OUP user menu

Individual hydrothermal vents at Axial Seamount harbor distinct subseafloor microbial communities

Andrew D. Opatkiewicz, David A. Butterfield, John A. Baross
DOI: http://dx.doi.org/10.1111/j.1574-6941.2009.00747.x 413-424 First published online: 5 November 2009


The microbial community structure of five geographically distinct hydrothermal vents located within the Axial Seamount caldera, Juan de Fuca Ridge, was examined over 6 years following the 1998 diking eruptive event. Terminal restriction fragment length polymorphism (TRFLP) and 16S rRNA gene sequence analyses were used to determine the bacterial and archaeal diversity, and the statistical software primer v6 was used to compare vent microbiology, temperature and fluid chemistry. Statistical analysis of vent fluid temperature and composition shows that there are significant differences between vents in any year, but that the fluid composition changes over time such that no vent maintains a chemical composition completely distinct from the others. In contrast, the subseafloor microbial communities associated with individual vents changed from year to year, but each location maintained a distinct community structure (based on TRFLP and 16S rRNA gene sequence analyses) that was significantly different from all other vents included in this study. Epsilonproteobacterial microdiversity is shown to be important in distinguishing vent communities, while archaeal microdiversity is less variable between sites. We propose that persistent venting at diffuse flow vents over time creates the potential to isolate and stabilize diverse microbial community structures between vents.

  • hydrothermal vents
  • subseafloor biosphere
  • microbial diversity
  • microbial ecology


Submarine hydrothermal vent environments are noted for their complex and dynamic habitats, high energy flow, steep thermal/chemical gradients, multiple microbial energy sources and generally high concentrations of carbon dioxide and reduced gases (Jannasch & Mottl, 1985; Karl, 1995). This results in a high physiological and phylogenetic diversity of microorganisms spanning most of the known microbial metabolisms. Diffuse-flow vents, in which hydrothermal vent fluid mixes with seawater at variable depth levels within the crust, show great variations in temperature and chemical composition (Edmond et al., 1979; Butterfield & Massoth, 1994; Butterfield et al., 1997, 2004). These chemical variations occur over short time intervals (weeks to years) within individual diffuse-flow vent sites due to processes affecting the hydrothermal heat supply or plumbing system. Seismic activity, magmatic intrusions and tidal processes can influence the composition of the source hydrothermal fluids (Butterfield et al., 1997, 2004; Huber et al., 2002; Lilley et al., 2003; Scheirer et al., 2006). Too few diffuse-flow vents have been sampled on a regular basis to know how consistent these variations are between vents within one vent field or in different vent fields, and whether or not the microbial community diversity follows changes in chemistry and temperature.

Previous work at the Marker 33 diffuse site on Axial Seamount located on the Juan de Fuca Ridge found that bacterial diversity increased over a 3-year period following an eruptive event (Huber et al., 2002, 2003). The Epsilonproteobacteria were found to be the most dominant group of Bacteria, whereas thermophilic Methanococcus spp., and uncultured species of Euryarchaeota were the dominant Archaea (Huber et al., 2002, 2003). In these studies, the Epsilonproteobacteria species diversity increased with time and it was concluded that the degree of mixing of seawater with hydrothermal fluid was the most important variable effecting microbial diversity (Huber et al., 2002, 2003). In a separate study it was found that species diversity within the Thermococcales from a range of diffuse-flow vents at Axial Seamount was correlated to the fluid chemistry of the vents, and it was proposed that both the fluid chemistry and microbial subpopulations reflect different fluid flow patterns in the subseafloor (Huber et al., 2006).

To gain a better understanding of microbial ecology in diffuse hydrothermal vents, it is necessary to examine an expanded dataset of microbial diversity and environmental factors in both the spatial and temporal domains. In this study, we used terminal restriction fragment length polymorphism (TRFLP) and 16S rRNA gene sequence analyses to determine the bacterial and archaeal diversity of hydrothermal fluid samples from five vents over the course of 6 years following the 1998 eruption at Axial Seamount (Fig. 1). In an effort to correlate differences in microbiology with differences in vent environmental conditions, we also performed chemical analyses on the vent fluids sampled, and used the statistical software primer v6 (Primer-E) to compare vent microbiology, temperature and fluid chemistry. Our results indicate thermophilic members of the Crenarchaeota and Euryarchaeota were abundant, as well as many epsilonproteobacterial groups. While considerable overlap in the temperature and chemical composition of the fluids was seen between vents, this work provides evidence that each vent studied harbors a unique microbial community structure.


Sample location maps adapted from Huber et al. (2006). (a) Location of Axial Seamount on the Juan de Fuca Ridge, NE Pacific Ocean. (b) Axial Seamount caldera with the location of the five vents included in this study, as well as other areas of known venting.

Materials and methods


Axial Seamount (45°58′N; 130°00′W) is an active submarine volcano located 300 miles off the coast of Oregon at the junction of the Cobb-Eickelberg hotspot and the Juan de Fuca Ridge (Johnson & Embley, 1990) (Fig. 1a). The 3 × 8 km caldera rises 700 m above the mean ridge level, with a caldera floor at 1520 m and steep caldera walls to the W, N and E. Known hydrothermal venting occurs near the caldera boundary fault and in the south rift zone (Butterfield et al., 1990, 2004; Embley et al., 1990). A volcanic eruption occurred in the SE caldera in January 1998 (Baker et al., 1999; Dziak & Fox, 1999; Embley et al., 1999). The volcano has been slowly inflating following the 1998 eruption (Chadwick et al., 2006), with very low levels of seismicity relative to the pre-eruptive period (Dziak & Fox, 1999; R.P. Dziak, pers. commun.).

Five vent sites were sampled: Bag City, a stable diffuse vent in lobate lavas outside of the 1998 eruption zone in the southern portion of the caldera; Cloud, an evolving, high flux, diffuse vent within a jumbled complex portion of the 1998 lava flow; Marker 33, a dynamic vent within a sheet flow portion of the 1998 lava flow zone at which the highest diffuse vent temperature (75 °C) was recorded in 1999; Gollum, a diffuse vent within the ASHES vent field; and Marshmallow, an anhydrite-dominated high-temperature vent in the ASHES field with a hot fluid source just beneath the seafloor (Fig. 1b). All vent samples are located in areas of hydrothermal venting which predate the 1998 eruption (Butterfield et al., 2004).

Fluids were collected over the course of 6 years following the 1998 eruption event using either the ROV ROPOS or the DSRV Alvin. Filtered and unfiltered fluids were sampled at the vents after finding a steady intake temperature on the hydrothermal fluid and particle sampler (HFPS) (Butterfield et al., 2004). In some cases, fluid was first pumped through a 47-mm Millipore (3-μm pore size) cellulose ester filter, before pumping through a Sterivex-GP (0.22-μm pore size) filter. These split filter samples are used to compare the ‘particle attached’ and ‘free-living’ microbial community diversity (Huber et al., 2002, 2003). In other cases, fluid was pumped directly through the Sterivex-GP filter. Temperature and the volume of fluid filtered were monitored throughout the 10–15-min sampling time required to obtain between 1–2 L of fluid. On shipboard, the filters were placed in 50-mL sterile Falcon tubes, frozen in liquid nitrogen, and stored at −80 °C. Chemical analyses were performed on unfiltered and filtered fluid samples.

Geochemical analysis

Analytical methods are described in Butterfield et al. (1997, 2004). Fluids collected with the HFPS were analyzed shipboard for H2S, pH, alkalinity and dissolved silica. On shore, fluids were analyzed for major, minor and trace elements.


DNA was extracted as described by Huber et al. (2002) and purified using the Qiaquick PCR purification columns according to the manufacturer's (Qiagen) protocol. DNA was stored at −80 °C.

TRFLP analyses of environmental DNA were performed using fluorescently labeled universal primers for archaeal and bacterial 16S rRNA gene. TRFLP primer sequences for 16S rRNA genes are as follows (Moyer et al., 1995): ARC21F (5′-[6-FAM]TTC [dP]GG TTG ATC C[dP]G CC[dK] GA-3′), ARC922R (5′-[dP]CCG GCG TTG A[dN]T CCA ATT-3′), BAC68F (5′-[6-FAM]T[dN]A [dN]AC ATG CAA GTC G[dK][dK] CG-3′) and BAC1492R (5′-[dK]G[dP] TAC CTT GTT ACG ACT T-3′), where [dP] indicates the pyrimidine analog P, [dK] indicates the purine analog K and [dN] is an equal mixture of the two analogs (Glen Research, Sterling, VA).

Each PCR reaction (50 μL) contained 1.5 mM MgCl2, 0.8 mM deoxynucleoside triphosphates, 0.5 μM of each (bacterial or archaeal) primer, 1 × Go Taq PCR Buffer (Promega) and 1 U of Go Taq DNA polymerase (Promega). An initial denaturation step of 2 min at 94 °C was followed by 30 cycles of 94 °C for 30 s, 48 °C (Bacteria) or 52 °C (Archaea) for 45 s, and 72 °C for 2 min. The final extension step was 10 min at 72 °C. To minimize PCR drift (Polz & Cavanaugh, 1998), 10 replicate amplifications were pooled, then concentrated and purified with the Qiaquick PCR purification columns according to the manufacturer's (Qiagen) protocol.

A 10 ng sample of the clean PCR product was digested in four separate 10-μL reactions with the restriction endonucleases HaeIII (0.34 U ng−1), MspI (0.5 U ng−1), RsaI (0.34 U ng−1) and BstUI (0.34 U ng−1). Restriction digest reactions were incubated for 6 h at 37 °C except for BstUI (60 °C). The reactions were then inactivated by heating the solutions containing MspI and RsaI to 65 °C, and HaeIII to 80 °C for 20 min, and freezing the solution containing BstUI at −20 °C. Digests were ethanol precipitated according to the manufacturer's instructions (Amersham Pharmacia Biotech Inc.) and analyzed on a MegaBACE 1000 (Molecular Dynamics). Electrophoretic profiles were visualized with the dax software (Van Mierlo Software Consultancy, the Netherlands).

Each vent time series consisted of between three and six time points. Some time points were represented by both 0.2 and 3.0 μm fractions of the microbial community. In these cases, results from both fractions were combined to compare with time points with only the 0.2-μm fraction.

Predicted terminal restriction fragments (TRFs) of clones (see Clone library construction) were obtained using sequencher (v4.1.2; Gene Codes Corporation). Actual TRF profiles of environmental samples and some clones were compared with the predicted TRFs. The presence/absence of phylogenetic groups in environmental TRFLP profiles was determined by the positive identification of a predicted TRF in the actual TRF profile of an environmental sample. In order to score a phylogenetic group as ‘present,’ we used three criteria: (1) the predicted TRF size matched the actual TRF size within ±1 base pairs, (2) the presence of a TRF was replicable and (3) at least two restriction enzyme digests yielded a positive identification of a predicted TRF of a clone belonging to that group. Clones that had fewer than two predicted TRFs within the 50–550 bp size window were not included in further analysis. See Supporting Information, Fig. S1, for examples of TRFLP profiles.

Clone library construction

16S rRNA genes were amplified from purified environmental DNA using PCR with universal bacterial and archaeal specific primers: BAC 8F (5′-AGACGTTTGATCCTGGCTCAG-3′), BAC 1492R (5′-GGTTACCTTGTTACGACTT-3′) and ARC 8F (5′-TCCCGGTTGATCCTGCCC-3′), ARC 1492R (5′-GGCTACCTTGTTACGACTT-3′). Each PCR reaction (20 μL) contained 1.5 mM MgCl2, 0.8 mM deoxynucleoside triphosphates, 0.5 uM of each primer (bacterial or archaeal), 1 × GO Taq PCR Buffer (Promega), and 1 U of Go Taq polymerase (Promega). An initial denaturation step of 5 min at 94 °C was followed by 22–26 cycles of 94 °C for 30 s, 45 °C for 30 s and 72 °C for 2 min. The final extension step was 72 °C for 10 min. PCR cycles were stopped while product concentration was still in the exponential phase, as visualized and quantified on 1% (w/v) agarose gels stained with SYBR green (Molecular Probes) at 15, 20, 25 and 30 cycles. To minimize PCR drift (Polz & Cavanaugh, 1998), six to 10 replicate amplifications were pooled, then concentrated and purified using Qiaquick PCR purification columns according to the manufacturer's (Qiagen) protocol.

Consolidated and cleaned PCR products were cloned with a TA cloning vector kit according to the manufacturer's protocol (Invitrogen). A total of 100–200 colonies for each library were selected and stored on agar plates. Each clone was grown in 100 μL of Luria–Bertani broth shaking at 200 r.p.m. for 1 h at 37 °C and PCR-amplified using the M13F and M13R primers. Each 50-μL reaction contained 1 μL of clone culture, 1.5 mM MgCl2, 0.8 mM deoxynucleoside triphosphates, 1 μM (each) primer, 1 × Go Taq PCR Buffer (Promega) and 1 U of Go Taq polymerase (Promega). The PCR conditions were as follows: 94 °C for 2 min, 28 cycles of 30 s at 94 °C, 45 s at 54 °C and 2 min at 72 °C. The final extension at 72 °C was for 10 min. PCR products were visualized on 1% (w/v) agarose gels stained with SYBR green.

PCR products were sent to the University of Washington High-Throughput Genomics Unit for bidirectional sequencing with the following bacterial and archaeal primers: BAC 8F (5′-AGAGTTTGATCCTGGCTCAG-3′), BAC/ARC 519R (5′-GWATTACCGCGGCCKGCTG-3′), BAC 515F (5′-GTGCCAAGCMGCCGCGGTAA-3′), BAC 907R (5′-CCGTCAATTCMTTTRAGTTT-3′), BAC 926F (5′-AAACTYAAAKGAATTGACGG-3′), BAC 1492R (5′-GGTTACCTTGTTACGACTT-3′), ARC 8F (5′-TCGGTTGATCCTGCC-3′), ARC 515F (5′-GTGGCASCMGCGCGGTAA-3′), ARC 907R (5′-CCGTCAATTCCTTTRAGTTT-3′), ARC 906F (5′-GAAACTTAAAKGAATTG-3′) and ARC 1492R (5′-GGCTACCTTGTTACGACTT-3′).

The sequencher program (Gene Codes Corporation) was used to assemble sequences, and they were subsequently checked for chimeras using the chimera-check program of the Ribosomal Database Project (RDP) website (Maidak et al., 1999). Sequences found to be nonchimeric were submitted for alignment to the rdp sequence alignment program, with common gaps conserved, and manually manipulated using the bioedit v4.7 software. To find closely related sequences for phylogenetic analysis, sequences were submitted to the blast search program (National Center for Biotechnology Information). Between 800 and 1400 nucleotide bases were used in phylogenetic analyses, with only homologous positions included in the comparisons. The phylip 3.5 package was used to construct maximum likelihood trees (dnaml). Bootstrap analysis (seqboot) was used to provide confidence estimates for tree topologies. Negative branch lengths were prohibited. Shannon–Wiener biodiversity index, species observed and rarefaction analysis were computed based on 97% similarity phylotypes using estimates v6.9. For each sample 70 randomizations were performed.

Nucleotide sequence accession numbers

The GenBank nucleotide sequence accession numbers for the sequences determined in this study are as follows:

primer v6

Matrices reporting the presence or absence (community structure) of all identified bacterial and archaeal groups based on TRFLP and clone library data, within each individual sample included in this study, as well as fluid chemistry data spreadsheets, were input to the statistical software primer v6.1.6 (Primer-E) (Clarke & Gorley, 2006). Resemblance matrices (Bray–Curtis for TRFLP and clone libraries, Euclidean distance for Chemistry), where each vent time point was compared with all other time points from all vents, were used in multidimensional scaling (MDS) analyses to create two-dimensional (2D) similarity plots. The purpose of MDS plots are to represent samples (here, by TRLP community structures and/or physio-chemical environmental parameters) as points in low-dimensional space (here 2D). The MDS plots were created such that the relative distances apart from all points are in the same rank order as the relative dissimilarities of the samples. Therefore interpretation is made straightforward in that points that are close together represent samples that are very similar in community composition (or environmental variables), so that the x and y axes on MDS plots are nondimensional similarity distances with no required or appropriate units or scale (Clarke & Warwick, 2001).

Primer can also be used to relate percentages of minimum similarity within similarity boundaries (dotted ellipses in plots). These represent the minimum likeness (% similarity) of all samples included inside the boundary. Decreasing the threshold percent similarities increases the 2D space and number of samples included within the ellipse. Similarity boundaries are used here to outline clusters of samples (Fig. 2), and relate the minimum similarity of every sample within each cluster.


Primer 2D similarity plot (MDS) comparing vent time points via the entire TRFLP dataset. Vents are abbreviated as follows: Marshmallow (MA), Gollum (GO), Bag City (BC), Marker 33 (33) and Cloud (CL). The presence or absence of 121 different bacterial and archaeal TRFLP groupings were considered in comparing vent time points. Time points grouping closely together have similar microbial community structures. With only one exception (Marker 33, 1999), community structures were more alike within the same vent than compared with time points from other vents. Dotted ellipses surrounding groups of time points represent similarity boundaries.

anosim and simprof tests were carried out to examine the statistical significance of relationships between vents and between samples. Primer best-fit analyses, which examined the correlation and forcing of environmental parameters on the microbial community structure, were also carried out. More information on the statistical software primer v6.1.6 can be found on the Primer-E website (http://www.primer-e.com).


Fluid temperature and chemistry

Diffuse vents do not have a well-defined point of exit from the seafloor, and, in difficult cases, changes in measured temperature from year to year cannot be conclusively interpreted to mean that the subseafloor reservoir temperature is increasing or decreasing. At Marker 33 in particular, the configuration of the vent (a long, wide crack in sheet flow), made it difficult to pinpoint and sample the strongest source; hence, the reported maximum temperatures were probably not the actual maximum temperatures. At the other end of the spectrum, Cloud vent has a large pipe-like opening with uniform temperature, where measurements were highly accurate, replicable and indicative of changes in the warm reservoir temperature. In all cases, the temperature measured does correspond to the fluid we collected and analyzed for chemistry and microbiology; hence, regardless of any uncertainty in the absolute trend of temperature over time in some vents, it is valid to correlate temperature, chemical composition and microbial community structure within our dataset.

The fluid temperature and selected chemistry data (recognized as being biologically relevant) is presented in abbreviated form in Table 1. Trends in the chemistry dataset are shown in Fig. S2. The vents studied ranged in temperature from 7 to 222 °C, and fluid temperature changes at individual vents showed different patterns after the 1998 eruption (Fig. S2a). Marker 33 showed an initial increase in temperature between 1998 and 1999 that was followed by a decrease and apparent leveling out between 2001 and 2003. Nearby Cloud vent showed a decay in temperature from July 1998 to August 2004. Bag City, located outside the 1998 lava flow area, was less affected, showing little change in temperature over the time series. The ASHES field vents, Gollum and Marshmallow, are approximately 3 km from the 1998 lava flow and may not be responding to the eruption. The highest temperature at Marshmallow was measured in 2001, while Gollum was relatively high in 2000, and was again rising in 2004 (Fig. S2a). The complete temperature time series for Marshmallow is not included in Fig. S1a for scaling purposes, but is shown here: 1998 (81.0 °C), 1999 (73.6 °C), 2000 (129 °C), 2001 (222 °C), 2002 (NA), 2003 (NA), 2004 (74.4 °C).

View this table:

Fluid temperature and chemical value ranges for the five vents included in this study

VentChemistry value ranges over time series*
Date rangeTmax (°C)H2/heat (nM J−1)CH4/heat (nM J−1)H2S/heat (nM J−1)Fe/heat (nM J−1)
Marker 331998–200412.5–74.60.003–0.140.035–0.421.2–514.6–220
Bag City1999–200412.9–31.30.0010–0.00200.072–0.110.44–5.34.2–59
  • In all, 16 thermo/chemical parameters were measured and used in the study. This table presents temperature and four parameters most closely linked to metabolic activity.

  • * In all 16 thermo/chemical parameters were measured and used in chemical and microbial evaluations: [Mg], pH, H2S/Si, Fe/heat, Fe/Mn, Mn/heat, CH4/heat, H2/heat, dAlk/dSi, H2S/heat, [Cl], ClEM, NH4+/heat, NH4++NH3, Tmax, Tavg.

Figure S1b–d show the trends of biologically relevant parameters, H2S, H2 and Fe(II), respectively. These plots were chosen for their significance as known electron donors for microbial populations, and because they show evidence of biologically mediated reaction in the subseafloor at this study site (Butterfield et al., 2004). Because every diffuse fluid sample collected may be affected by a variable degree of seawater mixing, it is most useful to compare ratios (e.g. component/heat or component/silica) that normalize out the effect of seawater mixing and yield information about the source characteristics. H2S/Si (Fig. S2b) shows the general trend of decreasing H2S concentrations over the time series in all vents, which is consistent with decreasing magma chamber degassing, decreasing vapor content or changing reaction zone temperature or redox state after 1998 (Butterfield et al., 1990, 1997, 2004; Von Damm, 2000; Lilley et al., 2003; Seyfried et al., 2003). Within each year, the H2S/Si ratio falls as vent temperature decreases, reflecting progressive oxidation of hydrogen sulfide as oxygenated seawater is entrained below the seafloor (Butterfield et al., 2004). The ratios of H2 (Fig. S2c) and Fe (Fig. S2d) to heat underwent more complicated changes and support individual vent reactions to the perturbation.

Primer MDS analyses were performed to compare vents by temperature and 14 fluid chemistry parameters ([Mg], pH, H2S/heat, Fe/heat, Fe/Mn, Mn/heat, CH4/heat, H2/heat, dAlk/dSi, H2S/Si, [Cl], ClEM, NH4+/heat and NH4++NH3) individually and in combination. We performed several targeted sets of MDS analyses using individual, selected groups and all of the above listed chemical parameters aimed at looking for trends within and between vents. Specifically we focused on those parameters recognizable as being biologically relevant: [Mg] as a proxy for hydrothermal fluid mixing with seawater, Mn/heat as a proxy for water/rock reaction, H2S/heat or H2S/Si, Fe/heat, CH4/heat and pH. The resultant MDS plots for three of such analyses are shown in Fig. S3. anosim results did not distinguish all of the vents based on the 16 temperature and chemical variables used (Fig. S3).

The highest temperature fluids (222 °C) of the five vents included in this study were sampled from Marshmallow in 2001. There is, however, no evidence that the microorganisms sampled from this fluid were active at 222 °C. In fact similarities in community structure (dominated by ‘high-temperature’ Archaea) between 2001 and the rest of the time points sampled from Marshmallow point to a lack of change in the microbial community structure between 2001 and lower temperature years (Fig. 2). This may have to do with the physical process by which microorganisms are delivered from their subseafloor habitat to the sampling device.

Clone libraries

Clone libraries were sequenced using universal 16S primers to identify the bacterial and archaeal peaks present in TRFLP profiles. A total of 1007 clones (574 Bacteria, 433 Archaea) were sequenced (≤1400 bp each). Using these clone sequences, we were able to identify between 70% and 100% of all TRFLP peaks in all profiles.


The dominant subsurface hydrothermal and seawater groups of Bacteria and Archaea that were detected using TRFLP, as well as information about their known metabolic pathways from cultured representatives, are shown in Table 2. Thermophilic members of the Crenarchaeota and Euryarchaeota are abundant, as well as many epsilonproteobacterial groups. Many uncultured groups were also identified, including 39 Epsilonproteobacteria. The seawater microbial groups (36 TRFLP groups) present were dominated by marine group I Crenarchaeota (seven TRFLP groups), marine group II Euryarchaeota (eight TRFLP groups), Alpha- (three TRFLP groups) and Gammaproteobacteria (10 TRFLP groups). Here, TRFLP groups are defined as TRFs identified from this study using clone library sequences (see Materials and methods).

View this table:

Dominant subseafloor hydrothermal bacterial and archaeal groups identified using TRFLP

TRFLP groupNo. of TRFLP phylotypesT (°C) range of growthAnaerobic/ aerobicMetabolic processElectron acceptorElectron donorChemical evidence
Desulfurococcales260–100AnaerobicSulfur reductionS0, thiosulfateH2H2 depletion, H2S elevation
Uncultured groups28
Archaeoglobus265–85AnaerobicSulfur reductionThiosulfate, SO42−H2H2 depletion, H2S elevation
Methanocaldococcus148–94AnaerobicMethanogenesisCO2H2CH4 elevation
Methanothermococcus140–75AnaerobicMethanogenesisCO2H2CH4 elevation
Thermoplasmatales12–40EitherIron oxidationS0, O2Fe(II)
Uncultured groups8
Sulfurimonas22–36EitherH, S oxidationO2, NO3H2S, H2, S, thiosulfateH2S depletion
Uncultured groups392–80EitherH, S oxidationO2, NO3H2S, H2, S, thiosulfateH2S depletion
Uncultured groups392–80EitherNitrate reductionNO3, NO2, N2OOrganics, H2SNO3 depletion, presence of NO2 and N2O
  • The metabolic pathways that these groups are thought to be carrying out are also listed. Chemical evidence is discussed in detail in Butterfield et al. (2004).

Primer MDS analyses were also carried out on TRFLP data, consisting of the presence or absence of 121 identifiable groupings of Bacteria and Archaea, to compare the microbial community structure at each vent over time. Each vent was seen to consist of a specific community, which, while changing over time, conserved a distinct community structure relative to the other vents (Fig. 2). With one exception (Marker 33, 1999, compared with Marshmallow), the community structure at each time point was >50% similar to all other time points taken from the same vent. 1999 was the hottest time point sample taken from Marker 33. This temperature peak corresponded to a higher abundance of ‘high-temperature’ Archaea sampled from Marker 33 in 1999 than other time points sampled from the same vent. anosim tests revealed that with the exception of Marker 33 compared with Marshmallow (the highest temperature vent of the five vents studied), the microbial community structures of all vents were statistically different from each other. When the 1999 time point from Marker 33 was treated as an outlier, anosim results showed all vents to consist of statistically distinct communities (data not shown).

The Epsilonproteobacteria made up the most important fraction (numerically, 41/121 TRFLP groups) of the microbial community detected using TRFLP. In addition to overall community structure analysis, the Epsilonproteobacteria portion, as well as other subsets (all Archaea, all Bacteria, high-temperature Archaea, seawater Archaea, seawater Bacteria) of the TRFLP dataset were analyzed separately. MDS results of the Epsilonproteobacteria community revealed similarly distinct structures between the vents (Fig. 3a). The high-temperature Archaea results also showed distinct community structures between most vents, but there was much more overlap between vents than in the Epsilonproteobacteria community structures (Fig. 3b). Lastly, the Epsilonproteobacteria sequences from eight clone libraries, used to identify TRFLP peaks, were grouped into 97% similarity groupings, and the presence or absence of each group was recorded over the time series. MDS analysis of this dataset substantiates the distinct community structures between vents documented from TRFLP data within the Epsilonproteobacteria portion of the dataset (Fig. 3c).


Primer 2D similarity plots (MDS) comparing vent time points via different subsets of the microbial dataset. Vents are abbreviated as follows: Marshmallow (MA), Gollum (GO), Bag City (BC), Marker 33 (33) and Cloud (CL). (a) MDS plot comparing vent time points via the Epsilonproteobacteria subset of the TRFLP dataset. Marshmallow time points were left out of the plot for viewing ease, as the vent has very low Epsilonproteobacteria diversity because it is a higher-termperature vent. Therefore, the Marshmallow time points were treated as outliers. (b) MDS plot comparing vent time points via the high-temperature Archaea subset of the TRFLP dataset. (c) MDS plot comparing vent time points via the Epsilonproteobacteria subset of the clone library dataset compiled to identify TRFLP peaks. Epsilonproteobacteria sequences were grouped into 97% similarity groupings. The presence or absence of 76 groups were considered in comparing vent time points. Some of the 97% similarity groupings (16S rRNA gene clone library clusters) were not detected using TRFLP. A single TRFLP group could contain as many as eight clone sequences and five 97% sequence similarity groupings.

Best-fit analyses were carried out to search for correlations between individual fluid temperature and chemistry parameters and the microbial community structure that would explain the distinct patchiness between vents. As expected, temperature and magnesium concentration, two proxies for the degree of subseafloor mixing between seawater and hydrothermal fluid, were determined to be the most influential temperature and chemical parameters on the overall microbial community structure (r2=0.61 and 0.58, respectively) (Table 3). However, when best-fit analyses were carried out searching for correlations between the same temperature and chemical parameters, and the Epsilonproteobacteria subset of the microbial community structure, H2/heat, Fe/heat and H2S/heat were determined to be the most influential of the temperature and chemical parameters (r2=0.59, 0.58 and 0.51, respectively) (Table 3).

View this table:

Primer best-fit analysis

EpsilonproteobacteriaTotal TRFLP
  • Primer best-fit analysis r2 values evaluating the influence of thermo/chemical parameters on the total and Epsilonproteobacteria subset of the microbial community structure detected using TRFLP. Table lists the most influential parameters in each case.


A molecular fingerprinting technique was used to compare the microbial communities associated with five hydrothermal vents over a 6-year period, within the Axial Seamount caldera following the 1998 eruption. We addressed fundamental questions surrounding the subsurface biosphere associated with hydrothermal vents, primarily, is there spatial and temporal variation in the microbial community structure associated with different vents, and, if so, is that variation linked to differences in vent fluid chemistry and temperature?

From our results it is clear that the chemical parameters measured from vent fluids, specifically those known to be important substrates and byproducts of hydrothermal vent microbial metabolisms, were not consistently distinct between different vents. Rather, the vent fluid chemical properties varied over time and there was considerable overlap between different vents (Table 1, Figs S1 and S2). Multiple vents lie within the 50% similarity boundaries in the statistical analysis of the chemical dataset. No matter which subset of the temperature and chemistry dataset was considered, MDS plots showed a large degree of overlap between most vent time series, and anosim did not statistically distinguish the vents studied.

In contrast, each vent maintained a consistent and statistically distinct microbial community structure, as determined by TRFLP (Fig. 2). Epsilonproteobacteria diversity was found to be important in distinguishing the overall microbial community structures. When considering the Epsilonproteobacteria microdiversity in isolation, MDS analysis showed patterns in the microbial community structure between vents (Fig. 3) that were similar to the overall microbial diversity patterns. However, in the case of the Epsilonproteobacteria subset, community changes were linked to changes in iron, hydrogen and sulfur fluid chemistry (Table 3). Moreover, based on MDS analyses, vents that are closer together (e.g. Marker 33 and Cloud compared with Gollum) are as different in microbial community structure as vents that are far apart (Fig. 2). Therefore, the distinguishing factor is not necessarily distance.

Microbial community composition and comparison

Although our clone libraries were able to identify between 70% and 100% of all peaks in all profiles, because of the molecular nature of the 16S rRNA gene, PCR-based TRFLP technique, analyses may have missed microbial groups making up numerically small fractions of the overall community sampled. Furthermore, without cultured representatives from all of the TRFLP groups detected, it is not possible to describe the physiological diversity, including the metabolic potential of the microbial community to affect the concentrations of measured electron acceptors and donors in vent fluid.

Correlations between Epsilonproteobacteria groups and H2, H2S and Fe concentrations indicate that there is a link between the microbiology and fluid chemistry of the vents studied, as expected (Table 3). However, there does not appear a strong correlation between overall microbial community structures, and the key environmental properties including temperature and 15 chemical parameters measured in the vent fluids (r2 data not shown). This can also be discerned by comparing Fig. 2, showing the microbial community structures between vents and Fig. S3, comparing the physicochemical signature between vents.

Metabolic activity should be linked to the chemical environment. However, sampling a diffuse-flow vent where the physical process of fluid-tearing microorganisms from the walls of the rock where they are living, and delivering them to the sampling device as fluids exit the seafloor averages the microbial community over the entire fluid flow path. Therefore, our sample set likely crosses boundaries between major metabolic groups, and physicochemical regimes, which lead to weak correlations between the physicochemical parameters and the microbial community structure.

In addition to the r2 data shown in Table 3, subtleties of the 2D MDS plots show that in some cases vent samples are distinct based on physicochemical parameters (e.g. Cloud and Bag City, Fig. 3). However, MDS patterns of the physicochemical dataset indicate that physicochemical properties overlap much more between vents than the microbial metrics, in part because the temporal variability in chemistry is large. Within the chemical variables that we considered, individual vents do not maintain a unique composition over time.

The microbial populations that are released from diffuse-flow vents are physiologically diverse and include different thermal groups, anaerobes and aerobes, and chemolithotrophs and heterotrophs (Huber et al., 2002). This is due to the plumbing system in diffuse vents in which hydrothermal fluid interacts with down-welling seawater resulting in the mixing of microorganisms from different subseafloor strata. Therefore, it is difficult to distinguish which groups of microorganisms inhabit the thermochemical gradients in the subseafloor. Furthermore, microscopic evidence of microbial communities from diffuse-flow vents show a high percentage of clumps of microorganisms vs. single cells, thus providing evidence that the microbial communities sampled may represent a steady-state flux of organisms derived from stable subseafloor biofilms. It is very likely that the disconnect between the overall microbial community and chemical parameters is a result of this homogenization of physiologically diverse Bacteria and Archaea in diffuse-flow fluids.

Importance and diversity of Epsilonproteobacteria

Epsilonproteobacteria, previously reported as wide spread in the hydrothermal vent environment (Haddad et al., 1995; Moyer et al., 1995; Cary et al., 1997; Lynch, 2000; Longnecker & Reysenbach, 2001; Lopez-Garcia et al., 2002), were found to play an important role in shaping the overall community structures sampled. Although it was difficult to correlate fluid chemistry with microbial community trends, the Epsilonproteobacteria fraction of the community structure showed a relationship with hydrogen, sulfur and iron chemistries. This is consistent with the metabolic properties of the Epsilonproteobacteria isolated from vent environments (Nakagawa et al., 2005; Campbell et al., 2006).

Epsilonproteobacteria are wide spread in the deep-sea hydrothermal vent environment associated with warm vent fluids, microbial mats and macrofauna (Moyer et al., 1995; Corre et al., 2001; Longnecker & Reysenbach, 2001; Reysenbach & Shock, 2002). Epsilonproteobacteria are physiologically diverse, having the ability to metabolize sulfur autotrophically or heterotrophically, use hydrogen as an electron donor and a variety of other elements, such as oxygen and nitrate, as electron acceptors under thermophilic conditions. Their metabolic flexibility allows them to colonize and proliferate in many vent microenvironments (Campbell et al., 2006). However, the presence of many uncultured groups of Epsilonproteobacteria points to our need for a better understanding of their phylogenetic and physiologic diversity in the vent environment. A quantitative study to estimate their abundance in this environment is yet to be carried out.

Geographical trends in the distribution of microorganisms

Geographical trends in the distribution of hot-spring microorganisms have been demonstrated previously in terrestrial vent environments (Papke et al., 2003; Whitaker et al., 2003). Holden et al. (2001) found that Thermococcus spp. isolates formed groups that correlated with their sampling location, and it has been suggested that distinct microbial populations may exist at individual vents owing to differences in the physical and geochemical properties of the vents (Lilley et al., 1984; Jannasch & Mottl, 1985). Studies have shown that vents that are only meters apart may have significantly different chemistries (Butterfield et al., 1997, 2004; Von Damm & Lilley, 2004). Nakagawa et al. (2005) found that variability in the composition of deep-sea hydrothermal vent microbial communities, between multiple vents in the same field, was linked to the chemistry of the sampled fluids. Gradients in fluid chemical composition at different vent sites showed evidence of microbial activity (Butterfield et al., 2004) and pointed to likely differences in metabolism as temperature decreases and various electron donors are consumed (Jannasch & Mottl, 1985; Butterfield et al., 1997). Relationships between environmental characteristics and microbial populations have been demonstrated in terrestrial hot springs as well (Blank et al., 2002; Ferris et al., 2003).

Within the Axial Volcano hydrothermal areas, a correlation in Thermococcus populations and vent chemistry has been noted (Huber et al., 2006). Specifically, it was found that vents with low hydrogen sulfide concentrations and evidence for iron reduction and iron mobilization at low temperature had subgroups of Thermococcus that did not appear in the warmer, more sulfide-rich, lower iron vents. (High levels of hydrogen sulfide in low-temperature vents keep dissolved iron concentrations low due to iron sulfide mineral solubility.) Cloud vent is unique, both in terms of chemistry and microbial community structure. Cloud vent has continued to decrease in temperature through 2007 (data not shown), indicating that it may be linked to the cooling of part of the 1998 volcanic dike, rather than to a persistent upwelling zone from a deeper heat source. In MDS plots for chemistry and microbial community structure (Figs 2 and 3), Cloud plots around the edges of the 2D perimeter, showing the least amount of overlap with other sites, especially in the high-temperature Archaea (Fig. 3b). Cloud and Bag City vents have no overlap in chemical or microbial community structure (Figs 2 and 3). Conversely, the chemical properties of Marker 33 and Marshmallow vent have considerable overlap in all of the 2D MDS plots (Figs 2 and 3) and also have the highest degree of overlap within the high-temperature archaeal subset (Fig. 3b).

Physical isolation

In addition, or in accompaniment with the fluid chemistry and temperature, subsurface geology, permeability and plumbing structure of a vent may also control how fluids flow through the crust (Butterfield et al., 2004; Scheirer et al., 2006) and therefore may be important in influencing the microbial community structure.

Individual eruptions can inflate, collapse and form networks of lava tubes. Heterogeneous permeability structure could result from dynamic volcanic stratigraphy and fissuring (Fornari & Embley, 1995). Dikes and eruptive fissures also channel fluids from the subsurface heat source, and their influence can vary horizontally over scales as small as tens of meters. The vertical permeability created by these features is important in determining the distribution of hydrothermal venting at the seafloor. And the localization of hydrothermal venting at discrete sites above a magmatic heat source indicates that impermeable features must exist, giving the vertical fluid pathways a limited lateral extent and thereby potentially isolating individual vents.

Further evidence that geologic structure controls hydrothermal fluid pathways within the Axial Seamount caldera can be found simply by mapping the extent of venting. All known venting is associated with the caldera boundary fault and rift zones (Butterfield et al., 1990, 2004; Embley et al., 1990). Much of the southern half of the caldera boundary fault is hydrothermally active, with long-term activity at specific sites before and after the 1998 eruption (Butterfield et al., 2004). In contrast, all known hydrothermal activity in the northern half of Axial caldera is limited to one relatively small area of venting known as the CASM vent field (Fig. 1). Areas with no known venting may represent lower permeability or established down-flow zones. The persistence of diffuse venting at specific sites over significant periods of time (>18 years observed) provides the opportunity for microbial communities to adapt to specific conditions at each site, thereby setting up divergence in the community structure, especially at mesophilic and moderately thermophilic temperatures.

This study is one of the very few to combine temporal and spatial information to compare microbial and chemical analyses of multiple vents within a single large hydrothermal area. The evidence that individual submarine hydrothermal vents from the same vent field harbor persistent and distinct microbial populations has not been shown previously, and has significant implications for future modeling efforts of the subseafloor biosphere connected to hydrothermal vents.

Supporting Information

Fig. S1. TRFLP profiles from different vents and time points throughout the course of this study.

Supporting Information

Fig. S2. Thermo/chemical trends. Fluid temperature and chemistry plots from each vent over the course of the time series. Vents are abbreviated as follows: Marshmallow (MA), Gollum (GO), Bag City (BC), Marker 33 (33) and Cloud (CL). A) Maximum observed temperature, B) H2S/Si (mol/mol) C)H2/heat (nmol/J) D) Fe/heat (nmol/J).

Supporting Information

Fig. S3. Primer 2D similarity MDS plots comparing vent samples via (a) dilution-dependent variables: Tmax, [Mg], [H2S], [Fe], and pH, (b) reaction-dependent variables: CH4/heat, H2S/heat, Fe/Mn and Fe/heat, and (c) all 16 thermo/chemical parameters. Vent time points closer together have similar environmental conditions (as measured by thermo/chemical parameters). Vents are abbreviated as follows: Marshmallow (MA), Gollum (GO), Bag City (BC), Marker 33 (33) and Cloud (CL).

Supporting Information

Table S1. Fluid temperature and chemical parameters measured and used as part of this study.

Please note: Wiley-Blackwell is not responsible for the content or functionality of any supporting materials supplied by the authors. Any queries (other than missing material) should be directed to the corresponding author for the article.


We are indebted to Dr Marv Lilley, Eric Olson and Benjamin Larson for their work on the gas chemistry data which was included in comparison studies. Kevin Roe performed much of the chemical analysis. We thank ROV ROPOS group for their efforts in collecting samples over the course of this work. This publication is partially funded by the NOAA/PMEL Vents Program and the Joint Institute for the Study of the Atmosphere and Ocean (JISAO) under NOAA Cooperative Agreement No. NA17RJ1232, JISAO contribution number 1594, PMEL contribution number 3237. This work also received support via an NSF IGERT grant (DGE-9870713) and NASA Astrobiology through the Carnegie Geophysical Institute.


  • Editor: Gary King


View Abstract