][102][103].Gazeau et al. [103] demonstrated a linearly declining calcification rate with increasing pCO 2 on adult and juvenile M. edulis, with significant negative effects starting at Ω Ar values similar to our treatment values.The effect sizes calculated from Gazeau et al. [103] for treatment Ω Ar values between 1.6 and 2.0 are within the 95% CIs in our study indicating that the variability in our calcification results may mask biologically relevant effects.However, it should be noted that Gazeau et al. [103] studied adults during two hour incubations, and these results may not be directly comparable to ours.
It is more likely that our results reflect a biologically relevant signal for the studied Ω Ar values.Calcification measurements were conducted 48-55 h after fertilization, when most larvae were in late trochophore stage with occasional early D-veligers (Figure 1).Although M. galloprovincialis larvae in late veliger stage are known to calcify [104], their calcification rate is likely much lower compared to D-veliger stage [32].Hence, D-veligers could be responsible for most of the calcification measured in the cylinders, and uneven number of them among cylinders could explain the high variability in the calcification data.Temperature, however, seems to have a negative effect on calcification starting already at the early stages in this study.Malone and Dodd [105] reported a positive effect of temperature on calcification rates of adult M. edulis, but also that variability among individuals increased strongly with increasing temperature.This variation was partly explained by some individuals that did not calcify during the 24-h experimental period.Large variation is also evident in our data, both for temperature and pCO 2 effects, although the increasing variation with temperature is not strongly supported (Figure A2).Our results are in conflict with Duarte et al. [106], who found a decreasing net calcification rate with increasing pCO 2 and no temperature effect on juvenile Mytilus chilensis.This suggests that there may be an interspecific or life-stage dependent response to temperature within the Mytilus genus.
The method we used to measure calcification (microdiffusion technique using radiolabeled 14 C bicarbonate as a substrate) was developed for phytoplankton [45] and relies on the assumption that calcifiers are evenly distributed in the sampling population.We extracted small subsamples (40 mL each) from each cylinder using two subsamples as replicates assuming an even distribution of bivalve larvae in the cylinders.This is a reasonable assumption as the bubbling design should have mixed the larvae evenly, but it is possible that our subsamples did not contain a similar number of larvae, as it was not possible to count the number of larvae in the subsamples.However, as calcification values varied ten-fold between replicates (Figure A2), it is unlikely that the subsampling explains all of the variation between replicates.

Conclusions
Although M. galloprovincialis is already likely to be exposed for short periods to pH values corresponding to the treatments in this study, we found a substantial reduction in sperm activity when exposed to the high pCO 2 treatment.Even though we cannot demonstrate that this reduction translates into lower fertilization success, modeling and experimental manipulations show that such a reduction typically leads to poorer fertilization ability of sperm in low sperm concentrations [65,68,69].Our results from the larval study indicate that ocean warming may have a larger negative effect on very early life-stages (until late trochophore-early D-veliger) of M. galloprovincialis than near-future OA.Nevertheless, the negative effects of temperature on larval survival and size are in contrast with earlier studies showing the optimal temperature range of M. galloprovincialis to be between 20 and 25 • C [94][95][96][97].These results may indicate variability among populations of M. galloprovincialis as a factor that should also be considered when assessing the impact of climate change on species and ecosystems [84].Mytilus species are naturally exposed to a highly variable pH regime [19,87], and spatial and temporal variability in pH may contribute to inter-population and site-specific variation in this mortality.Ocean acidification, even though not the main source of reduced performance and decreased growth in our experiments, might act synergistically with the stress of expected temperature increases to negatively affect the early life-stages of M. galloprovincialis.As the pelagic phase is crucial for maintaining populations and dispersal of this species [107], these impacts may have repercussions at the population level in the future.Because genetically inherited traits play a key role in recruitment processes for benthic species with planktonic larvae [108], genotypes more resistant to OA and warming will be favored in the future, which, in turn, could alter the population sensitivity to future environmental fluctuations.Our results provide an insight into a subset of a M. galloprovincialis population from a commercial farm in the Mediterranean.Considering the conflicting results by studies on Mytilus, more research is needed to provide a robust synthesis on climate change effects on this wide-spread, ecologically and commercially important genus.The future studies should test the individual variability with respect to increased pCO 2 and temperature, and consider habitat-specific variation in pH values for the study organism.(1) Ambient air was collected via aquarium pumps and stripped of CO 2 by soda-lime columns; CO 2 gas was obtained from tanks (2) Gasses were mixed in marble filled containers using mass flow controllers to achieve the precise ratio of stripped air and CO 2 gas and derive pCO 2 concentrations of 380 (control, CSW) and 1000 ppm (treatment, TSW); (3) Three replicate cylinders and one 10 L tank per treatment were bubbled with CO 2 controlled air.Sensors measuring pH were deployed in two cylinders for each treatment combination; (4) Cylinders were submerged into two separate tanks containing fresh water, which was passed through temperature control units set at 16

Figure 1 .
Figure 1.Mytilus galloprovincialis larvae from the end of the experiment [58 h after fertilization (a.f.)].(a) Late trochophore stage larvae incubated at control conditions (pCO 2 = 380 ppm, T = 16 • C) showing the univalved shell; (b) The occasional early D-veliger larva with a bivalved shell among the majority of late trochophores.

Figure 2 .
Figure 2. Seawater pH in experimental cylinders over the larval experiment.Points with error bars show mean pH T (±0.5σ) derived from spectrophotometer measurements 0.5 h and 47.5 h after fertilization (left y-axis scale).Lines show average pH NBS measured with sensors from two cylinders per treatment combination (shading: ±0.5σ; right y-axis scale).Light orange refers to the high pCO 2 treatment (TSW) and dark blue to the control (CSW).Dots and solid line refer to the low, and triangles and dotted lines to the high temperature treatment.Dot-dashed vertical line shows the time of fertilization (A), solid line the time when embryos were introduced (B), dark gray shading between dashed lines the time frame of calcification measurements (C), light gray shading between dotted lines the time of metabolism measurements (D; C and D overlapped, hence the dark grey zone), and solid line both, the end of the larval experiment, and the time of larval size and survival measurements (E).

Figure 3 .
Figure 3.Effect size (R) for a range of sperm activity and early life-stage parameters of M. galloprovincialis: (a) sperm motility and sperm swimming speed between treatment and control pCO 2 at control temperature; (b) larval parameters (survival rate and size 58 h after fertilization (a.f), larval respiration rate 51-57 h a.f., and calcification rate 48-55 h a.f.) between treatment and control pCO 2 ; and (c) larval parameters between treatment and control temperature.Dots and triangles represent mean effect sizes ( R), and error bars give the 95% CIs.CIs crossing the black line at 100% indicate no significant difference between pCO 2 treatment and control (a and b) or high and low temperature treatments (c).Values less than 100% indicate that either the elevated pCO 2 or the high temperature treatment resulted in lower values compared to the respective control.Asterisks ( * ) indicate statistically significant effects, i.e., when the 95% CIs did not cross the 100% line.Note the different logarithmic scales on y-axes.

Figure 4 .
Figure 4. Mean effect size of pCO 2 on growth of Mytilus larvae 48-360 h after fertilization extracted from five experimental studies.Colors and symbols represent different studies.Abbreviation after the reference in legend indicates the studied species (MG = M. galloprovincialis, ME = M. edulis, MC = M. californianus, and MT = M. trossulus).Double point for a single pH difference in a study indicates the effect measured at different times during the experiment (darker hue pinpoints the later measurement).Error bars show the 95% CIs estimated assuming a normal distribution.Rectangles with grey shading illustrate the range of IPCC SRES projected global average surface pH pathways (2100 range)[5,85] and projected IS92a maximum pH reduction following the burning of the remaining fossil-fuel resources (Max)[1].Red points[82] have been recalculated to reflect proper levels of replication (see Section 4.3.1.).

Figure A1 .
Figure A1.Schematic representation of the setup for the larval development experiment.(1)Ambient air was collected via aquarium pumps and stripped of CO 2 by soda-lime columns; CO 2 gas was obtained from tanks (2) Gasses were mixed in marble filled containers using mass flow controllers to achieve the precise ratio of stripped air and CO 2 gas and derive pCO 2 concentrations of 380 (control, CSW) and 1000 ppm (treatment, TSW); (3) Three replicate cylinders and one 10 L tank per treatment were bubbled with CO 2 controlled air.Sensors measuring pH were deployed in two cylinders for each treatment combination; (4) Cylinders were submerged into two separate tanks containing fresh water, which was passed through temperature control units set at 16 • C and 20 • C. See Table 1 for actual experimental temperatures and pCO 2 values.

Figure A2 .
Figure A2.Raw data used for calculating the effect sizes (Figure 3): (a) sperm swimming speed and (b) sperm motility measured at 16 • C. Lines between points mark individuals; (c) survival rate; (d) larval size; (e) respiration and (f) calcification rate of larvae measured in low (≈17 • C) and high temperature (≈20 • C) treatments.Dots indicate mean values for replicates (n = 13 for (a,b), 3 for (c-e), 2 for low temperature TSW, otherwise 3 in (f)) and error bars show SEX.Measurements without error bars lack measurement replication in (f).Dark blue indicates control pCO 2 treatment, while light orange high pCO 2 treatment.

Table 2 .
Analysis of Covariance parameters for sperm swimming speed (Speed) and percentage of motile sperm (Motility), and regression analysis parameters for Ln response ratio Ln( R) of sperm swimming speed and percentage of motile sperm.Time (h) means hours from spawning, and pCO 2 (TSW) is the independent variable and states the estimated difference from control pCO 2 conditions.Df means degrees of freedom and SE standard error.
• C and 20 • C. See Table 1 for actual experimental temperatures and pCO 2 values.