Unique Prokaryotic Consortia in Geochemically Distinct Sediments from Red Sea Atlantis II and Discovery Deep Brine Pools

The seafloor is a unique environment, which allows insights into how geochemical processes affect the diversity of biological life. Among its diverse ecosystems are deep-sea brine pools - water bodies characterized by a unique combination of extreme conditions. The ‘polyextremophiles’ that constitute the microbial assemblage of these deep hot brines have not been comprehensively studied. We report a comparative taxonomic analysis of the prokaryotic communities of the sediments directly below the Red Sea brine pools, namely, Atlantis II, Discovery, Chain Deep, and an adjacent brine-influenced site. Analyses of sediment samples and high-throughput pyrosequencing of PCR-amplified environmental 16S ribosomal RNA genes (16S rDNA) revealed that one sulfur (S)-rich Atlantis II and one nitrogen (N)-rich Discovery Deep section contained distinct microbial populations that differed from those found in the other sediment samples examined. Proteobacteria, Actinobacteria, Cyanobacteria, Deferribacteres, and Euryarchaeota were the most abundant bacterial and archaeal phyla in both the S- and N-rich sections. Relative abundance-based hierarchical clustering of the 16S rDNA pyrotags assigned to major taxonomic groups allowed us to categorize the archaeal and bacterial communities into three major and distinct groups; group I was unique to the S-rich Atlantis II section (ATII-1), group II was characteristic for the N-rich Discovery sample (DD-1), and group III reflected the composition of the remaining sediments. Many of the groups detected in the S-rich Atlantis II section are likely to play a dominant role in the cycling of methane and sulfur due to their phylogenetic affiliations with bacteria and archaea involved in anaerobic methane oxidation and sulfate reduction.


Introduction
The Red Sea represents a 450,000-km 2 inlet of the Indian Ocean located between the African continent and the Arabian Peninsula. Among the most intriguing ecological niches in the Red Sea are its deep-sea brines, which exhibit unique and diverse geochemical conditions [1]. Twenty-five brine pools have been described to date in the Red Sea [2], however, our knowledge of the microbial communities that inhabit these pools, and how these organisms are affected by and contribute to the geochemical properties of their extreme environments, remains sparse.
For example, the hot brine system at Atlantis II Deep (ATII) is the largest and best characterized pool in the Red Sea. Located at a depth of approximately 2200 m near the Central Rift (at around 21u209 N), it is less than 100 m thick, and covers an area of 60 km 2 [3][4][5]. The combination of anoxia and high pressure, increased temperature (68uC) and salinity (250 parts per thousand or 7.5 times that of normal seawater [6]) in the brine pool makes it an ideal setting for the study of extremophiles. The Atlantis II Deep is the largest marine sulfide deposit currently known and contains high concentrations of manganese, iron, molybdenum, cadmium, cobalt, copper, nickel, lead and zinc, as well as methane and other hydrocarbons [7][8]. The brine is stratified, with the lower convective layer (LCL) exhibiting the highest temperature and salinity, while the three upper convective layers (UCL1, 2 and 3) display step-wise drops in salinity and temperature [4]. Discovery Deep (DD) is likely linked to Atlantis II via subsurface connections [9,10]. However, the lack of documented long-term temperature variations, suggests the absence of hydrothermal activity at this site [11]. The conditions in the Discovery Deep brine are less harsh and not as well described as Atlantis II, with relatively high concentrations of metals and a milder temperature of 44.8uC [12].
The microbial life found in the hot brine pools in the Red Sea has been little studied owing to technical constraints [reviewed in [12]. Microbial diversity studies based on cultivation-dependent and cultivation-independent approaches have focused mostly on the colder, northern brine pools of the Red Sea. These surveys have revealed an unexpected microbial diversity, including the identification of several new deeply branching phylogenetic lineages, and the isolation and description of several new unusual extremophiles [13,14]. Studies on the hot brines of the Red Sea were much more limited, and include the isolation and description of Flexistipes sinusarabici (phylum Deferribacteres) from the Atlantis II brine pool [15]. A recent 16S rDNA survey suggested the presence of similar microbial communities in Atlantis II and Discovery Deep brines [16], while a comparative metagenomic analysis revealed profound differences in the dominant bacterial groups in these environments [17]. The latter study also identified genes associated with degradation of aromatic compounds in Atlantis II [17].
Whereas these previous studies have yielded insights into the composition and metabolic capabilities of microbial organisms in brine pools, the microbial assemblages that inhabit the sediments underlying the hot brine-pools in the Red Sea and it's relationship to sediment geochemistry have not been reported previously. Here we analyze and compare the prokaryotic microbiomes in several chemically and geologically distinct sediment sections of Atlantis II (ATII), Discovery (DD), Chain Deep (CD) and an adjacent brineinfluenced (BI) site ( Fig. 1) using 16S rDNA pyrotag sequencing. In this poorly explored marine environment, distinctively unique prokaryotic communities were observed in a sulfur-rich Atlantis II and a nitrogen-rich core section of Discovery Deep brine pool, providing insight into how geochemical components allocates 'polyextremophiles'.

Results
Chemical profiling of the sediments from brine pools Sediment were collected from Atlantis II (ATII), Discovery (DD), Chain Deep (CD) and an adjacent brine-influenced (BI) site ( Fig. 1 and Table 1) as described in materials and methods. Carbon Hydrogen Nitrogen & Sulfur (CHN&S) profiling showed that the analyzed brine sediments and the adjacent BI sediments differed greatly in C and S content, whereas little variation was found in H (1.4-1.9% w/w) or N (,0.10 to 0.12% w/w) content (Table 2). Sulfur content was highest in the ATII sediment sections (mean 8.3% w/w) with the highest value found in the deepest analyzed section (ATII-1; 19% w/w). The sulfur content was on average much lower in the other brine (0.6% in DD and ,0.10% in CD) and adjacent (BI; 0.3-1.0% w/w) sediments. SO 4 22 concentrations varied more between ATII sediments than % sulfur, but were also found to be highest in the deepest ATII-1 section (20% w/w). Far lower SO 4 22 concentrations were measured in the other brine and BI sediments ( Table 2). Carbon content on the other hand was highest in DD (,7% w/w) and BI (8.3-8.7% w/w) followed by one order of magnitude lower values in ATII and CD sediments (,0.8% w/w). The highest level of carbon was found in DD-2. Interestingly nitrogen (0.22% w/w) was only detected in the deepest section of DD (DD-1). CHN&S results are tabulated in Table 2.

Pyrotag 16S rDNA data sets
A total of 1,012,178 reads were generated from the 16 brine and adjacent sediment samples and the overlying ATII water column (Table 3), ranging from a total of 14,006 for ATII-1 to 53,665 for ATII-6 ( Table 3). The sediment reads were compared to 388,659  reads from the water column. Of the total sediment pyrotag  library, 242,366 pyrotags represented bacteria and 377,425 represented archaea. The number of bacterial reads varied between 12,125 in ATII-1 and 22,540 in ATII-6. The number of archaeal reads generated varied between 1,624 in ATII-1 and 37,451 in DD-5. Unassigned reads (unknown) represented a relatively small proportion of the total pyrotags.

Taxonomic assignment of OTUs to major archaeal groups
The overall archaeal community structure differed greatly between the S-rich ATII-1, N-rich DD-1, and the remaining sediments (Fig. 2, Table 4). Out of 39 archaeal OTUs analyzed, 10 were detected in both water column and sediments (Table S1), 6 were detected only in the water column (data not shown) and 23 were only detected in the sediments (Table S1, Table 4). To exclude potential contamination from the overlying water column we focus on the sediment only taxa, that doesn't exclude the fact that taxa shared in sediment and water may also be present in sediments. In the S-rich ATII-1, euryarchaeota clearly dominated the archaeal community (96%) with 50% of the archaeal tags in this section being assigned to Marine Benthic Group E (MBG-E; Thermoplasmata), note that MBG-E was not detected in the water column. Other abundant euryarchaeota in ATII-1 were Methanomicrobia belonging to ANME-1b (25%) and the Methanosarcinales clade GOM ArcI. Pyrotags of the latter groups were not detected in any of the other sediment sections or the water column (Fig. 2, Table 4).
Similarly, euryarchaeota predominated the archaeal community in the N-rich DD-1 section (65%). Here, an unclassified family of the ST-12K10A lineage (SA1 group) represented the most abundant archaeal group (42%), whereas pyrotags of this archaeal family were very rare (0.1% in ATII-1) or absent in all other analyzed sediment sections and the water column. Uncultured Thermoplasmata of the South African Goldmine (SAG) Group, were found to be ten times more abundant in DD-1 (15%) than in the Srich ATII-1, whereas MBG-E were 250 times less abundant than in the ATII-1 sections. Marine Group III (MG-III) Thermoplasmatales, which were not recovered from ATII-1, represented 0.2% of the total archaeal pyrotags in DD-1 (Table 4). Although not highly abundant, we detect a diverse and exclusive Halobacteriales-related pyrotags only in the N-rich Discovery Deep brine-pool sediments (Fig. 2, Table 4).
Crenarchaeota assigned to Marine Group I (MGI) comprised only 1.9% in the S-rich ATII-1 and 34.3% in the N-rich DD-1 sediments, whereas members of this group represented the most abundant archaea in all other sediment sections (81.668.2%) including the water column (Table 4). Methanomicrobiales-Methanospirillum, Halobacteriales Deep Sea Hydrothermal Vent Group 6 and Terrestrial Hot Spring Crenarchaeota were detected in the remaining sediments but not in the water column (Table 4).
Relative abundance-based hierarchical clustering of the 16S rDNA pyrotags assigned to major taxonomic groups allowed us to categorize the sediment archaeal communities into three major and distinct groups (Fig. 3A, Table S1) and to detect a set of sediment/ brine specific archaeal OTUs (23) that were not detected in the overlying water column (Table S1). Group I (12 OTUs, of which 10 are sediment specific) was unique to the S-rich Atlantis II sample (ATII-1), group II (13 OTUs, of which 10 are sediment specific) was characteristic for the N-rich Discovery sample (DD-1), and group III (8 OTUs, of which 4 are sediment specific) reflected the composition of the remaining sediment samples.
Taxonomic assignment of OTUs to major bacterial groups Similar to archaea, the bacterial community structure greatly differed between the S-rich ATII-1, N-rich DD-1, compared to the remaining overlying and adjacent sediments. Out of 171 bacterial OTUs, 75 were detected in both water column and sediments (Table S1), 42 were detected only in the water column (data not shown) and 48 were only detected in the sediments (Table S1,  Table 4).
In the S-rich ATII-1, Chloroflexi accounted for 32% of the assigned bacterial OTUs, with the order Anaerolineales predominating (Table 4, Fig. S1). The distribution of Chloroflexi was  Table 1 for sampling and sample details. ATII (Atlantis II; with a total of six distinctive sediment sections), DD (Discovery Deep; with seven sections), CD (Chain Deep; one section), and adjacent brine sections (BI; two sections). doi:10.1371/journal.pone.0042872.g001 in all other analyzed sediment sections. Fusobacteria (20%) and the OP1 division (15%) were two of the major bacterial phyla in the Srich ATII-1 sample. The latter group also comprised 5% of the bacterial OTUs in DD-1 but was insignificant elsewhere (including the water column). In contrast to their high abundance in all the other samples, Alphaand Gammaproteobacteria constituted only 1% and 4% of bacterial OTUs in ATII-1, respectively. In the N-rich DD-1, Alphaproteobacteria dominated the bacterial community constituting 25% of bacterial OTUs, followed by Cyanobacteria with 22%. In other DD sediments, the latter phylum was far less prominent, and this variability is largely due to the contributions of an unassigned family related to Prochlorococcus (21%). Prochlorococcus constituted 4% (68) of water column OTUs, and therefore the overlying water column could have compromised this sample.
Based on the relative abundance of 16S rDNA pyrotags and their assignment to major bacterial groups a set of sediment/brine specific bacterial OTUs (48) were identified (Table S1). Additionally, three distinct sediment groups were observed in our samples. Group I (41 OTUs, of which 27 are sediment specific) are enriched in the S-rich Atlantis II sample (ATII-1), group II (31 OTUs, of which 15 are sediment specific) corresponds to the Nrich Discovery sample (DD-1), group IIIa (20 OTUs, of which 3 are sediment specific) in samples ATII-2 to ATII-5, while group IIIb (34 OTUs, of which 3 are sediment specific) reflects the combination of taxa found in all other samples (Fig. 3B, Table S1). Rare bacterial communities were determined by selecting OTUs that accounted for less than 1% of reads in all the samples. Comparative analysis of the less abundant bacterial genera leads to the same basic division into four consortia (Fig. S2, Table S2). The distinctive bacterial communities in the S-rich Atlantis II and Nrich Discovery samples are discussed in relation to sulfur and methane cycling below.

Comparative analysis of microbial community in the Red Sea sediments with diverse oceanic systems
To assess the uniqueness and diversity of the microbial communities present in the Atlantis II brine-system sediments, we compared the pyrotags recovered in this study with a global survey of sub-seafloor microbial communities. The 16S rDNA tag sequences that were used for this comparison are publicly available through the VAMPS website http://vamps.mbl.edu (project title ICM_GMS). We compared our data with oxic sediments from the White Oak River estuary in the North Atlantic Ocean [18], anoxic deep-sea hydrocarbon seeps from both a microbial mat and sediments [19] in the Gulf of Mexico, a microbial mat from the oxic Storegga seep enrichment on the continental shelf in the Norwegian Sea, an anoxic microbial mat from a carbonate reef in the Black Sea, and a microbial mat from an oxic hydrothermal vent in the Guaymas Basin in the Gulf of California. These oceanic sediments were surveyed as part of the 'Microbial Census of Methane and Hydrocarbon Seep Sediments in Relation to Geochemical and Temperature Gradients'. Heat map analysis revealed four sample groups based on the unique microbial consortia present: I) ATII-1, II) DD-1, IIIa) ATII-2 to ATII-5 and IIIb) ATII-6, DD-2 to DD-7 and CD and BI (Fig. 3). A clustering analysis was used to compare and contrast our data with 16S rRNA (V6) obtained from the other oceanic systems. With respect to both Archaea and Bacteria, groups II and III clustered in one clade ( Fig. 4A and 4B). Both the bacterial and archaeal assemblages at ATII-1 showed greater similarity with the consortia in the methane and hydrocarbon seep systems. Notably, ATII-1 showed a closer resemblance to the communities at marine hydrothermal vents in the Gulf of California (Fig. 4A).

Discussion
Unique microbial taxa in the brine pool sediments that are enriched in the S-rich Atlantis II and the N-rich Discovery brine-pool sediments We identify the unique microbial communities in sediments from the three Red Sea brine pool sites. To exclude potential contamination from the overlying water column, we focus on the 48 sediment/brine specific bacterial taxa and 23 archaeal taxa. Although, the microbial communities in sediments from three Red Sea brine pool deeps and a brine-influenced site displayed very similar distributions of abundant and rare microbial groups, the deepest S-rich sediment section from Atlantis II and an N-rich section from Discovery Deep each revealed unique archaeal and bacterial communities. Clustering of 16S rDNA pyrotags from each sediment sample into major bacterial and archaeal groups on the basis of relative abundance and taxonomic assignment (Fig. 3, Table S1 and Fig. S2), allowed us to discern three major assemblages, comprising (I) taxa that are unique for S-rich ATII-1, (II) taxa unique for N-rich DD-1, (IIIa) taxa that are restricted to samples ATII-2 to ATII-4, and (IIIb) taxa that are found in all the remaining samples (i.e., ATII-5 and ATII-6, CD, BI and DD-2 to DD-7 (Fig. 3A). Note, that the water column is clustered into a fourth group, with a closer resemblance to group III (data not shown). Such clustering suggests vertical and horizontal variability in the prokaryotic community structure within the Atlantis II deep-sea brine system. Comparison with data from other oceanic sites revealed that the prokaryotic community in the ATII-1 is more closely related to prokaryotes found in methane and hydrocarbon seep sediments (Fig. 4). The detection of OTUs from known typical marine prokaryotic groups (e.g., MGI, Cyanobacteria and SAR11) might result from accumulation in the sediments of non-active or dead cells or even cellular debris previously active in the water column. It may also be a result of contamination from the overlying water column. However, we attempted to exclude such potential contaminants by subtracting ATII water column taxa from sediment taxa, and we have stressed on the sediment/brine-specific taxa in this study. However, this does not exclude that these OTUs may belong to sediment-clades or possibly new Atlantis II brine system clades. For example, members of the MGI seem to dominate the archaeal community in most sediment layers. Members of this group have been previously found within the brine-systems, but seemed to be restricted to the brine-seawater interface [12], probably reflecting the standard communities found in seawater above the brine. MGI is dominant in deep-sea water and surface (oxic) sediments, although populations in these two biotopes should not be assumed as being identical [20,21] There are specific sediment phylotypes, distinct from seawater relatives [20].
Although we have only utilized one sediment core at each sample point, our results can be generalized to the entire brine because the physical conditions within the spatial dimensions of the brine pool are uniform. Performing a single DNA extraction protocol may also be a limitation in our study, suggesting a representation of, rather than the entire, microbiota in each site.
The composition of the prokaryotic community in the S-rich section of Atlantis II sediments suggests metabolism based on sulfate reduction and methane consumption. On the other hand, the diversity and exclusive detection of selected Halobacteriales in the N-rich Discovery Deep sediment may be related to its characteristic elemental profile (discussed below).

The S-rich subsection of the Atlantis II brine-pool sediment sample harbors a unique consortium of prokaryotes putatively involved in sulfate-reduction coupled with anaerobic oxidation of methane
The oxidation of methane in anoxic marine sediments is mediated by consortia of methane-oxidizing and sulfate-reducing archaea and bacteria [22], in a process described in previous studies [23]. The exclusive combined presence of pyrotag related to archaeal and bacterial sulfate-reducing genera (Archaeoglobus [24], and Desulfovermiculus sp. (Deltaproteobacteria)) and anaerobic methane-oxidizing archaeal clusters ANME-1 (Methanosarcinales and Methanomicrobiales) [25], ANME-2 and ANME-3 (Methanosarcinales), strongly argues for an exclusive inhabitance of microbial sulfate reduction coupled with anaerobic oxidation of methane in the S-rich ATII section (Fig. 3B).
The distribution of Methanomicrobia in the S-rich Atlantis II sample is also unique, and quite distinct from those in the remaining samples. Methanosarcinales species were exclusively identified in the S-rich ATII-1. Methanomicrobia were identified in all the sediments analyzed, but their abundance and detailed composition varied widely. Methanomicrobia were most abundant in the S-rich ATII-1 and N-rich DD-1 samples, but accounted for ,1% of reads in the remaining samples. The ANME-1b clade and the Methanosarcinales clade GOM-Arc-1 constituted the majority of Methanomicrobia in the S-rich ATII-1, while ST-12K10A (SA1) dominated the N-rich DD-1 sample, which was previously identified in Shaban Deep and brine-sea water interface of Discovery Deep [14]. Interestingly, in these two samples, hydrogenotrophic Methanomicrobiales (DD-1) and acetoclastic Methanosarcinales (ATII-1) were mutually exclusive. Presence of sulfate reducers would explain these differences as they would likely be partially involved in anaerobic methane oxidizing consortia with ANME and would simultaneously outcompete and exclude all methanogenic groups with the exception of Methanosarcinales [25,26]. This implies distinctive styles of methane Table 3. Pyrotag 16S rDNA data sets. production in the S-rich Atlantis II and N-rich Discovery sediment sections.
Overall, our comparative taxonomic assignments of OTUs that defined the four distinct microbial assemblages (Fig. 3B) reveal a unique consortium of methanogenic, ANME, and sulfate-reducing microbes that are distinct in the environments of S-rich ATII-1, Nrich DD-1 and the remaining samples.

Diversity of denitrifying microbes in the N-rich subsection of the Discovery Deep brine-pool sediment
Nitrification requires several groups of microorganisms and play a significant role in the nitrogen cycle [27].
Group II archaea were predominantly detected in DD-1 (Table 4 and Figs. 2B, 3B) and included potential ammonium oxidizers such as Crenarchaeum of the terrestrial hot spring group (Table 4) [28][29][30]. On the other hand, archaea known to be capable of reducing oxidized nitrogen species were also exclusively identified in DD-1, such as Halobacteriaceae; Natronomonas sp., Halococcus sp., Halobacterium sp., Halorabdus sp., Halosimplex sp., and an unassigned MSP41 genus (Fig. 2). A previous study documented an abundance of Halobacteriaceae in the Atlantis II and Discovery Deep overlying water column [16]. However, our analysis did not identify such abundance of Halobacteriaceae. These are capable of reducing nitrate to nitrite during anaerobic respiration [28][29][30]. Note that, in contrast to other analyzed brine and non-brine sediments, the archaea in Group II, which are present in DD-1, all cluster among ammonium oxidizers. Interestingly, the S-rich Atlantis II sections (ATII-1), was the only sample that had no detectable Halobacteriaceae related to nitrogen cycling. The diversity and detection of archaea previously known to play a role in nitrogen cycling in the N-rich Discovery sample implies an active role specifically in nitrogen regulation in this sediment section.

Conclusion
This study identifies three major microbial consortiums in sediments of brine and brine-influenced sites in the Red Sea. Based on taxonomic assignment of 16S rDNA OTUs to major bacterial and archaeal groups, the characteristic distributions of both the abundant and the rare taxa was revealed. A sulfur-rich Atlantis II and a nitrogen-rich Discovery Deep section harbored distinct prokaryotic assemblages that differed furthermore from those found in all other sediment samples examined. The exclusive combined presence of sulfate-reducing and anaerobic methaneoxidizing prokaryotes strongly argues for an exclusive importance of microbial sulfate reduction coupled with anaerobic oxidation of methane in the S-rich ATII section. Additionally, the diversity of denitrifying microorganisms in the N-rich Discovery section implies an active role in nitrogen cycling. More in depth characterization of the physical and chemical properties of these sediment sections combined with further metagenomics and metatranscriptomes could shed further insights into microbial metabolic processes and the cycling of (in)organic matter in these under-sampled extreme environments.

Sample collection
The sediment samples from the Atlantis II, Discovery Deep, and Chain Deep brines, as well as from the brine-influenced site used for this study were collected during the 2 nd KAUST/WHOI Red Sea Expedition of the RV Aegaeo in April 2010 (see Table 1 for details about the samples and sampling locations). All necessary permits were obtained for the described field studies. Between 3.5 and 4 m-long gravity cores (Benthos Instruments, KC Denmark) were obtained from the brine sediments, whereas the short BI core was obtained by multicoring. The water depth of the brine settings varied only between 1937 and 2168 m, but the brines differed greatly in temperature, being highest at ATII (68uC), followed by DD (45uC) and CD (32uC) ( Table 1). The BI site was slightly shallower (1856 m water depth) and the temperature (22uC) was lowest of all settings. Water samples were collected from the overlying ATII water, at different depth 50 m, 200 m, 700 m and 1500 m (Ferreira et. al., submitted). Water depth in the Atlantis II Table 4. Taxonomic assignment of OTUs to major archaeal and bacterial groups.   Deep area was obtained during the cruise from a bathymetric survey with single-beam echo sounder on R/V Oceanus (Fig. 1A). Cores were divided into sections (numbered from bottom to top) and different sediment sections were studied separately (Fig. 1B). In ATII, six distinct sections of the sediment core were macroscopically identified (Fig. 1B). The bottom 50 cm of the ATII core (ATII-1) consisted of black slurry. ATII-2 was dominated by brown, ATII-3 by reddish-brown, and ATII-4 by brown clay. ATII-5 was composed of black slurry, while brownish slurry sediments (ATII-6) were found at the top of the core facing the lower convective layer of the pool. The seven DD sediment samples were less macroscopically distinct than the ATII sections (Fig. 1B). For the brine-influenced sediments, BI-1 comprised the top half (0-5 cm) and BI-2 bottom 5 cm of the core, although the two samples were not macroscopically distinct. All sediment samples were immediately frozen upon recovery, and kept at 220uC until further analysis.

Chemical analysis
Profiling of the elements C, H, N and S was performed by oxidation of the samples by flash combustion using a Thermo FlashEA 1112 elemental analyzer. The combusting gases were then separated by passage through a chromatographic column using He as a carrier gas, and were detected with thermal conductivity detector. Sulfate level was measured by ion chromatography.
DNA extraction, pyrosequencing of 16S rDNA amplicons and computational analysis Genomic DNA was extracted 0.5-1 gm of sediment with the PowerSoilH DNA Isolation Kit (MO-BIO, Calsbad, CA), the total amount of DNA isolated from the sediments is presented in To infer phylogenetic relationships among the identified Operational Taxonomic Units (OTUs), an initial taxonomic classification into bacterial or archaeal groups was performed. The OTUs were estimated by clustering the 910,068 raw 16S reads with an identity threshold .0.97% using USEARCH (http://www.drive5.com/ usearch/). The seed (longest sequence that is at least 0.97% similar to all other sequences in the cluster), was utilized as the representative sequence (i.e., the OTU) of that cluster, resulting into a non-redundant set of 837 bacterial OTUs and 63 archaeal OTUs. For each domain, all OTUs across the different samples were aligned using MUSCLE [32]. The two generated multiple sequence alignments were visually inspected and manually tuned. A maximum-likelihood phylogenetic tree was then estimated based on the refined alignment under the GTR model [33] using FastTree [34]. Finally, the estimated tree was taxonomically annotated, and the normalized abundance levels of each OTU in the different samples were graphically represented on the trees as bars stemming from the corresponding OTU using iToL [35].
Heat maps were generated using data statistically filtered by a two-tailed Fisher's exact test (FET). The normalized read value for each taxon in a sample was compared to those of other samples, and to the totals of remaining taxa. Statistical analyses were performed using R (http://cran.r-prroject.org) [36]. Results were filtered for a p-value threshold ,0.05. The frequency-normalized values of the filtered statistically significant data were used to create heat maps using the 'enhanced heat map function' from the R package gplots (http://cran.r-project.org/package = gplots). Complete linkage hierarchical clustering was used across the samples. The Pearson correlation distance matrix was calculated between the taxa. We also performed Euclidean distance Ward clustering approach and Spearman's correlation, which did not alter the clustering of the data (data not shown).  than 1% of bacterial reads. The taxonomically assigned OTUs are represented in Table S1. (TIF) Table S1 The archaeal assigned OTUs (from Fig. 3A) and the bacterial assigned OTUs (from Fig. 3B) are tabulated. OTUs detected only in sediments are in bold. (DOCX)