The Amundsen Sea is located in western Antarctica between the Ross Sea and Bellingshausen Sea (69°S–74°S; 100°W–135°W, Figure 1), and is characterized by a large polynya from November to February (Arrigo and van Dijken, 2003). The Korean Amundsen Sea Expedition was conducted during the austral summer, from February 9 to March 10 in 2012, aboard the Korean icebreaker research vessel RV Araon. Water depth ranged from 530 to 1,064 m, and bottom water temperature ranged from −1.8 to −1.1°C (Table 1).

Sediment samples were collected using a box corer at four stations at three contrasting sites: the MIZ (Stn 83), inside the polynya (Stns 10 and 17), and in direct proximity to the ice shelf site (Stn 19) (Table 1). P. antarctica was the major planktonic algae in this highly productive polynya area, whereas diatoms were more abundant in the relatively less-productive MIZ (Stn 83) (Figure 1B) (Yang et al., 2018). Subsamples for DNA extraction were taken from the center portion of the box corer using acryl sub-core liners (6 cm in diameter). Cores were sliced at 1-cm intervals to a depth of 11 or 18 cm, and immediately frozen at −80°C.

Total genomic DNA was extracted from the different sediment layers using a PowerMax DNA Isolation kit (Mo Bio Laboratories, Carlsbad, CA, United States), following the manufacturer’s instructions. Quantitative real-time polymerase chain reaction (q-PCR) was used to determine the copy number of 16S rRNA genes and archaeal ammonia monooxygenase A (amoA) genes using a TaqMan assay and a SYBR Green I assay, respectively (Supplementary Table S1). Details are described in Supplementary Material.

For each of the 48 sampled sediment layers at the polynya and non-polynya sites, we generated a 16S rRNA gene amplicon library, with one primer set covering the V5–V8 region of both bacterial and archaeal taxa (Jorgensen et al., 2012). PCR amplification of the 16S rRNA genes was performed in triplicate using a primer set of Uni787F (Roesch et al., 2007) and Uni1391R (Lane et al., 1985) according to Jorgensen et al. (2012). Resulting amplicons were sequenced by Macrogen Corporation (Korea) using the 454 GS FLX + system (Roche). The raw data have been deposited to the NCBI SRA database under accession number SRX3405376. Raw flowgrams of pyrosequencing reads were filtered and de-noised by a PyroNoise algorithm (Quince et al., 2011) implemented in MOTHUR (ver. 1.36.1) (Schloss et al., 2009). Singleton OTUs were removed prior to analysis. Chimeric sequences were identified and removed by ChimeraSlayer. A total of 132,914 pyrosequencing reads from 48 samples were qualified for further processing. The sequences were then clustered into operational taxonomic units (OTUs) that met the criteria of a 97% similarity threshold and a minimum cluster size of 2 using a QIIME pipeline (ver. 1.9.1) (Caporaso et al., 2010). Taxonomy for each OTU_0.97 was assigned using the RDP classifier method (Wang et al., 2007) with the Greengenes database (ver. 13_8) (McDonald et al., 2012). To avoid the effects of different sample sizes for estimating diversity, comparison sequences were randomly subsampled to the smallest library size (Kirchman et al., 2010), which were 1,628 sequences in the present study. Chao1 estimates were created using QIIME software to assess diversity (Chao, 1984).

Geochemical constituents (NH4+, NO_X [NO3- + NO2-], PO43-, and Fe²⁺) in the pore water, oxygen penetration depth (OPD), total oxygen uptake (TOU), sulfate reduction rates, and N₂ removal rates by denitrification and anaerobic ammonia oxidation (anammox) were adopted from the results reported by Choi et al. (2016) and Kim et al. (2016). Total organic carbon (TOC) and total nitrogen (TN) were analyzed with Elemental Analyzer (Carlo Erba, NA-1500) (Verardo et al., 1990) (see Supplementary Material for detailed methods).

Spatial difference of geochemical constituents was assessed using a Mann–Whitney U-test, with a statistical significance level of 0.05. The relative abundance of microbial populations was determined using a Kruskal–Wallis test. Tukey’s honest significance difference (HSD) test was used for multiple comparison of means at a 95% confidence interval. The OTU table from QIIME and geochemical measurements were analyzed using R (v. 3.3.2) (R Core Team, 2016) with custom scripts and several packages including vegan (v. 2.4-2) and lmtest (v. 0.9-35). Exploratory data analysis was carried out for both microbial communities and geochemical measurements data using non-metric multidimensional scaling (NMDS), diversity measures, and hierarchical clustering. Microbial community was analyzed at the OTU_0.97 level as well as at the phylum and order levels. Compositional difference of microbial communities among stations was tested by multivariate analysis of variance–like non-parametric tests (analysis of similarities [ANOSIM] and permutational multivariate analysis of variance [PERMANOVA]). Ordination of microbial communities was fitted with geochemical measurements by vector fitting. Constrained ordination models were constructed by redundancy analysis (RDA) in an iterative fashion considering collinearity among constraining geochemical variables. Community structures were compared among stations, and between bacteria and archaea using a Procrustes test on RDA ordination configuration and a Mantel test.

The distribution of geochemical constituents and community structures showed distinct spatial variation between the polynya sites and non-polynya sites of the Amundsen Sea. Contents of TOC and TN were approximately 1.25 to 2.1 times higher at the polynya than at the ice shelf (Stn 19) and sea-ice zone (Stn 83) (Figure 2 and Table 1) (P < 0.001). Accordingly, OPD in the polynya sites (1.8–2.0 cm) was shallower than in non-polynya sites (3.5–3.6 cm) (P < 0.001). Pore-water analysis revealed that the concentration of NO_X at the polynya sites decreased with depth from approximately 30 μM at the top to 8 μM at 3–5 cm depth, and then remained constant down to a depth of 10–20 cm (Figure 2). In contrast, NO_X concentration at the non-polynya sites (ice shelf and sea-ice zone) was high (>20 μM) at all depths. NH4+ concentration was higher at the polynya sites than at the non-polynya sites (Figure 2) (P < 0.001). Concentration of Fe²⁺ in the pore-water was low at all sites (<10 μM), but the average concentration of dissolved Fe²⁺ was higher at the polynya sites (4.9 μM) than at the non-polynya sites (1.98 μM) (P = 0.003). Metabolic activities such as TOU rate and anaerobic respiration by sulfate reduction were consistently higher in the polynya sites with relatively higher C_org content compared to those measured in non-polynya sites (Figure 2 and Table 1) (P = 0.002).

Total prokaryotic abundance determined by q-PCR of 16S rRNA genes ranged from 2.3 × 10⁵ to 3.3 × 10⁸ copies cm^–3 per each sample (Table 2). Although the 16S rRNA gene copy numbers appeared higher at Stn 17 than those at Stn 10, the variation between polynya sites was not significant (P = 0.06). However, 16S rRNA gene copies in the sediment of Stn 17 were clearly higher than those of Stn 83 (P < 0.05). Total prokaryotic abundances were higher in the surface sediments and decreased with depth at all sites (Figure 2). The relative abundance of bacterial 16S rRNA gene copies to total prokaryotic cells was highest at Stn 10 (96.7 ± 3.3%), and then decreased to 84.8 ± 3.7% (Stn 17), 77.6 ± 7.0% (Stn 19) and 70.3 ± 9.7% on average (Stn 83). In contrast, archaeal 16S rRNA gene proportion comprised 3.3 ± 0.7% at Stn 10, and then gradually increased with distance from Stn 10 to 15.2 ± 3.7% (Stn 17), 22.4 ± 7.0% (Stn 19), and 29.7 ± 9.7% on average (Stn 83) (Table 2). The percentage of bacterial and archaeal 16S rRNA gene copies of the total 16S rRNA genes determined by q-PCR (Table 2) corresponded with the relative abundance of Bacteria versus Archaea as estimated from pyrosequencing data (Figure 2). On the other hand, copy numbers of archaeal amoA gene ranged from 2.8 × 10⁴ to 5.7 × 10⁶ and from 1.7 × 10⁴ to 8.3 × 10⁶ cm^–3 for polynya sites and non-polynya sites, respectively. The depth profiles of amoA gene copy numbers in the ASP sediments showed a similar trend to the archaeal 16S rRNA gene copy numbers (Supplementary Figure S1).

In total of 132,914 pyrosequencing reads with an average length of 601 bp, 9,852 reads were unique. Chao1 indices are as high as an average of 1,909 in the 7–8 cm depth of Stn 17, and were lowest at 704 in the 14–16 cm depth of Stn 83 (Table 2). Chao1 and observed OTU_0.97 counts of the polynya and ice-shelf sites were higher than those of the sea-ice zone. These two indices were estimated to be highest within the sub-oxic layers (3–8 cm depth) of all sites.

Non-metric multidimensional scaling of OTUs resulted in a separate clustering between polynya and non-polynya samples, and the separation was mainly driven by differences in geochemical constituents (NH4+, Fe²⁺, and PO43-), TOC, TN, and sulfate reduction rate (Figure 3). Both archaeal and bacterial communities were quite distinctive between polynya sites and non-polynya sites (P < 0.001 from both PERMANOVA and ANOSIM) (Supplementary Figure S2). Similar patterns in community structure were observed at order-level resolution (Supplementary Figure S2D). From RDA analysis (Supplementary Figure S3), NO_X concentrations correlated with surface microbial communities at all stations. C_org contents correlated with microbial communities in surface sediments of Stn 10 and at intermediate depths of Stns 17 and 19, respectively. Archaeal communities were more distinctive among stations, while bacterial communities were more similar overall (Supplementary Table S2). Compositional similarity at the phylum level better reflected ecological or geographic settings.

The relative contributions of each phylum were dramatically different between sites (Figures 2, 4). At the polynya site (Stn 10), the most dominant phyla were Planctomycetes (35–71%), Proteobacteria (4.8–22%), Thaumarchaeota (1.2–12.1%), and Chloroflexi (1.9–10.8%). Candidate Division GN02 (0.5–4.4%), Candidate Division SBR1093 (0.2–8.3%), Acidobacteria (1.2–5.0%), and Bacteroidetes (0.4–3.2%) were of minor abundance. At Stn17, the proportion of Planctomycetes decreased slightly (14.8–43%), whereas the percentage of Thaumarchaeota (13–51%) appeared to be more abundant than at Stn 10. Other phyla included Proteobacteria (10.6–31%), Chloroflexi (2.1–8.7%), Candidate Division SBR1093 (0.5–8.3%), and Acidobacteria (2.6–4.1%). The proportion of archaeal sequences prominently increased at Stns 19 and 83, which is consistent with the archaeal cell abundance estimated from 16S rRNA gene quantification (Figure 2 and Table 2). Relative abundance of Thaumarchaeota accounted for 41–59% and 41–67% at Stn 19 and Stn 83, respectively. Proteobacteria (4.1–27.8%), Planctomycetes (5.8–24%), Chloroflexi (1.6–8.1%), Acidobacteria (2–8%), and Candidate Division SBR1093 (0.3–8.1%) appeared in similar proportions at both non-polynya sites (Figure 2). The relative abundance of Planctomycetes in the polynya sediments showed a high positive correlation with NH4+ (r = 0.64, P < 0.001), Fe²⁺ (r = 0.46, P < 0.001), TOC (r = 0.75, P < 0.001), TN (r = 0.73, P < 0.001), and SRR (r = 0.48, P < 0.001) (Figure 5). In contrast, the relative abundant of Thaumarchaeota in the non-polynya sediments showed a negative correlation with TOC (r = −0.75, P < 0.001) (Supplementary Figure S4).

The most abundant Planctomycetes in the microbial communities were divided into three clades: a Pirellula-like group, candidate order MSBL-9 (Mediterranean Sea Brine Lake-9) (Pachiadaki et al., 2014), and Candidatus Brocadiae (Supplementary Figure S5). The Pirellula-like group and the candidate order MSBL-9 were the two most dominant bacterial groups at Stn 10, representing 34% and 32% of the total 16S rRNA gene sequences, respectively (Figure 4 and Supplementary Figure S5). In contrast, at Stn 83, the relative abundance of the Pirellula-like group decreased to < 17%, and the presence of the candidate order MSBL-9 group was not discernible. The members in Candidatus Brocadiae, which are capable of anaerobic ammonium oxidation (anammox) using nitrite as an electron acceptor (Schmid et al., 2003), were most abundant at Stn 17, and comprised a maximum of 9.2% of the total prokaryotic sequences at the 9–10 cm depth (Figure 4 and Supplementary Figure S5).

The relative abundance of Proteobacteria did not exhibit significant spatial variation among sites [Kruskal–Wallis, χ²(3) = 1.375, P = 0.711]. Similarly, the spatial distribution of Gammaproteobacteira was similar between sites. Most gammaproteobacterial sequences in all sites were affiliated with the orders Thiotrichales and Chromatiales (Supplementary Figure S6). On the other hand, based on the order level, delta- and alphaproteobacterial compositions showed spatial variations (Supplementary Figure S6). Desulfobacterales and Desulfuromonadales, which are well-known as sulfate- and sulfur (S⁰)-reducing bacterial groups in marine sediments, appeared to be most abundant at Stn 10, but were rarely detected at Stn 83. In contrast, the deltaproteobacterial sequences in the candidate order NB1-j appeared higher at Stn 19 and Stn 83, but showed low relative abundance at Stn 10 (Figure 4 and Supplementary Figure S6). Members of the Alphaproteobacteria made up 0.4–6.9% of total reads and were divided into three major orders (Supplementary Figure S6): Rhodobacterales, Rhodospirillales, and Rhizobiales. The sequences falling into the Rhodobacteraceae appeared more frequently in Stn 10, while members of Rhodospirillacaea were detected more frequently at non-polynya sites (Stns 19 and 83).

The OTUs in Chloroflexi and Bacteroidetes appeared to vary between polynya sites and non-polynya sites (Supplementary Figure S7). The Anaerolineae were a major member of Chloroflexi at Stn 10, whereas sequences clustered in class SAR202 and TK17 were more abundant at Stns 17, 19, and 83. Members of Bacteroidetes that have been reported as major organic matter decomposers in SO sediment (Carr et al., 2013; Ruff et al., 2014; Learman et al., 2016) were substantially low with relative abundance less than 5% of total 16S rRNA gene sequences in the polynya (Figure 2).

The Thaumarchaeota were a predominant microbial group in sediments from the Amundsen Sea, except at the polynya center (Stn 10) (Figure 2). Most Thaumarchaeota sequences could be classified into three subgroups termed Alpha, Theta, and Upsilon (Durbin and Teske, 2011). The Thaumarchaeota subgroup Alpha was dominant in the top layers (0–5 cm) at all sites. The community composition of the Thaumarchaeota subgroup changed gradually from the top to the bottom at the non-polynya sites (Supplementary Figure S8). The relative abundance of archaeal sequences associated with Theta and Upsilon subgroups in Thaumarchaeota increased with increasing depth, especially in the marginal sea ice zone (Stn 83). Marine Benthic Group B (MBGB) and Miscellaneous Crenarchaeotic Group (MCG), which have been proposed to be Thorarchaeota (Seitz et al., 2016) and Bathyarchaeota (Rinke et al., 2013), respectively, were only detected at Stn 10 (∼9.5% and ∼5% of total sequences, respectively). They have been reported as putative heterotrophic microorganisms (Biddle et al., 2006; Jorgensen et al., 2012; He et al., 2016). Likewise, archaeal sequences related to the methane cycle, including Methanobacteria, Methanomicrobia, and Thermoplasmata, were only detected at Stn 10 (<2% of the total sequences).

One of the prominent features revealed by 16S rRNA genes pyrosequencing was that the members of Planctomycetes were the most abundant microbial members detected in the highly productive polynya sites, especially at Stn 10 (Figure 2). Many Planctomycetes have been found attached to sinking marine aggregates in the water column (DeLong et al., 1993; Fuchsman et al., 2012). At the ASP, however, none were detected in the water column (Delmont et al., 2014; Kim J.-G. et al., 2014). Thus, the highly abundant Planctomycetes sequences in the ASP sediment could not have originated from the water column, confirming the results reported by Probandt and coworkers who showed that Planctomycetes in subtidal, sandy sediments differed from those in the overlaying water column (Probandt et al., 2017). Pirellula members in Planctomycetes have been reported often as heterotrophic bacteria that degrade organic matter produced by algae (Glöckner et al., 2003; Morris et al., 2006; Bižić–Ionescu et al., 2015), while little information is available to speculate about the ecological role of uncultured Planctomycetes (Figure 5). Most cultivated Planctomycetes are known as aerobes or facultative aerobes (Schlesner et al., 2004). However, environmental sequences from Planctomycetes are often retrieved from anoxic zones, such as methane hydrate-bearing sediment (Inagaki et al., 2006) and subsurface seafloor sediment (Jorgensen et al., 2012). Therefore, detection of these sequences in oxygen-depleted layers is unsurprising (Figure 4).

In fine-grained marine sediment receiving high organic material input, Delta- and Gamma-proteobacteria have been reported as the predominant bacterial groups (Rooney-Varga et al., 1997; Bowman and McCuaig, 2003; Bissett et al., 2006; Liu et al., 2015). Similarly, in most benthic systems of the SO, where organic materials primarily originate from diatom blooms in the water column, proteobacterial groups such as Delta- and Gamma-proteobacteria occupy ecological niches as primary mineralizers of organic materials in the sediments (Bowman and McCuaig, 2003; Baldi et al., 2010; Ruff et al., 2014; Learman et al., 2016). Therefore, our observation of high Planctomycetes abundance, an average of 40% of total sequences (Figure 2) in ASP sediments, is intriguing. Further discussion on the dominance of Planctomycetes is presented in next section.

To the best of our knowledge, this is the first report suggesting the predominance of Planctomycetes in the ASP, which contrasts with the microbial communities in other benthic systems underlying diatom-dominated water columns. Major controls regulating the distribution of Planctomycetes in the sediments remain unknown, largely because no culture-based studies are available (Schlesner et al., 2004; Lee et al., 2013). However, our statistical analysis (Figure 6 and Supplementary Figure S9) revealed that relative abundance of Planctomycetes in the ASP showed a significant positive correlation with TOC contents and inorganic constituents (NH4+, PO43-, and Fe²⁺), presumably resulting from the benthic C_org mineralization. Most labile C_org produced by P. antarctica blooms are rapidly decomposed by heterotrophic bacteria in the water column before reaching the sediment of the ASP (Kirchman et al., 2001). The export flux of C_org formed by a Phaeocystis bloom is twofold slower than that formed by a diatom bloom (DeJong et al., 2017), and thus the contribution of P. antarctica cells to total export flux below the photic zone declines dramatically (Reigstad and Wassmann, 2007). Indeed, in the ASP, most photosynthetically-produced particulate organic carbon (POC) (>95%) exported from the surface layer is converted to suspended POC and/or dissolved carbon within the top 400 m of the water column (Ducklow et al., 2015; Lee et al., 2017). In addition, direct rate measurement conducted in this study also demonstrated that, despite high primary production (110 mmol C m^–2 d^–1) in the Phaeocystis-dominated water column of the ASP, benthic C_org mineralization (average, 2.1 mmol C m^–2 d^–1) accounted for only 1.9% of primary production, which was strikingly lower than that measured in other less productive polar regions (Kim et al., 2016). Therefore, organic materials accumulated in surface sediments of the deep ASP (>700 m) would be intrinsically recalcitrant. Meanwhile, it has been well-established that Phaeocystis colonies excrete a mucous matrix containing both carboxylated and sulfated heteropolysaccharides as main constituents (van Boekel, 1992; Alderkamp et al., 2007). Based on metagenomic information from Namibia and Oregon coastal upwelling systems, Woebken et al. (2007) revealed that all marine Planctomycete genomes, except for Candidatus Kuenenia stuttgartiensis, possess a high number of sulfatase genes, which suggests that marine Planctomycetes may be able to break down recalcitrant sulfated heteropolysaccharides (Glöckner et al., 2003; Wegner et al., 2013). Consequently, the high relative abundance of Planctomycetes at the center of the polynya (Stns 10 and 17) suggests that the members of Planctomycetes constitute a significant heterotrophic bacterial group utilizing recalcitrant organic materials produced by Phaeocystis blooms in the water column of the ASP. Recently, Probandt et al. (2017) also suggested that Planctomycetes play a key role in the degradation of high molecular weight compounds and recalcitrant materials entering surface sediment from the water column of the Wadden Sea.

In addition to the Planctomycetes-dominated bacterial communities in the sediments of the polynya sites (Stns 10 and 17), another interesting finding revealed from the quantification of 16S rRNA gene was that archaeal abundance occupied more than half (30–71%) of the total prokaryotic abundance in the MIZ (Stn 83), in which most archaeal 16S rRNA gene sequences were assigned to Thaumarchaeota (Figure 2 and Table 2). Members of Thaumarchaeota are major contributors to aerobic ammonia and nitrite oxidation in aquatic environments (Könneke et al., 2005). To examine the potential of sedimentary Thaumarchaeota to oxidize ammonia, we quantified the archaeal amoA gene, which is known for a genetic marker for the ammonia oxidation. The depth profiles of archaeal amoA and 16S rRNA gene copy numbers were very similar to each other in the samples except for some layers (Supplementary Figure S1), which implies that most archaeal members at all depths have a gene encoding ammonia monooxidase.

Both the proportion of archaeal abundance in total prokaryotic abundance (Supplementary Figure S10) and the relative abundance of Thaumarchaeota in total 16S rRNA gene sequences (Supplementary Figure S4B) showed a negative correlation with TOC contents. Previously, the environmental members of Thaumarchaeota have been reported in oligotrophic marine sediments as an important chemolithotrophic microbial assemblage (Inagaki et al., 2006; Durbin and Teske, 2011). Similarly, cultivated thaumarchaeal ammonia oxidizers have been shown to be adapted to oligotrophic conditions (Martens-Habbena et al., 2009). Recently, based on continuous succession of an active microbial group belonging to Thaumarchaeota in the absence of an external C_org supply, Sebastián et al. (2018) suggested that Thaumarchaeota play a significant role in sustaining oligotrophic bathypelagic ecosystems where the C_org supply is limited. Likewise, metaproteomic data indicate that MG-I are abundant and metabolically active at the surface water of the West Antarctic Ocean during the ice-covered winter (Williams et al., 2012), and chemoautotrophic carbon fixation by Thaumarchaeota contributes up to 9% of bacterioplankton production during the Antarctic winter (Tolar et al., 2016). Indeed, in the present study, primary production was low in the water column of the ice-covered sea-ice zone (Stn 83) (Lee et al., 2012; Hyun et al., 2016), and the C_org content in the sediment was consequently lowest (<0.5%, Table 1). Our results thus suggest that the Thaumarchaeota predominating in the sediment of non-polynya sites are a significant chemolithoautotrophic group, sustaining the benthic ecosystem where input of C_org from the water column is limited.

The structure and function of benthic heterotrophic microbial communities are ultimately controlled by the quantity and quality of organic matter from the water column (Danovaro et al., 2000; Bissett et al., 2006; Jamieson et al., 2013; Learman et al., 2016; Probandt et al., 2017). Thus, the predominance of Planctomycetes in ASP sediment provides ecological and environmental baseline information for assessing the responses of benthic ecosystems to transitions in phytoplankton communities that may result from ongoing and future climate change in the Amundsen Sea. Due to the inflow of warm CDW, the western Antarctic Ocean including the Amundsen Sea has experienced significant surface warming and loss of sea ice (Stammerjohn et al., 2008; Thoma et al., 2008; Jenkins et al., 2010; Jacobs et al., 2011). In particular, glaciers near the Amundsen Sea are undergoing the highest rates of melting and thinning on the Antarctic continent (Rignot, 2008). Sea ice melting increases the supply of nutrients and intensifies water column stratification (Arrigo et al., 1999, 2012; Poulton and Raiswell, 2005), which consequently results in increased primary productivity (Arrigo et al., 2017; Oliver et al., 2018) and shifts in phytoplankton community structure from P. antarctica to diatoms (Arrigo and van Dijken, 2003; Arrigo et al., 2008, 2012). As discussed in previous sections, while C_org originating from Phaeocystis blooms settles slowly, the particulate organic material produced by diatoms possesses a relatively faster sinking rate and a lower C:N ratio (i.e., they are relatively more labile) than those produced by Phaeocystis (Alderkamp et al., 2007; DeJong et al., 2017). Therefore, the transition of dominant phytoplankton from Phaeocystis to diatoms would deposit more labile organic matter into the benthic system of the ASP, which ultimately results in a shift in the benthic microbial community structure from Planctomycetes to Proteobacteria. Likewise, in the marginal sea-ice zone, the loss of sea ice due to global warming may cause an increase in primary production in the water column (Arrigo et al., 2017; Oliver et al., 2018). As a result, benthic microbial communities in the MIZ can undergo a shift from Thaumarchaeota to other heterotrophic bacteria. Consequently, in the Amundsen Sea, a spatial distribution of the benthic microbial community reflecting the composition and production of dominant phytoplankton groups in the water column can act as a relevant proxy for detecting possible variation in primary productivity and phytoplankton community that is associated with sea ice melt due to the global warming.