Previous acoustic seafloor surveys identified several 10s of SMVs in the northwestern Pacific Ocean at water depths of 1200–3500 m off Tanegashima Island, Japan. For one of these, geochemical information, including stable isotopic compositions of methane and water (i.e., interstitial water in sediment) is available (Nakayama et al., 2010), whereas there are as yet no microbiological studies. In 2015, we investigated the distribution and diversity of microbial communities inhabiting two of the SMVs in this area: MV#1 and MV#14 (Figure 1). MV#1 has a diameter of about 3000 m, a height of 300 m, and is situated at a water depth of 1419 m. MV#14 has a diameter of about 3000 m, a height of 250 m, and is situated at a water depth of 1691 m. Intensive seafloor surveys around MV#1 using the autonomous underwater vehicle Urashima and remotely operated vehicle Hyper-Dolphin (Japan Agency for Marine-Earth Science Technology [JAMSTEC]) showed well-preserved mud-flow channels suggestive of a recent eruption.

Sediment and seawater samples were obtained at two stations, VTF2 (MV#1) and VTF5 (MV#14), during cruise KH-15-2 of R/V Hakuho Maru in 2015. Sediment samples were collected using the Navigable Sampling System (Atmosphere and Ocean Research Institute, The University of Tokyo). At each station, hydrographic parameters were recorded with a CTD-carousel system (SBE 9, Sea-Bird Scientific, Bellevue, WA, United States), which was deployed with simultaneous water sampling using Niskin bottles (General Oceanics, Inc., Miami, FL, United States). For comparison in the DNA analysis, we used stratified sediment core samples from the Peru Margin, Juan de Fuca ridge flank, and offshore of the Shimokita Peninsula, Japan, which were collected during Ocean Drilling Program (ODP) Leg 201 in 2002, Integrated Ocean Drilling Program (IODP) Expedition 301 in 2004, and shakedown cruise CK06-06 of JAMSTEC’s deep-sea drilling vessel (D/V) Chikyu in 2006, respectively (Table 1). For DNA extraction, all SMV sediments and the stratified sediments were frozen immediately after sampling and stored at -80°C until use. For cell count of the SMVs sediment, 2 g of sediment was collected by using a tip-cut syringe and transferred to a 15-ml falcon tube. Subsequently, 8 ml of 4% paraformaldehyde was added to the sediment for fixation. After overnight fixation at 4°C, the fixed sediment was washed twice with PBS buffer and stored in 50% PBS/ethanol at -20°C.

After core recovery, wet sediment was immediately sampled at 20- to 50-cm intervals and pore water samples were extracted onboard with a stainless-steel squeezer (Manheim and Sayles, 1974). The sulfate content of pore water was analyzed using an ion chromatograph (Dionex ICS-500) with an isocratic carbonate/bicarbonate eluent coupled with suppressed conductivity detection. Analytical precision was estimated to be within 0.4% from repeated measurements of the same samples.

Sediment samples for the determination of methane concentrations were collected using tip-cut plastic syringes immediately after core recovery. The syringe end was tightly capped with a silicone-rubber stopper. The sediment sample was transferred from the syringe into a glass vial in a N₂-flushed glove bag, and then the vial was capped with a Teflon-coated rubber septum and aluminum seal for subsequent geochemical analyses. The glass vials were frozen at -20°C and stored until analysis. In a laboratory on shore, the glass vials were heated in an oven at 60°C for 30 min. From the glass vials, 0.1–0.5 mL of the gas phase was extracted with a gas-tight syringe and introduced into a gas chromatograph-isotope ratio monitoring mass spectrometer (Tsunogai et al., 2002). The methane content was calculated by comparing the ⁴⁴CO₂ output of the isotope-ratio-monitoring mass spectrometer with those measured during analyses of a working-standard gas containing 5000 ppm methane in He. The standard deviation of repeated analysis of the laboratory standard gas was <4%.

For quantification of methane in seawater, two replicate samples were collected from each Niskin bottle: two 120-ml serum vials were filled to overflowing after flushing with approximately two vial-volumes to avoid air contamination. The seawater samples were fixed by the addition of 500 μl saturated HgCl₂ solution to inhibit microbial activity during storage, sealed with gray butyl-rubber septa and aluminum crimp-caps, and then stored in the dark at 4°C until analysis in the laboratory at the University of the Ryukyus (Okinawa, Japan).

Dissolved methane was analyzed by purge–trap gas chromatography (GC) (GC-2014, Shimadzu) equipped with a flame ionization detector. Hydrocarbons were chromatographically separated on a Porapak Q packed column (80–100 mesh; 2-m length) operated at 40°C using ultra-high-purity He as the carrier gas (Air Liquide, purity 99.9999%). The system was calibrated using a gas standard with a certified methane concentration (9.72 ppm) provided by Japan Fine Products (Kawasaki, Japan). The precision of the method was 10%, as estimated from the standard deviation of replicate analysis of water samples (n = 4).

Cell concentrations in sediment samples were determined using a previously described procedure (Morono et al., 2013). Briefly, after washing fixed sediment with a NaCl solution, a mixture of detergent and methanol was added and the sediment slurry was shaken for 60 min at 500 rpm using a Shake Master (Bio Medical Science, Tokyo, Japan). The slurry was then sonicated and carefully layered onto a high-density cushion consisting of multiple density layers of Nycodenz and sodium polytungstate solutions. Following centrifugation at 10,000 × g for 60 min, the supernatant was carefully transferred to a new vial. The detergent and methanol mixture was added to the remaining pellet, and the mixture was again sonicated and centrifuged with the high-density cushion solution. The supernatants were pooled and the combined recovered suspension was filtered onto a polycarbonate membrane. For seawater samples, 300 mL of formalin-fixed seawater was directly filtered onto a polycarbonate membrane. The number of microbial cells stained by SYBR Green I was determined either using an automated cell counting system with analysis of acquired images using MetaMorph software (Molecular Devices, Sunnyvale, CA, United States) (Morono and Inagaki, 2010), or counted manually under fluorescence microscopy.

DNA was extracted from 5 g wet sediment or 300 mL seawater using the PowerMax Soil DNA Isolation Kit (Qiagen, Tokyo, Japan) according to manufacturer’s instructions. For the seawater samples, soon after recovery of the Niskin bottles 300 mL seawater was filtered using a disposable MicroFunnel 300 Filter Unit with 0.2-μm pore size (Pall Corporation, Tokyo, Japan) and stored at -20°C prior to extraction. For a negative control, 5 mL Milli-Q water (Millipore) was used for extraction in place of sediment or seawater. The extracted DNA solution (approximately 5 mL) from sediment was further concentrated by ethanol precipitation. To increase the DNA yield, 10 μL of linear acrylamide (Thermo Fisher Scientific, Kanagawa, Japan) was added before centrifugation. The DNA precipitate was dissolved in TE buffer (10 mM Tris-HCl, 1.0 mM EDTA; pH 8.0) and stored at -20°C until use.

The abundance 16S rRNA genes was quantified by microfluidic digital PCR (dPCR) using the BioMark Real-time System and 12.765 Digital Array (Fluidigm Corporation, South San Francisco, CA, United States) as described elsewhere (Hoshino and Inagaki, 2012). For each sample, 6 μL of PCR mixture was prepared as follows: 3.0 μL of 2× MightyAmp buffer (TaKaRa Bio, Shiga, Japan), 0.15 μL of Binding Dye sample loading reagent (Fluidigm), 0.12 μL of MightyAmp DNA polymerase (TaKaRa Bio), 0.3 μL of EvaGreen (Biotum, Fremont, CA, United States), 0.015 μL of Rox dye, 0.15 μL each of forward and reverse primer, 1.0 μL of extracted DNA, and 1.115 μL of water. We used the domain-specific primers B27F-B357R and A806F-A958R (Table 2) for the amplification of bacterial and archaeal 16S rRNA genes, respectively.

The V3–V4 hyper-variable region of the 16S rRNA gene, which is often used for analyzing microbial diversity and community structure, was amplified by PCR using universal primers U515F/U806R (Table 2). The 25-μl PCR mixture consists of 0.3 μM of each primer, 1 μl of the template DNA, 1× MightyAmp Buffer Ver.2, and 0.5 μl of MightyAmp DNA polymerase (TaKaRa Bio). The PCR started with 2 min at 98°C followed by 30 cycles of denaturation at 98°C for 10 s, annealing at 55°C for 15 s, and elongation at 68°C for 30 s. After purification of the desired PCR products by agarose gel electrophoresis, index and adapter were added to the purified product during the eight cycles of second-round PCR using KAPA HiFi HotStart Ready mix (Kapa Biosystems) with 250 pg of the purified PCR product as described elsewhere (Caporaso et al., 2012). The indexed PCR products were purified twice by AMPure XP (Beckman Coulter), and then the PCR product was sequenced by the MiSeq platform with MiSeq Reagent Kit v3, 600 cycles (Illumina, San Diego, CA, United States). All the purified PCR products were sequenced in a sequencing run of Miseq and account for two fifth of total clusters. Raw sequences were demultiplexed, and the sequence quality was filtered using the BaseSpace app (Illumina). Further processing, including primer trimming, sequence length screening, taxonomic identification, clustering and diversity analysis, were performed using the ‘mothur’ software package (v. 1.35.0) (Schloss et al., 2009) and USEARCH 64-bit version¹. After the screening, the average number of sequence reads per sample was 54,410 (Supplementary Table 2). The operational taxonomic units (OTUs) obtained were used for non-metric multidimensional scaling (nMDS) analysis using the ‘vegan’ package of R software as described elsewhere (Dixon, 2003; Doi and Okamura, 2011) for beta-diversity analysis of microbial communities. To exclude confounding results due to potential contaminants from the extraction process and chemicals used for PCR and sequencing, the top 10 families in the negative controls, and previously reported core-human microbiomes (Li et al., 2013) were arbitrary removed (listed in Supplementary Table 1). The removed top 10 families are generally consistent with laboratory and reagent contaminants in microbiome analyses (Salter et al., 2014). We performed all the analyses in this study without removing those potential contaminant and confirmed that removal of those did not significantly affect the results (Supplementary Figures 1, 2 and Table 2).

To detect and visualize atribacterial cells in microbial communities in the MVs, we designed a new oligonucleotide probe Atri578 (Table 2), which can be used to detect a broad range of atribacterial sequences in the ARB Silva Database by using ARB software and mathFISH (Ludwig et al., 2004; Yilmaz et al., 2011). The coverage of Atri578 to atribacterial sequences determined by TestProbe3.0 with the SILVA Database SSU128 is 70.6% and only two sequences in the outgroup are no-mismtach to Atri578. The horseradish peroxidase (HRP)-labeled Atri578 probe was purchased from Biomers GmbH (Ulm, Germany). The optimal formamide concentration in hybridization buffer was determined by clone-FISH (Schramm et al., 2002) using a synthesized vector (Eurofins Genomics, Tokyo, Japan) containing the full-length 16S rRNA gene sequence of Atribacteria bacterium SCGC AAA255-N14 (ASPC01000002). We tested formamide concentrations of 0% to 60% at 10% intervals, at 35°C and found 10% to be the optimal formamide concentration. Using this concentration, Atribacteria on the membrane filters were detected by CARD-FISH for the water column samples. The sediment samples from SMVs had very low cell number and the number of cells on the filter membrane after cell separation described above was not enough to detect Atribacteria.

Catalyzed reporter deposition-fluorescence in situ hybridization was performed according to a previously described protocol (Hoshino et al., 2008, 2015). Briefly, hybridization was performed in hybridization buffer containing 10% formamide and 0.5 μM of the HRP-labeled Atri578 probe at 35°C for 2 h. After washing away excess probe, the membrane filter was incubated with Alexa555- or Alexa488-labeled tyramide (Thermo Fisher Scientific) at 35°C for 30 min for signal deposition, followed by counterstaining with SYBR Green I or DAPI.

Sequence data obtained in this study have been submitted to the DDBJ database under accession numbers DRA005491 and DRA005492.

Sulfate concentrations in sediment pore water from both SMVs showed maxima around 25–30 mM in the uppermost sediment, with concentrations rapidly decreasing with depth (Figure 2). In sediments deeper than 1 m below the seafloor (mbsf), sulfate was almost depleted at both SMVs. The concentration of methane was generally higher in pore water from MV#1 than MV#14, ranging from 0.5 to 13 mM and from 0.06 to 5.1 mM, respectively. Although measured concentrations varied widely probably because of depressurization of sediment, methane was depleted near the sediment surface at around 1 mbsf. The depth profiles of sulfate and methane indicate that the SMTZ was located at around 1 mbsf at both SMVs. In the overlying water column, the concentration of methane peaked to 0.71 nM at 29 m above MV#1 and to 1.93 nM at 70 m above MV#14. Those peaks indicate the occurrence of methane plume and thus advective flow from the SMVs because methane is immediately oxidized in water column and its concentration almost never reached to that high in normal water column at the similar depth (Toki et al., 2016).

Cell densities in seawater, based on counts of cells stained with SYBR Green I, were estimated to be in the range of 4.7 × 10⁴ to 1.4 × 10⁵ cells/ml at MV#1 and MV#14, respectively (Figure 3). The gene abundance of 16S rRNA genes in seawater, as sum of archaeal and bacterial 16S rRNA gene (Supplementary Figure 3) quantified by microfluidic dPCR, were overall slightly lower than the cell densities. Cell counts and gene abundance from the seawater above the SMVs were relatively constant throughout the water column at 5 × 10³ copies/ml and 6 × 10⁴ cells/ml (Top panel of Figure 3) up to around 400 m above the summit, the upper limit of sample collection.

The cell concentrations in the shallowest sediment samples from the seafloor were 2.4 × 10² and 1.3 × 10⁵ cells/g-wet sediment for MV#1 and MV#14, respectively. Cell concentrations at the two SMVs dropped sharply to only 20 and 13 cells/g-wet sediment at 0.58 and 0.51 mbsf, respectively. The 16S rRNA gene abundance for MV#1 in the uppermost sediment sample was 3.3 × 10³ copies/g-wet sediment, which is an order of magnitude higher than the cell concentration, whereas the cell counts in the uppermost sediment at MV#14 were higher than the 16S rRNA gene abundance (8.0 × 10³ copies/g-wet sediment). The 16S rRNA gene abundance in sediment from depths greater than 1.5 mbsf were 10–100 times the cell concentrations.

Throughout the sediment cores, MV#1 generally had lower 16S rRNA gene abundance than MV#14, and the gene abundance at MV#1 decreased to about 1 × 10² copies at 1.1 mbsf. At MV#14, the 16S rRNA gene abundance gradually decreased with increasing depth to 4.7 × 10² copies/g-wet sediment at 3.0 mbsf, the deepest sediment sampled. The estimated 16S rRNA gene abundance as well as cell concentrations in the deeper mud volcano sediments were markedly lower than those in shallow seafloor sediments and in other stratified subseafloor sediments on Pacific margins collected during previous IODP expeditions. In general, the abundance of 16S rRNA genes in the SMVs varied from 10¹ to 10⁵ copies/g-wet sediment, which is two to seven orders of magnitude lower than in stratified sedimentary environments (Figure 3).

The cell counts are higher than the gene abundance in the water column in this study, whereas in the sediment the trend was opposite to the water column. The reasons of formar discrepancy can be partially explained by biases as low DNA extraction efficiency or false positive counts of fluorescence-emitting particles (Lloyd et al., 2013). The latter might be caused by low recovery of cells from very sticky texture of the MV sediment. In any case, the results from both methods are consistent with the general trends; (1) Higher microbial abundance in the overlying seawater than mud volcano sediment: (2) Decrease of microbial abundance with sediment depth.

We obtained around 1.6 million sequences in total by Miseq sequencing of SMV sediment and overlying seawater samples (Supplementary Table 2). Phylum-level classification showed the predominance of Proteobacteria in all seawater samples, comprising more than 50% in all sequence libraries (Figure 4). The taxonomic composition of microbial communities in seawater was generally constant at both sites and at different depths. The second most representative phylum was Marinimicrobia (formerly SAR405), comprising 10–14% of the total community, followed by Thaumarchaeota comprising 5–15%.

In contrast, the community composition of mud volcano sediments was markedly different from that in the water column, even at the phylum level; for example, Marinimicrobia was nearly absent in sediment samples. The MV community composition was also distinct from those of the reference stratified sediment (Supplementary Figure 4). In addition, the community compositions at the two mud volcano sites were notably different: Proteobacteria was dominant at MV#1, comprising between 21 and 58% of the total, whereas Atribacteria (formerly OP9 and JS1) comprised 21–88 and 0.7–39% of total sequence reads at MV#14 and MV#1, respectively. Some phyla, such as Firmicutes, Aerophobetes (formerly CD12), and Aminicenantes were prevalent only in the sediment samples.

The degree of similarity between microbial communities in sediment and seawater was determined by nMDS analysis applied to OTUs with >97% similarity (Figure 5). Microbial communities in the overlying seawater from 200 to 300 m above the mud volcano summits were generally similar. Compared to the seawater communities, microbial communities in sediments were more diverse, even across very small changes in depth at 2 and 6 mbsf for the SMVs and reference stratified marginal sediments. Microbial communities in stratified sediments were grouped into one cluster despite being collected from geographically distinct sites, whereas the microbial communities in SMV sediments were different from those in both the stratified sediments and in the overlying water column. A gradation in microbial composition is evident from deep to shallow sediment within both SMVs. These trends suggest that physical and chemical conditions may affect microbial composition more than can be explained by the conditions in well-stratified sediment. The nMDS plots (Figure 5) also indicate that shallower SMV sediment communities, especially at MV#14, are similar to those in normal stratified marine sediments. This difference might suggest differences in activity of the SMVs: if MV#1 is more active than MV#14, the seafloor at MV#14 would be covered with marine sediments freshly deposited since the last eruption; consequently, shallower sediments at MV#14 would harbor the microbial communities that typically occur in the surrounding stratified sediments.

We identified the common OTUs in all SMV sediment and seawater samples to determine whether mud volcano activity dispersed some of the community members of the subseafloor sedimentary habitats at SMVs to the overlying seawater. We found only one OTU (i.e., OTU-1) common to all of the samples examined for both sites (MV#1 and MV#14). OTU-1 is the most abundant OTU in the SMV sediment samples, comprising 1.4–38 and 22–85% of the sequence reads from MV#1 and MV#14, respectively. OTU-1 sequences are affiliated with Atribacteria, which is well-known as a predominant taxonomic clade of marine sedimentary bacteria (e.g., Inagaki et al., 2006; Webster et al., 2006b; Blazejak and Schippers, 2010); almost all Atribacteria-related sequences detected in this study are included in this OTU (Supplementary Figure 5). The relative abundance of OTU-1-related sequences was generally low in seawater samples, comprising less than 0.1% of total reads. This abundance peaked, however, to more than 0.7% at 39 m (MV#1) and 52 m (MV#14) above the summits of both SMVs and returned to around 0.1% above this depth (Figure 6). Interestingly, the depth profile of methane similarly peaked at 19 m (MV#1) and 72 m (MV#14) above the summit, indicating the presence of Atribacteria in methane plumes emanating from the SMVs.

Because the relative abundance of OTU-1 in sequence libraries are not necessarily quantitative, we verified the presence and abundance of Atribacteria in seawater samples by CARD-FISH using the newly designed probe Atri578, which specifically targets the Atribacteria-related sequences detected in this study and related sequences in publicly available databases (Figure 7). CARD-FISH analysis showed that Atribacteria in the water column were most abundant at 39 m (MV#1) and 52 m (MV#14) above the respective summits; results overall in good agreement with the relative abundance of OTU-1 sequence. Consequently, we infer that Atribacteria inhabiting the subseafloor sediments at SMVs are dispersed into the overlying seawater along with the methane plume.