Surface sediments were sampled at the central station of the LTER HAUSGARTEN, located in the Fram Strait west of Spitsbergen (79° 03.86′ N, 4° 10.85′ E; 2,470 m water depth). A TV-guided multiple corer (MUC; Barnett et al., 1984) was used for retrieving undisturbed sediments during RV Polarstern cruise PS93.2 (PS93/050-5 and -6; Soltwedel, 2015). Surface sediments were fully oxic (300 μmol L^-1 O₂) and consisted of fine clays (Kanzog et al., 2008). Immediately after sediment retrieval, a control sample ‘day 0’, representing the in situ bacterial community structure, was taken and frozen at –80°C for RNA and at –20°C for DNA extractions.

For the feeding experiment, the first two centimeters of ten cores were combined into one sterile glass bottle and transferred into a cold room (0°C, in situ temperature). The sediment was 3.5-fold diluted with sterile-filtered and air saturated bottom water (35 PSU; 0°C), in order to supply enough oxygen for at least one community duplication. The well-mixed slurry was divided into six sterile 1 L glass bottles. One bottle with slurry remained as unfed control. The other five bottles were amended with substrates to final concentrations of 0.2 mg organic carbon per mL undiluted sediment, which corresponds to 16.6 mmol L^-1 organic carbon, and is comparable to an annual input of particulate organic carbon at this site (Bauerfeind et al., 2009; Kanzog and Ramette, 2009; Lalande et al., 2013). All treatments were divided into ten replicates of 50 mL each, transferred and sealed into sterile, gas-permeable polyethylene (PE) bags with no headspace or air bubbles.

Slurry incubations were started within 24 h after retrieval of the samples, using pressure vessels described in Boetius and Lochte (1994). Ten pressure vessels were filled with fully oxygenated sterile-filtered bottom water, and six PE bags were added to each of them (one replicate bag of each of the five treatments, and one control bag). Five of these pressure vessels were incubated at atmospheric pressure (1 atm), the other five at in situ pressure (250 atm). Pressure was applied as hydrostatic (water) pressure, using a mechanical pump. The PE bag material was oxygen permeable, so that oxygen diffusion into the bags prevented the slurries from becoming anoxic (Supplementary Table S1; Supplementary Data). All incubators were stored at 0°C in the dark for 23 days.

We selected five different, naturally occurring, complex substrates to test functional and structural responses of Arctic deep-sea bacterial communities. These were: chitin (CHI), Thalassiosira weissflogii (TWEI), Emiliania huxleyi (EHUX), Bacillaria sp. (BCLA), and Melosira arctica (MARC).

Chitin is one of the main components of, e.g., fecal pellets, exoskeletons of worms, arthropods, and mollusks, and thus constitutes a key source of carbon and nitrogen to deep-sea life (Wakeham and Lee, 1993). Here, we used a commercially available product (flakes from shrimp shells, Sigma–Aldrich, Germany). As more complex detritus we added different algae: the centric diatom TWEI, representing a major fraction of sinking particulate organic matter at HAUSGARTEN observatory, as well as the coccolithopore EHUX that has been reported to migrate from the Atlantic towards the Arctic (Bauerfeind et al., 2009; Winter et al., 2014). Furthermore, we provided detritus of the sea-ice algae BCLA (Stecher et al., 2015), a pennate diatom, and MARC, a centric diatom. As a consequence of sea-ice melting, these sea-ice algae can cause massive sedimentation events that strongly impact local sediment turnover rates (McMahon et al., 2006; Boetius et al., 2013). In reference to the incubations, the term phytodetritus includes all algae used (EHUX, TWEI, BCLA and MARC), of which EHUX and TWEI belong to phytoplankton, and BCLA and MARC to the sea-ice algae.

TWEI, EHUX, and BCLA were collected during several Arctic cruises, and were cultivated in F/2-medium (Guillard and Ryther, 1962; Guillard, 1975). Algal mats dominated by MARC (but also containing few other species of pennate diatoms and Chaetoceros sp.) were collected directly from the ice during Polarstern cruise PS86 in July 2014 (Boetius, 2015) and frozen at –20°C.

The phytoplankton species (TWEI and EHUX) were incubated in 2.5 L F/2 medium at 18°C and 15 μE using 5 L beakers, and were resuspended twice a week by gentle shaking. TWEI and EHUX were transferred into new media every 10 days. 100 mL of the 2.5 L culture were used to inoculate fresh medium, while the remainder was centrifuged at 1811 × g, and the pellet was frozen at –20°C. BCLA was grown at 0°C and 10 μE in 400 mL medium in a 1 L beaker. Due to slower growth, it was transferred and harvested once a month as described above.

For inactivating algal-attached bacteria, we sterilized the algae with microwave radiation at 600 W for three minutes before adding them to the slurry. Sterilization may alter organic matter lability, as has been shown for the process of autoclaving (Turley and Lochte, 1990; Wolf and Skipper, 1994). This may, to some extent, cause changes similar to the natural aging of organic matter, but was not quantified in more detail here. This treatment also led to a disintegration and homogenization of the algae cultures. The effective sterilization was tested by measurements of enzyme activity, oxygen uptake, bacterial cell numbers, and bacterial genetic diversity, which were all negative or without significant increase or change. The carbon and nitrogen content of the different food sources were determined according to Grasshoff et al. (1983), using a Fisons element analyzer type NA 1500 (Fisons plc, UK). C:N ratios were 8:1 for CHI, 10:1 for TWEI, 11:1 for EHUX, 12:1 for BCLA and 23:1 for MARC.

To determine prokaryotic cell numbers, 1 mL sediment slurry was fixed with sterile filtered formalin/seawater at a final concentration of 2% and stored at 4°C. Each sample was filtered on a 0.2 μm polycarbonate filter, stained with acridine orange as DNA dye (0.01% final concentration), and counted using an epifluorescence microscope (Axiophot II Imaging, Zeiss, Jena, Germany). For each sample, 30 random grids from two replicate filters (technical replicates) of three biological replicates, representing three out of five replicate PE-bags per treatment, were counted (Meyer-Reil, 1983). Bacterial biomass was estimated using a conversion factor of 3 × 10^-13 g C μm^-3 biovolume (Børsheim et al., 1990), and assuming an average cell volume of 0.07 μm³ (Boetius and Lochte, 1996).

By measuring the potential hydrolytic activity of two abundant enzymes, N-acetylglucosaminidase (chitobiase) and beta-glucosidase, in a modification of the method by Hoppe (1983) we tested, which of the supplied macromolecular organic substrates induced a production of extracellular enzymes in the community. Two 11 mL subsamples were taken from each of the five replicate PE-bags, and mixed with the methylumbelliferon (MUF)-labeled artificial substrates 4-methylumbelliferyl ß-D-glucopyranoside (MUF-beta-glucosidase) and 4-methylumbelliferyl-N-acetyl-ß-D glucosaminide (MUF-chitobiase), at saturation levels (100 μmol L^-1) according to the method described in Boetius and Lochte (1994; Supplementary Table S2). The assay was run at 0°C for three hours at atmospheric pressure. We measured the release of the fluorochrome (MUF) after one and three hours, using a RF-5300 spectrofluorophotometer (Shimadzu Scientific, USA; emission: 455 nm, excitation: 365 nm). Fluorescent activity was calibrated with MUF standards ranging between 0 and 6 nmol mL^-1. Extracellular enzymatic activity (EEA) was then calculated per volume of sediment and time (μmol mL^-1 d^-1). Since substrate concentrations are at saturation level within the incubation time, enzyme activities represent maximum velocities (Vmax). Enzymatic activity measurements may be confounded by competitive inhibition between the fluorogenic substrate and the added carbon (Arnosti, 2011). However, previous studies have reported an induction of enzyme production for beta-glucosidase and chitobiase by their respective substrates (Boetius and Lochte, 1994; Boetius and Lochte, 1996).

Oxygen concentrations were measured at the end of the incubation in each of the five replicate PE-bags per treatment as an indicator of community activity over the duration of the experiment. Optodes were connected to the FireStingO₂ fiber-optic oxygen meter (both obtained from Pyro Science GmbH, Germany), and calibrated at in situ temperatures by a two-point calibration using air-purged seawater and oxygen-free seawater by adding sodiumhydrogensulfite (Sigma Aldrich, Germany). The measuring accuracy was between 0.2% at 20% oxygen and 0.01% at 1% oxygen (FireSting O₂, Fiber-optic Oxygen Meter manual, Pyro Science GmbH, Germany). The PE-bags were gently shaken to mix the slurries and punctured immediately after retrieval from the incubator with the needle-protected optode fiber. Oxygen concentrations were measured in the center of the slurries. Because the PE-bags were oxygen-permeable, the measured oxygen concentration at the end of the 23 days of incubation was in steady state, i.e., the microbial respiration in the bag was balanced by the diffusive transport of oxygen across the PE-foil. From previous tests, an oxygen diffusion coefficient of 2.65 × 10^-12 m^-2 s^-1 was determined and in combination with the oxygen difference across the PE-foil, the bulk flux and thus the respiration rate in each of the bags were determined (Supplementary Data; Supplementary Table S1; Supplementary Figure S1).

The analysis of 16S rRNA-genes (16S rDNA) reveals all taxa present in a given sample, while the analysis of the 16S rRNA identifies those taxa actively transcribing RNA under a given condition. From the latter, the metabolically active bacterial community can be inferred, and community shifts may be more easily detected (Moeseneder et al., 2005; Gaidos et al., 2011).

For DNA extraction, 7 mL slurry of three replicate PE-bags per treatment were sampled, centrifuged (3 min at 3000 rpm to remove the overlaying water) and stored at –20°C. Total DNA was extracted from 0.5 g sediment with the UltraClean Soil DNA Isolation Kit (MoBio Laboratories Inc., Carlsbad, CA, USA), and quantified using a microplate spectrometer (Infinite^®200 PRO NanoQuant, TECAN Ltd, Switzerland). For assessing the active fraction of the bacterial community, 20 mL slurry of one replicate per treatment were sampled, immediately frozen in liquid nitrogen and stored at –80°C. Total RNA was extracted from 5 g sediment with the MoBio PowerSoil RNA extraction Kit (MoBio Laboratories Inc., Carlsbad, CA, USA). After DNAse digestion and purification using the RNeasy MinElute Cleanup Kit (QIAGEN, Germany), the purity and quantity were determined by electrophoresis using the 2100 Bioanalyzer (Agilent Technologies, Inc., Santa Clara, CA, USA). The extracts were translated into cDNA using the qScriptTM cDNA SuperMix Kit (Quanta Biosciences, Gaithersburg, MD, USA).

For 16S cDNA amplicon library preparation, the standard instructions of the 16S Metagenomic Sequencing Library Preparation protocol (Illumina, Inc., San Diego, CA, USA) were followed. The hypervariable V3–V4 region of the bacterial 16S cDNA and rDNA was sequenced using the bacterial primers S-D-Bact-0341-b-S-17 (5′-CCTACGGGNGGCWGCAG-3′) and S-D-Bact-0785-a-A-21 (5′-GACTACHVGGGTATCTAATCC-3′; Klindworth et al., 2013). Sequences were obtained on the Illumina MiSeq platform in a 2 × 300 bp paired-end run as well as in a 2 × 250 bp paired-end run on the Illumina HiSeq platform (CeBiTec Bielefeld, Germany). Raw paired-end reads were primer-trimmed using cutadapt (Martin, 2011). For quality trimming, a sliding window of four bases and a minimum average quality of 15 was applied in trimmomatic v0.32 (Bolger et al., 2014), and the reads were merged using PEAR v0.9.5 (Zhang et al., 2014). Clustering into OTUs was done with the swarm algorithm using default parameters (v2.0; Mahé et al., 2014). One representative sequence per OTU was taxonomically classified with SINA (SILVA Incremental Aligner; v1.2.11; Silva reference database release 123) at a minimum alignment similarity of 0.9, and a last common ancestor consensus of 0.7 (Pruesse et al., 2012). OTUs that were classified as chloroplasts, mitochondria, Archaea, and those not classified at the domain level were excluded from further analyses, as well as OTUs that occurred with only a single sequence in the whole dataset.

Enzyme activity, cell abundances, and respiration rates based on oxygen measurements were calculated per volume of undiluted sediment. Differences between treatments were calculated based on ANOVA (analysis of variance) with Tukey HSD post hoc tests at a significance threshold of 0.05 (Hothorn et al., 2008). For cell counts, where technical replicates were available, mixed model ANOVA and the implementation of Tukey HSD in the R package multcomp were used. EEA and AODC data were square-root transformed to meet the assumptions of ANOVA.

Observed richness (number of OTUs) and evenness (inverse Simpson index) of the communities, were calculated via repeated random subsampling to the minimum library size of the amplicon data set (i.e., 26,712 sequences). The change in community structure (beta-diversity) between samples was assessed by calculating Bray–Curtis dissimilarities from relative OTU abundances [%], to produce non-metric multidimensional scaling (NMDS) plots, and to visualize community similarity between the 16S rRNA and rDNA datasets as well as between the different treatments.

The R-package ALDEx2 (ANOVA Like Differential Expression) was used to identify differentially abundant OTUs and families between treatments in the rDNA data set. Only OTUs that were present in two out of three replicates were used for the ALDEx2 analysis. Prior to analysis, proportional 16S rDNA based OTU abundances were centered log ratio (clr-) transformed using the aldex.clr function in R with 128 Dirichlet instances (Fernandes et al., 2014). Taxa were classified as differentially abundant at an adjusted parametric significant threshold of 0.05 (Benjamini and Hochberg, 1995), and at an unadjusted, non-parametric significance threshold of 0.05. Furthermore, the pairwise average difference and the effect size between the treatments at day 23 compared to the start of the incubation were calculated in the rDNA data set (Fernandes et al., 2014). Taxa were identified as responding to the treatments at an effect size of more than 4 and an average difference between treatments of more than log₂2 or less than log₂0.5, i.e., one duplication or decrease by half in relative abundance. Since only one sample per treatment was available for the rRNA data set, no statistical analyses could be performed. Here, only abundant taxa with a relative sequence abundance of at least 1% (on OTU level) in at least one sample were screened for responses to the carbon treatments, and only taxa exhibiting at least one duplication or a decrease by half after 23 days are reported. On family level, only differentially abundant taxa with relative sequence abundances above 2% in at least one sample will be discussed.

All statistical analyses were conducted in R using the core distribution (version 3.3.0; R Development Core Team, 2014) and the following packages: vegan (Oksanen et al., 2015), ALDEx2 (Fernandes et al., 2014), nlme (Pinheiro et al., 2016), and multcomp (Hothorn et al., 2008).

Data are accessible via the Data Publisher for Earth & Environmental Science PANGAEA¹ (Hoffmann et al., 2016). Raw paired-end sequence, primer-trimmed reads are available on ENA Accession Number: PRJEB17614. The data were archived using the brokerage service of GFBio (Diepenbroek et al., 2014).

Prokaryotic cell numbers in sediments retrieved from the central HAUSGARTEN station were on average 1.4±0.2 × 10⁹ cells (mL sediment)^-1 (n = 3). After 23 days of incubation, cell numbers increased in all treatments, and in the unfed control, with highest cell numbers of about 4 × 10⁹ cells mL^-1 in the CHI treatment (ANOVA, F_1,9 = 152.7, p < 0.05) (Figure 1A; Table 1). Accordingly, bacteria in CHI treatments showed the highest estimated assimilation of carbon into biomass, i.e., 17% of the supplied carbon, while bacteria assimilated around 8% of the supplied carbon in MARC treatments, and ≤5% in the remaining algae treatments (Table 1). Based on the cell numbers, estimated duplication times ranged between 33 days for the control incubation and 12 days for the CHI-treated samples. The unfed control may serve as an approximation of community doubling times in Arctic deep-sea surface sediments, which can thus be estimated to about four weeks. The corresponding biomass calculations represent rough estimates of carbon conversion into bacterial biomass, as we did not quantitatively determine changes in cell volumes.

Chitobiase and beta-glucosidase enzymatic activities increased significantly in all carbon-amended samples, while enzymatic activities in the unfed control remained low (approximately 0.01 μmol mL^-1 d^-1), and were close to starting conditions (chitobiase: ANOVA, F_11,48 = 5,811, p < 0.05; beta-glucosidase: ANOVA, F_11,48= 5,297, p < 0.05; Figures 1B,C). Highest chitobiase activity was measured in the BCLA-fed sediments with 0.25 μmol mL^-1 d^-1, which was 16-fold higher than in the unfed control, followed by CHI amended samples with 0.16 μmol mL^-1 d^-1. The remaining algae-fed communities showed a moderate increase in chitobiase activity (0.10–0.13 μmol mL^-1 d^-1). All carbon treatments were significantly different from the unfed control and from each other (Tukey HSD, p < 0.05; Figure 1B). Highest beta-glucosidase activity was induced in the EHUX treatments with 0.15 μmol mL^-1 d^-1, which represents a 30-fold higher activity compared to the unfed control. Lowest beta-glucosidase activity was induced in the CHI and MARC treatments, turning over only 0.04 μmol mL^-1 d^-1. TWEI and BCLA addition resulted in moderate beta-glucosidase activities. All carbon treatments were significantly different from the unfed control and from each other (Tukey HSD, p < 0.05; Figure 1C), except for beta-glucosidase activity in the CHI and MARC treatments. Calculation of the total hydrolysis potential of both measured enzymes over the duration of the experiment showed that communities in BCLA and EHUX treatments had the highest potentials (52 and 38 μmol C mL^-1 d^-1, respectively), whereas MARC treatments had the lowest (17 μmol C mL^-1 d^-1).

Over the 23 days of incubation, oxygen concentrations declined significantly in all treatments (ANOVA, F_6,28 = 3,381, p < 0.05; Figure 1D; Table 1). Inferred oxygen uptake among the treated sediment slurries ranged between 6.1 μmol O₂ mL^-1 and 4.4 μmol O₂ mL^-1 sediment over 23 days for CHI and MARC treatments, respectively (Table 1). The unfed control showed lowest oxygen uptake of 2.9 μmol O₂ mL^-1 sediment in 23 days. Based on the calculated oxygen uptake, 19 and 9% of the added carbon have been respired at the end of the incubation period for CHI and MARC treatments, respectively (Figure 1D; Table 1). Oxygen uptake calculated for the addition of the planktonic algae and BCLA were equally moderate, with approximately 10% of the added carbon being respired (Table 1).

For both, atmospheric and in situ pressure incubations, an increase in bacterial biomass in control treatments indicated that the presence of sedimentary organic carbon provided sufficient energy for an increase of the bacterial standing stock. This may be partly due to the onset of a settling phytoplankton bloom in the area during the time of sampling (Soltwedel, 2015), but also to the presence of refractory organic material. The initial (in situ) sediment bacterial community (Figure 3) was very similar to previous reports from this site (Jacob et al., 2013), and from other deep-sea surface sediments globally (Zinger et al., 2011; Bienhold et al., 2016; Learman et al., 2016). OTUs affiliating with the gammaproteobacterial families Oceanospirillaceae and Colwelliaceae showed a strong increase in relative sequence abundance in the unfed control treatments after 23 days of incubation, but also in CHI treatments. These groups are known to include polymer degraders (Methé et al., 2005; Weiner et al., 2008). Additionally, the genera Colwellia and Marinomonas sp. (Oceanospirillaceae) have been reported to produce polyhydroxyalkanoate compounds, which serve as intracellular carbon and energy reserves (Methé et al., 2005; Cai et al., 2011). This may be a beneficial strategy for organisms living in organic-poor sediments of the Arctic deep sea.

We also observed a conspicuous increase in relative rRNA sequence abundance of deep-branching acidobacterial groups and taxa belonging to the Sphingomonadaceae in the unfed controls after 23 days of incubation. We speculate that the increased relative rRNA sequence abundance may represent a strategy, where in lack of nutrients, energy is invested in maintaining the metabolism rather than increasing biomass (Morita, 1982; Watson et al., 1998 and references therein). However, this pattern was not reproducible in incubations at 250 atm, and it remains unknown what caused this discrepancy. Due to the lack of fresh organic material in control incubations, bacteria adapted to more oligotrophic conditions and the breakdown of refractory, aged organic material may become more important. Indeed, Sphingomonadaceae are described to play important roles in oligotrophic environments, and in the degradation of recalcitrant polyaromatic compounds (Ohta et al., 2004; García-Romero et al., 2016). In spite of their frequently observed oligotrophic character, members of this family are widespread in nature, occurring in soils, freshwater, and marine habitats (Cavicchioli et al., 1999; Yabuuchi and Kosako, 2005; Vaz-Moreira et al., 2011). Acidobacteria have also been reported to cope well with low organic matter availability (Fierer et al., 2007; Bienhold et al., 2012). These groups may thus be representatives of a background community, surviving in the oligotrophic conditions that prevail at the deep Arctic seafloor for most of the year.

Chitin addition caused strong functional changes in the sediment microbial community, i.e., the largest increase in net biomass (highest bacterial growth efficiency), highest oxygen consumption rates, and a strong increase in chitobiase activity (Figure 1; Table 1). Despite these effects, and even though overall diversity decreased significantly (Table 2), community composition remained very similar to its initial structure and to control incubations (Figure 2; Supplementary Figure S4). The sediment community thus seemed to be well adapted to the degradation of this type of polymeric organic matter. Since chitin is the most abundant biopolymer in the ocean (Gooday, 1990), and bacteria are the main decomposers of this material in the deep-sea benthos (Christiansen and Boetius, 2000; Kanzog et al., 2009), benthic communities may constantly be exposed to relatively high levels of chitin, and our results may therefore reflect an inherent adaptation to the degradation of this carbon source. The decrease in community diversity probably reflects the loss of taxa that were overgrown by chitin degraders.

In contrast to CHI treatments, the addition of algae caused a much more moderate response in bulk community function, including lower estimated bacterial growth efficiencies. This is in line with reports by Mayor et al. (2012), who showed that benthic bacteria displayed lower growth efficiencies when respiring diatoms compared to fecal pellet material, and confirms that resource quality may influence carbon uptake and retention potential in natural settings.

There are several potential explanations for the lower bacterial growth efficiencies in algae treatments. Some of the tested algae may contain compounds that suppress growth of certain bacterial groups (Guedes et al., 2011; Qin et al., 2013). As one example, the production of dimethylsulphopropionate (DMSP)-related substances by EHUX has been suggested to act as protection against grazers (Hansen et al., 1996; Fileman et al., 2002), and may also inhibit bacterial attachment (Saha et al., 2012). Viral lysis might act as another structuring factor (Proctor and Fuhrman, 1990; Fuhrman and Noble, 1995), as viruses are also considered to be favored by high energy input and biological productivity (Cochlan et al., 1993; Bratbak et al., 1994). Furthermore, meiofauna, e.g., nematodes, may be direct or indirect beneficiaries of the fresh phytodetritus, by either feeding on the algae or on the bacteria (Guilini et al., 2010; Ingels et al., 2010). Numerous studies have shown that fauna may be rapid consumers of fresh phytodetritus at the seafloor (e.g., Blair et al., 1996; Moodley et al., 2002, 2005, Witte et al., 2003a), in particular in the Arctic, following the deposition of ice algae or phytoplankton (McMahon et al., 2006; Morata et al., 2011, 2015). To further resolve the substrate degradation cascade, labeled substrates could be used in future experiments (Witte et al., 2003a,b; Guilini et al., 2010).

Compared to CHI treatments, we observed a much stronger decrease in bacterial diversity as well as shifts in community structure in all phytodetritus treatments, indicating the response of a few adapted groups outcompeting others. This suggests that changes in community composition needed to precede changes in community function (Schindler, 1987; Niederlehner and Cairns, 1994). This may be due to the fact that the input of relatively fresh algae used in this study is less common than the ubiquitously available chitin. Also, while chitin is solely composed of N-acetyl-glucosamine subunits, and only a small set of enzymes is necessary to degrade it, the degradation of algae is more complex, involving a cascade of multiple enzymes and bacterial groups (Xing et al., 2015; Teeling et al., 2016).

Changes in community structure and function were very similar between the phytoplankton (TWEI, EHUX) and sea-ice algae (BCLA, MARC) treatments, despite presumable differences in their nutritional value (McMahon et al., 2006; Sun et al., 2007). However, the input dynamics of these organic matter sources in the natural system would differ profoundly. Specifically, a slow rain of phytoplankton cells that are already being degraded on their way through the water column represent a stark contrast to a highly concentrated and localized input of large ice algae aggregates sinking rapidly to the deep seafloor. These scenarios would likely result in differential impacts on community structure, functioning, and carbon turnover.

Previous studies have already identified links between organic matter quantity and bacterial community structure in marine sediments (Franco et al., 2007; Bienhold et al., 2012; Learman et al., 2016). Here, we specifically addressed the response of bacterial communities to different types of organic matter when provided in a single, large pulse. Overall, we identified 12 bacterial families on both rRNA and rDNA level that increased in relative sequence abundance in the organic matter treatments, and that were consistent between incubations under atmospheric and in situ pressure conditions. These opportunistic bacterial groups may represent the specific community fraction that contributed to the increase in enzyme activity, biomass, and oxygen respiration, and that might thus be of key importance for initial polymer degradation at the deep-sea floor. The gammaproteobacterial families Colwelliaceae, Psychromonadaceae, and Oceanospirillaceae were found in highest abundances in CHI treated sediments. Gammaproteobacteria are a globally ubiquitous group in marine environments that have been described as versatile opportunists and copiotrophs (Glöckner et al., 1999; Bienhold et al., 2012). Indeed, representatives of the family Oceanospirillaceae, e.g., the genera Marinomonas and Neptunomonas, have been described to degrade chitin (Nogi et al., 2004; González and Whitman, 2006; Jung, 2006). Also representatives of the families Colwelliaceae and Psychromonadaceae appear to have a broad substrate spectrum including complex organic compounds (Methé et al., 2005; Auman et al., 2006). Several members of the Colwelliaceae have been reported to hydrolyze chitin (Huston et al., 2000, 2004; Ivanova et al., 2004). We therefore identify these bacterial groups as potential opportunists, especially in scenarios of high chitin input, such as mass sedimentation events of crustaceans (Sokolova, 1994; Christiansen and Boetius, 2000).

For Psychromonadaceae, only some species are described as being able to degrade chitin (e.g., strain Psychromonas ingrahamii 37; Riley et al., 2008), whereas other species, e.g., Psychromonas arctica, appear to lack this ability (Groudieva et al., 2003). This supports our findings of a strong increase in relative sequence abundance for diverse OTUs affiliated with Psychromonas in CHI amended slurries on the one hand, and a decrease of other Psychromonas-associated OTUs on the other hand, which instead showed a stronger response in algae treatments (SI text). This deviation is consistent with previous reports for this genus in a similar environment (Bienhold et al., 2012), and demonstrates the usefulness of community information at a high taxonomic resolution. In the future, genomic and physiological studies may reveal in more detail the different metabolic niches that these taxa occupy.

Several other families were identified with most pronounced shifts in the algae treatments. Flavobacteria from Arctic sediments were recently described to exhibit a strong positive correlation with energy availability (Bienhold et al., 2012), consistent with the association of copiotrophy to the phylum Bacteroidetes (Fierer et al., 2007). Also, Flavobacteria and Alteromonadales, in particular Pseudoalteromonadales, have shown strong responses to phytoplankton blooms (Teeling et al., 2016; Xing et al., 2015). Representatives of the order Pseudoalteromonadales, e.g., Pseudoalteromonas haloplanktis, which shares closest sequence identity to the responsive clade in our study, can produce anti-biofilm molecules (Parrilli et al., 2016). It may be able to degrade extracellular polymeric substances produced by phytoplankton and ice algae like MARC by using its diverse set of enzymes, including peptidases, amylase, and alpha-glucosidase (Médigue et al., 2005). In the future, extended monitoring periods after the addition of organic matter, combined with metagenomic and metatranscriptomic approaches, may yield further insights into succession patterns, species competition, and niche adaptation, e.g., as observed for bacterial communities associated with phytoplankton blooms in the water column (Teeling et al., 2012).

Most of the opportunistic bacterial groups identified in this study contain psychrophilic and psychrotolerant cultured representatives (Médigue et al., 2005; Methé et al., 2005; Auman et al., 2006; Riley et al., 2008; Lee et al., 2012) with mostly heterotrophic lifestyles (e.g., Médigue et al., 2005; Methé et al., 2005), the potential for motility (e.g., Médigue et al., 2005; Auman et al., 2006; Lee et al., 2012), and biofilm formation (e.g., Médigue et al., 2005; Methé et al., 2005; Riley et al., 2008; Baek et al., 2015). The latter two features may aid the association with, and directed degradation of particulate organic matter in the deep sea (O‘Toole et al., 2000; Golyshin et al., 2002), but the relevance of such mechanisms remains speculative and will require further investigations. Most of the groups are not specific or exclusive for the sediment environment, but cover a wide range of natural habitats, and appear to have a rather cosmopolitan character (e.g., Médigue et al., 2005; Methé et al., 2005; Zhao et al., 2005; Auman et al., 2006; Xing et al., 2015). Nevertheless, many of the closely related cultured representatives were isolated from polar marine environments, and are described to produce cold-adapted enzymes, capable of degrading high-molecular-weight organic compounds, e.g., Alteromonadales (Médigue et al., 2005), Colwelliaceae (Methé et al., 2005), Shewanellaceae (Zhao et al., 2010), Psychromonadaceae (Riley et al., 2008), Moritellaceae (Lee et al., 2012; Malecki et al., 2013), Flavobacteriaceae (Pati et al., 2011; Baek et al., 2015; Xing et al., 2015), and Oceanospirillaceae (Weiner et al., 2008). Future research may therefore identify explicit Arctic and/or sediment ecotypes within these groups, which need to be further metabolically characterized using genomic and physiological methods, in order to assess their role in the turnover of organic matter in Arctic marine sediments.