Microbial community composition was evaluated within 88 total samples from the Mariana and Kermadec trenches. This included 72 samples belonging to 14 intact, depth-fractioned cores, 3 from the Mariana Trench at depths of 6,844–7,942 m and 11 from the Kermadec Trench ranging from 6,011 to 9,177 m (Supplementary Table 1). Kermadec Trench samples were collected on the R/V Thompson from April to May 2014, and Mariana Trench samples were collected during a cruise on the R/V Falkor from November to December 2014. Bathymetry was obtained from NOAA (Amante and Eakins, 2009) and plotted using the R package marmap (Pante and Simon-Bouhet, 2013). Sediment in the Kermadec Trench was collected using push cores with the HROV Nereus (Fletcher et al., 2009). Samples in the Mariana Trench were recovered using untethered free-falling/ascending landers (Free Vehicle Coring Respirometer (FVCR) and Rock Grabber (Schmidt Ocean Institute, https://schmidtocean.org/technology/elevators-landers/)). Large-diameter (10 cm) megacores were inserted into the seabed by the FVCR using an oil-compensated motor at a steady and slow speed 2 h after landing (Nunally et al., in prep). Megacore tubes were visibly inspected using a task camera on the FVCR to assure sediment integrity was maintained during coring operations. The Rock Grabber lander collected sediments using a Van-Veen Grab. After shipboard recovery, samples were immediately moved to a 4°C cold room to minimize the effects of temperature stress.

Sediment samples were depth fractioned from 0–1, 1–2, 2–5, and 5–10 cm in the Kermadec Trench and down to 10 cm at one-centimeter increments in the Mariana Trench. For samples collected using the Rock Grabber, subsamples were obtained using a sterile syringe inserted up to 10-cm depth, after which subsamples from the top 2–3 cm were extruded into KAPAK bags (Komplete Packaging, Grand Prairie, TX) and homogenized. Samples for DNA extraction were then frozen at −80°C. To determine the effects of long-term pressurization on microbial communities, ~5-g samples were also incubated at in situ pressure in KAPAK bags without amendment for 1 (Mariana) or 1.5 (Kermadec) years. Rock Grabber and other samples not part of intact cores were not included in our analysis of the in situ community as these samples were likely disturbed and mixed with deeper sediment layers and overlying water during ascent. Therefore, they were only used for culturing and experiments under long-term high-hydrostatic pressure conditions.

Sediment (5-g wet-weight) samples from either frozen samples or long-term in situ high-hydrostatic pressure incubations were used for DNA extraction. DNA was extracted using a modified version of Lysis Protocol II described by Lever et al. (2015). 2.5 V of lysis solution (30 mM EDTA, 30 mM Tris–HCl, 800 mM guanidine hydrochloride, 0.5% Triton X-100, final pH 10) and 500 μmol pyrophosphate was added to each sample and the mixture briefly vortexed. Samples were then subjected to two 15-minute freeze–thaw cycles at −80°C, followed by incubation at 50°C with shaking at 150 rpm for 1 h. Samples were centrifuged and the supernatant was treated twice with 1V 24:1 chloroform:isoamyl alcohol. DNA was precipitated using 5 M NaCl and 70% ethanol for 2 h at room temperature and resuspended in nuclease-free water. Extracted DNA was cleaned again using a Quick-gDNA MiniPrep kit (Zymo Research, Irvine, CA). Extraction blanks, consisting of all reagents but no sediment material, were performed in concomitance with every extraction.

Sequencing and processing were conducted as previously described (Peoples et al., 2018a). Briefly, the V4-V5 16S rRNA gene region between 515f-926R was amplified (Parada et al., 2015) for 30 cycles and tagged with Illumina barcodes using a secondary PCR procedure. Samples were pooled at equimolar concentrations and sent for sequencing on an Illumina Miseq at the Institute for Genomic Medicine Genomics Center (University of California, San Diego, La Jolla, CA). Paired-end reads were assembled using FLASH (Magoc and Salzberg, 2011) and OTUs picked at 97% similarity using Uclust in QIIME 1.9.1 (Caporaso et al., 2010). Chimeras were identified and removed using VSEARCH (Rognes et al., 2016) and taxonomy assigned against the SILVA 123 release database (Quast et al., 2012). Sequences found within sequenced extraction blanks and which showed similarity to known contaminants were discarded. The exclusion of contaminants was done manually to avoid removing any autochthonous taxa that may have resulted from cross-contamination between samples and controls. Removed genera included Acinetobacter, Stenotrophomonas, Methylobacterium, Achromobacter, Rhizobium, Tetragenococcus, Escherichia-Shigella, Ferribacterium, Massilia, Curvibacter, Ralstonia, Variovorax, Deinococcus, Anoxybacillus, Aeribacillus, Brevundimonas, Enterobacter, Streptococcus, Burkholderia, Staphylococcus, Sphingomonas, Sphingobium, Sphingobacterium, Bradyrhizobium, Paenibacillus, and the family Comamonadaceae. Raw sequence data can be accessed at the Sequence Read Archive under the Biosample Accession numbers SAMN07732241-SAMN07732328.

Sequencing reads were processed with the R package phyloseq (McMurdie and Holmes, 2013). OTUs were removed if they did not have an abundance of at least three reads in at least four samples across the entire dataset. Samples were rarefied to even sampling depth to account for differing sequencing depths between samples. Alpha diversity was calculated using the package vegan (Oksanen et al., 2017) and comparisons between samples were performed using the beta-diversity metric Bray-Curtis in phyloseq. Permutational analysis of variance with adonis was used to determine if trench of collection, sediment horizon depth, and water column depth significantly explained community variation within and between trenches. DESeq2 (Love et al., 2014) was used to identify taxa that were differentially abundant between trench of collection, sediment horizon depth (0–5 or 5–10 cm), water column depth (Kermadec only; 6,000 and 7,000 m vs. 8,000 and 9,000 m), and high-pressure comparison (samples maintained under long-term high hydrostatic pressures and the same samples extracted immediately). DESeq2 analyses were performed on the un-rarefied dataset (McMurdie and Holmes, 2014) with low-abundance (at least <500 total reads per OTU or < 1,000 reads per class or phylum) reads removed. For phylogenetic analysis, representative sequences were aligned with the SINA Aligner (Pruesse et al., 2012; https://www.arb-silva.de/aligner/) and a phylogenetic tree built with FastTree (Price et al., 2010). Trees were visualized with the Interactive Tree of Life (Letunic and Bork, 2007) and GGTREE (Yu et al., 2017). For calculating OTU and sequence percentages shared between trenches, samples were, in order, (1) rarefied to even sequence depth across samples, (2) pooled by trench of collection, and (3) the pooled groups rarefied again to account for different numbers of samples and sequencing depths between trenches. Because of the possibility of losing rare OTUs through pruning and rarefaction (McMurdie and Holmes, 2014), an analysis was also performed where low-abundance OTUs were not removed as described above (at least three reads in at least four samples) but instead only had to be in abundances greater than three reads across the entire dataset. Shared OTU and sequence percentages were then calculated by pooling samples within each trench and rarefying the combined trench communities to even total sequencing depth. This analysis therefore maintained rare OTUs and did not discard large numbers of sequences through rarefaction of each individual sample, but ignored the influence of certain samples having higher sequencing depths than others. For calculating shared OTU percentages between cores, samples were, in order, (1) rarefied to even sampling depth, (2) combined within 0–5 and 5–10 cm groupings within the same core of collection, (3) rarefied to even sampling depth again, and (4) combined by core. This was done to account for different sediment horizon depths, number of samples, and sequencing depth from influencing the abundances of OTUs, although at the cost of losing sequencing depth through rarefaction.

Bacteria were cultured from sediment following plating onto agar or inoculation into pressurizable bulbs (Yayanos, 2001) and incubated at 4°C. Sediment samples from the top 1 cm of sediment, or a homogenized sample from the Rock Grabber, were diluted into 0.1-μm filtered trench water. Samples from the Kermadec Trench were incubated in 2216 Zobell Marine Medium (BD Difco™), A1 Medium (5 g potato starch, 2 g yeast extract, 1 g peptone, 0.5 g NH₄Cl, 0.01 g L-methionine, 500 ml 0.2 μm-filtered autoclaved seawater collected from the Ellen Browning Scripps Pier), or a seawater minimal medium (10 mM NH₄Cl, 14 mg Na₂HPO₄, American Type Culture Collection (ATCC) vitamin supplement (ATCC, Manassas, VA), ATCC trace mineral supplement, 1 L 0.2 μm-filtered Scripps Pier seawater), while samples from the Mariana Trench were inoculated in 2216 Marine Medium only. For samples to be incubated at high pressure, the media described above was inoculated, mixed with gelatin (pH 7 in sterile seawater) to a final concentration of 4% (Yayanos, 2001), transferred into polyethylene transfer pipettes (Samco, Thermo Fisher Scientific), heat sealed, and placed inside pressure vessels and incubated at the desired pressure at 4°C. Kermadec Trench samples were incubated at 100 MPa while those from the Mariana Trench were incubated at in situ pressure (40–110 MPa). After approximately 2 months, colony-forming units (CFUs) were calculated and representative isolates selected for 16S rRNA identification. Colonies arising at high pressure in bulbs were picked using a sterile syringe needle, inoculated into liquid medium, and grown until turbid at the original pressure of incubation at 4°C. Selected colonies were boiled and lysate used as template for PCR using the primers 27F and 1492R (Lane, 1991). Partial sequenced reads of at least 500 bp were identified using the SILVA database and compared to known sequences within NCBI.

Hadal trenches are geographically isolated and can have different overlying primary productivity and water mass inputs that may lead to variations in community composition. Therefore, we evaluated the microbial communities within sediments from the Kermadec and Mariana trenches using high-throughput 16S rRNA gene sequencing of 88 samples. Seventy-two of these were from 14 pristine depth-fractioned cores, while 16 were from potentially disturbed sediments and were not included in the comparative analysis of the in situ communities here (Supplementary Figure 1, Supplementary Table 1). Rarefaction resulted in 7,231 sequences per sample and 6,317 total operational taxonomic units (OTUs) at 97% 16S rRNA V4-V5 gene sequence similarity. Phyla with high relative sequence abundances within the sediment samples included the Proteobacteria, Thaumarchaeota, Bacteroidetes, Planctomycetes, and Chloroflexi (Figure 1). While the majority of OTUs were seen in only a few samples each, 27 OTUs were present in all samples in both trenches (Figure 2). These shared OTUs included members related to the Marine Group I (MGI) within the Thaumarchaeota, the JTB255 clade and the genus Marinicella within the Gammaproteobacteria, the Rhodobacterales and Rhodospirillales within the Alphaproteobacteria, the class JTB23 within the Proteobacteria, and BD2–11 within the Gemmatimonadetes (Figure 2, Supplementary Table 2). On average, these OTUs made up 23 ± 4.5% (range 13.7–35.9%) of the total reads in each community and were closely related to sequences from previous studies of abyssal and hadal sediments (Supplementary Figure 2).

Sediment community diversity was structured by trench of collection (Figures 3A,B, Supplementary Figure 4; PERMANOVA, p < 0.001, R² = 0.23). Fifty-eight percent of OTUs, representing 95% of all sequences, were seen in both trenches (Figure 3C). When the only criterion for maintaining reads was that they had an abundance of at least three across the entire sediment dataset and samples were not rarefied, 40% of the OTUs were shared and still represented greater than 90% of all sequences. More OTUs from any given core could be found within cores from the same trench than cores from the other trench (Figure 3D; minimum found within a comparison core 32%, maximum 58%). Within both trenches assemblages varied by sediment horizon depth (Figure 3B; Kermadec, p < 0.001, R² = 0.16; Mariana, p < 0.001, R² = 0.21). Community richness was lower in deeper sediment horizons (Supplementary Figure 3). Community composition also varied by sampling site within each trench (Kermadec, p < 0.001, R² = 0.25; Mariana, p < 0.001, R² = 0.18) and water column depth (Kermadec, p < 0.001, R² = 0.15; Mariana, p < 0.001, R² = 0.15). Overall these parameters explained half of the variability within each trench.

Taxa that showed the largest differences in relative sequence abundances between trenches, sediment horizon depths, and water depths were identified. The Kermadec Trench was enriched in the phyla Bacteroidetes, Hydrogenedentes, Planctomycetes, and Proteobacteria, while the Mariana had higher abundances of the Marinimicrobia, Thaumarchaeota, Woesearchaeota, and Chloroflexi (Figure 4A, Supplementary Figure 5). The OTUs that varied the most between the Mariana and Kermadec trenches belonged to the MGI Thaumarchaeota, with specific OTUs showing enrichment in the Mariana Trench (Supplementary Figure 5). An OTU related to the genus Aquibacter reached abundances up to 11% of certain sediment horizons and was enriched in the Kermadec Trench. This Aquibacter OTU is identical to that which was found to be more abundant in the Kermadec Trench pelagic community than in the Mariana waters (Peoples et al., 2018a) and likely represents a new, hadal-adapted species (Peoples et al., 2018b).

Comparisons of surficial (0–5 cm) and deep (5–10 cm) sediments showed that taxa belonging to Woesearchaeota, Marinimicrobia, and Chloroflexi, including members of the clades S085, JG30-KF-CM66, SAR202, and Anaerolineae, were enriched in the deeper sediments (Figure 5A, Supplementary Figure 5, Supplementary Table 3). Phylogenetic analysis of Marinimicrobia OTUs enriched in the deeper sediment horizons showed they are related to sequences predominantly affiliated with deep-ocean sediments (Supplementary Figure 6). In contrast, the shallower sediments were enriched in classes belonging to the Verrucomicrobia, Planctomycetes and Nitrospirae, and OTUs belonging to the Proteobacteria, including Colwellia, and Bacteroidetes. MGI Thaumarchaeota showed strong variation by sediment horizon depth, with certain OTUs reaching up to 45% of some sediment horizon depths. Phylogenetic analysis indicated differentially abundant OTUs belonged to distinct clades, with members of the MGI alpha (α) subgroup more abundant in the upper sediments, while those belonging to the eta (η), upsilon (υ), and zeta (ζ) subgroups were enriched in deeper sediments (Figure 5B).

Sediment community composition also varied with water depth. Within the Kermadec Trench, higher percentages of OTUs within a given core could be identified within other cores from similar water depths (Figure 3D). When comparing cores as a function of water depth in the Kermadec Trench, the Chlorobi, Nitrospirae, and Actinobacteria were more abundant in the 8- and 9-km samples, while the Chloroflexi, Acidobacteria, and uncharacterized lineages were enriched at 6 and 7 km (Figure 4B). Cyanobacteria, which could be deposited through topographical funneling, were present, but not found at high abundances regardless of depth. However, an OTU related to the ML635J-21 group of Cyanobacteria was identified at relative sequence abundances of less than 0.1%. This OTU showed highest similarity to taxa from the Japan Trench (99%) and deep-ocean sediments from the Okinawa Trough (99%), with the next closest sequences being only 94% similar.

Bacteria were isolated under both low and high hydrostatic pressures to determine the diversity of culturable isolates. Fifty isolates obtained at high pressures under the facultatively anoxic conditions that develop in bulbs included members of the genera Colwellia, Shewanella, Moritella, and Psychromonas (Figure 6A, Supplementary Table 4). Isolates were also obtained from the genus Psychrobium within the Gammaproteobacteria, Arcobacter within the Epsilonproteobacteria, and a member of the Flavobacteriaceae, although none of these isolates survived cryopreservation or repeated subculturing. Qualitatively, the number of isolates obtained per sample was higher from the Kermadec than the Mariana Trench. This is consistent with the pristine sediment community sequence data where OTUs related to previously culturable piezophilic taxa represented ~0.20% of Kermadec samples but less than 0.05% in the Mariana Trench (Supplementary Figure 7; t-test, sediment only, p < 0.115; sediment and water from Peoples et al., 2018a, p < 0.001). One sample from the Kermadec Trench had relative sequence abundances of these taxa in excess of 4.8%. Relative sequence abundances of taxa related to culturable piezophiles were highest in the Kermadec Trench 0–1-cm fraction and lower in deeper sediments (t-test, Kermadec Trench, p < 0.003). No trend was seen among the altogether low relative sequence abundances present in the Mariana samples. Isolates obtained at atmospheric pressure on plates, rather than at high pressure, were related to the genera Pseudoalteromonas, Pseudomonas, Halomonas, and Shewanella. On average in both trenches OTUs related to these genera made up 0.18% relative sequence abundances within each sample.

Sediment samples were maintained at 4°C under static, unamended, in situ pressures for longer than 1 year (hereafter called “long-term” samples) to test the effects of long-term sample incubation on community composition. These analyses were performed with the cores previously described and a number of other samples, including those collected using Van-Veen grabs, that may have been mixed with deeper sediment horizon depths and the overlying water (Supplementary Table 5). The resulting communities were compared to those from the same samples immediately frozen upon collection (“immediate” samples; Supplementary Table 5). DNA gram sediment⁻¹ and alpha diversity were significantly lower in long-term samples relative to immediate samples (Supplementary Figure 8; t-test, DNA gram sediment⁻¹, p < 0.035; Chao1, p < 0.004; Shannon, p < 0.001). OTUs related to the genera Psychrobium, Colwellia, Arcobacter, and Shewanella were enriched after long-term incubation, in some cases representing above 50% of the community despite starting abundances ranging from 0.0 to 0.2% (Figure 6). These OTUs were most similar to cultured piezophiles, including Colwellia marinimaniae (Kusube et al., 2016), Colwellia piezophila, and Shewanella benthica KT99, and to a member of the Psychrobium previously enriched after long-term pressurization (Aoki et al., 2014). Other long-term enriched taxa included members of the Gammaproteobacteria, Rhodobacteraceae, Planctomycetes, and Bacteroidetes (Supplementary Figure 9). In contrast, the long-term pressurized samples showed relative decreases in taxa belonging to uncharacterized and uncultivated lineages, including members of the clades BD2–11, BD7–8, NS72, SAR324, SAR202, Marinimicrobia, and Omnitrophica (OP3), and putative deep-ocean taxa such as Defluviicoccus and Rhodospirillaceae.

In this study, we evaluated the microbial community composition within trench sediments to test the hypothesis that hadal zones have distinct microbial communities from one another. While sediment communities were observed to be distinct between trench of collection and sampling site, there was significant overlap in the abundant taxa found in the Mariana and Kermadec trenches. Shared OTUs within both trenches represented greater than 90% of all reads. Therefore trench endemic OTUs made up small proportions of the communities when evaluated at 97% similarity. Many abundant taxa were closely related to members identified within other abyssal and hadal samples. These taxa belonged to lineages previously identified as having cosmopolitan members at bathyal and abyssal depths (Bienhold et al., 2016; Mußmann et al., 2017), such as sequences related to JTB255, BD2–11, JTB23, and the genus Marinicella. These results indicate that members within these lineages are present even at hadal depths. Similarly, the isolates obtained at high hydrostatic pressures were related to the genera Colwellia, Shewanella, and Moritella, consistent with previous studies. Many of the partial 16S rRNA genes of isolates from the present study were more than 97% similar to those of other piezophilic isolates. New strains were also obtained, including isolates related to the genus Psychrobium within the Gammaproteobacteria, Arcobacter within the Epsilonproteobacteria, and a member of the Flavobacteriaceae. Unfortunately, none of these isolates survived cryopreservation or repeated subculturing. Still, this Psychrobium isolate was related to that previously enriched after long-term pressurization (Aoki et al., 2014). Altogether, our results demonstrate the possible dispersal of OTUs between two widely separated trenches. Deep-ocean currents may lead to the dispersal and deposition of microorganisms in sediments (Müller et al., 2014), such as Antarctic Bottom Water flowing between the Kermadec and Mariana trenches. It is also possible some members do not require dispersal between trenches, but may originate within abyssal sediments, a possibility not yet evaluated. In this scenario, the microbes in question would possess high fitness in both abyssal and trench zones, potentially spreading between the two environments via bottom currents, or perhaps through earthquake-induced mass-wasting deposition down slope (Oguri et al., 2013). If taxa endemic to specific trenches exist in the hadal settings examined, they must exist at the strain rather than the species level, be rare, were lost during sampling, or are present in patchy or sample-specific distributions and were missed by our sampling. Comparisons with microbial communities in more distant trenches, such as the Puerto Rico or Atacama trenches, may show higher sequence abundances of endemic taxa.

Differences in organic matter due to primary production in overlying surface waters and its deposition through topographical funneling may be one of the most important factors structuring communities within trenches (Danovaro et al., 2003; Glud et al., 2013; Ichino et al., 2015; Wenzhöfer et al., 2016). The Kermadec Trench may have higher concentrations of organic matter than the Mariana Trench because of differences in primary productivity at the surface (Longhurst et al., 1995; Jamieson, 2015). While we do not have organic matter concentrations to report here for each core, preliminary data from the 0–1-cm fraction at 8000 m in each trench suggest that percent total organic carbon is higher in the Kermadec Trench (~0.5%) than the Mariana Trench (~0.4%; Grammatopoulou et al., in prep). Therefore, it is reasonable to hypothesize that organic matter is in part responsible for the differences between the Mariana and Kermadec trench communities. Consistent with this, the Kermadec Trench was enriched, relative to the Mariana Trench, in the Bacteroidetes, Actinobacteria, and Proteobacteria, lineages that have been found to correlate with higher concentrations of organic matter (Bienhold et al., 2012; Learman et al., 2016). In contrast, the Mariana Trench had higher proportions of Thaumarchaeota, Chloroflexi, and other uncharacterized lineages. Although Archaea are ubiquitous in marine sediments, organic matter concentrations are thought to negatively correlate with abundances of ammonia-oxidizing Archaea (Luo et al., 2015; Learman et al., 2016), potentially due to differences in electron acceptor availability or Archaea outcompeting Bacteria under energy-starved conditions (Valentine, 2007; Hoehler and Jorgensen, 2013). The Chloroflexi may also be adapted to degrading recalcitrant organic matter (Landry et al., 2017). Furthermore, the number of piezophilic isolates and related sequences in the high-throughput community data were higher within the Kermadec than the Mariana Trench and were enriched in the upper layers of sediment. Their enrichment in the Kermadec Trench pelagic community and on larger size fractions has been suggested to be a result of adaptations to higher concentrations of organic matter, including particulate forms (Peoples et al., 2018a). Intra-trench variability may also be influenced by organic matter due to topographic funneling into the axis of the trench. Within the Kermadec Trench the Chloroflexi and Acidobacteria were enriched at 6- and 7-km depths whereas the Nitrospirae and Actinobacteria are more abundant at 8- and 9-km depths. Enrichments of Nitrospirae within the hadal water column have been attributed to a more eutrophic environment (Nunoura et al., 2015). Altogether, these findings suggest the Kermadec Trench may be enriched in organic matter relative to the Mariana Trench. Ultimately, deep-sea microbial communities are governed by a myriad of variables that may contribute to these inter- and intra-trench differences, such as hydrography (Hamdan et al., 2013), specific location of sample collection, temporal variability and season of sample collection (Lampitt, 1985), sediment lithology (Probandt et al., 2017), or the quality, not just the quantity, of organic matter present. Our findings support the notion that organic matter can contribute to the spatio-temporal variability in deep-sea microbial communities.

While trench sediments may become mixed and suspended due to tectonic activity and topographic instability (Kawagucci et al., 2012; Oguri et al., 2013; Nunoura et al., 2016), the sediment communities studied here were stratified and compositionally distinct from those in the water column above them (Peoples et al., 2018a). Because the oxygen penetration depth can be limited to the top 10 cm in trenches (Glud et al., 2013; Wenzhöfer et al., 2016), we compared the microbial communities present in the shallower and deeper sediment horizons. Community richness and diversity varied with sediment horizon depth, consistent with previous comparisons of surficial and deep subsurface environments (Durbin and Teske, 2010; Teske et al., 2011; Shulse et al., 2016; Walsh et al., 2016a,b). Our findings suggest that specific deep-sea lineages are enriched within deeper marine sediments. Within the Thaumarchaeota, MGI-α showed enrichment in the upper sediment while those belonging to the MGI-η subgroup were more abundant in the deeper sediments. Such ecotype differentiation has been previously noted in abyssal sediments (Durbin and Teske, 2010; Tully and Heidelberg, 2013; Lauer et al., 2016) and may be due to changes in organic matter abundance or oxygen availability (Durbin and Teske, 2010). Abundances of alternative electron acceptors, such as nitrate, sulfate, or iron, influence community composition (D’Hondt et al., 2009; Durbin and Teske, 2010, 2012; Nunoura et al., 2013; Jessen et al., 2017) and therefore likely play an important role in hadal sediments (Nunoura et al., 2013, 2018) as the oxygen penetration depth can be shallow (Glud et al., 2013; Wenzhöfer et al., 2016). Relative sequence abundances of the Woesearchaeota increased with increasing sediment horizon depth. This archaeal phylum has members that are specifically enriched in anoxic niches and can have fermentative or symbiotic lifestyles (Castelle et al., 2015; Lazar et al., 2017; Liu et al., 2018). Differences in genomic potential and niche separation of Woesearchaeota suggest that unexplored diversity exists within this phylum (Shcherbakova et al., 2016; Narrowe et al., 2017). OTUs related to the Woesearchaeota within our dataset were remarkably distinct from other previously identified studies, showing highest similarity to sequences from abyssal and hadal sites and low (<95%) similarity to representative sequences from other habitats (Supplementary Figure 10). The Marinimicrobia, which are abundant within the deep-ocean water column (Nunoura et al., 2015; Tarn et al., 2016; Peoples et al., 2018a), also increased with sediment horizon depth. Based on phylogenetic analysis (Supplementary Figure 6) the Marinimicrobia OTUs identified here belong to deep-sea sediment-associated clades that have not been previously identified (Hawley et al., 2017).

Although a large diversity of Bacteria and Archaea were found within the sediments, very few taxa were successfully isolated. Many of those that were belong to the same genera as known piezophiles, consistent with most previous isolation attempts from the deep sea. Interestingly, after unamended, long-term static pressurization of sediment samples, higher relative sequence abundances of taxa related to the cultured piezophiles Psychrobium, Colwellia, Arcobacter, and Shewanella were found. In shallow-ocean communities, Colwellia is one of the first responders to microcosm conditions (Stewart et al., 2012; Mayali et al., 2016) and enrichments of Colwelliaceae, Moritellaceae, Psychromonadaceae, and Shewanellaceae have been observed within enrichments of Arctic bathyal-depth sediment samples under high pressure (Hoffmann et al., 2017). Therefore, regardless of location, depth of collection (surface, bathyal, and in this study hadal), nutrient enrichment, or temperature or pressure incubation conditions, taxa belonging to the same heterotrophic and copiotrophic genera within the Gammaproteobacteria are enriched under mesocosm conditions. These taxa may contain metabolic versatility for colonizing various ecological niches during fluctuating environmental conditions (Stewart et al., 2012). In contrast, microbial populations from immediately recovered samples were enriched in taxa belonging to uncharacterized and uncultivated lineages representative of the in situ diversity. Taken together with the enrichment of culturable taxa, these findings highlight the difficulties in obtaining pure cultures of representative and novel deep-ocean lineages. This study is the first to show that long-term incubations in pressure vessels, regardless of nutrient amendment or collection location, clearly select for the same piezophilic taxa. Decompression during sample retrieval, static incubation conditions in pressure vessels leading to lack of oxygen or other important nutrients, or eukaryotic predation (Tsagaraki et al., 2018) may significantly bias our ability to obtain cultures of piezophiles.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.