Our survey combined conventional Sanger sequencing with 454 tag sequencing; the first approach provided more precise information on the taxonomy of the sequences, while the second allowed rapid generation of large numbers of sequences without the bias introduced by the cloning step, thereby enabling a deeper survey and better coverage of bacterial diversity and the detection of rare species. Furthermore, while only semi-quantitative, 454 tag sequencing provides a more accurate estimate of the relative abundance of OTUs in the microbial communities (Sogin et al. 2006).
Neither of the two molecular approaches used in this study yielded any archaeal sequences. Alain et al. (2004) pointed out that while mesophilic Archaea are known to be ubiquitous, only thermophilic and hyperthermophilic strains have been detected in hydrothermal vent environments. Furthermore, these strains were mostly anaerobic or microaerophilic organisms, while diffuse flow habitats are zones of mixing between hydrothermal fluids and well-oxygenated background seawater (Alain et al. 2004). Roussel et al. (2011) used nested PCR to improve yield of 16S rRNA gene PCR product from samples from three hydrothermal vent sites on the Mid-Atlantic Ridge. Even with this approach, results were mixed with not all samples yielding sufficient PCR product for cloning.
The use of 454 tag sequencing technology allowed the detection of 29 different bacterial phyla, more than four times the number of phyla detected by Sanger sequencing. The substantially greater information yield from the 454 tag sequencing-based survey increased the power of the statistical analyses, thereby reducing the probability of type II errors, the failure to reject a false null hypothesis, that could have prevented detection of patterns related to habitat type. Yet, the relative abundance of the major taxonomic groups detected in each sample was comparable with the Sanger sequencing survey. Unsurprisingly, the Epsilonproteobacteria, which play a major role in the cycling of nitrogen and sulfur, were dominant in all libraries. This has been observed in numerous microbial diversity surveys of hydrothermal habitats, including venting fluids (Huber et al. 2010), mats covering surfaces in the vicinity of vents (Zhou et al. 2009; Flores et al. 2011; Lanzen et al. 2011), or in association with vent fauna (Alain et al. 2002, 2004; Lopez-Garcia et al. 2002). Both of our surveys found the facultative anaerobic mesophilic sulfur-oxidizing genus Sulfurovum to be dominant in all samples (Inagaki et al. 2004). The genus Sulfurimonas, whose members are also mesophilic chemolithoautotrophs using similar electron donors but in aerobic conditions (Inagaki et al. 2003), was mostly represented in High Flow samples. Other relatively abundant Epsilonproteobacterial genera included the mesophilic to thermophilic and strictly chemolithoautotrophic Thioreductor, Hydrogenimonas, Nitratifractor, and Nitratiruptor, all using molecular hydrogen as electron source in anaerobic to microaerophilic habitats (Takai et al. 2004; Nakagawa et al. 2005a,b). These genera have in common metabolisms well adapted to vent conditions and are therefore frequently detected in hydrothermal environments (Campbell et al. 2006).
Both surveys also agreed on the high relative abundance of the class Gammaproteobacteria in Low Flow libraries, but a large proportion of the OTUs could not be classified to the genus level, and the majority of the identified genera accounted for <1% of the sequences. However, the Sanger survey showed that most of these OTUs were closely related to uncultured bacteria collected from vent chimneys or living in association with vent fauna that were also sampled from the Endeavour Segment of the Juan de Fuca Ridge, and described, respectively, as methanotrophs and decomposers involve in the degradation of organic debris (Alain et al. 2002; Wang et al. 2009). These descriptions are consistent with the most abundant genera that were confidently identified in this study, such as Leucothrix, Endozoicomonas, Methylosarcina, and Dasania from the pyrosequencing survey and Granulosicoccus and Methylobacter from the Sanger sequencing survey (Harold and Stanier 1955; Bowman et al. 1993; Wise et al. 2001; Kurahashi and Yokota 2007; Lee et al. 2007a,b). The occurrence of methanotrophs in our samples suggests the presence of CH4 in hydrothermal fluids emitted at these sites, confirmed by the elevated concentrations measured by de Angelis et al. (1993). Free-living methanotrophs have been reported from many hydrothermal sites (Takai et al. 2006) and are also part of the epibiotic fauna found on the Mid-Atlantic shrimp R. exoculata (Zbinden et al. 2008; Guri et al. 2012). Interestingly, only two relatively abundant Gammaproteobactera genera found in our samples, Ectothiorhodosinus and Thiohalophilus, have been described as sulfur oxidizers (Gorlenko et al. 2004; Sorokin et al. 2007). The dominance of Epsilonproteobacteria at all sites would suggest that metabolisms using sulfur compounds as electron donors should also be common among the metabolically diverse Gammaproteobacteria. It is possible that diffuse flow conditions at these sites provided sulfur-metabolizing Epsilonproteobacteria with a competitive advantage over their Gammaproteobacteria counterparts.
The most abundant genera detected within the class Deltaproteobacteria included strictly anaerobic and heterotrophic genera, such as Desulfobulbus, Desulfocapsa, and Desulforhopalus detected by both methods, and Desulfonema, Desulfuromusa, and Desulfoluna detected only by 454 tag sequencing (Widdel and Pfennig 1982; Widdel et al. 1983; Liesack and Finster 1994; Isaksen and Teske 1996; Janssen et al. 1996; Suzuki et al. 2008). Members of these taxonomic groups are sulfur reducers, and their presence along with sulfur oxidizers from the Epsilonproteobacteria and the Gammaproteobacteria classes might suggest internal sulfur cycling within these habitats, as proposed by Forget et al. (2010) and Lanzen et al. (2011) in studies of other low-temperature hydrothermal environments.
The OTUs that belonged to the other detected major phyla, including the Bacteroidetes, Actinobacteria, Chloroflexi, and Firmicutes, were all members of heterotroph genera, and most likely involved in the decomposition of organic matter. The minor groups were metabolically diverse, and a relatively large proportion could not be classified to the genus level, but identified genera were also heterotrophs.
Diversity indices and community composition
Our results showed a significant relationship between microbial community composition and habitat type, suggesting a predictable pattern in High Flow and Low Flow environments. In the tree constructed from the cluster analysis, all libraries sampled from a same habitat at different vent sites, except one, grouped together, and the spatial separation of the High Flow and Low Flow libraries in the NMDS ordination plot was highly significant. The Low Flow library from Clam Bed did not cluster with any other habitat, indicating that the microbial communities from this Low Flow site were distinct from all other communities investigated. Furthermore, the spatial distribution of the Low Flow libraries on the NMDS plot, more widely spread along both axes, suggested that the overall diversity within this habitat, when considering all samples, was higher than that of the High Flow habitat. The distinct composition of the Low Flow sample from Clam Bed was primarily responsible for the apparently higher diversity in the Low Flow habitat. Removing the Clam Bed Low Flow library from the NMDS analysis resulted in there being no significant difference between High Flow and Low Flow libraries in terms of overall diversity. The High Flow library collected from the same site did not diverge from the other libraries in terms of composition. However, both samples from Clam Bed had significantly lower Chao1 richness and Shannon–Wiener diversity indices compared to most other samples, and high coverage values, suggesting that the microbial communities at this vent site were less complex than the other sites.
This result could be explained by a difference in sources of energy available between sites, especially between High Flow habitats, possibly limiting the diversity of the bacterial taxonomic groups adapted to the environmental conditions at Clam Bed. The observation of chemical gradients in the composition of hydrothermal fluids observed along the Endeavour Segment (Butterfield et al. 1994), together with the location of the sites (while the three other sites sampled are all within 150 m of each other, Clam Bed is ~1750 m away from Grotto and Smoke & Mirrors) supports this explanation. This is also coherent with the results obtained by Flores et al. (2011) showing the composition of the microbial communities to be more similar within vent fields than between them, suggesting that the effect of distinct environmental conditions between vent fields is more important than any other variable within a vent field. However, the chemical components of fluids in Low Flow habitats are very diluted. Therefore, the difference in the composition of the fluids between sites is unlikely to explain the lower diversity in the Low Flow sample from Clam Bed. Also, the composition of the microbial communities showed no statistically significant difference within or between vent field relationships (data not shown). This was also the case in a recent study of the patterns of distribution of Epsilonproteobacteria in hydrothermal fluids along the Mariana Arc, where the communities from individual vents seemed to be unique (Huber et al. 2010). These authors suggested that the high diversity of chemical conditions in hydrothermal environments might explain the distinctness of the epsilonproteobacterial populations.
Another fact to take into account is the temporal stability of the sampled habitats. Except for the Low Flow sample from Clam Bed, all samples were collected directly on active sulfide edifices, with the High Flow samples typically found adjacent to vigorous venting and the Low Flow samples further away, on the same edifice, where little or no shimmering flow was visible. These structures are known to be very unstable environments that rapidly evolve at small spatial scales, affecting the distribution and structure of the macrofaunal communities (Sarrazin et al. 1997, 2002). In contrast, the Low Flow sample from Clam Bed was collected from a basaltic substratum in an area where diffuse flow has supported communities of R. piscesae since at least 1991 (Reyes et al. 1995; Tivey et al. 1999; Durand et al. 2006). The stability of this site compared with the others may explain the development of a distinct microbial community.
In order to understand other features distinguishing the High Flow and Low Flow habitats, we explored the taxonomic affiliation of the OTUs that were differently represented between habitats. OTUs that were significantly more represented in our Low Flow libraries were over three times more abundant than in our High Flow libraries. This observation suggests that bacterial groups representing only a small proportion of High Flow communities might play a more important ecological role in Low Flow communities. Also, several species were significantly more abundant in the High Flow libraries than in the Low Flow libraries. These include mesophilic to thermophilic microaerophilic sulfur and hydrogen oxidizers from the class Epsilonproteobacteria, with one organism classified to the species level, S. lithotrophicum, as well as a few heterotrophs from the phylum Bacteroidetes and sulfur reducers from the Deltaproteobacteria. Different species from the same genera having similar metabolisms also proved to be more represented in Low Flow samples, but most of the OTUs significantly more abundant in this habitat belonged to heterotrophic groups, including methanotrophs of the class Gammaproteobacteria. Other relatively abundant heterotrophic genera from the phyla Actinobacteria, Planctomycetes, and Chloroflexi were also more represented in Low Flow libraries.
Differences in temperature, available energy for metabolism, and stability between High Flow and Low Flow habitats potentially explain their distinctive bacterial communities. Even though sulfur and hydrogen oxidizers dominate both types of habitat, the fact that some species were significantly more abundant in High Flow or Low Flow samples suggests that environmental conditions that control tubeworm morphotypes and the composition of their associated macrofauna can also shape the composition of free-living bacterial communities. Environmental shaping of hydrothermal vent microbial communities has also been observed in other studies (Schrenk et al. 2003; Byrne et al. 2009; Kato et al. 2010; Flores et al. 2011), although it is not usually considered in the broader ecological context of factors controlling the composition of vent faunal assemblages. The abundance of heterotrophic organisms in Low Flow libraries (vs. High Flow) indicates that detrital organic matter may be a more important source of energy and carbon for the microbial communities in this habitat. Perner et al. (2007) also point out the potential importance of heterotrophic metabolism in microbial assemblages at low-temperature vents, although their study did not include analysis of available organic matter. Our carbon and nitrogen analyses consistently showed High Flow sites to have fresher, less degraded detritus (lower C/N ratios), which should normally provide more energy for heterotrophic growth than the more refractory organic material (higher C/N ratios) sampled from the Low Flow sites. This is consistent with findings by Venkitachalam (2005) that both autotrophic and heterotrophic microbial enzyme activity were more intensive in detrital samples from High Flow tubeworm colonies than in samples from Low Flow sites. Despite the fact that Low Flow habitats represent lower energy environments (compared to High Flow sites), where the availability of reduced sulfur species is limited (Urcuyo et al. 2003) and fresh organic matter is scarce, there were no major differences in overall bacterial diversity between the two habitats. All conditions being equal, one would expect the higher energy habitat to support a greater level of bacterial diversity. One important difference between the High Flow and Low Flow habitats may be related to their relative stability and longevity. High Flow habitats have been observed to be short lived (Sarrazin et al. 1997; Tunnicliffe et al. 1997) while at least one of the Low Flow sites sampled here (Clam Bed) is known to have been active for at least two decades. The greater stability of the Low Flow sites may have therefore favored a diversification of both autotrophic and heterotrophic bacteria, despite the less favorable growth conditions.
Hydrothermal vents are extremely dynamic environments where physical and chemical conditions vary rapidly both in time and space, and this is known to affect the composition of the macrofaunal communities (Sarrazin et al. 1999, 2002). To date, logistic considerations have limited the use of replicate sampling in studies of the dynamics of vent faunal communities, thus limiting the scope of any interpretation of observed differences between samples. For similar practical reasons, replication and statistical comparisons have been even less common in comparative studies of hydrothermal vent microbial communities. As the field progresses from general descriptive studies to probing relationships between microbial diversity and habitat dynamics, replicate sampling and statistical comparisons will become essential to identifying trends and testing hypotheses. In this study, we have provided some examples of the insights that can be derived from a combination of systematic, replicated field collections, and molecular analysis of microbial diversity. Even if the small number of cultured microorganisms known from hydrothermal vent habitats, combined with the limitation of taxonomic assignment by 454 tag sequencing data prevented us from investigating which species defined the bacterial communities of High Flow and Low Flow habitats, our survey revealed interesting general characteristics about these communities, such as their high overall diversity and a probable relationship of composition with habitat type. Furthermore, results underline the ubiquity of Epsilonproteobacteria in hydrothermal vent environments, which were mainly related to sulfur- and hydrogen-oxidizing genera. Other approaches, such as metagenomic and proteomic analysis, could help to define more specifically the ecological role of the microorganisms detected by providing information on genes and enzymes specific to each habitat.