Interactive comment on “ Response of bacterioplankton community structure to an artificial gradient of p CO 2 in the Arctic Ocean ” by R

Response to general comments of referee 2: We appreciated the detailed and constructive comments from referee 2. As a part of team work in EPOCA, our study used DNA fingerprinting technique (T-RFLP) to analyze and compare large amount of samples (e.g. 159 samples of 19 sampling points from all 9 mesocosms). In two companion papers, Sperling et al. and Roy et al. presented size fraction (particle-attached and free-living) bacterial community dynamics. These three papers provided the most detailed and systematic description about bacterial community changes under artificial pCO2 gradient in marine ecosystem so far. For the concern of water used in our


Introduction
Microorganisms are a key component in marine planktonic food webs (Azam et al., 1983).Eukaryotic phytoplankton and cyanobacteria contribute significantly to photosynthesis and primary production in the ocean.Heterotrophic bacteria and archaea process approximate half of the primary production and recycle organic and inorganic nutrients.These microbial functions are crucial components in several major concepts of marine food web structure and organic matter cycling, e.g., the microbial loop and microbial carbon pump (Azam et al., 1983;Jiao et al., 2010).
Generally, microbial diversity and community structure are controlled by abiotic and biotic factors.For example, bacterial communities are affected by the spatial and temporal dynamics of environmental parameters.Nutrient availability and water characteristics (e.g., temperature, salinity and pH) contribute to microbial biogeography in the ocean (Martiny et al., 2006).Considering the changing oceanic conditions induced by anthropogenic impacts (including climate change), it is necessary to investigate whether and how the microbial communities would respond to the emerging changes, and to evaluate the possible ecological consequences of the microbial responses.Such hitherto unknown information is crucial to our understanding and predictions of the possible effects of global climate changes to the biosphere (IPCC, 2007).
A consequence of increasing CO 2 in the atmosphere is a decrease of pH in the water (ocean acidification), which has been confirmed by long-term field observation and is related to anthropogenic CO 2 emission (Dore et al., 2009).It is estimated that the pH of global surface seawater will decrease by 0.2 to 0.4 units by the end of this century based on the general emissions scenarios and circulation models (Caldeira and Wickett, 2003).This predicted rate of ocean acidification is greater than those which have occurred over the past 300 million years (Hönisch et al., 2012).Such changes would fundamentally alter ocean chemistry from the surface water to the deep sea.Our understanding of the biological and ecological effects of changing seawater carbonate chemistry is still in its infancy.While it is well recognised that ocean acidification will have serious negative impacts for most marine calcifiers, the increasing pCO 2 effects on non-calcifiers are not clear.For example, positive effects of elevated pCO 2 concentrations were observed for marine autotrophs in the phytoplankton (Riebesell, 2004), suggesting a "CO 2 fertilisation" phenomenon (Hutchins et al., 2009).While similar fertilisation effects occurred for the cyanobacteria Synechococcus, the other major cyanobacteria Prochlorococcus showed no response to pCO 2 increase (Fu et al., 2007).It appears that autotrophic nitrogen fixers and anaerobic denitrifiers directly or indirectly benefit from ocean acidification (Hutchins et al., 2007(Hutchins et al., , 2009)).However, the abundance and/or activity of nitrifiers might be repressed, since nitrification rates declined with rising pCO 2 level in the ocean (Beman et al., 2011).
Thus far, most of our knowledge of ocean acidification on microbes is derived from laboratory experiments with a small number of cultured microbial taxa (Liu et al., 2010), and the knowledge about the response of natural heterotrophic bacterioplankton communities to ocean acidification is scarce (Weinbauer et al., 2011).A general observation was that ocean acidification had no significant influence on bacterial abundance, but affected various activity parameters like bacterial production, extracellular enzyme activities, etc. (Allgaier et al., 2008;Arnosti et al., 2011;Grossart et al., 2006;Newbold et al., 2012).However, the responses of bacterial community composition (BCC) to pCO 2 levels are complex.Newbold and coworkers did not observe a clear change of bacterial community composition under elevated pCO 2 (Newbold et al., 2012), while data from two mesocosms of the PeECE II experiment showed different BCC in low and high pCO 2 treatments (Arnosti et al., 2011).In addition, there is evidence that community structure of freeliving bacteria, but not particle-attached ones, changed with the pCO 2 level (Allgaier et al., 2008).Therefore, in the Svalbard 2010 mesocosm experiment of the European Project on Ocean Acidification (EPOCA) (Gattuso and Hansson, 2009), the largest ocean acidification experiment performed so far, BCC was investigated for a better understanding of bacterioplankton response to different pCO 2 treatments.
In the present study, we used terminal restriction fragment length polymorphism (T-RFLP) analysis and clone library analysis to investigate the dynamics of bacterial diversity and community structure during the 30 day mesocosm experiment in Arctic waters.The automatic capillary running, rapid data production and high reproducibility of T-RFLP analysis allowed analysing and comparing large amount of samples (e.g., 159 samples of 19 sampling points from 9 mesocosms in our study).Two companion publications (Roy et al., 2013;Sperling et al., 2013) present the size fraction (particle-attached and free-living) bacterial community dynamics in same mesocosm experiment by using automated ribosomal intergenic spacer analysis (ARISA) and 454-based 16S rRNA amplicon sequencing.These three papers provide the most detailed and systematic description of bacterial community changes in an artificial pCO 2 gradient in marine ecosystem so far.

Experimental set up and sampling
As a part of EPOCA, the ocean acidification mesocosm experiment was performed in King's Bay (Kongsfjorden), Spitsbergen (78 • 56.2 N, 11 • 53.6 E).Detailed information of the experimental setup is provided elsewhere (Riebesell et al., 2012).Briefly, nine KOSMOSs (Kiel Off-Shore Mesocosms for Future Ocean Simulations) were deployed on 31 May, 2010 (experimental date: t-7) and closed on experimental date t-5.Each mesocosm was equipped with a 17 m long polyurethane bag (2 m above sea level and 15 m below), containing about 45 m 3 of seawater.CO 2 -saturated seawater was added stepwise during t-1 to t-4 and nine pCO 2 concentrations were used in the mesocosms: low pCO 2 levels including 175 (M3), 180 (M7) and 250 (M2) µatm; intermediate pCO 2 levels including 340 (M4), 425 (M8) and 600 (M1) µatm; high pCO 2 levels including 675 (M6), 860 (M5) and 1085 (M9) µatm.The setup of a gradient of pCO 2 levels allows the use of regression statistics and the detection of possible threshold levels for CO 2 sensitive processes.The control (with ambient pCO 2 level) was duplicated (M3 and M7) to reduce the risk of losing the control for the experiment (Riebesell et al., 2012).Nutrients (5 µmol L −1 Nitrate, 0.3 µmol L −1 phosphate and 2.5 µmol L −1 silicate) were added into all mesocosms prior to the sampling on t13 in order to induce the development of a phytoplankton bloom (Schulz et al., 2013).
Environmental parameters (temperature, conductivity, pH, oxygen, fluorescence, turbidity and light intensity) in the mesocosms were measured daily with a CTD60M (Sun and Sea Technologies).Depth-integrated seawater samples (5 L; from surface to 12 m depth) were collected from the mesocosms using HYDRO-BIOS water samplers (Kiel, Germany).The measurements of chemical and biological core parameters are described in Schulz et al. (2013).Microbial cells in freshly collected seawater samples (2 L each) for bacterial community structure analysis were harvested using membrane filtration (0.22 µm-pore-size Isopore membrane, Millipore).In the present study, sampling was performed on nineteen days during the experiment, with a relatively dense sampling in first half of 30 day period, followed by regular sampling every other day.

DNA extraction and PCR amplification of bacterial 16S rRNA genes
DNA was extracted following the previously published protocol (Zhang et al., 2007) with some modifications.The filter membranes were deep frozen and thawed three times in liquid N 2 and at 65 • C, respectively.8 µL of proteinase K (10 mg mL −1 in TE buffer) was then added, followed by incubation for 30 min at 37 • C.After that, 80 µL of 20 % SDS was added, followed by incubation for 2 h at 65 • C.After vortexing, an equal volume of phenol-chloroformisoamylalcohol (25 : 24 : 1 by volume) was added and centrifuged at 10 000 g for 5 min.The upper aqueous layer was transferred into a fresh tube and treated with an equal volume of chloroform-isoamylalcohol (24 : 1 by volume).The mixture was centrifuged at 10 000 g for 10 min, and the upper aqueous phase was transferred into a new tube.Subsequently, 0.6 times volume of isopropanol was added to the aqueous solution and the mixture was incubated at −20 • C for 20 min, and then centrifuged at 12 000 g for 15 min.The DNA was washed with 70 % ethanol and dissolved in 100 µL of sterilized ultrapure water (produced by the Milli-Q academic A10 system).The quality and quantity of the environmental genomic DNA were evaluated by NanoDrop (Thermo Scientific).

Terminal restriction fragment length polymorphism (T-RFLP) analysis
Purified PCR products were double-digested with restriction enzymes Msp I and Rsa I (New England Biolabs, Ispswich, MA, USA) according to the manufacturer's protocol.Selected samples were digested only with Rsa I so as to verify that the T-RFLP patterns obtained were the result of dou-ble digestion.Aliquots of digested PCR products were mixed with 0.125 µL of the internal size standard (Bioventures Map-Marker 1000, Cambio, Cambridge, UK) and analysed using the MegaBACE platform (Amersham).All T-RF analyses were performed using the Genetic Profiler in the MegaBACE software package (Amersham) as described previously (Lau et al., 2005).T-RFs < 50 bp and > 900 bp were excluded from the analysis to avoid detection of primers and uncertainties of size determination, respectively.To normalise the variation of the amount of DNA loaded on the capillary, the percentages of each T-RF peak area compared to the total peak area of each sample were calculated and T-RFs with a peak area less than 1 % of the total peak area were removed for statistical analysis.An improved binning strategy was applied for the remaining T-RFs data matrix following the protocol described previously (Hewson and Fuhrman, 2006).

Clone library construction and phylogenetic analysis
The bacterial 16S rRNA genes in the t30 samples were PCR amplified using the primer set 27F and 1492R (5 -GGC TAC CTT GCC ACG ACT TC-3 ) with the same PCR programme as described above (Lane, 1991).The PCR amplicons (∼ 1500 bp in size) were cloned into the vector with a TOPO TA Cloning Kit (Invitrogen, Carlsbad, California, USA) according to the manufacturer's instructions.The 16S rRNA genes were sequenced from both ends using the primers M13F (5 -GTTGTAAAACGACGGCCAGTG-3 ) and M13R (5 -CACACAGGAAACAGCTATG-3 ) on an ABI DNA autosequencer.Phylogenetic affiliations of the sequences were determined using the BLASTN programme on the NCBI (http://www.ncbi.nlm.nih.gov) and RDP (http://rdp.cme.msu.edu/)website.The 16S rRNA gene sequences were clustered as operational taxonomic units (OTUs) at an overlap identity cut-off of 97 % sequence similarity.Pairwise comparisons of each clone library were performed using LIBSHUFF (version 0.96; http://www.mothur.org).The DNA sequences obtained in this study are available from GenBank under the accession numbers JN975970-JN976712.
The possible phylogenetic assignment of T-RFs from T-RFLP analysis of environmental samples was performed with in silico digestion of the DNA sequence of clone library.A variation of 1-3 bp, depending on the size of the T-RF, was applied for the comparison between T-RF positions of community-based and sequence-based in silico analysis (Fabrega et al., 2011).

Statistic analysis
The resultant T-RFLP data matrix (percentage of peak area) were analysed using PRIMER 5 software (Clarke and Gorley, 2001).The similarity between T-RF patterns was calculated using the two-way crossed analysis www.biogeosciences.net/10/3679/2013/Biogeosciences, 10, 3679-3689, 2013 of similarities (ANOSIM).The differences were considered significant when P < 0.05.The dynamics of bacterial communities were displayed by MDS (multidimensional scaling) analyses, which were based on Log (x + 1) transformation of T-RF data and Euclidean distance.The bacterial taxonomic richness was defined as the number of peaks of T-RFLP analysis.The Shannon-Weaver index (H ) was calculated as: where P i = n i N −1 , n i is the area of a peak and N is the sum of all peak areas in each sample.The relationship between the measured environmental parameters and the bacterial community structure as revealed by T-RFLP was assessed by the BIOENV analysis provided in PRIMER 5 software.The BIOENV analysis selects the environmental parameters that may best explain the community pattern (presence/absence and area of T-RFs) by maximising the correlation between their respective similarity matrices with the application of a weighted Spearman's correlation coefficient.Due to the different frequencies for sampling for specific parameters and the analysis of T-RFLPs, we could not include all T-RFLP and environmental parameters in BIOENV analysis.Therefore, the following environmental parameters were used for BIOENV analysis: Salinity, temperature, density anomaly, pH, DO (dissolved oxygen), Chl a measured by CTD and fluorometer, A T (total alkalinity), C T , pH CO2SYS, pCO 2 CO2SYS, HCO 3 CO2SYS, CO 3 CO2SYS, Omega Ca CO2SYS, Omega Ar CO2SYS, total bacterial abundance, high DNA bacterial abundance, low DNA bacterial abundance, total and four different DNA viral abundance, VBR (viruses-bacteria ratio), POC (particulate organic carbon), PON (particulate organic nitrogen), POP (particulate organic phosphate), nutrient concentration (NH + 4 , NO − 3 , NO − 2 , PO 3− 4 , and biogenic Si), DMS (dimethylsulfide) and TEP (transparent exopolymer particles).These factors were analysed individually and in combination (Table 1).The potentially collinear factors were excluded in BIOENV analysis with combined factors to avoid possible confounding effects (Zuur et al., 2010).

Bacterial community dynamics revealed using T-RFLP analysis
Four phases were defined based on the experimental setup and the temporal dynamics of Chl a concentration during the 30 day mesocosm experiment (Fig. 1): Phase 0 (CO 2 manipulation period, t-5 to t3), Phase 1 (the first bloom period, t4 to t12), Phase 2 (the second bloom period, t13 to t21) and Phase 3 (the third bloom period, t22 to t30).A detailed description of the development of Chl a is presented in Schulz et al. (2013).Bacterial abundance showed a distinct temporal development with an abruptly drop before t7 followed by gradual increase until the end of experiment (Brussaard et al., 2013).
Apparent taxonomic richness (number of T-RFLPs in each sample) and Shannon-Weaver diversity index calculated using the T-RFLP matrix showed three major phases (Fig. 1).The development of Chl a showed also three phases, however, they occurred on slightly different time scales.Firstly, the apparent bacterial taxonomic richness and the diversity index increased rapidly during the pCO 2 manipulation procedure, peaked at around t3-t5 and then decreased until  concentration.Afterwards, taxonomic richness and diversity index increased again until t20-t24.The second peak and trough of apparent bacterial diversity were delayed compared to Chl a.During the last days, the values of these parameters increased again until the termination of the experiment (Fig. 1).Similar to the Chl a pattern, the minimum apparent bacterial diversity and taxonomic richness gradually increased during the whole experimental period.
The temporal development of the maximum taxonomic richness (S max ) and diversity index (H max ) are important parameters for a description of the dynamics of bacterial community diversity.In our study, S max and H max appeared at various dates in Phase 3 after the second Chl a minimum and varied from 27-35 and 2.725-3.180,respectively, in the nine mesocosms (Fig. 2).Significant differences of S max (independent t test, p = 0.039) and H max (independent t test, p = 0.005) were observed between the three highest pCO 2 treatments (M6, 5, 9) and otherpCO 2 treatments.The S max (27-29) and H max values (2.725-2.906) in the three highest pCO 2 treatments were lower than in the other treatments.
The bacterial community composition (T-RFLP patterns) obtained for the nine mesocosms at each time point of sampling were generally similar (e.g., see Fig. 3 for the T-RFLP patterns of t30 samples).The numbers and positions of T-RFs were similar among mesocosms, although the peak heights varied.The MDS analysis of the T-RFLP patterns with double enzyme digestion (Msp I and Rsa I) showed no clear difference between samples from each individual mesocosm (Fig. 4).This was verified using T-RFLP analysis with single enzyme digestion (Msp I, Fig. S1).On the contrary, the MDS plots of T-RFLP analysis with double or single enzyme digestion revealed a temporal development of bacterial communities during the 30 day incubation (Figs. 4 and S1).The bacterial community structure gradually changed with the incubation period.Samples collected at adjacent dates grouped together on the MDS plot while the most distinctive comparisons were those from the onset and termination of the experiment (Figs. 4 and S1).This pattern, i.e., a clear temporal, but no pCO 2 effect, as observed in MDS plots was also demonstrated by the ANOSIM analysis based on the matrix of T-RF position and peak area (Tables S1 and S2).However, MDS plotting of the samples with maximum bacterial diversity showed that samples from three highest pCO 2 treatments grouped together, but were separated from the other samples (Fig. 2).The BIOENV analysis was applied to investigate the potential effects of environmental parameters on the observed bacterial community structure (Table 1    the correlation between bacterial community structure and measured environmental factors was low.The highest correlation (0.452) was obtained with combined environmental factors including salinity, dissolved oxygen (DO) concentration, bacterial abundance (BA), viruses-to-bacteria ratio (VBR) and dimethylsulfide (DMS)/particulate organic nitrogen (PON) concentration.Other correlated parameters included particulate organic carbon (POC) and total alkalinity (A T ) (Table 1).Considering individual factors, the five parameters, which correlated best with community composition, were salinity, DO, BA, A T and DMS.

Bacterial community composition revealed by clone library analysis
In total, 743 clones were obtained from eight clone libraries (74-100 clones each) constructed from the samples collected on t30 to investigated the final responses of bacterial community to 30 day artificial CO 2 manipulation (Table 2).Sequences of Proteobacteria, Actinobacteria and Bacteroidetes dominated all clone libraries.At the family level, the Microbacteriaceae, Rhodobacteraceae and Alteromonadaceae were the major groups.Based on a definition of operational taxonomic units (OTU) as 97 % similarity of the 16S rRNA gene sequence, there were 13-27 OTUs in each clone library (Fig. S2).The phylogenetic assignment and number of clones of each OTU are shown in Table 2. Cryobacterium, Haliea, Ilumatobacter, Sulfitobacter and Loktanella related OTUs were the five most abundant ones.We did not ob-serve a clear trend of apparent bacterial diversity (number of OTUs) versus pCO 2 concentrations (data not shown).Although LIBSHUFF analysis showed significant difference in various pair comparisons of the clone library, there was no statistical evidence suggesting a pCO 2 -related effect (Table S3).The difference among clone libraries was probably related to the fact that the relatively small sampling size (< 100 clones) in the clone library analysis was insufficient to capture the true diversity, as shown by the unsaturated asymptotic rarefaction curves in Fig. S2.Similar to the T-RFLP analysis, rarefaction analysis did not reveal clear diversity changes along the pCO 2 gradient.Although it is hard for us to predict responses of specific bacterial genotypes to ocean acidification based on clone library analysis, some OTUs seemed to be affected by pCO 2 manipulation (Table 2).For example, Polaribacter OTUs (Bacteroidetes) were only found in the four lowest pCO 2 mesocosms, but not in the four highest pCO 2 treatments.In contrast, we observed highest abundance of Pelagibacter OTU (Alphaproteobacteria) in the higher pCO 2 mesocosms.The obtained sequences were in silico digested using the restriction enzymes Msp I and Rsa I for possible phylogenetic assignment to the T-RFs observed for t30 samples.In total, we successfully attributed six major T-RFs (see Fig. 3 for their positions), accounting for 73.8-86.6 % of the total peak area of each sample (Fig. 5), to Alphaproteobacteria (2 T-RFs), Bacteroidetes (2 T-RFs), Actinobacteria (1 T-RF) and Gammaproteobacteria (1 T-RF).A T-RF of 72-75 bp was the largest peak (44.2 ± 14.9% of total peak area, n = 8) in all t30 samples, especially in M9 (Figs. 3 and 5).The e-digestion analysis of clone library sequences suggested that two Actinobacteria groups (Ilumatobacter and Cryobacterium) produced this T-RF.This was supported by the finding that the clone libraries were dominated by clones affiliated to these two Actinobacteria groups (Table 2).Several Bacteroidetes sequences (Polaribacter, Ulvibacter, Haliscomenobacter and Owenweeksia) produced two major T-RFs located at 86-92 bp and 308-313 bp.A negative linear correlation (R 2 = 0.62) between the relative abundance of Bacteroidetes T-RFs and pCO 2 levels was observed.The T-RFs at 420-424 bp and 436-440 bp probably originated from the alphaproteobacterial genera Sulfitobacter and Loktanella, respectively.The second most abundant OTU (Haliea of the Gammaproteobacteria) in the clone library contributed to the T-RF of 485-488 bp (Figs. 3 and 5).There was no clear trend of relative abundance of Alphaproteobacteria (individual or combined) and Gammaproteobacteria along the gradient of pCO 2 levels.

Temporal development of the bacterial community
Generally, MDS analysis of T-RFLP patterns revealed significant temporal variations of BCC and the trend of apparent bacterial diversity and taxonomic richness seemed to be influenced by the variation of Chl a during the whole experimental incubation (Fig. 1).The temporal dynamics of bacterial community was confirmed by other two companion studies with different sampling and DNA fingerprinting strategies for both particle-attached and free-living bacterial populations (Roy et al., 2013;Sperling et al., 2013), This suggests that in our mesocosm experiment, the general heterotrophic bacterial community was tightly related with ecosystem productivity.Coupling of the bacterial community and the phytoplankton development is frequently observed in natural or experimental systems (Allgaier et al., 2008;Azam et al., 1983;Duarte et al., 2005).Heterotrophic bacterial activities, including protein production and extracellular enzyme activity, were closely coupled with phytoplankton productivity in the Svalbard mesocosm experiment (Piontek et al., 2013).In the PeECE II ocean acidification experiment, Arnosti et al. (2011) observed diverged BCC in the late-bloom phase of phytoplankton development and related this to extracellular enzyme activities associated with phytoplankton development with different pCO 2 treatments.In the present EPOCA ocean acidification experiment, bacterial activity co-varied with organic production, followed by a breakdown of picophytoplankton population, in post-bloom phase (Brussaard et al., 2013;Engel et al., 2012;Piontek et al., 2013).Therefore, the dynamics of bacterial community (including diversity and BCC) in our incubation could reflect the competition between different bacterial groups for labile dissolved organic matter released by phytoplankton.Furthermore, considering that there is no similar temporal trend of the development of heterotrophic bacterial abundance (Brussaard et al., 2013), strong top-down control of bacterioplankton biomass likely occurred in the mesocosm, which should contribute to the variation of BCC.This is supported by the fact that multiple biological parameters, including those associated with viruses (VBR) and phytoplankton (DMS), correlated to BCC dynamics in the BIOENV analysis (Table 1).This was verified by the investigation of particle-attached and free-living BCC in two accompanying studies (Roy et al., 2013;Sperling et al., 2013).Therefore, these studies showed that the cascading trophic interactions were likely a key driver of the response of heterotrophic bacterial to pCO 2 perturbation.

The effects of pCO 2 on bacterial community structure
The lack of a clear pCO 2 effect on general bacterial community composition was demonstrated by different molecular techniques (this study and Sperling et al., 2013;Roy et al., 2013) in the EPOCA experiment and in a similar mesocosm experiment performed in Bergen (Newbold et al., 2012).Our BIOENV analysis also showed that organic and inorganic parameters contributed in a complex way to BCC dynamics (Table 1).This indicates that the increase of pCO 2 levels and its chemical consequences in the ocean might not directly affect the main structure of the marine bacterial assemblage and seem to support the null hypothesis (Joint et al., 2010) that the general bacterial community is not fundamentally different under high CO 2 /low pH conditions.In our study, however, possible pCO 2 effects on BCC were observed for the developing process of bacterioplankton during incubation (Fig. 2).High pCO 2 treatments significantly reduced S max and H max and changed BCC during bacterioplankton development.Our study also indicated a threshold of the pCO 2 level between 600-675 µatm which caused observed response of bacterioplankton development to ocean acidification.Such a change of bacterioplankton community behaviour can probably only be documented in long-term incubations with high sampling frequency, which was not carried out in previous studies (e.g., Allgaier et al., 2008;Newbold et al., 2012).In addition, a negative relationship between pCO 2 level and relative abundance (peak area) of Bacteroidetes T-RFs was observed in T-RFLP analysis.Bacteroidetes is an important consumer of high molecular weight (HMW) DOM in the ocean (Kirchman, 2002).They are major heterotrophic inhabitants on marine particles as well, contributing to the degradation of particulate organic matter (POM) and TEP (Zhang et al., 2007;Kirchman, 2002).The possible stimulation of POM/TEP production by phytoplankton at high pCO 2 (Liu et al., 2010) and the decline of their heterotrophic consumers (e.g., Bacteroidetes) might influence the biological carbon pump and subsequent carbon sequestration in the ocean sediment.Microbial communities with a reduced abundance of Bacteroidetes might result in a reduced HMW DOM consumption and transformation in sea water, thus, potentially affecting the efficiency of the microbial carbon pump (Jiao et al., 2010).It is noteworthy that Roy et al. (2013) did not observe a similar negative relationship between Bacteroidetes sequences and pCO 2 level using sequencing of barcode bacterial 16S rRNA gene PCR products.No significant response of Bacteroidetes to pCO 2 treatment was detected in the Bergen mesocosm experiment (Newbold et al., 2012).Therefore, further quantitative and Bacteroidetes specific analysis (e.g., fluorescence in situ hybridisation or real time PCR) is required.
Metagenomic analysis along a vertical profile in the oligotrophic ocean showed that representatives of the Alphaproteobacteria and Gammaproteobacteria are distributed in waters covering a wide range of natural pH conditions (DeLong et al., 2006).In our study, the relative abundance of Alphaproteobacteria and Gammaproteobacteria, revealed using T-RFLP analysis, did not vary with the level of acidification, suggesting that these two dominating marine bacterial phylogenetic groups had enough genetic or metabolic plasticity to compensate for lower pH conditions.However, considering the contrasting distribution of Polaribacter and Pelagibacter in the clone library analysis, there is the possibility that ocean acidification might change the relative contribution of some phylogenetic lineages in the Alphaproteobacteria and Gammaproteobacteria.In addition, particle-attached and free-living bacteria responded differently to pCO 2 treatments in the EPOCA experiment (Roy et al., 2013;Sperling et al., 2013), as well as in other ocean acidification investigations (Allgaier et al., 2008;Arnosti et al., 2011).These observations demonstrated divers responses of marine bacterioplankton to ocean acidification and urged further investigations on different phylogenetic and ecological groups.

Outlook
Our study, as well as the other two paralleled studies (Roy et al., 2013;Sperling et al., 2013), shed light on the effects of ocean acidification on bacterial community structure in the Arctic Ocean, demonstrating a general resilience of the major bacterial phylogenetic groups under various high pCO 2 conditions.However, the potential restructuring of the BCC during bacterioplankton development as a result of ocean acidification could have a direct or indirect influence on marine organic carbon cycling.Although there is evidence showing that ocean acidification potentially affects the production of DOM, TEP and POM (Liu et al., 2010), few studies have investigated the fate of organic matter in the acidified ocean (de Kluijver et al., 2010;Kim et al., 2011).As the major component incorporating organic substrates in the marine ecosystem, heterotrophic bacteria should be given more attention in the ocean acidification studies.Bacterial extracellular enzyme activity, which directly affects the degradation of organic matter, was shown to be related with pCO 2 levels (Arnosti et al., 2011;Piontek et al., 2013).The responses of specific bacterial phylogenetic or functional lineages and the development of bacterial community to different pCO 2 conditions, as revealed by the present study, might have contributed to the dynamics of extracellular enzyme activity during the incubation period (Piontek et al., 2013).Future studies linking the diversity, activity and ecological function of marine bacterioplankton have to be conducted, in order to improve our understanding of microbial mediated marine carbon cycling (e.g., the biological pump and microbial carbon pump) under high pCO 2 condition.Furthermore, considering the tight coupling of heterotrophic bacteria and phytoplankton in our mesocosm study, pCO 2 manipulation experiments performed without autotrophs are needed to elucidate the direct response of the bacterioplankton to ocean acidification.

Fig. 5 .
Fig. 5. Bacterial community structure after 30 day incubation with different pCO 2 treatments revealed by combined analysis of clone library and T-RFLP analysis.The phylogenetic assignment of major peaks in T-RFLP profiling was performed by in silico digestion of sequences from clone library.The relative abundance of different phylogenetic groups in each sample was calculated based on the percentage of their peak area to total peak area of the T-RFLP profile.See the Materials and Methods for details.

Table 1 .
BIOENV analysis of similarity matrices of bacterial community structure based on T-RFLP analysis and environmental factors.BIOENV analysis was performed for combined environmental factors to obtain the five highest ranked correlations between similarity matrices of community fingerprints and environmental data.
The five individual environmental parameters with highest correlation coefficient values were also shown.DO: dissolved oxygen; BA: bacterial abundance; VBR: viruses-to-bacteria ratio; DMS: dimethylsulfide; particulate organic nitrogen; POC: late organic carbon; A T : total alkalinity.

Table 2 .
Major OTUs (operational taxonomic units) and clone numbers observed in clone library analysis.The taxonomic assignment of each OTU was based on RDP Classifier and GenBank Blastn analysis.