Simulated leakage of high pCO2 water negatively impacts bivalve dominated infaunal communities from the Western Baltic Sea

Carbon capture and storage is promoted as a mitigation method counteracting the increase of atmospheric CO2 levels. However, at this stage, environmental consequences of potential CO2 leakage from sub-seabed storage sites are still largely unknown. In a 3-month-long mesocosm experiment, this study assessed the impact of elevated pCO2 levels (1,500 to 24,400 μatm) on Cerastoderma edule dominated benthic communities from the Baltic Sea. Mortality of C. edule was significantly increased in the highest treatment (24,400 μatm) and exceeded 50%. Furthermore, mortality of small size classes (0–1 cm) was significantly increased in treatment levels ≥6,600 μatm. First signs of external shell dissolution became visible at ≥1,500 μatm, holes were observed at >6,600 μatm. C. edule body condition decreased significantly at all treatment levels (1,500–24,400 μatm). Dominant meiofauna taxa remained unaffected in abundance. Densities of calcifying meiofauna taxa (i.e. Gastropoda and Ostracoda) decreased in high CO2 treatments (>6,600 μatm), while the non - calcifying Gastrotricha significantly increased in abundance at 24,400 μatm. In addition, microbial community composition was altered at the highest pCO2 level. We conclude that strong CO2 leakage can alter benthic infauna community composition at multiple trophic levels, likely due to high mortality of the dominant macrofauna species C. edule.

Scientific RepoRts | 6:31447 | DOI: 10.1038/srep31447 of a continuous diffuse seepage up to > − 1.0 pH units in a continuous point source leakage scenario 7,8 . The latter represents the extreme case of a major geological fault break or an unrecognized pipe failure with outgassing of high amounts of CO 2 into the marine environment [9][10][11][12] . Experimental release of ca. 4 tons of CO 2 via sediment injection during the 37 day long QUICS experiment in a Scottish Sea Loch resulted in maximum decreases in sediment pH by > 0.8 pH units 12 .
Studies that have investigated the influence of elevated seawater pCO 2 on benthic organisms primarily used short-to medium-term exposures, single species experiments and pCO 2 levels relevant for ocean acidification (OA) forecasts. The influence of medium-term exposure to high pCO 2 on communities is not well investigated and different threshold levels need to be identified. The sensitivity of marine organisms to varying pCO 2 levels and accompanying seawater pH variations differs between taxa, resulting in very diverse responses to elevated internal CO 2 concentrations (hypercapnia) 13 . The general response of animals to hypercapnia is related to disturbance and regulation of the intra-and extracellular acid-base balance [13][14][15] . Those processes are energy consuming and can result in growth reduction, decreased metabolic activity, low reproduction rates, behavioural changes or ultimately death 13,16,17 . Calcifying organisms are particularly susceptible to acidified seawater due to reduced aragonite and calcite saturation of the water which can lead to enhanced dissolution of unprotected skeleton components and reduced availability of inorganic carbon for calcification [18][19][20] . Non-calcifying, infaunal invertebrates (e.g. meiofauna, and more particular the dominant nematodes) with a low mobility and bacterial communities are naturally exposed to large fluctuations of pore water pH and pCO 2 21,22 , and are therefore likely to be more tolerant to high pCO 2 (OA) 13 .
Only a handful of studies have exposed sediment communities in their natural composition to elevated seawater pCO 2 , primarily due to the great logistic effort necessary to collect and maintain such communities in the laboratory and to control the carbonate system with sufficient accuracy during experiments. Negative effects of medium-term (20 weeks) exposure to high seawater pCO 2 (pH 7.3-5.6) have been reported on macro-and meiofauna diversity and community structure 23 . In addition, Widdicombe and Needham 24 could show that even though seawater acidification did not alter nereid worm burrow size or structure, sediment nutrient fluxes were changed after five weeks of acidification exposure, supposedly due to changes in bacterial communities. Moreover, since the burrowing activity of e.g. echinoderms has a strong influence on the biogeochemistry of sediments and the composition of meiofauna communities 25 , it seems likely that changes in macrofauna abundance in response to elevated seawater pCO 2 can have strong repercussions on infaunal ecosystem processes 23,24 .
Up to date, no comprehensive study has addressed the impact of elevated seawater pCO 2 and acidification on benthic bivalve dominated communities (e.g. bivalves, meiofauna and bacteria) thriving in shallow, sandy coastal sediments in the North Atlantic region. Bivalves are of high ecological importance as they represent the defining macrobenthic organisms in these habitats, with a key nutritional role for many fish and migratory bird species 26,27 . As they filter large amounts of water, deposit particulate organic matter onto and into the sediment and provide settlement habitat for other organisms, they are irreplaceable key players in the benthic-pelagic coupling 28,29 .
A range of recent studies on several (mainly epibenthic) bivalve species could establish that calcification, growth, filtration and metabolism can be negatively impacted by elevated seawater pCO 2 , often already at moderately decreased pH > 7.5 (ca. < 5,000 μ atm). In addition, increased occurrence of oxidative and acid-base regulatory stress has been observed in bivalves 16,[30][31][32] . Detrimental external and internal shell dissolution occurs in several species when these are exposed to acidified seawater for prolonged time intervals (weeks), with less severe effects observed in species that possess a thick and intact periostracum [33][34][35][36] . However, biological impacts such as bioturbation 22,24,25 or interactions between and among different taxa and trophic groups 23,25,37,38 can result in very diverse responses of the biological community when compared to single species/taxon studies. Mesocosm experiments which address the effect of seawater acidification on shallow water benthic species report an overall higher sensitivity of macrofauna while meiofauna, with nematodes as dominant and commonly reported taxon, remain unaffected or even increase in abundance 23,25,37,39 . Laboratory experiments confirm a strong resistance of dominant meiofauna taxa to pH changes as expected to occur as a result of OA by the year 2300 39 but a reduced survival of nematodes has been observed when the pH is further reduced to values that could potentially occur in response to a severe CCS leakage scenario (< pH 7) 40 . Microbial communities are important for the coastal ecosystem as they are primarily responsible for organic matter remineralisation, nutrient cycling and function as food source for grazing marine organisms [41][42][43][44] . Investigations about the direct effects of lowered pH are still inconsistent 41,45 . While some studies observed differences in abundance of a few genera or rare taxa, others found a change in bacterial biomass production with increasing pCO 2 [46][47][48][49] .
In this study, we conducted a mesocosm experiment using natural C. edule dominated sandy communities from the Western Baltic Sea utilizing a flow-through seawater design with optimized food supply. Three of the dominant infauna bivalve species from the western Baltic and North Sea are the cockle Cerastoderma edule, the soft-shell clam Mya arenaria, and the Baltic tellin Macoma balthica 50 . C. edule and M. balthica live within the top two to five centimetres of the sediment, while M. arenaria occurs to sediment depths of up to 50 cm 51 . Generally, all three species are widespread in tidal flats and shallow coastal areas and serve as an important link between primary producers and consumers. Owing to their very high densities in the sediment (often > 1,000 individuals m −2 ) and considerable mobility and activity 52 they exert a strong influence on infaunal communities and biogeochemical processes, e.g. by increasing primary production and fertilization of microphytobenthos through NH 4 + excretion 52,53 . To predict marine benthic ecosystem vulnerability to potential chronic leakages from CCS storage sites, experiments need to utilize a high pCO 2 range, realize a multispecies approach and intermediate exposure durations. In search for an indicator species useable in future CCS site assessments, experiments focused on the most abundant bivalve C. edule. We hypothesized that C. edule would be very sensitive to elevated pCO 2 , as it lives close to the sediment surface. C. edule mainly occurs in relatively unpolluted areas while e.g. M. balthica can also be found in polluted and hypoxic areas 54 , suggesting a higher resistance of M. balthica to unfavourable environmental conditions. To extend the common single-species approach to an ecosystem investigation and to test for the role of C. edule as an important ecosystem engineer, the impact of high pCO 2 on benthic meiofauna and bacterial communities was also assessed. We hypothesized that elevated pCO 2 and changes in bivalve abundance and fitness would impact meiofaunal and microbial communities.

Materials and Methods
Experimental setup. Sandy communities were exposed to six different seawater pCO 2 regimes for a total of three months (17.12.2011-06.03.2012) in a climate -controlled room. Six header tanks were continuously supplied with filtered seawater from Kiel Fjord, each one connected to six experimental units (EU) ensuring continuous seawater supply (Supplementary Figure S5). This design is a randomized design (B4) according to Cornwell and Hurd 55 . Each EU consisted of a round plastic container with a volume of 12.5 L containing ca. 9.5 L of sediment and an overlying water column of ca. 3 L. The lower 10 cm of the sediment consisted of sieved sand taken from a local beach (Kiel, Falckenstein: 54°23,66 N; 10°11.56 E) while the upper 10 cm consisted of surface sediment from the station at which the experimental animals were sampled to resemble natural conditions as well as to provide naturally occurring microbial and meiofauna communities. Bivalves and sediment were sampled in Kiel Fjord at Falckenstein with a Van Veen grab in 1-2 m depth using the vessel FK Polarfuchs on November 21 st 2011 and kept in holding basins at 9 °C before being placed in EUs. Density of infauna bivalves was determined during the sampling process. 1 m 2 of sediment at Falckenstein was found to contain 146 M. arenaria, 9 M. balthica, and 1,040 C. edule. In order to simulate a natural size distribution in our laboratory experiment, we sieved the collected sediment (1 mm mesh width). All bivalves were taken out and replaced by a defined number per EU: 5 M. arenaria (size classes: 0.5-1 cm: 2 animals; 1-1.5 cm: 2 animals; 2-2.5 cm: 1 animal), 1 M. balthica, and 40 C. edule (size classes: 0-0.5 cm: 3 animals; 0.5-1 cm: 18 animals; 1-1.5 cm: 11 animals; 1.5-2 cm: 7 animals; 2-2.5 cm: 1 animal). Small gastropods (exclusively Hydrobia spp.) were abundant with ~10 individuals per EU. Due to their small size (< 0.5 mm) they were randomly distributed within all EUs with the sieved sediment. Due to the natural low diversity of the Baltic, the density of other macrofauna individuals was < 1 individuals per m 2 . These low abundant species (e.g. nereid polychaetes, pharid bivalve species) were excluded from the experiment. The experimental units were kept in a seawater flow-through system for two weeks under control conditions prior to the experiment to allow proper acclimatization of biogeochemistry and the faunal community. Seawater pH was maintained in the header tanks using a pH feedback system (IKS Aquastar, iksComputersysteme GmbH, Karlsbad, Germany). Treatment levels were achieved through continuous addition of acidified water from the header tanks into the overlaying seawater of each EU and included levels of 900 μ atm (control, pH 7.8 NBS scale), 1,500 μ atm (pH 7.7), 2,900 μ atm (pH 7.4), 6,600 μ atm (pH 7.0), 12,800 μ atm (pH 6.7), and 24,400 μ atm (pH 6.4) (Supplementary Table S1). 900 μ atm was used as a control due to the high background pCO 2 in Kiel Fjord 31,35 . To support the bivalve nutritional needs unicellular algae (Rhodomonas sp.) were cultured as described in Thomsen et al. 35 and added continuously into the header tanks via a peristaltic pump, thus maintaining a stable concentration of 3,500-4,000 (Supplementary Table S1) cells ml −1 within header tanks (Supplementary Table S1).
A flow rate of 100 ml min −1 was provided to each EU from the respective header tank via gravity feed. Throughout the experiment, pH, salinity, temperature, and flow rate were measured daily in each replicate. Salinity and temperature fluctuated in accordance with naturally occurring changes in Kiel Fjord seawater (14.6-20.5 psu and 4.3-8.9 °C, respectively). Light conditions were similar for all EUs. Dead animals were removed daily and behaviour of bivalves (presence/absence on the sediment surface) was noted every other day starting in the third experimental week. Carbonate chemistry and algae concentration in the EUs were measured weekly. Dissolved inorganic carbon (C T ) was measured using an Automated Infrared Inorganic Carbon Analyzer (AIRICA, Marianda, Kiel, Germany). Seawater chemistry (pCO 2 and calcium carbonate saturation state) was then calculated according to the guide to best practices for ocean CO 2 measurements 56 , using CO2SYS 57 with pH (NBS scale) and C T , temperature, salinity, and first and second dissociation constants of carbonic acid in seawater 47 . Bivalve sampling and processing. At the end of the experiment, all but four C. edule specimen per EU were frozen at − 20 °C. Shell-free dry mass was measured according to Thomsen et al. 31 . To survey shell dissolution, five randomly selected C. edule from each treatment were analyzed using a stereomicroscope (40-fold magnification). In addition, scanning electron microscopy (SEM) was used to examine external shell dissolution at a higher resolution. For that purpose, three shells of the 900-6,600 μ atm CO 2 treatments (as signs of dissolution were obvious at higher levels) were mounted separately on SEM pedestal stubs, coated with gold-palladium and examined using scanning electron microscopy (Nanolab 7, Zeiss, Oberkochen, Germany and HitachiS4800, Hitachi High -Technologies Europe, Krefeld, Germany).
The extent of oxidative stress in the lipid fraction of the whole body tissue was estimated by analysing the malondialdehyde (MDA, a marker of lipid peroxidation) content in four C. edule specimens per EU, which were stored frozen at − 80 °C prior to measurement. The entire tissue mass of each bivalve was separately processed via grinding in liquid nitrogen. Fifty milligrams of ground tissue powder of each of the four specimens from the same EU were pooled together to obtain sufficient tissue amounts for the analysis (one pool per EU). MDA concentration of each replicate was determined following the protocol of Uchiyama and Mihara 58 . Approximately 100 mg tissue-powder was homogenized with 0.2% phosphoric acid in a 1:5 ratio (sample/phosphoric acid). In the next step an equivalent volume of 2% phosphoric acid was added. One blank (homogenate + 0.2 ml 3 mM hydrogen chloride) and two samples (homogenate + 0.2 ml TBA solution) were adjusted to a pH of 1.6 (with HCl or NaOH) and were incubated at 100 °C for one hour. After cooling, 0.5 ml of butanol was added, samples were vortexed for 40 seconds and centrifuged at 1,000 g for 5 minutes. The supernatant was collected into fresh vials and centrifuged for 5 minutes at 14,000 g. Samples and blanks were transferred to a 96 well plate and the extinction was measured in a plate reader (Plate Chameleon, Hidex, Turku, Finland) at 532 and 600 nm. Tissue concentration of MDA was calculated following equation (1).
Scientific RepoRts | 6:31447 | DOI: 10.1038/srep31447 Meiofauna sampling and processing. The uppermost centimetre of each EU was sampled after six and twelve weeks using a small corer (Ø 2.5 cm) and stored in 4% buffered paraformaldehyde until further extraction of the meiofauna. For this purpose, samples were washed on two stacked sieves in order to separate the macrofauna fraction (on a 1mm sieve) from the meiofauna fraction (on a 38 μ m sieve). Subsequently, meiofauna was extracted by triple density gradient centrifugation (3000 rpm) with colloidal silica gel LUDOX HS40 Dupont (specific gravity 1.18) as a flotation medium 59 , fixed in 4% buffered formaldehyde solution and stained with Rose Bengal. Meiofauna was counted under a stereo microscope (50x magnification) and identified to higher taxonomic level consulting e.g. Higgins and Thiel 60 . Only nematodes that were in a good shape and showed no clear signs of degradation (loose cuticle, lack or damage of internal structures, biofilm, flattened or dehydrated habitus) were counted. For nematode identification, 100 nematodes were picked out per sample, transferred to glycerine following the protocol of De Grisse 61 and subsequently mounted on paraffin-ring glass slides. Nematodes were identified to genus level following a pictorial key 62 and the NeMys database 63 under a microscope (Leica DMR, 10x-100x magnification).
Microbial community sampling and processing. After six and twelve weeks, one corer sample (Ø 2.5 cm, 1 cm depth) per EU was collected, transferred into 50 ml plastic tubes and stored frozen at − 20 °C for bacterial community analysis. The samples were subjected to total community DNA extraction using the FastDNA SPIN Kit for Soil (Qbiogene, Carlsbad, CA) including an additional heating step to increase yield and final elution of the DNA in TE-buffer. Benthic bacterial community structures were determined by means of the high-throughput fingerprinting technique ARISA, following a previously published procedure by Ramette 43  To test for differences in C. edule behaviour, a permutational ANOVA (PERMANOVA), with the factor EU nested in treatment, was used as a non-parametric solution to a repeated measures analysis. C. edule mortality between treatments and the fraction of dissolved shells between treatments was tested using a Kruskal-Wallis test combined with a Kruskal multiple comparisons test since assumptions of normality (Shapiro-Wilk test) and homogeneity of variances (Levene's test) were not met. Mortality between different size classes within each treatment was tested with a Kruskal-Wallis test. For this, animals were pooled in two groups, large (1-2, 5 cm shell width) and small (0-1 cm shell width). The influence of pCO 2 on malondialdehyde (MDA) content was tested using a one-way analysis of variance (ANOVA) and a Tukey HSD post hoc test. The difference in shell-free dry mass between treatments was tested with an analysis of covariance (ANCOVA), using log shell -free dry mass and log shell width. As no significant interaction was found for replicate and treatment (nested ANOVA, EU nested in CO 2 treatment, p-value = 0.8411), all measured animals (n = 479) of each treatment were used in the regression analysis.
Meiofauna community analysis was based on calculated densities (individuals 10 cm −2 ) while nematode community analysis was based on relative abundances and calculated densities. Due to an unbalanced and non-normal distributed dataset, Permanova was chosen as an ANOVA approach for repeated measures to test for differences in total meiofauna densities and meiofauna and nematode community structure between treatments, time and interaction of both, with EU nested in treatment. When the number of unique permutations was lower than ten a Monte Carlo (MC) test was applied to calculate the p-value. The data were square root transformed and Bray-Curtis resemblance matrices were calculated. A pairwise-comparison test was executed for factors that differed significantly. The PermDisp tests always assured homogeneity of data dispersion unless mentioned otherwise. A multidimensional scaling (MDS) plot was created to visualize the results. Permanova was repeated for the meiofauna dataset after excluding Nematoda and a SIMPER analysis was performed to reveal the taxa that contributed most to the differences between treatments. The taxa with the most impact on dissimilarities were tested each by a univariate Permanova analysis as described above. For taxa with empty samples a dummy variable was added to avoid undefined values when calculating the Bray-Curtis resemblance matrix as suggested by Scientific RepoRts | 6:31447 | DOI: 10.1038/srep31447 Clarke et al. 67 . Meiofauna diversity indices (Shannon-Wiener index, Simpson index and Pielou's evenness index) were calculated and tested for significant differences between treatments with Permanova.
Quality assessment of raw bacterial community profiles and binning (2-3 replicates per sample) were done as previously reported 43 . Merged community profiles were generated in R (v.2.13.2 64 ); by using a custom script and considering Operational Taxonomic Units (OTUs) that occurred at least twice 43 . Non-metric Multidimensional Scaling (NMDS) was used to represent dissimilarity matrices based on Bray-Curtis or Jaccard coefficients into a reduced space 68 . Analysis of similarity (ANOSIM) was conducted with the PAST software (Version 1.76 69 ). To test for the effect of pCO 2 treatment and time on bacterial community composition, multivariate ANOVA and variation partitioning were conducted in R.

Results
Bivalve community response. Out of the three investigated bivalve species, the cockle C. edule was the most abundant and most sensitive to acidified seawater. As the abundance of the other two bivalve species was very low and no mortality was recorded, we focus on C. edule in the following sections. Throughout the experiment, cockles were either buried with open siphos or lying on the sediment surface. Under control conditions and at 1,500 μ atm, cockles were mostly buried (Fig. 1). In the highest CO 2 treatment (24,400 μ atm), cockles migrated towards the surface (Fig. 1). At 12,800 μ atm, some cockles were observed on the surface, however always at a lower abundance than in the highest treatment. The distribution of cockles on the surface was significantly influenced by CO 2 treatment (p(perm) = 0.0001; Supplementary Table S2). Although time and the interaction between time and treatment were significant (p(perm) = 0.0001), 63% of the variance were explained by the factor treatment. In the 24,400 μ atm treatment, 50% of cockles were located on the sediment surface on day 50 (Fig. 1).
Mortality of C. edule increased with seawater pCO 2 with 50% mortality in the highest treatment already after 68 days (Fig. 2). A significantly elevated mortality could be shown for the 24,400 μ atm group compared to the control, 1,500 μ atm, and 2,900 μ atm groups (Kruskal-Wallis multiple comparison test, p < 0.05, Supplementary Table S3, S3a). Mortality, when averaged over all size classes, tended to increase in 12,800 μ atm as well (total mortality ca. 15%). However, this increase was not significant by the end of the experiment. Smaller individuals reacted more sensitively towards high pCO 2 (Fig. 3). There were no differences in mortality between size classes in the control, 1,500 μ atm, and 2,900 μ atm treatments. At 6,600, 12,800 and 24,400 μ atm, mortality in the smaller size class (0-1 cm) was significantly higher than mortality of cockles in the larger size class (1-2.5 cm) (Kruskal-Wallis multiple comparison test, p < 0.05, Supplementary Table S4).
The comparison of intact shells against shells with signs of dissolution showed an increase in shell corrosion of C. edule with increasing pCO 2 (Figs 2 and 4). Shells from the control treatment had an intact periostracum and were not characterized by shell dissolution (Fig. 4, N = 3 of 3 observations). Shells from ≥ 1,500 μ atm were characterized by signs of external dissolution (SEM analysis, N = 3 of 3 observations). Stereo microscopic images demonstrated visible signs of shell dissolution for all treatments above 2,900 μ atm, with increasing severity in higher pCO 2 treatments (Fig. 4). For some cockles (< 5%), exclusively small animals (< 0.7 cm) holes were already visible at 2,900 μ atm. Cockles maintained under high pCO 2 (6,600, 12,800 and 24,400 μ atm) were characterized by severe shell damage. In the 6,600 μ atm treatment, 72% of dead cockles, mostly small individuals, contained holes in their shells. In the two highest treatments (12,800 μ atm and 24,400 μ atm) > 85% of deceased cockles were characterized by holes in their shells. This rate of shell dissolution was significantly higher when compared to the control and 1,500 μ atm treatment (Kruskal-Wallis, p < 0.05, Supplementary Table S3, S3b).
Whole body malondialdehyde (MDA) concentrations, an indicator for oxidative stress, were significantly lower in the 6,600 μ atm, 12,800 μ atm and 24,400 μ atm groups when compared to the control (Fig. 5a, Tukey HSD, p < 0.05). Additionally, MDA values of the 24,400 μ atm treatment were significantly lower than values measured in 1,500 μ atm and 2,900 μ atm animals.
M. arenaria and M. balthica remained burrowed during the entire experimental duration in all treatments and no mortality was observed for these two species.

Meiofauna community response.
A total of twelve meiofauna groups were found. The four most abundant taxa were Nematoda (72.5 ± 7.5%), Gastrotricha (11.0 ± 7.1%), Copepoda (5.9 ± 4.5%), and crustacean nauplii (2.6 ± 3.2%). Total meiofauna densities ranged from 218 to 988 ind. 10 cm −2 . Towards the end of the experiment, significantly higher total meiofauna densities were found in the 24,000 μ atm treatment compared to all other treatments, except for the 1,500 and 2,900 μ atm treatment (Pairwise tests (Treatment * Time),   The densities of nematodes (352.43-1733.64 ind. 10cm −2 ), did not consistently differ over time or between treatments. It only occasionally differed between the 6,600 μ am and 1,500 μ atm treatment after six weeks, and between the 24,400 μ atm and 12,800 μ atm treatment after twelve weeks (see Supplementary Table S6a). A total of 36 nematode genera were found. The overall most abundant genera (> 5%) were Ascolaimus (34.7% ± 2.2%), Metachromadora (13.2% ± 1.0%), Hypodontolaimus (9.4% ± 1.0%), Microlaimus (5.9% ± 0.5%) and Enoplolaimus (5.6% ± 0.6%). Multivariate analyses of relative and total densities of nematode genera did not reveal any significant difference between samples according to the pCO 2 level, time or a combination of both. Samples exhibited an overall high evenness (Pielou's evenness index) ranging from 0.60 to 0.89 but univariate measures based on densities and relative abundances (genus richness, Shannon-Wiener index, Simpson index, Pielou's evenness index) did not differ between samples based on the factors treatment, time or a combination of both.

Discussion
To our knowledge, this is the first experiment to examine the impact of elevated seawater pCO 2 on benthic infaunal communities in coastal sandy sediments from the Baltic Sea under near-natural conditions. Our study indicates that persistent (3 month) strong leakage from a sub-seabed CCS site and the subsequent distribution of a high pCO 2 bottom water plume could lead to massive accumulation of C. edule on the seafloor and strong reductions in survival, with a high sensitivity especially for smaller bivalve size classes. Changes in microbial and meiofaunal community composition at the highest pCO 2 level indicate a direct or cascading effect of elevated CO 2 concentration on the entire benthic assemblage in a coastal sandy sediment ecosystem. However, we also demonstrate pronounced sub-lethal effects at lower treatment levels that warrant further research attention.
High pCO 2 changes coastal bivalve communities and bivalve behaviour. The observed behavioural changes of C. edule as a result of exposure to high seawater CO 2 correspond well with responses observed for the same species during exposure to hypoxia 70,71 . With increasing pCO 2 , moribund or weakened C. edule accumulated on the sediment surface. The fraction of C. edule on the surface of the sediment increased with duration of the experiment. Behavioural responses such as gathering of bivalves on the seafloor could be used as cost-efficient tool for future monitoring of sub-seabed CCS storage sites, e.g. by towing camera systems across large sea floor areas, when appropriate software processing and detection tools are being developed. Quite similarly, Widdicombe et al. 23 observed emersion of infauna echinoderms from the sediment during a high -CO 2 mesocosm incubation when pH values dropped below 6.5.
We observed substantial mortality of smaller size classes of C. edule in all treatments > 6,000 μ atm. This corresponds with previous work on other bivalve species where smaller juveniles appeared to be more sensitive to elevated seawater pCO 2 72,73 . This effect might be related to less favourable area to volume ratios, as smaller animals have to protect a relatively larger surface area from acid-base disturbance and relatively larger shell area from (internal) dissolution. Shell production costs and inorganic carbon demand for calcification are also much higher in smaller bivalves 31,74 . Size-dependent mortality implies that leakage could lead to an alteration of the demographic structure of bivalve communities. In our study, shell corrosion was evident from 1,500 μ atm to 24,400 μ atm and signs of severe dissolution and presence of holes were found in most animals exposed to 12,800 μ atm and 24,400 μ atm. These findings indicate that even moderate degrees of acidification can already lead to non-reversible shell damage. Outer shell corrosion has been observed in a number of gastropod and bivalve molluscs 34,75 and has often been linked to absence of a thick and intact periostracum 34 . C. edule is characterized by a very thin periostracum (ca. 2 μ m) 76 and a shell that is exclusively composed of aragonite, the polymorph of calcium carbonate that is most prone to dissolution 50,77 . In contrast, mytilids are protected by a periostracum of > 20 μ m 76 and can live in seawater which is strongly undersaturated for calcium carbonate (Ω arag < 0.2) as long as their periostracum is intact. Corrosion starts to occur when the periostracum is mechanically damaged 35 , leading to complete dissolution of the shell in extreme cases, e.g. in deep-sea hydrothermal vent mussels that live in strongly acidic waters with pH < 6 33 . While bivalves are able to repair holes and fractures in their shells 78 , it is apparent from our results that C. edule in the highly acidified treatments were unable to allocate sufficient resources to shell repair, or that the rate of dissolution simply overwhelmed their repair capacity. Holes in their shells probably lead to massive and energy intensive stimulation of the immune system due to invasion of foreign microorganisms and loss of valuable proteins from the extrapallial fluid, the fluid which is in contact with the inner side of the shell 79 . In summary, progressive shell corrosion, even in the lower treatment levels (i.e. 1,500 μ atm and higher), constitutes a large problem as shell integrity is essential for fitness of bivalve species.
One of the toxic effects of elevated oxyradical formation in cells is an increase in lipid peroxidation levels 80 . A commonly employed mode of detection is the reaction of lipid peroxidation intermediates with thiobarbityric acid in the so called TBARs assay. One of the major products of lipid peroxidation, but by far not the only one, is malondialdehyde (MDA), which serves as a marker for oxidative stress 81 . Oxidative stress is frequently also related to metabolism 82,83 . Previous studies on bivalves suggest a good relationship between high MDA accumulation and elevated metabolic rate 84,85 . MDA concentration was low in bivalves exposed to pCO 2 > 6,600 μ atm suggesting that animals exposed to such high CO 2 concentrations suffer from metabolic depression. While intermediate levels of acidification (i.e. < 4,000 μ atm) have repeatedly been shown to cause elevated metabolic rates 86,87 and increased oxidative stress 32 in a number of bivalve species, studies using treatment levels of > 4,000 μ atm have generally found metabolic reduction 86,88 .
When employed as a strategy in the long run, metabolic suppression will lead to consumption of endogenous energy stores and reduced fitness. It has been demonstrated in a range of marine invertebrate species that acidification primarily impacts energy allocation processes and that ultimately, sensitivity is defined by depletion of available energy (scope for growth) by basal metabolism or reduced energy uptake 17,89 . In support of this view, all high-CO 2 exposed C. edule in our experiment (1,500-24,400 μ atm) were characterized by a reduction in body condition, as indicated by reductions in shell-free dry mass in relation to shell width. This effect becomes more pronounced at the higher treatment levels (Fig. 5b,c) and correlates with reduced MDA accumulation. These findings indicate a negative scope for growth, which is unsustainable in the long -run. In a next step, it will be important to investigate aerobic metabolism and energy uptake in this species to better understand the underlying processes leading to energy budget disturbance.
M. arenaria and M. balthica survived the complete experimental duration. Even though abundance of these two species was low, the observations confirm our original hypothesis that C. edule, the species most sensitive to hypoxic stress, also is the most vulnerable to ocean acidification 90 . M. balthica is a species which is generally quite tolerant towards stressful conditions. It occurs deep in the sediment where oxygen availability is low 91 and tolerates polluted sediments 54 . Resistance of M. arenaria could be due to a thick, protective periostracum (20 μ m) 76 . Additionally, a greater burrowing depth 92 might render this species less sensitive to acidified seawater. All bivalves used in this study have a shell consisting of aragonite, the more soluble calcium carbonate polymorph 50,76 . While the mineralogy is an important factor in defining susceptibility of bivalves 34 , there is a high variation in vulnerability between species with the same shell mineralogy 30 . Crystal size and the proportion of organic matrix within the shell play an additional factor in resistance against environmental stress such as acidified seawater 76 . Very high pCO 2 induces shifts in meiofaunal community structure. The findings of this experiment indicate that when very high seawater carbon dioxide levels occur, meiofauna community structure at higher taxon level can change as a result of differential sensitivity. The changes in densities were most severe in some of the less abundant meiofauna taxa that are not often reported. While most acidification studies focused on the dominant taxa such as Copepoda and Nematoda 23,25,37,39 , only few other studies emphasized the importance of changes in taxon composition, particularly including less abundant taxa as a proxy for the status of benthic environments [93][94][95] .
In this study, densities of three meiofauna taxa were affected by changing pCO 2 (i.e. Gastrotricha, Gastropoda, and Ostracoda). In accordance with the results of the macrofauna analysis, the calcifying Gastropoda and Ostracoda suffered from decreased densities in treatments with high pCO 2 (at > 6,600 μ atm and > 24,000 μ atm, respectively). Densities of Gastrotricha on the other hand increased significantly at the highest treatment level. It is, however, impossible to differentiate whether this is the direct result of the seawater acidification having a positive effect on their physiology (growth and reproduction) or whether this group was favoured by indirect factors, such as changes in food availability or space occupation by the disappearance of macrofauna or other taxa. Gastrotricha, as well as many other meiofauna taxa, feed on bacteria and protozoa 96 . The divergent microbial community in the highest acidification treatment could be the driver of the change in meiofaunal community structure with a strong increase in Gastrotrich densities. Despite all uncertainties, our study shows that Gastrotricha act as a physiologically tolerant, opportunistic taxon, apparently benefitting from severe seawater acidification.
In a laboratory experiment without macrofauna, Kurihara et al. 39 found no differences in the abundance and biomass of nematodes, harpacticoid copepods or harpacticoid nauplii between treatments with 380 ppm CO 2 Scientific RepoRts | 6:31447 | DOI: 10.1038/srep31447 (control) and 2,400 ppm CO 2 (acidified) over a time frame of 56 days. Several single-species studies reported decreased reproduction and slowed larval development of copepods when seawater pCO 2 is elevated [97][98][99] . Interestingly, the decline in densities of nauplius larvae over time in our mesocosm experiment occurred both in the control and acidified treatment indicating that the reproductive success of the copepods was not influenced by increased pCO 2 .
Nematodes, the most abundant taxon in our experiment, seemed to be unaffected by the different treatments in terms of community composition and abundance with no significant changes over time. Several experimental studies on nematodes indicate that a certain threshold exists above which nematodes do not suffer from pH reduction in short-to medium-term (days to weeks) experiments. Once pH drops below this threshold (~pH 6), a decline of nematode densities has been observed that can be assigned to direct physiological responses 40 . Widdicombe et al. 23 found significant changes in nematode community structure only after a 20 weeks exposure and only at pH values lower than 6.0, while macrofauna decreased in abundance at much less severe pH changes. Similarly, in a different mesocosm experiment, nematodes remained unaffected following a seven week exposure to seawater pH of 7.5 25 . In the study of Takeushi et al. 40 , nematode survival was monitored over a time frame of one week and only in the most severe treatments with a pH of 5.4 and 5.1 survival rates were significantly decreased with a species specific response related to the activity of the nematodes. Sub-lethal measures such as reproductive success and scope for growth might be more appropriate variables to assess sensitivity during medium-term exposure 39 . In our case nematode densities remained stable in all treatments and between time steps, suggesting that reproduction had most likely taken place. This was supported by the occurrence of juveniles and gravid nematodes.
High pCO 2 impact on bacterial communities. Significant differences between bacterial communities became evident when comparing the highest treatment (24,400 μ atm, severe leakage scenario) to the control (900 μ atm) and slightly increased CO 2 level (1,500 μ atm). This suggests the presence of bacteria adapted to a wide range of natural fluctuations in pCO 2 , but also the risk of substantially influencing ecosystem functionality at very high pCO 2 . Part of the potentially occurring changes in bacterial community structure might have been masked by the high spatial variability of the bacterial communities, as suggested by low R-values (PAST analyses) and a low percentage of shared OTUs among the EUs of a given treatment and time. The overall average number of OTUs in this study (186 ± 14 OTUs, 0-1 cm sediment depth, all treatments and time points) was slightly higher than the average value obtained from samples taken during a previous study on shallow subtidal sands in the North Frisian Wadden Sea 100 (145 ± 45.7 OTUs, 0-5 cm sediment depth, different seasons). Both pCO 2 level and time played a significant role in the observed community shifts, but as the two factors did not significantly interact, at this point, the prolonged pCO 2 treatment can neither be confirmed nor disregarded as a causal factor. Overall, influential factors might, besides a direct CO 2 influence, include a change in sediment nutrient composition or a change in sediment bio-irrigation rates mediated by dying cockles. While C. edule has previously been shown to significantly influence the microphytobenthic primary production due to release of NH 4

+53
, bacterial abundance was not significantly influenced by bio-diffusing activities of C. edule. The comparatively minor effects of substantial decrease in C. edule density at the highest CO 2 treatment level on the microbial community structure could be explained by a less pronounced impact of C. edule on sediment oxygen and nutrient fluxes compared to polychaetes (Nereis diversicolor) that construct elaborate burrows and strongly shape microbial communities 101 . However, the loss of an important ecosystem engineer poses unknown risks. A progressive loss of C. edule could affect settlement success of other macrobenthic species (e.g. polychaetes) that are otherwise competitively impacted by C. edule 52 . These species could more strongly impact microbial and meiofauna communities or alter ecosystem functionality during long-term CO 2 leakage. Similar to findings in our experiment, Kerfahi et al. 46 demonstrated shifts in sediment microbial communities (top 2 cm of the sediment) along CO 2 clines (600-1,600 μ atm) in the Mediterranean. The authors found abundances of most dominant genera to be unaffected by CO 2 , while 5% of genera differed in abundance along the CO 2 cline. Whether these changes can alter the biogeochemical functions of marine sediments still needs to be investigated in future experiments.
We can conclude that experimental medium-term exposure to high seawater pCO 2, which mimics strong leakage from sub-seabed CCS sites, can cause high mortality of C. edule and shifts in the composition of meiofauna and microbiota in a shallow sandy sediment environment. The response of bivalves was strongly species-specific and size-dependent. C. edule was found to be sensitive to most treatment levels with signs of shell corrosion at pCO 2 ≥ 1,500 μ atm and reduced body condition at CO 2 ≥ 1,500 μ atm, which points at a negative, unsustainable energy budget. This is supported by reduced tissue MDA accumulation at ≥ 6,600 μ atm. Thus, even lower levels of leakage resulting in seawater pCO 2 values of > 1,500 μ atm could strongly impact C. edule dominated sediments in the long -run as they might lead to reduced growth, enhanced mortality and reduced reproductive output. Longer -term ecological studies on C. edule are needed to investigate these processes in more detail. The dominant meiofauna taxa tolerated the highest acidification levels. Rare taxa on the other hand were affected by very high pCO 2 illustrating a decrease in calcifying organism abundance and an increase in other, more opportunistic taxa. While the calcifying organisms where likely directly affected by seawater acidification, it is unclear whether the response of other taxa is directly related to acidification or rather a result of cascading effects due to high C. edule mortality. Severe leakage from storage sites that result in very high seawater pCO 2 (> 20,000 μ atm) on a longer time scale could thus lead to significant changes in the overall benthic community structure, with unforeseen consequences for ecosystem health and functions. Lower levels of leakage (< 5,000 μ atm) have sub -lethal impacts on the dominant macrofauna species and may impact benthic ecosystem function in the long -term. Clearly, more research effort needs to be devoted into studying longer term consequences of exposure to lower seawater and sediment pCO 2 levels to better constrain the risks that are associated with sub -lethal species responses, particularly with respect to potential impacts on sediment meiofauna and bacterial communities.