OUP user menu

Isolated communities of Epsilonproteobacteria in hydrothermal vent fluids of the Mariana Arc seamounts

Julie A. Huber , Holly V. Cantin , Susan M. Huse , David B. Mark Welch , Mitchell L. Sogin , David A. Butterfield
DOI: http://dx.doi.org/10.1111/j.1574-6941.2010.00910.x 538-549 First published online: 1 September 2010

Abstract

Low-temperature hydrothermal vent fluids represent access points to diverse microbial communities living in oceanic crust. This study examined the distribution, relative abundance, and diversity of Epsilonproteobacteria in 14 low-temperature vent fluids from five volcanically active seamounts of the Mariana Arc using a 454 tag sequencing approach. Most vent fluids were enriched in cell concentrations compared with background seawater, and quantitative PCR results indicated that all fluids were dominated by bacteria. Operational taxonomic unit-based statistical tools applied to 454 data show that all vents from the northern end of the Mariana Arc grouped together, to the exclusion of southern arc seamounts, which were as distinct from one another as they were from northern seamounts. Statistical analysis also showed a significant relationship between seamount and individual vent groupings, suggesting that community membership may be linked to geographical isolation and not geochemical parameters. However, while there may be large-scale geographic differences, distance is not the distinguishing factor in the microbial community composition. At the local scale, most vents host a distinct population of Epsilonproteobacteria, regardless of seamount location. This suggests that there may be barriers to exchange and dispersal for these vent endemic microorganisms at hydrothermal seamounts of the Mariana Arc.

Keywords
  • hydrothermal vents
  • biogeography
  • bacterial diversity
  • 454 tag sequencing

Introduction

Historically, biological studies of seamounts have focused on endemism and the hypothesis that some seamounts may host isolated macrobiological communities, similar to islands (Whitaker, 1998). While this theory remains controversial for macrofauna (Samadi et al., 2006; McClain, 2007; Stocks & Hart, 2007; Brewin et al., 2009), few studies have examined the issue of microbial endemism at seamounts. In the last 15 years, the microbiology of hydrothermally active seamounts has grown considerably (Emerson & Moyer, 2010), with the first reports of bacteria and archaea in microbial mats at Loihi Seamount near Hawaii (Moyer et al., 1994, 1995) to diverse studies of different microbial habitats at Axial in the northeast Pacific (Huber et al., 2007; Opatkiewicz et al., 2009), Vailulu'u in the central Pacific (Staudigel et al., 2006), Suiyo in the northwest Pacific (Higashi et al., 2004), South Chamorro in the western Pacific (Mottl et al., 2003), and seamounts of the Mariana Arc in the western Pacific (Davis & Moyer, 2008). The chemistry of hydrothermally active seamounts can vary considerably depending on the geological setting (forearc, backarc, volcanic arc, mid-ocean ridge, hotspot), making comparisons between different volcanoes challenging. Davis & Moyer (2008) presented the most comprehensive view of microbial diversity in hydrothermal seamounts to date based on their collection of microbial mats from Axial, Loihi, and volcanoes of the Mariana Arc and backarc. They found either Epsilon- or Zetaproteobacteria (as assessed by terminal restriction fragment length polymorphism) at most of the sites, and a higher diversity on the arc systems compared with Axial and Loihi (mid-ocean ridge and hotspot). No biogeographical patterns were observed, but links between the microbial community structure and chemistry, depth, or temperature of the sample were suggested, demarcating three clusters of microbial communities. Often, multiple samples from within a seamount fell into more than one cluster, highlighting the heterogeneity of the microbial mats and hydrothermal systems they sampled (Davis & Moyer, 2008). They concluded that arc and backarc hydrothermal systems may represent bacterial hotspots of diversity in the ocean.

Here, we focus on volcanoes of the Mariana Arc using venting fluids as an access point to the subseafloor microbial communities of seamounts (Schrenk et al., 2010). Our previous work in diffuse vents from mid-ocean ridges showed remarkable microbial diversity (based on 16S rRNA gene sequencing) and evidence that vents with different geochemical characteristics harbored distinct subseafloor microbial populations (Huber et al., 2006, 2007). This study uses a 454 tag sequencing approach to examine the distribution and diversity of the Epsilonproteobacteria in the subseafloor of hydrothermally active seamounts of the Mariana Arc. The Epsilonproteobacteria are one of the most dominant groups of microorganisms found in almost all hydrothermal vent biotopes, including microbial mats, sulfide chimneys, diffuse fluids, and associated with animals (Reysenbach et al., 2000; Corre et al., 2001; Huber et al., 2003; Higashi et al., 2004; Nakagawa et al., 2005b; Takai et al., 2005a; Campbell et al., 2006; Moussard et al., 2006; Huber et al., 2007). After initial detection by molecular methods, the first members of this ubiquitous group were isolated from hydrothermal vents in 2001, and since then, have been cultured from vents around the world (Campbell et al., 2001; Takai et al., 2005a). They display wide metabolic and phylogenetic diversity, but in general, most are hydrogen- and sulfur-oxidizing chemolithoautotrophs, including both mesophiles and thermophiles (Campbell et al., 2006). All Epsilonproteobacteria studied to date from deep-sea vents fix carbon via the reductive or the reverse tricarboxylic acid cycle (Campbell et al., 2001; Takai et al., 2005a). Because they are not found in high abundances in oxygenated deep seawater and are very abundant in hydrothermal vents, the Epsilonproteobacteria serve as an ideal indicator organism for examining vent-specific microbial community structure and distribution patterns.

Materials and methods

Field sites and sampling

In 2006, as part of the NOAA Ocean Explorer Ring of Fire program (Embley et al., 2004), we sampled 14 vents from five different seamounts along the Mariana Arc (Fig. 1, Table 1) using the hydrothermal fluid and particle sampler (Butterfield et al., 2004) on the ROV Jason 2. Fluids for DNA-based analyses passed through a Sterivex-GP (0.22-μm pore size) filter, which was frozen at −80 °C upon recovery of the vehicle. Whole unfiltered fluids for chemical analyses were also collected and analyzed according to Butterfield et al. (2004). Fluid samples were preserved for cell enumeration using epifluorescence microscopy with DAPI as described previously (Huber et al., 2002), as well as cultivation of anaerobic thermophiles and hyperthermophiles.

1

Map of the Mariana Arc and five volcanoes where vent fluids were collected. (Map courtesy of R. Embley and S. Merle, NOAA/PMEL.)

View this table:
1

Characteristics of the samples used for analysis

SamplesSeamountsVent namesT max (°C)T avg (°C)Depth (m)pHLatitude (DecDeg)Longitude (DecDeg)Volume (mL)Cells mL−1 (±95% CI)
FS431ForecastSnail Scrum16.06.014486.5613.395143.92020503.38 × 105 (3.92 × 104)
FS432ForecastHomer Vent19.06.514516.2313.395143.92020152.37 × 105 (2.70 × 104)
FS445NW Rota-1Brimstone27.919.75601.6414.601144.7753004No data
FS446NW Rota-1Iceberg53.848.05343.7514.601144.77630078.54 × 104 (8.84 × 103)
FS447NW Rota-1Sandy Saddle35.629.05212.6214.601144.77619601.23 × 105 (1.78 × 104)
FS448NW Rota-1Fault Shrimp25.925.05845.2514.601144.77730103.49 × 105 (2.62 × 104)
FS449NW Rota-1Scarp Top17.615.15685.0914.601144.77820003.07 × 105 (3.17 × 104)
FS467NW EifukuSulfur Flow52.742.916125.2621.487144.04240001.34 × 106 (1.16 × 105)
FS468NW EifukuCliff House54.145.115785.2821.487144.04250011.95 × 106 (1.64 × 105)
FS473DaikokuNE Pit15.615.34385.8921.325144.19320192.17 × 105 (1.46 × 104)
FS475DaikokuAlka Seltzer64.545.54143.3621.325144.19113171.67 × 105 (2.40 × 104)
FS479NikkoNorth Vent96.180.24585.5423.081142.32540004.18 × 104 (7.33 × 103)
FS480NikkoTubeworm Hangover26.124.14455.0523.079142.32635224.95 × 105 (3.56 × 104)
FS481NikkoTop Vent35.432.64136.2123.08142.32730592.11 × 105 (6.21 × 104)

Quantitative PCR (qPCR)

qPCR TaqMan assays described previously by Nadkarni et al. (2002) and Takai & Horikoshi (2000) were used to determine the relative abundance of bacterial and archaeal 16S rRNA genes in environmental vent fluids. Plasmid DNA was extracted from Axial Seamount low-temperature diffuse vent clone libraries, purified, and linearized using the WizardPlus SV Minipreps DNA Purification System (Promega). Standards were constructed by mixing equal amounts of four bacterial plasmids for the quantification of bacterial 16S rRNA gene and two archaeal plasmids for the quantification of archaeal 16S rRNA gene. A 1 : 10 dilution series of the plasmid mixtures beginning with an initial concentration of approximately 0.008 ng μL−1 was used to produce standard curves with R2 values >0.998 and efficiencies ranging from 86% to 99%. Each 20-μL reaction contained TaqMan Gene Expression Master Mix (Applied Biosystems), forward and reverse primers at optimized concentrations of either 9 μM (Bacteria) or 8 μM (Archaea), optimized probe concentrations of either 1.5 μM (Bacteria) or 2.0 μM (Archaea), DEPC-treated water, and 2 μL of DNA template. Reactions were performed in triplicate including no template controls on a StepOne Plus Real Time PCR System (Applied Biosystems). Cycles began with initial denaturation for 2 min at 50 °C and 10 min at 96 °C, followed by 40 cycles of 15 s at 96 °C and 3 min at 59 °C. stepone software version 2.0 (Applied Biosystems) was used to analyze the results.

454 tag sequencing

Total genomic DNA was extracted according to Sogin et al. (2006). DNA was quantified using a Nanodrop ND-1000 spectrophotometer, and PCR amplicons of the bacterial V6 region were generated using the pool of five forward primers and four reverse primers using a barcoding approach as described in Huber et al. (2007). All sequencing was carried out on the Roche GS20.

Sequence and statistical analysis

Pyrosequencing reads were passed through quality filters to reduce the error rate (Huse et al., 2007). Any read shorter than 50 nucleotides or containing one or more ambiguous nucleotides was discarded. The expected sample barcode and primer sequences were trimmed from the proximal and distal ends of the reads. Reads lacking a perfect match to the forward primer or a recognizable match to the reverse primer sequence (including rare variants) at either end, reads covering only a portion of the variable region, and reads that could not be assigned unambiguously to a sample were discarded. The resulting datasets contained 229 913 V6 tags. Sequences were assigned taxonomy using the global alignment for sequence taxonomy (GAST) process (Huse et al., 2008), which incorporates the RDP II taxonomy (Cole et al., 2007). Seventy-four thousand seven hundred and sixty-seven tag sequences mapped to the Epsilonproteobacteria. Sequences have been deposited in the VAMPS database (https://vamps.mbl.edu) and the National Center for Biotechnology Information Sequence Read Archive (SRA) under the accession number SRA012278.

A variety of analyses were then carried out using internal SQL database queries, slp pwal (Huse et al., 2010), mothur (Schloss et al., 2009), and primer-e (Clarke & Gorley, 2006). All Epsilonproteobacteria sequences from each sample were pooled and operational taxonomic units (OTUs) were calculated via slp pwal and mothur. To reduce the residual effects of sequencing error and noise, slp pwal uses pairwise alignments of the V6 tags based on the Needleman–Wunsch implementation of ESPRIT (Sun et al., 2009) to precluster the tags at the 2% difference level (corresponding to a single difference in the V6 region) using a modified single-linkage method. Each precluster is represented by its most abundant sequence in the final clustering step, which uses the ESPRIT pairwise alignment distances and the average-linkage method of mothur. The resulting matrix of OTUs shared between samples was normalized to total and transformed via square root and used to calculate a distance matrix with the Morisita–Horn measure (Horn, 1966). Distance matrices were imported into primer-e for a variety of analyses, including hierarchical cluster analysis, principal component analysis (PCA), analysis of similarity (anosim), similarity profile (simprof), and nonmetric multidimensional scaling (MDS) analysis. Analyses were also repeated with all singletons removed. The entire chemical dataset, including temperature, pH, gas (H2S, H2, CH4), Na, Li, Fe, Mn, and Si, was also included for PCA and Spearman correlation analyses.

Results

Site descriptions

The Mariana Arc is located in the western Pacific from about 14–24°N (Bloomer et al., 1989) (Fig. 1). We analyzed vent fluids from five hydrothermally active seamounts along the Arc: NW Rota-1 and Forecast are located in the Southern Seamount Province, while Daikoku, NW Eifuku, and Nikko lie in the Northern Seamount Province (Stern et al., 1988; Embley et al., 2004) (Fig. 1, Table 1). All of the seamounts occur along the active Mariana Arc front, with the exception of Forecast, which may be more closely affiliated with the backarc spreading axis, although this remains unknown (Embley et al., 2004). At all five sites, we sampled fluids that were venting directly out of the seafloor as opposed to those venting from sulfide structures. There are a number of excellent geological descriptions and plume water surveys of the volcanoes sampled elsewhere (Embley et al., 2004, 2007; Baker et al., 2008; Resing et al., 2009); therefore, here, we will only summarize the specific vent sites sampled.

Moving south to north, the first seamount sampled was Forecast, discovered previously by the Japanese in 1993 (Gamo & The Shipboard Scientific Party of the Y9204 Cruise, 1993; Ishibashi & Urabe, 1995). At this site, we found chimneys vigorously venting clear fluids up to 195 °C, one of the highest temperature vent systems known in the Mariana Arc. The two samples used in our analysis were from low-temperature vents located 80 m apart on the summit. Snail Scrum (FS431) was taken in a group of snails about 1 m away from a high temperature structure, and Homer Vent (FS432) was taken in a rock crack, also a few meters away from a different high-temperature chimney. Both of these vents had relatively high pH values (Table 1) and low gas (H2, H2S, and CH4) concentrations (M. Lilley, unpublished data). The second site sampled was NW Rota-1, an actively erupting volcano with abundant diffuse venting (Embley et al., 2006; Chadwick et al., 2008). We sampled the eruptive Brimstone Pit (FS445) during a quiet period, as well as four diffuse vents (FS446–FS449) trending east and downslope from the eruptive area, many of which had abundant white microbial mat, as well as shrimp present. The lowest pH values were found both at the eruptive pit (pH 1.64) and at nearby diffuse vents Sandy Saddle and Iceberg (pH 2.62 and 3.75, respectively). The concentrations of H2 and CH4 were the highest seen on the arc (M. Lilley, unpublished data). In the Northern Seamount Province, we sampled Daikoku, an unusual site with ponds of molten sulfur and thick sulfur crusts, as well as abundant macrofauna, including tube worms, snails, and flatfish (Embley et al., 2007). Alka Seltzer vent (FS475) was located near the molten sulfur cauldron and gas bubbles were visible during sampling; NE Pit (FS473) was at the summit, where a plume of white smoke was sampled (FS473). Nikko, the northernmost site, was similar in venting style and macrofauna to Daikoku and hosted pillars of native sulfur, small ‘pots’ of molten sulfur bubbling out of the seafloor, and extensive communities of tube worms and other fauna (Embley et al., 2007). Site North Vent (FS479) was the highest temperature vent sampled, close to 100 °C, and was a plume of white smoke; Tubeworm Hangover (FS480), an area of dense microbial mat and animals; and Top Vent (FS481), a clear venting fluid with abundant microbial mat and shrimp. Both Nikko and Daikoku fluids had pH values around 5 (although Alka Seltzer had a pH of 3.36) with relatively low gas and iron concentrations (M. Lilley, unpublished data). The final site was NW Eifuku, the deepest of the sites, with venting fluids rich in both liquid and gas CO2, as well as large mussel beds and dense microbial mats (Lupton et al., 2006; Embley et al., 2007; Tunnicliffe et al., 2009). Both samples of sites here, Sulfur Flow (FS467) and Cliff House (FS468), were taken near areas where liquid CO2 droplets were venting with high concentrations of CO2 and H2S as well as minor CH4 and H2 (Lupton et al., 2006). The gas data (CO2, H2S, H2, CH4) are presented in Lupton et al. (2006) and further discussed in Lupton et al. (2008).

Microbial communities in fluids: cell counts, qPCR, and 454 sequencing

All fluids sampled were enriched in cell densities compared with background seawater at the same depth, with the exception of North Vent (FS479) on Nikko and Iceberg (FS446) on NW Rota-1 (Table 1). qPCR results indicate that all fluids were dominated by bacteria, with archaea compromising <3% of the total community, although Brimstone Pit (FS445) on NW Rota-1 had almost 8% archaea (Table 2). Cultivation of thermophiles and hyperthermophiles in anaerobic media was attempted on most samples; out of approximately 250 inoculations, 16 tubes were verified for positive growth via microscopy. These cultures are being examined as part of a separate study. Total DNA was extractable from all fluids, and amplification of the bacterial V6 region was successful. Overall, 229 913 bacterial sequences were obtained from the 14 samples, 74 767 of which mapped to the Epsilonproteobacteria after the GAST process (Huse et al., 2008) (Table 2). The relative abundance of Epsilonproteobacteria in the total bacterial sequences per sample (Table 2) ranged from 3.7% in NE Pit (FS473) from Daikoku to 87.4% in Tubeworm Hangover (FS480) from Nikko. No correlation was found between relative abundance and location or chemical parameter. All subsequent analyses were carried out on this dataset of 74 767 sequences.

View this table:
2

Sequencing statistics and qPCR results for each sample

Total number of bacterial V6 tag sequencesTotal number of Epsilonproteobacteria V6 tag sequencesEpsilonproteobacteria V6 tag sequences (%)Bacteria (%)Archaea (%)
FS43115219393025.8098.33 (±0.26)1.67 (±0.26)
FS43216393424825.9098.61 (±0.16)1.39 (±0.16)
FS4458411146417.4092.04 (+1.21)7.96 (+1.21)
FS44612478868469.6097.31 (±0.07)2.69 (±0.07)
FS4479982251525.2097.02 (±0.42)2.98 (±0.42)
FS44819343303315.7096.79 (+0.08)3.21 (+0.08)
FS44911707391533.4097.64 (±0.24)2.36 (±0.24)
FS46712279549844.8098.80 (±0.19)1.20 (±0.19)
FS468625211728327.6098.54 (+0.46)1.46 (+0.46)
FS473214597863.7097.55 (±0.04)2.45 (±0.04)
FS47510804849378.6099.53 (±0.06)0.47 (±0.06)
FS47911890306325.8097.51 (+0.21)2.49 (+0.21)
FS4809255809387.4099.46 (±0.00)0.54 (±0.00)
FS4818946376242.1099.32 (±0.00)0.68 (±0.00)
  • * Trimmed reads that passed quality control as described in Materials and methods.

  • Based on qPCR as described in Materials and methods.

Taxonomy and total diversity of Epsilonproteobacteria

The taxonomic breakdown at the genera level for each sample is shown in Fig. 2. Dominant genera included Thioreductor, Lebetimonas, Caminibacter, Sulfurovum, Sulfurimonas, and Arcobacter. Sequences that were found<1000 times across all samples were binned into ‘Other’ and included Nautilia, Sulfuricurvum, Hydrogenimons, Nitratifractor, and Sulfurospirillum. Because sites were unequally sampled, rarefaction was used to compare sampling effort and richness across all sites at the 3% difference level (Fig. 3). Results indicate that eight of the 14 sites cluster together, whereas the two vents from Forecast have the highest richness (FS431 and FS432) and two of the vents from NW Rota-1 along with Alka Seltzer from Daikoku have the lowest richness (FS445, FS446, and FS475). It is only in these latter three sites that sampling of Epsilonproteobacteria appears to be near complete. Vents from Forecast had the highest pH and samples FS445, FS446, and FS475 had among the lowest with respect to pH values (Table 1).

2

Taxonomic breakdown and relative abundance at the genus level for the individual tag sequences of Epsilonproteobacteria in each vent fluid sample. Only those genera that occurred >1000 times across all datasets are labeled. All others are lumped into ‘Other.’

3

Rarefaction curves at the 3% difference level for each of the vent fluid samples.

Patterns of Epsilonproteobacteria distribution

All sequences were pooled together to generate OTUs, resulting in 739 OTUs at the 3% difference level. Only five OTUs occurred in all 14 datasets, compromising <0.7% of the total OTUs, but over 10% of the total sequences. Approximately 40% of the OTUs were found in 2–13 samples, comprising 88% of sequences. Over half of the OTUs (59%, representing about 0.6% of the total sequences) were only found in one sample. simprof test of the data indicated that there were significant community structure differences among the individual samples (π=6.23, P value <0.001). At the 3% difference level, nMDS and cluster analysis yielded the pattern seen in Fig. 4. The two vents from Forecast (FS431 and FS432) group together, the five vents from NW Rota-1 group together (FS445–FS449), and the vents from Nikko, Daikoku, and Eifuku group together. One-way anosim test using the distance matrix and seamount resulted in a global R value of 0.667 (P<0.0002), indicating that the groupings by seamount are significant. Pairwise R values indicate that the major differences are found between the two southern seamounts (Forecast and NW Rota-1) and the southern seamounts as compared with all of the northern seamounts (NW Eifuku, Daikoku, and Nikko, Table 3). Nikko is the only seamount that has a significant overlap with the other two northern seamounts, with North Vent and Tubeworm Hangover (FS479 and FS480) clustering with Daikoku and Top Vent (FS481) clustering with NW Eifuku (Fig. 4).

4

(a) Cluster diagram and (b) nMDS 2D similarity plot comparing vent samples at the 3% difference level of Epsilonproteobacteria, with each sample labeled according to seamount. The circles represent similarity boundaries from the cluster diagram.

View this table:
3

Pairwise R values from the anosim test between groupings and seamounts

ForecastNW Rota-1NW EifukuDaikokuNikko
Forecast
NW Rota-10.95
NW Eifuku10.64
Daikoku10.730.5
Nikko0.670.820.08−0.33

Sequences from each seamount were pooled and analysis was repeated with grouping by seamount rather than individual vent site. A UPGMA tree calculated from Morisita–Horn distances between datasets (Fig. 5) supports the finding that Forecast and NW Rota-1 are distinct from one another and from the northern seamounts. No significant groupings or correlations were found based on chemical parameters, including temperature, pH, gas (H2S, H2, CH4), Na, Li, Fe, Mn, and Si (data not shown).

5

Similarity of pooled samples at the 3% difference level from each seamount based on the Morisita–Horn measure. The length of the reference bar represents a distance of 0.10.

Considerable concern has been raised recently over pyrosequencing errors and their contribution to diversity estimates (Quince et al., 2009; Reeder & Knight, 2009; Kunin et al., 2010). In order to account for potential errors, all samples were run on the same sequencing platform (GS20) within a few weeks of one another, and all samples were treated identically from DNA extraction all the way through to sequence analysis. We used stringent quality control measures based on our knowledge of known GS20 sequencing errors (Huse et al., 2007). In addition, here, we report the most conservative methodology to date for reporting OTUs via the slp pwal pipeline (Huse et al., 2010). The slp pwal approach is thought to more accurately predict expected OTUs than multiple alignment and complete-linkage clustering methods and reduces the number of OTUs almost by half compared with the multiple alignment complete-linkage clustering methods (Huse et al., 2010). We also repeated all analyses with singletons removed to insure that potentially erroneous singletons were not skewing community patterns.

Discussion

In this study, we used 454 tag sequencing of the V6 bacterial region to determine the diversity and distribution of Epsilonproteobacteria in rock-hosted diffuse vent fluids from five hydrothermally active seamounts along the Mariana Arc. While universal bacterial primers were used and a large taxonomic diversity of bacteria was detected, we chose to only analyze a subset of the sequences that were classified to Epsilonproteobacteria. Diffuse vents represent a mixture of seawater and vent fluids that mix in the crust before venting on the seafloor (Butterfield et al., 2004). We sampled vent fluids from a large range of depths and locations (∼400–1600 m), and we did not want background seawater organisms from different regions and depths complicating the interpretation of vent-specific patterns. Epsilonproteobacteria are rarely detected in nonplume, nonvent background oxygenated seawater and are considered endemic to sulfidic environments such as deep-sea hydrothermal vents (Campbell et al., 2006).

Taxonomically, over 90% of the sequences that mapped to Epsilonproteobacteria were categorized down to the genus level, leaving only 9% of the sequences unclassified beyond Epsilonproteobacteria, although many were classified down to order and family. Members of Thioreductor spp. were found the most frequently. There is only one named species within this genus, Thioreductor micantisoli, and it was isolated from hydrothermal sediments in the Mid-Okinawa Trough (Nakagawa et al., 2005a). Thioreductor micantisoli is a strict autotrophic mesophile that oxidizes hydrogen while reducing sulfur, and sequences related to Thioreductor spp. have been retrieved from hydrothermal habitats throughout the world's oceans, including the Mid-Atlantic Ridge associated with venting fluids (Reysenbach et al., 2000), the East Pacific Rise in microbial mats (Longnecker & Reysenbach, 2001), and the Indian Ocean associated with gastropods (Goffredi et al., 2004). The phylogenetic position of this genus remains undetermined and the physiological diversity completely unexplored. Sulfurimonas spp. and Sulfurovum spp. were also detected at high relative abundances. These are among the most frequently detected Epsilonproteobacteria in hydrothermal vent habitats, and both are mesophilic microaerobic autotrophs that oxidize elemental sulfur (Campbell et al., 2006). These three genera –Thioreductor, Sulfurovum, and Sulfurimonas– were found at every site, although some at very low relative abundances. The remaining dominant genera, Caminibacter and Lebetimonas, both belong to the family Nautiliaceae and were not found at every site. This family has a number of cultured isolates, all of which are thermophilic autotrophs (or mixotrophs) that use hydrogen as an electron donor and both sulfur and nitrate as electron acceptors. They have been isolated from the East Pacific Rise, Mid-Atlantic Ridge, and the Mariana Arc, where the first acidophilic member was isolated, Lebetimonas acidiphila (Takai et al., 2005b). The largest group of unclassified sequences was found in sample FS446, where two tag sequences were found close to 3500 times and were only resolved to the family Nautiliaceae. The best match revealed by the GAST process is a sequence from Davis & Moyer (2008) from microbial mats from the exact same vent, Iceberg on NW Rota-1. Their sequencing of full-length 16S rRNA gene amplicons indicates that this sequence falls into an unresolved group within the Nautiliaceae (accession number EU574679, IBB_14). The taxonomic breadth found in the vent fluids also included rare members that occurred less than a total of 1000 times across the datasets, such as Nautilia, Sulfuricurvum, Hydrogenimons, Nitratifractor, and Sulfurospirillum.

Because samples were unequally sampled, we used rarefaction analysis rather than absolute OTU numbers or nonparametric estimators to compare richness and sampling effort among samples. As illustrated in Fig. 3, we found that the Epsilonproteobacteria in three of the samples (FS445, FS446, and FS475) appear to be the least diverse as indicated by the flattening of the rarefaction curve. Sample FS445 from NW Rota-1 has the lowest pH (1.64) of all vents sampled in this study, and samples FS446 and FS475 had low pHs as well (3.75 and 3.36, respectively). The other vent with a low pH (FS447, pH 2.62) did not display the same flattening trend. The two vents with the highest community richness as measured by rarefaction, FS431 and FS432, also had the highest pH values in this study (6.56 and 6.23, respectively). While the trend of pH and total richness is not consistent across the entire dataset, there is some evidence to suggest that Epsilonproteobacteria diversity may be constrained by pH, similar to what is seen in soils (Lauber et al., 2009). However, no statistically significant relationship was found between pH and community composition. We used nMDS, cluster, PCA, correlation, simprof, and other statistical tools to examine the distribution patterns of Epsilonproteobacteria and how those patterns may be linked to environmental parameters, including 15 chemical species. Statistical tests (simprof) indicate that there is evidence of significant group structure beyond what one would expect randomly, and the only significant relationship found was between seamount and individual vent groupings (anosim, Table 3). Forecast and NW Rota-1 each formed distinct clusters, with very little to no overlap with each other or with vents found in the northern province of the arc (Fig. 4, Table 3). However, the three sites in the northern province group together, with subgroupings within. This suggests that community membership may be linked to geographical isolation, but it is important to note that at a finer resolution and higher similarity values, many of the individual vents appear to be isolated. While there may be large-scale geographic differences between the northern and the southern seamounts, at the local scale, each vent is practically an island, with little community overlap with other vents, regardless of the seamount location.

The concept of geographical isolation, rather than environmental parameters such as geochemistry, controlling microbial community structure has been observed frequently in microorganisms from terrestrial hot springs (Papke et al., 2003; Whitaker et al., 2003; Takacs-Vesbach et al., 2008; Reno et al., 2009), as well as soils and lakes (see Papke & Ward, 2004; Whitaker, 2006; Green et al., 2008 for reviews). In the marine hydrothermal environment, there is also some evidence of geographic isolation of microbial communities. For example, recent work at Axial Seamount found that individual vents within the caldera maintain a distinct composition over time (Opatkiewicz et al., 2009), and no correlation was found between total bacterial or archaeal community structure and geochemical parameters. However, the authors did find trends in Epsilonproteobacteria community composition that were linked to changes in hydrogen, sulfur, and iron chemistry. Similarly, Huber et al. (2003) found that Epsilonproteobacteria diversity increased over time in vent fluids from Axial Seamount following the 1998 eruption, and this increase followed changes in the mixing of seawater and temperature of the vent. Other qualitative relationships between chemistry and microbial community structure have been observed at deep-sea hydrothermal vents. Nakagawa (2005c) found that there were more chemolithoautotrophs, such as hydrogenotrophic methanogens, in gas-enriched fluids as compared with gas-depleted or normal vent fluids. Again, working at Axial, Huber et al. (2006), found a correlation between the populations of Thermococcales (as defined by the internal transcribed spacer region) and vent chemistry. With the exception of Opatkiewicz et al. (2009), none of these studies used statistical tools to analyze the relationships between chemistry and microbiology, and most performed a very shallow level of sequencing in comparison with this study. The fact that no pattern was seen in our study may be due to a number of factors. It is possible that the bulk measurements carried out here for both chemistry and microbiology may obscure any meaningful relationships, and it is also well known that vent fluid chemistry changes over time. In addition, the level of mixing in arc seamounts is difficult to constrain because of the lack of zero-end member magnesium fluids (or other conservative tracers) and the extremely high volume of water rapidly moving through the rock. Another important factor is that we analyzed only the DNA fraction of the community, meaning we measured all organisms present, regardless of activity. Patterns in the community composition of active Epsilonproteobacteria, as studied by RNA, may reveal more significant relationships with environmental parameters. In addition, the V6 region of the rRNA may be too short or may lack enough variability to detect fine-scale patterns. Using other portions of the genome, such as the intergenic spacer region or genes actively used for metabolism, may help resolve patterns further. Finally, other microorganisms may exhibit different patterns in their distribution and diversity, such as the Zetaproteobacteria (Rassa et al., 2009; Emerson & Moyer, 2010). However, our analysis shows that the biogeographic pattern of Epsilonproteobacteria at these seamounts is statistically significant.

Along the Mariana Arc, it is possible that the central province is a geographical barrier to distribution and exchange between the northern and the southern provinces. The central province is mostly subaerial, with little submarine volcanism found (Embley et al., 2007). However, the two seamounts in the southern province (Forecast and NW Rota-1) were as different from one another as they were from seamounts to the north. Similarly, NW Eifuku and Daikoku have different community compositions, despite the fact that they are located closer together than any of the other seamounts. Clearly, distance is not the distinguishing factor in microbial community composition, and chemically, no statistically significant relationships were observed. While the chemistry of most of these arc volcanoes is dominated by magmatic gas and sulfur, there is still considerable chemical diversity at the vents, with wide ranges of pH, gas concentrations, and temperature, even within a single seamount. Therefore, it may not be surprising that each individual vent hosts its own distinct and somewhat isolated population of Epsilonproteobacteria. Telford et al. (2006) hypothesized that endemic organisms in rare or isolated habitats have a lower probability of successful dispersal, thus allowing for niche specialization and genetic divergence. Hydrothermal vents are certainly isolated and often ephemeral habitats in the deep sea. At the level of resolution provided by 16S rRNA gene 454 tag sequencing, it appears that the pattern seen here is driven mainly by geographical isolation. Our data suggest that there may be barriers to the exchange and dispersal for Epsilonproteobacteria at hydrothermal seamounts of the Mariana Arc. Further studies, including characterization of isolates, manipulative physiological experiments, and genomic characterizations, will help elucidate how to interpret the patterns seen in situ. With the increasing amounts of sequencing data, we are now poised to carry out a much larger census of vent and seamount microorganisms to address the important issue of whether or not microbial endemism exists in benthic marine microbial communities.

Acknowledgements

We thank the NOAA Ocean Explorer Ring of Fire Program, the ROV Jason 2, and S. Murdock for field support and sample collection. This work was supported by a National Research Council Research Associateship Award and L'Oréal USA Fellowship (J.A.H.), NASA Astrobiology Institute Cooperative Agreement NNA04CC04A (M.L.S.), the Alfred P. Sloan Foundation's ICoMM field project, and the W.M. Keck Foundation. This publication is (partially) funded by the Joint Institute for the Study of the Atmosphere and Ocean (JISAO) under NOAA Cooperative Agreement No. NA17RJ1232, Contribution #1814. This is NOAA Pacific Marine Environmental Laboratory Contribution #3542.

Footnotes

  • Editor: Patricia Sobecky

References

View Abstract