CO2 leakage can cause loss of benthic biodiversity in submarine sands

One of the options to mitigate atmospheric CO2 increase is CO2 Capture and Storage in sub-seabed geological formations. Since predicting long-term storage security is difficult, different CO2 leakage scenarios and impacts on marine ecosystems require evaluation. Submarine CO2 vents may serve as natural analogues and allow studying the effects of CO2 leakage in a holistic approach. At the study site east of Basiluzzo Islet off Panarea Island (Italy), gas emissions (90–99% CO2) occur at moderate flows (80–120 L m−2 h−1). We investigated the effects of acidified porewater conditions (pHT range: 5.5–7.7) on the diversity of benthic bacteria and invertebrates by sampling natural sediments in three subsequent years and by performing a transplantation experiment with a duration of one year, respectively. Both multiple years and one year of exposure to acidified porewater conditions reduced the number of benthic bacterial operational taxonomic units and invertebrate species diversity by 30–80%. Reduced biodiversity at the vent sites increased the temporal variability in bacterial and nematode community biomass, abundance and composition. While the release from CO2 exposure resulted in a full recovery of nematode species diversity within one year, bacterial diversity remained affected. Overall our findings showed that seawater acidification, induced by seafloor CO2 emissions, was responsible for loss of diversity across different size-classes of benthic organisms, which reduced community stability with potential relapses on ecosystem resilience.


Introduction
The combustion of fossil fuels and industrialization are the main causes of the current exceptionally high rates of increase in atmospheric CO 2 , which in turn contributes to climate change (Blanco et al., 2014). In the short term, CO 2 Capture and Storage (CCS) in sub-seabed geological formations is considered as the most significant means of reducing net carbon emissions into the atmosphere (Turkenburg, 1997;Bachu and Adams, 2003), while the search for other profitable and renewable energy sources is ongoing. Since it is, however, difficult to predict long-term underground storage security, the inevitable risk of a CO 2 leak from storage reservoirs is one of the arguments raised to dispose CO 2 in aquifers located off-shore instead of land inwards (Turkenburg, 1997;House et al., 2006). Nevertheless, in the advent of a leak, the escaping CO 2 will rapidly dissolve into sediment porewater and into bottom-seawater, reducing the pH and carbonate saturation state and increasing acidity of either (Chen et al., 2005;Millero, 2007). Modeling and experimental studies have shown that the majority of biological impacts from CCS leakage are likely to occur in benthic or epibenthic communities of less mobile organisms, with the magnitude of the impact depending on the duration, the scale and the intensity of the leak, and on the local hydrodynamic regime (Blackford et al., 2009Zeppilli et al., 2015).
The majority of CCS leakage impact studies that measured the response of benthic, infaunal or microbial communities applied acidified seawater (pH ≥ 5.6) above sediments in laboratory-based, circulation system experiments (e.g. Dashfield et al., 2008;Ingels et al., 2018;Widdicombe et al., 2009;Schade et al., 2016;Thistle et al., 2006), injected liquefied CO 2 into corrals deposited on the deep-sea floor (porewater pH: 5.4; Carman et al., 2004;Barry et al., 2005) or released CO 2 gas via a borehole at 11 m below the seafloor (porewater pH: 7.5; Widdicombe et al., 2015). Those experiments are very relevant in determining the effects of more or less severe, acute (max. 140 days; Rastelli et al., 2015) CO 2 leakage from CCS sites or injection pipelines. However, they also recognize the inability to predict the impact of more chronic CO 2 leakage or to generalise the results to a more complex https://doi.org/10.1016/j.marenvres.2019.01.006 Received 20 August 2018; Received in revised form 7 January 2019; Accepted 11 January 2019 natural environment, where other environmental and ecological processes can affect the observed responses (Widdicombe et al., 2013). Studies at natural submarine CO 2 vents, where escaping gas is mainly composed of CO 2 and lacks toxic compounds (e.g. sulphide) or temperature increases, may serve to gain this complementary information as they allow to study ecological consequences of relatively long-term (e.g. multi-decadal; Gambi et al., 2003;Jones et al., 2014) exposure to high pCO 2 and low pH porewater conditions in seafloor surface sediments, where porewater pH exhibits a steep decline within the first few millimeters (e.g. pH T range: 7.7-5.5; Molari et al., 2018). So far, only few studies have used natural gas seeps to investigate the effect of seabed CO 2 leakage on benthic biodiversity (Dias et al., 2010;Pettit et al., 2013;Johnson et al., 2015;Raulf et al., 2015;Hassenrück et al., 2016), focusing only on specific size-class of organisms. Studies at such natural analogues may provide answers to whether species from originally undisturbed natural environments are able to persist and tolerate, and eventually adapt to the highly acidified conditions, or whether they are being replaced by colonizing tolerant species and what consequences this might have for the biodiversity and the role they fulfil within an ecosystem under high biochemical pressure (Molari et al., 2018;Zeppilli et al., 2015). Molari et al. (2018) found that long-term, high CO 2 conditions at the vents East of Basiluzzo Islet (N of Sicily, SE Tyrrhenian Sea, 14-21 m water depth) led to substantial structural and functional shifts in the bacterial and invertebrate communities, accompanied by a decrease in biomass and abundance, which led to a reduced efficiency in carbon transfer along the food web. Based on a sediment transplantation experiment, Molari et al. (2018) observed that CO 2 leakage impacts on the composition of the benthic communities were already apparent after one year. The recovery of the system after one year of release of exposure to acidified conditions was, however, far from complete. In this study, we focus on the effect of CO 2 leakage on the biodiversity of the vent system at Basiluzzo Islet and explore the relation between biodiversity and temporal stability in structural aspects of the benthos (i.e. community composition, abundance, biomass; data presented in Molari et al., 2018). While we focus on bacteria at the level of operational taxonomic units (OTUs) (i.e. internally transcribed spacer or ITS phylotypes corresponding to binned ARISA peaks), we specifically addressed the most dominant meio-and macrofauna taxa (i.e. nematodes and polychaetes) at species level and tested the following hypotheses: 1) Both one year and at least 3 years of exposure to acidified porewater conditions reduces OTU or species diversity of benthic organisms belonging to different size classes (i.e. microbia, meio-and macrofauna). 2) Benthic bacterial OTU and nematode species diversity do not recover after one year of release of exposure to acidified conditions.

Study site and sampling
The study site is located east of the Basiluzzo Islet, ca. 4 km NE of the Aeolian Island Panarea (Italy), in the SE Tyrrhenian Sea (Fig. 1A). The mainly submarine Aeolian volcanic structure that lies along the NEorientated faults (Beccaluva et al., 1982;Gabbianelli et al., 1993) is responsible for the visible release of gas from the seabed at several sites around Panarea. Gas emissions at Panarea submarine fumarolic fields have been reported to occur since two thousand years (Italiano and Nuccio, 1991, and references therein;Lucchi et al., 2013, and references therein). Specifically, at the south side of Basiluzzo Islet, shallow (8-13 m water depth) gas emissions have been observed since more than 25 years (Calanchi et al., 1995). This vent area east of Basiluzzo Islet was selected as natural analogue as it fulfills the following criteria: (i) continuous, dispersed degassing of CO 2 through sand causing low pH, (ii) similar oxygen availability and negligible co-emission of toxic substances or microbial energy sources such as sulfide and methane, (iii) no significant temperature anomalies from hydrothermalism (Molari et al., 2018). The sampled locations include two CO 2 vent sites ("CO2-R", N 38°39.749′ E 15°07.123′, 15-17 m water depth; "CO2-G", N 38°39.820′ E 15°07.137′, 21 m water depth) with rather evenly distributed gas leakage (density of 2-3 gas bubble strings per m 2 , 120 L m −2 h −1 and 97-99% CO 2 at CO2-R, 80 L m −2 h −1 and 90-97% CO 2 at CO2-G) and one reference site ("REF", N 38°39.827′ E 15°07.118′, 14-17 m water depth), with no gas emissions and similar hydrological and sedimentological characteristics as found at the vent sites. CO 2 venting through the sandy sediments resulted in a loss of solid phase carbonate and a decrease in porewater pH by 1-2 units relative to the reference site. A detailed description of the biogeochemistry of the bottom seawater, porewater and, sediments of the selected sampling sites is provided by Molari et al. (2018).
The scuba-diving team collected natural sediment samples in the month June of 2011, 2012 and 2013 for the investigation of the effects of at least 3 years of continuous CO 2 -emissions and porewater acidification on the benthos. Macrofauna samples were collected with a push core with an inner diameter of 6.4 cm (only in 2012, n = 5 per site). The upper 5 cm of the sediment was retained and stored unsieved in seawater-buffered formalin (final concentration of 4%). Meiofauna samples were gathered with pre-cut and taped push cores with an inner diameter of 4.7 cm (2011, n = 3 per site) or 5 cm (2012, 2013, n = 3 per site). The sediments were vertically sectioned in 2 cm slices, down to 8 cm depth. All sample sections were preserved in a 4% seawaterbuffered formalin solution. Sediment samples for investigating bacterial community composition and diversity, via Automated Ribosomal Intergenic Spacer Analysis (ARISA; Ramette, 2009), were obtained with push cores with an inner diameter of 5.4 cm (2011, n = 2 per site; 2012-2013, n = 3 per site), vertically sectioned in 2 cm slices down to 10 cm depth (n layers = 5), and surficial sediment (0-2 cm) collected with 50 ml plastic tubes (2011, n = 15 per site; 2012-2013, n = 20 per site). Samples were stored frozen at −20°C. Additionally, 454-Massively Parallel Tag Sequencing (454-MPTS; Sogin et al., 2006) was used for better characterization of the low abundant and rare bacterial taxa in the 0-2 and 4-6 cm sediment layers collected in 2012 (n = 3 per site), as representative for oxic/suboxic and anoxic sediment environments, respectively (Molari et al., 2018).
Additionally, the response of benthic organisms to one year of CO 2venting or reference conditions was assessed with an in situ sediment transplantation experiment. In 2012, sediment samples were transplanted in situ within and between REF and CO2-R sites: (i) reimplanted at the same site (within habitat: REF in REF and CO2-R in CO2-R) to control for transplantation effects, and (ii) transplanted to the other habitat type (across habitat: REF in  to assess the effect of the different environmental setting (Molari et al., 2018, Fig. 1B). Each transplant was collected after pushing a cylindrical TUBO device into the sediment for 15 cm. Thirty liters of sand were scooped out and transferred into a mesh bag (vinyl-coated fiberglass mosquito net: fiber diameter 280 μm and 1.8 mm × 1.6 mm mesh) covered by an additional plastic bag to protect the sediment from the surrounding seawater during transportation of the bags. At the site of interest, the bags were inserted into an empty TUBO hole (n = 5 per treatment). The TUBO and the plastic bag were removed, and the mesh bags were closed (Molari et al., 2018). Samples for microbial and meiofaunal communities analyses were collected one year later in the same manner as described above. Background sediment samples collected in 2013 from either site were used as control.

Sample processing
The bacterial community structure was determined with the highthroughput fingerprinting technique ARISA and by applying 454-MPTS on microbial DNA extracts from 1 g of sandy sediment per sediment layer (down to 10 cm for natural and transplanted sediments). A total of 450 OTUs (i.e. ITS phylotypes corresponding to binned ARISA peaks) were detected from 280 sediments samples. 454-MPTS sequencing data of the hypervariable V6 region of the bacterial 16 S rRNA gene were obtained according to the standardized sequencing pipeline applied previously and analyzed with mothur standard operating procedure (Version 1.29.2; Schloss et al., 2009). Sequences were clustered into operational taxonomic units at a 3% nucleotide difference (hereafter referred to as OTU 0.03 ). The whole dataset comprised a total of 169,490 high-quality bacterial sequences, which were clustered into 9674 OTUs 0.03 . From the whole 454-MPTS subset only OTUs0.03 > 0.1% of total bacterial sequences per sample were used for analysis of dominant classes.
The meiofauna was extracted from the sediment through decantation (5 sessions) and washing on stacked 1-mm and 32-μm mesh sieves.
The metazoan organisms were quantified and classified at higher taxon level under a stereoscopic microscope. Fifty to sixty nematodes per sediment layer (down to 8 cm for the natural sediments, down to 4 cm for the transplant experiment) were randomly handpicked with a fine needle, transferred to glycerine (De Grisse I, II and III; Seinhorst, 1959), mounted on glass slides and identified to species (natural sediment) and genus (transplant experiment) level based on original species descriptions which are available on the Nemys website (Guilini et al., 2017a). Nematode biomass (μg dry weight 10 cm −2 ) was determined for all identified nematodes based on Andrassy (1956) and assuming a dry-towet-weight ratio of 0.25 (Heip et al., 1985).
The macrofauna was extracted from the sediment through sieving and washing on a 1-mm mesh sieve, followed by handpicking all organisms with a forceps or needle under a binocular microscope. All macrofauna organisms were counted and identified to higher taxon level (down to Class), whilst Polychaeta were further identified to species level based on the World Polychaete Database (Read and Fauchald, 2016) and Gil (2011).

Data analysis
Alpha diversity was assessed as species richness, exponential of Shannon index and inverse of Simpson index, corresponding to Hill's numbers of order q = 0 (H 0 ), q = 1 (H 1 ) and q = 2 (H 2 ), respectively (Hill, 1973;Chao et al., 2014). The diversity measures spread along Hill's continuum provide us with a more complete understanding of shifts in rare and abundant species and a simplified interpretation of results because units are always in effective number of species (Jost, 2006). These effective species numbers behave as one would intuitively expect when diversity is doubled or halved, while other standard indices of diversity do not (Jost, 2006). For the natural sediments, the overall alpha diversity was visualized in interpolation and extrapolation species diversity curves based on replicates' average of the sum of sequences/individuals per layers. Additionally, for Bacteria, richness estimates (Chao1) were calculated with 100 random re-sampling runs to the smallest number of sequences per sample in the dataset (n = 4346), to account for differences in sequencing depth between samples. Diversity indices were also calculated for bacterial dominant classes, representing together more than 80% of total OTUs 0.03 at each sites (Molari et al., 2018). For testing the effect of space ("Site" and sediment "Layer") and time ("Year") on variations observed in bacterial, nematode and polychaete alpha diversity from natural sediments, multifactorial ANOVA analyses were performed based on a three factor design with Layer nested in Site. The nested design was chosen to take into account the site-specific environmental gradients due to presence/ absences of acidified porewater effluxes at CO2-venting sites and reference, respectively (for more details see Molari et al., 2018). In order to account the effect due to transition between aerobic and anaerobic environments, the diversity indices of single layer were grouped in oxic/suboxic (i.e. 0-2 cm) and anoxic (below 2 cm) layers. Moreover, for bacteria we had more observations for 0-2 cm layer (n = 15-20) than for other layers (n = 3), thus grouping together layers below 2 cm evened sampling effort. In order to account for the dependency of observation from the same sediment core, a mixed effects model with "Sediment Core" as random factor was carried out for Nematode data. Three-way ANOVA was applied to Bacteria and one-way ANOVA to Polychaete data due to high contribution of independent observations for layer 0-2 cm and one observation per core, respectively. To test the effect of short-term exposure to acidified porewater conditions on alpha diversity of bacteria and nematodes, a two-way split-spot ANOVA was performed with the factors "Treatment" (within habitat transplants and across habitats transplants) and sediment "Layer" (nested in Fig. 1. A) Location of the study area near Basiluzzo Islet (Panarea, Italy), adapted from Guilini et al. (2017b), and B) a scheme illustrating the experimental design of the transplantation experiment. The pictures show (1) how a mesh bag (1.8 mm × 1.6 mm mesh) containing 30 L of sediment was inserted in a TUBO device that was first pushed into the sediments (15 cm depth) and emptied, (2-3) the removal of TUBO and the plastic bag, and (4) the closure of the mesh bags.
"Treatment"), and "Sediment Core" as random factor. Transplants pairwise comparison was carried out using non-parametric Kruskal-Wallis tests with Benjamini-Hochberg (BH) adjusted P-values.
Temporal stability (S), which is an index of the stability of a community or population over time (Tilman, 1999;Lehman and Tilman, 2000;Tilman et al., 2006), was considered for bacterial and nematode density, biomass and community composition (data which are presented in Molari et al., 2018). S was estimated as the ratio between the temporal mean (μ) of community-level density, biomass and Bray-Curtis similarity values for species composition and their standard deviation (σ), per sampling site. Larger S values indicate higher temporal stability. The relationship between temporal stability and biodiversity could only be visually stated since three sample sites did not allow performing simple linear regressions.
Beta diversity (i.e. OTUs/species turnover) was calculated as the Jaccard dissimilarity distance on presence/absence data for each of the bacteria, nematode and polychaete datasets. Turnover of OTUs or species were quantified between the layers and between the years at each site for Bacteria (ARISA) and nematodes, and between sites for bacteria (ARISA and 454-MPTS), nematodes and polychaetes. The Jaccard dissimilarity distance was used to calculate beta similarity (i.e. the number of shared OTUs/species), and similarity percentage beakdown procedure (SIMPER; Clarke, 1993) was carried out to assess the average percentage of shared OTUs/species within and between the groups (sites and layers per each year, and years per each site) for Bacteria (ARISA and 454-MPTS) and Nematoda datasets. One-or twoway ANOVA, after verifying that the assumptions of normality and homoscedasticity were met, was applied to test significant differences between average of shared OTUs/species. Following significant ANOVA between the investigated sites and/or years, Tukey post-hoc comparison tests were applied. In case the assumptions for ANOVA were not met, Kruskal-Wallis tests and post-hoc Dunn-tests were performed instead, with BH-adjusted P-values.
Multivariate Redundancy Analyses (RDA) were performed to investigate which environmental variable or set of variables (Table A0.1) could best explain the patterns in bacterial and nematode diversity (Hill's numbers) observed with depth in the sediment (0-10 cm and 0-8 cm, respectively) and in the oxic/suboxic layer (0-2 cm) for the years 2012 and 2013. Prior to the analyses, the explanatory environmental variables were standardized (i.e. Z-scored) and assessed for collinearity based on the variance inflation factor (VIF). The variables pH, chlorophyll a (Chl-a), total organic carbon (TOC), median grain size (MGS), silicate, and dissolved inorganic carbon (DIC) were retained based on a VIF < 3, indicating no or little collinearity among these variables. Additionally, variation partitioning (VP) analysis was used with multiple partial RDAs to assess the explanatory power of each significant explanatory variable and proportion of variance explained by variables multicollinearity. All statistical analyses were performed in R (v. 3.3.1) (R Development Core Team, 2014; https://www.R-project. org) using packages iNEXT (Hsieh et al., 2016), vegan (Oksanen et al., 2015), usdm (Naimi et al., 2014), lme4 (Bates et al., 2015), and ggplot2 (Wickham, 2009).

Results
The rarefaction curves indicated that bacterial (OTUs), nematode (species) and polychaete (species) richness in the natural sediments was well captured for the abundant taxa in all three study sites, while diversity of the less abundant taxa and rare biosphere remained largely underestimated (Fig. 2, Fig. A1 and A2). REF contributed the largest fraction of absolute singletons (OTUs 0.03 with single-sequence; SSO abs ) and unique OTUs 0.03 , followed by CO2-G and then by CO2-R (Table  A.2). Hill numbers ( Fig. 3 and A.3) showed a general reduction in diversity at the CO 2 -venting sites compared to the reference site (ANOVA, p < 0.001, redundancy ≥ 63.1%; Table A.3), with no more than 1.5% and 21% of the variance explained by "Year" (2011-2013) and sediment "Layer" (0-2 cm and 2-10 cm for bacteria, 0-2 cm and 2-8 cm for nematodes), respectively, for both bacteria and nematodes. Bacterial diversity (OTUs) was generally the lowest at CO2-R with almost all  (Fig. A.4). Diversity of polychaete species was equally low at both vent sites for all indices (ANOVA pairwise test, p ≥ 0.72) while nematode species diversity was equally low at both vent sites for H 0 (ANOVA pairwise test, p = 0.35) and lower at CO2-R compared to CO2-G for H 1 and H 2 (ANOVA pairwise test, p ≤ 0.030). Redundancy analyses and variation partitioning revealed that alpha diversity was mainly determined by pH (29-56%; Table 1). Other biogeochemical variables explained no more than 11% individually (Table 1, Fig. A.5).
Interestingly, with the decreasing of species diversity (Hill's numbers) at vent sites, there was a clear trend of decreasing of temporal stability (S) in bacterial and nematode density and community composition, as well as in nematode biomass (Fig. 4).

CO 2 leakage affects bacterial diversity
The results from few previous studies investigating effects of acidified porewater on bacterial alpha diversity in soft sediments at CO 2 venting sites showed contradictory responses (Yanagawa et al., 2013;Kerfahi et al., 2014;Raulf et al., 2015;Hassenrück et al., 2016), likely as a result of the inherent high heterogeneity of environmental gradients at natural vents (German and Seyfried, 2014). Where no pure CO 2 release occurs, fluid and gas emissions of reduced elements provide potential energy sources for specialized microorganisms which leads to a shift in composition of the benthic microbial communities, favoring specific functional groups (i.e. chemolithotrophs; Price et al., 2015) and consortia (e.g. anaerobic methanotrophs; Boetius et al., 2000) that may mask the microbial responses to CO 2 . In the CO 2 -venting sites here investigated, Molari et al. (2018) did not find presence of hydrothermal endemism, and we observed a lower beta diversity compared to those reported for other ocean acidification natural analogues (Raulf et al., 2015;Hassenrück et al., 2016) and typically described for seeps and hydrothermal vents (Ristova et al., 2015;Raulf et al., 2015). Besides, the temporal variation in community composition was comparable with that observed in other coastal sandy sediments (Böer et al., 2009;Gobet et al., 2012). Direct comparison of beta diversity values between studies could be biased by application of different sequencing techniques (e.g. different sequencing depth, different primers) and methods for beta diversity estimation. However, here, our results were compared with previous studies applying the same methods to investigate microbial diversity (i.e. ARISA and/or 454-MPTS) and to assess beta diversity (i.e. Jaccard dissimilarity distance). Both long-(at least three years) and short-term (one year) exposure to CO 2 emissions resulted in a decrease Table 1 Results of variation partitioning and partial redundancy analyses illustrating which predictor variables explain the variance in Bacteria (ARISA) and Nematoda diversity (Hills numbers) for 2012 and 2013, considering the entire sediment profile (all layers) and restricted to the upper 2 cm of the sediment. Proportion of variance explained (%) and p values (ANOVA) are reported for significant variables, as identified by RDA (single source). Proportion of variance explained by variables multicollinearity is also reported, as identified by variation portioning analysis (multiple sources). RDA and partial RDA details are reported in Table A0

Table 2
Beta similarity between sites for three years and between years within study sites, presented as the average percentage ( ± standard deviation) of shared OTUs/species derived from Jaccard dissimilarity index based on presence/absence data for bacterial OTUs (ARISA, n = 3; 454-MPTS, n = 9), nematode (n = 3) and polychaete (n = 5) species. 454-MPTS was only applied on samples collected in 2012, therefore replicates' pairwise comparisons were used to test differences in beta similarity between site, and beta similarity between years is not available (na). Remarkably, the decrease in richness and evenness was observed within almost all dominant bacterial classes, even those favored at CO 2 vent sites (Molari et al., 2018). Together, these findings suggest a stable and univocal response of dominant bacterial functional groups to CO 2 emissions here investigated. Molari et al. (2018) showed that alongside the lowering of pH, the CO 2 emissions have an indirect effect on productivity (i.e. stimulating microphytobenthos) and sand mineralogy (i.e. dissolution of carbonate fraction). The trophic status and sediment composition are known to be factors shaping microbial communities (e.g. Schöttner et al., 2011;Bienhold et al., 2012). However, our results showed that bacterial diversity was highly related with pH rather than with others environmental factors, suggesting that at venting sites CO 2 emissions led to the selection for those bacteria physiologically tolerant to acid stress. Neutralophilic bacteria, growing at pH range of 5-9, can use different mechanisms for pH homeostasis (Booth, 1985;Slonczewski et al., 2009). Specifically for respiratory bacteria, which dominated microbial communities at our study sites, acid stress is met by direct active efflux of protons by redox potential-driven pumps, light-driven pumps or bond energy-driven pumps (Krulwich et al., 2011). Recently, a 9-days mesocosm experiment showed that a pH reduction of 0.2 units stimulated expression of different pH homeostasis genes in marine heterotrophic bacterioplankton (Bunse et al., 2016). These stimulated mechanisms to export protons across the cell membrane are energy demanding, which implies that the physiological acclimation of bacteria to seawater acidification may increase energy demand for cell maintenance, potentially resulting in a reduction of growth efficiency. Thus, under exposure to constant and moderate flows of acidified seawater, and depending on the extent of acidification, we can expect that an increase of metabolic cost could select for those bacteria with more versatile metabolism (e.g. resource exploitation, regulation of growth efficiency), with consequences for diversity and carbon cycling. This is exactly what we have observed at acidified Basiluzzo sites, where the unchanged bacterial abundance was accompanied with a loss of diversity and an increase of organic carbon remineralization (Molari et al., 2018). Future studies need to be carried out to look closely at the effects of seawater acidification on microbial physiology and energy allocation for elucidating the consequence on carbon and energy fluxes in the microbial food web. Nevertheless, for the fist time transplantation experiments and three years of field observations provide robust evidence that seawater acidification, induced by seafloor CO 2 emissions, can be responsible for the loss of bacterial diversity.

CO 2 leakage affects invertebrate diversity
The results of this study demonstrate that at least 2 years of exposure to diffuse CO 2 leakage reduced polychaete species diversity. These findings complement the results from laboratory and field experiments investigating more or less severe, acute CCS leakage impacts on benthic macrofaunal diversity (Christen et al., 2012;Hale et al., 2011;Ingels et al., 2012;Widdicombe et al., 2009Widdicombe et al., , 2015. The threshold at which impacts were observed on infaunal macro-invertebrate diversity in laboratory experiments that mimic a CO 2 rich plume, was identified at seawater pH levels below 7, but only after five weeks of continuous exposure . An experimental subseabed release of CO 2 caused a reduction in macrobenthic diversity within a few days, although porewater pH did not drop below 7.5 . In both scenarios, impact thresholds are likely determined by natural processes, such as carbonate buffering and permeability, which influence the carbon chemistry of the sediment . Moreover, the vulnerabilities to high pCO 2 / low pH conditions differ between and within phyla (Christen et al., 2012 and references therein). The reduction in biodiversity in these relatively short-term CO 2 leakage simulations result from the loss of species that hamper the ability to regulate the acid-base balance of internal fluids to maintain a number of key pH sensitive physiological processes (Kroeker et al., 2013) or in case of calcifying organisms, fail to maintain important physiological processes (e.g. growth, reproduction, immune function) as a result of an energetically more demanding calcification process under reduced calcite or aragonite saturation states (Pörtner, 2008;Wood et al., 2008). In general, annelids were classified among the taxa most resilient to highly elevated pCO 2 /low pH (Christen et al., 2012;Hale et al., 2011). The relatively long-term Table 3 Beta similarity between sediment depth layers at each study site, presented as the average percentage ( ± standard deviation) of shared OTUs/species derived from Jaccard dissimilarity index based on presence/absence data of three years (2011-2012) for bacterial OTUs (ARISA) and nematode species (n = 3); a is the percentage of bacterial OTUs or nematode species shared with 0-2 cm, b is the percentage of bacterial OTUs or nematode species shared with previous layer.

Shared bacterial OTUs (%)
Shared nematode species (%) perspective and natural environment in our study allows to consider whether the vacant niches of species that disappear on the short term are being occupied by colonizing tolerant species and what consequences this might have for the biodiversity. The low number of shared polychaete species between the reference and vent sites and the severely reduced species diversity at the vent site leaves no doubt that polychaete communities cannot cope with diffuse CO 2 leakage over a period of at least several years. This reduction in species diversity coincides with a reduction in functional diversity (i.e. feeding modes), density and biomass (Molari et al., 2018) and therefore warrants for a reduction in the optimal functioning of the benthic ecosystem. Nematodes, generally the numerically dominant and species-rich members of the meiobenthos, have most often been the focus of meiobenthic response measures to acute seawater acidification (Ingels et al., 2012(Ingels et al., , 2018Kurihara et al., 2007). This is the first study, however, that indicates that one to at least three years of exposure to acidified porewater conditions (pH T minimum: 5.5) in a natural environment severely reduces nematode species diversity. Earlier acute acidification exposure experiments testing the effect of seawater pH ≥ 6 over a maximum of 12 weeks on the meiobenthos found no negative effects on nematode abundance, community composition and diversity (Takeuchi et al., 1997;Dashfield et al., 2008;Ingels et al., 2018;Schade et al., 2016). Merely once an increase in nematode abundances occurred, which was attributed to the reduction of ecological constraints (predation, competition) resulting from a decrease in macrofaunal abundance (Hale et al., 2011). Only when seawater acidity was decreased below what appears to be a threshold of 6 pH units for a maximum of 20 weeks, nematode community structure changed and species diversity decreased , abundance decreased (Takeuchi et al., 1997;Barry et al., 2004;Ishida et al., 2005), or high mortality was suggested based on changed morphometrics (Fleeger et al., 2010). The nematodes' impermeable proteinaceaous cuticle is suggested to buffer them relatively well to short-term natural fluctuations of porewater pH in the upper few centimeters of sediments (typical pH range of 6.5-8.2; Brusca and Brusca, 1990;Widdicombe et al., 2011). Persistently high CO 2 conditions probably resulted in extracellular acidosis, a condition which might exert lethal physiological stresses on nematodes Widdicombe et al., 2009). The highly distinct and species-poor nematode communities found at the vents at Basiluzzo Islet, however, also demonstrate that there are nematode species which potentially adapted and/or exhibit tolerance to extreme and chronic high pCO 2 /low pH conditions, with few opportunistic species demonstrating strong competitive advantages in exploiting these harsh environments (e.g. Microlaimus compridus, Microlaimus honestus, Oncholaimus campylocercoides, and Daptonema Microspiculum; Molari et al., 2018). At the same time, these species-poor vent communities seem prone to a relatively high temporal variability in different structural community characteristics (i.e. community composition, abundance, biomass), which suggests reduced resilience to additional environmental perturbations. These results emphasize the importance of investigating synergistic impacts of environmental factors on multi-species assemblages to enable more informative predictions (Jones et al., 2014;Mevenkamp et al., 2018).
Similar susceptible changes to CO 2 exposure were also observed for other meiobenthic groups (Thistle et al., 2005). For example, for many species of harpaticoid copepods, alterations in CO 2 concentrations resulted in general mortality rates of around 80% of the total community (Thistle et al., 2006), but it also indicated that some species are less susceptible than others to changes in the environment (Thistle et al., 2006(Thistle et al., , 2007. In general, the meiobenthos is less affected by environmental changes compared to the macrobenthos (Ingels et al., 2012;Zeppilli et al., 2015). Some species can even be favored by extreme environmental conditions. Within the meiobenthos, copepods are considered particularly susceptible to stress in comparison with nematodes, suggesting the high versatility and different responses of this size class in predicting major global changes by identifying sensitive and tolerant species (Zeppilli et al., 2015).

Recovery of benthic diversity after CO 2 leakage has ceased
Transplanting sediments from the CO 2 vent site to the reference site allowed evaluating the recovery potential of high pCO 2 /low pH-impacted nematode communities. Our results show that nematode species diversity could re-establish within one year, while Molari et al. (2018) additionally demonstrates a recovery of the nematode community composition, though with reduced overall population densities. These findings are in agreement with the results presented by Widdicombe et al. (2015), where the macrofauna community composition had recovered from the impact of a subseabed CO 2 injection 18 days after the leakage was stopped. Neither the bacterial community composition (Molari et al., 2018), nor the diversity of bacteria recovered within one year, suggesting that colonization time exceeded one year. However, as mentioned above, the CO 2 leakage promoted dissolution of carbonate sands, an effect that was still recorded in the sediments transplanted to the reference site (Molari et al., 2018). Without carbonate sands the sediment lost its buffering capacity, resulting in porewater with a pH value 0.4-0.7 units lower than what is typically found in undisturbed sediments. Thus, we cannot exclude that bacterial diversity may be affected by variations of pH values even smaller than those observed at CO 2 -venting sites.
Although both experiments have shown that the chemical effects of CO 2 leakage on benthic invertebrate diversity are not long lasting, we need to consider that both study areas were spatially relatively small (35-200 m 2 ). The time needed for recovery from CO 2 leakage events will eventually depend on the area of the impacted site, the flux of CO 2 and the potential physical disruption of the sediments, the species dispersal capabilities, their rate of recruitment, and the distance between the disturbed site and the unaffected source populations.

Conclusions
In summary, this study proved that porewater acidification, induced by seafloor CO 2 emissions, was responsible for reducing 30-80% species diversity of dominant benthic organisms across size-classes (Bacteria, Nematoda and Polychaeta). With the transplantation experiment we provided evidence that the impact on microbe and nematode diversity was already apparent after one year of exposure to CO 2 emission. Additionally, the transplantation experiment showed that recovery of an impacted area relatively small in size was fairly quick (i.e. within one year), at least for nematodes. The lack of recovery of bacterial communities pointed out that effects of CO 2 leakage on sediment properties may last for a long time after the disturbance is ceased, with consequences on benthic biodiversity that may vary according to the size and ecology of organisms (e.g. dispersion, sessile/ motile) and to the scale of the impacted area.
High microbial and nematode species richness at the reference site revealed more stable communities over time, while loss of species at the vent sites resulted in higher temporal variability in community characteristics, suggesting reduced resilience to additional environmental perturbations at impacted locations. These results contribute to the growing body of evidence that highlights the importance of biodiversity for ecosystem stability.

Authors' contributions
K.G. processed the macro-and meiofauna samples, identified the nematodes to species level, analysed the faunal data, and wrote the manuscript; M.M. processed the bacterial samples, analysed the bacterial data, and wrote the manuscript; L.L. processed the meiofauna samples and wrote the manuscript; A.V., A.R. and K.G. designed the study. All authors revised the article critically and gave final approval of the version to be published.

Acknowledgments
We thank Antje Boetius for having inspired and supported this work. We greatly appreciate the efforts of the HYDRA scientific diving team from Elba (Italy) to organize the logistics of the sampling campaign and to collect the samples. We also acknowledge Niels Viaene and Bart Beuselinck for sample decantation and nematode morphometric measurements; Annick Van Kenhove and Guy De Smet for preparing the glass slides; Francisco Sedano Vera and Nair De Jesús for their help with processing the meiofauna samples; and Daniel Martin and Marta Segura for identifying the macrofauna to species level.        Details of the statistical analyses on the Hill's numbers (H 0 , H 1 , H 2 ) from the transplant experiment. Bacteria (ARISA) and Nematode species diversity was analysed with three-factor split-plot ANOVA with "sediment core" as random factor (model: Hill number ∼ Treatment + Layer-nested-Treatment, random = ∼1|Core, data).