Microbial analysis of Zetaproteobacteria and co-colonizers of iron mats in the Troll Wall Vent Field, Arctic Mid-Ocean Ridge

Over the last decade it has become increasingly clear that Zetaproteobacteria are widespread in hydrothermal systems and that they contribute to the biogeochemical cycling of iron in these environments. However, how chemical factors control the distribution of Zetaproteobacteria and their co-occurring taxa remains elusive. Here we analysed iron mats from the Troll Wall Vent Field (TWVF) located at the Arctic Mid-Ocean Ridge (AMOR) in the Norwegian-Greenland Sea. The samples were taken at increasing distances from high-temperature venting chimneys towards areas with ultraslow low-temperature venting, encompassing a large variety in geochemical settings. Electron microscopy revealed the presence of biogenic iron stalks in all samples. Using 16S rRNA gene sequence profiling we found that relative abundances of Zetaproteobacteria in the iron mats varied from 0.2 to 37.9%. Biogeographic analyses of Zetaproteobacteria, using the ZetaHunter software, revealed the presence of ZetaOtus 1, 2 and 9, supporting the view that they are cosmopolitan. Relative abundances of co-occurring taxa, including Thaumarchaeota, Euryarchaeota and Proteobacteria, also varied substantially. From our results, combined with results from previous microbiological and geochemical analyses of the TWVF, we infer that the distribution of Zetaproteobacteria is connected to fluid-flow patterns and, ultimately, variations in chemical energy landscapes. Moreover, we provide evidence for iron-oxidizing members of Gallionellaceae being widespread in TWVF iron mats, albeit at low relative abundances.


Introduction
Biological communities in hydrothermal systems are driven by chemotrophic primary producers, utilizing the energy available in chemical disequilibria, forming when reduced hydrothermal fluids mix with oxic seawater. The geological and geochemical setting varies between and within hydrothermal systems, giving rise to shifting chemical energy landscapes [1,2]. Revealing how this variation shapes the distribution of functional groups of microorganisms is not only important for understanding the microbial ecology of hydrothermal systems, but also for our general understanding of how geochemistry is linked to microbiology. Hydrothermal systems are considered to play a significant role in the marine iron cycle, with estimated releases PLOS  smoker hydrothermal chimneys, composed of anhydrite, barite and pyrite. The chimneys are surrounded by patches of white microbial mats [17,18]. The macrofauna present at TWVF consists mostly of calcareous sponges, sea anemones and sea stars [19].

Sampling
Five iron mats (Fig 1), comprising 13 discrete samples in total (S1 Table), were collected from the TWVF during cruises with R/V G.O. SARS between 2011 and 2014. Samples were taken with a remotely operated vehicle [20] equipped with a 1L hydraulic suction pump ('biosyringe'). Parallel samples from each iron mat were subjected to independent DNA extractions and generation of 16S rRNA gene amplicons. As a reference, a background seawater sample (11ROV11) was collected 50 m above the seafloor (approximately 500 mbsl) and approximately 100 m away from active, high-temperature venting. Zetaproteobacteria and co-colonizers in the Troll Wall Vent Field

Chemical analysis
The pH and alkalinity of the fluids in the biosyringes were measured shipboard immediately after sampling by a portable pH-meter and an autotitrator (Titrando 888), respectively. Ammonium, nitrate+nitrite, and phosphate were measured photospectrometrically by a Quaatro continuous flow analyser (Seal), directly after sampling. Aliquots for later analyses of major and minor elements by inductively coupled plasma optical emission spectrometry (ICP-OES) (Thermo elemental IRIS) were acidified to 3 vol% of ultrapure HNO 3 and stored in acidcleaned HDPE bottles. Both aliquots for ICP-OES and aliquots for anion analysis by Ion Chromatography (Metrohm) were stored at 4˚C until onshore analyses.

Scanning electron microscopy
Aliquots of iron mat samples for scanning electron microscopy were fixed in 2% glutaraldehyde. Dehydration steps were performed on a 0.2 μM pore filter and exposed to, sequentially, 50%, 75% and three times 100% ethanol solutions, with a settling time of 15 minutes between each step. Samples were glued to Al-stubs and coated with Ir using a Gatan 682 Precision etch- The default Needleman-Wunsch aligning algorithm was applied, with k-mer template searching and using +1 per match and mismatch penalties -1, -2 and -1 for each mismatch, opening and extension of a gap, respectively. The alignment was cropped to a minimum length of 210bp and further optimized by selecting both start and end positions by which 90% of the sequences started or ended. Sequences with homopolymers longer than 6 bp were removed. Further filtering and preclustering of sequencing data was carried out as described in [24]. Chimeras were removed from the dataset using default settings of the UCHIME program [25] as implemented in MOTHUR. More detailed information on the filtering of reads is given in the S2 Table. Classification of reads and clustering into operational taxonomic units (OTUs) at 97% sequence similarity was carried out as described in [24]. A total of 5573 OTUs were identified of which 1877 remained after removal of single-and doubletons. Rarefaction plots were generated using the MOTHUR command 'rarefaction.single'. In addition, richness [26] and diversity (inverse Simpson) indices were generated, with the 'summary.single' command.
Non-metric multi-dimensional scaling (NMDS) plots were generated using the Vegan package in R [24] and were based on Bray-Curtis distances, obtained using the average neighbour clustering algorithm. Analysis of similarities (ANOSIM) [27] was used to assess the significance of differences between microbial mats from the rift valley and the rift margin, in which the R-test statistic provides a measure for the separation of community structures; with R = 0 designating no separation, R < 0.25 as barely separable, R > 0.5 as separated but overlapping and R > 0.75 indicating well-separated community structures [28]. Heatmaps showing the compositional profile of the different iron mat communities were generated in R with the 'heatmap.2' function within the gplots package version 2.11.0.1 [29] where samples (columns) were clustered hierarchically, by complete linkage with Euclidean distance measure, and OTUs (rows) clustered phylogenetically, by incorporating a phylogenetic tree from MEGA version 5.2.2. This neighbour-joining tree, comprising the 50 main OTUs, was built with default settings of the incorporated maximum composite likelihood algorithm; using default ClustalW parameters (gap opening penalty = 15 and gap extension penalty = 6.66).
To contribute to the effort of mapping the global distribution of Zetaproteobacteria, all Zetaproteobacteria reads from TWVF were assigned to predefined OTUs (ZetaOtus), using the curated ZetaHunter application (https://github.com/mooreryan/ZetaHunter, last accessed 20/ 06/2017) [11]. ZetaHunter uses the SILVA v123 phylogenetic reference database in order to assign query sequences to ZetaOtus [23, 25, 30-32]. For comparison, all MAR sff-files provided by [14] were reanalysed using MOTHUR and ZetaHunter as described above. Due to differences in the targeted 16S rRNA gene region, we were not able to directly compare de novo ZetaOtus constructed for TWVF with those constructed from the other MAR samples.

Data submission
Original sff-files were submitted to the European Nucleotide Archive with study accession number PRJEB11309, sample accession numbers ERS903619-39, experiment accession numbers ERX1135963-83, and run accession numbers ERR1055648-68.

Environmental settings, estimated cell numbers, and electron microscopy
Sampling sites close to the TWVF white smokers at the eastern rift margin (11ROV3 and 12ROV5), were characterized by higher temperatures, lower pH, higher alkalinity and lower concentrations of Fe 2+ , SO 4 2-, NO 3 - and PO 4 3-, compared to sites in the ultra-slow, diffuseflow systems in the central rift valley (11ROV6, 12ROV9 and 14ROV13) (S1 Table) (Fig 1C  and 1D). Estimated cell numbers from DNA extraction yields are similar to cell numbers reported for iron mats at MAR [14], and were 1-2 orders of magnitude higher in the iron mats than in background seawater (S1 Table). Using electron microscopy, we identified twisted stalks as well as Fe-rich particles with uncharacteristic biological morphologies in all mats ( Fig  1E and 1F).

Community structures and diversity
Considering all iron mats, six phyla dominated (>1% total relative abundance); Proteobacteria (43.1%), Thaumarchaeota (16.7%), Bacteriodetes (9.6%), Planctomycetes (8.5%), Euryarchaeota (7.4%) and Chloroflexi (6.7%). However, we observed a large variability in class-level community structure between the different iron mats (Fig 2). In an NMDS ordination based on Bray-Curtis distances, we found that parallel samples from the same iron mat formed distinct clusters (Fig 3). Anaerolinae as well as Methanomicrobia (11ROV3) and Gammaproteobacteria (12ROV5). Most of the Gammaproteobacteria detected in 12ROV5 were further assigned to Methylococcales, whereas most of the Gammaproteobacteria in other mats were Alteromonadales, Oceanospirillales, Marinicellales, or not assigned on the order level. The most abundant Epsilonproteobacteria OTUs (OTUs 25 and 46), were relatives of sulphur-oxidizing genera, and closely related to Sulfurimonas autotrophica (97%) and Sulfurovum lithotrophicum (94%), respectively. They were found in all iron mats except for the 14ROV13 samples. Deltaproteobacteria were present with similar relative abundances across the different mat samples, and were mostly assigned to genera comprising sulphate reducers-i.e. Desulfuromonadales and Desulfarculales (Fig 2, S1 Fig). Most Thaumarchaeota (90%) were assigned to the genus Nitrosopumilus, which comprises ammonium oxidizers [33]. In order to identify patterns on a OTU-level resolution, we looked at variations in relative abundances of individual OTUs ( Fig  4). This revealed some variation in the presence and abundance among OTUs assigned to the same class. For example, among Zetaproteobacteria, 12ROV9 was dominated by OTU18, as opposed to all other mat samples, which were dominated by OTU3 or OTU12. These latter two OTUs were both 98% similar to the 16S rRNA gene of the iron-oxidizing Mariprofundus ferrooxydans PV-1. OTU18 was only 94% similar to OTU3 and 12, but showed 100% sequence identity with a sequence obtained from the Levantine continental margin [KF199324], a site without hydrothermal activity, where ferrous iron originates from turbation of soft sediments at a depth of 800 mbsl [34]. Notably, the most dominating OTU in the background seawater, assigned to the Nitrosopumilus genus, was also dominating among Thaumarchaeota in the iron mat samples. From rarefaction analyses (S2 Fig) and Shannon-index values (Fig 2), we found that the 11ROV3 samples had the highest diversity and that the diversity in all mat samples was higher than in background seawater.

Mapping of TWVF Zetaproteobacteria to ZetaOtus
Using ZetaHunter, we identified 18 different ZetaOtus in the TWVF samples. This includes ZetaOtus found in geographically separated environments, such as the Loihi Seamount, the South Pacific Ocean (Vailulu'u Seamount, Tonga Arc, East Lau Spreading Center, Kermadec Arc), the Mariana Trough, and the Mid-Atlantic Ridge (MAR) (Fig 5). Among these, ZetaOtus 1, 2, and 9 are considered as cosmopolitan [11,35]. It should also be noted that ZetaOtu18, the most dominant ZetaOtu specific to MAR/TWVF in Fig 5, has previously been detected at the Juan de Fuca ridge in the Pacific and in the Gulf of Maine [11,12]. ZetaOtu24, detected at TWVF but so far undetected at other MAR sites, also includes Guaymas Core B clone B03R022 [AY197408] from the Gulf of California [36]. Finally, ZetaOtu33 includes clone T13J-B63 [JNH860378] obtained from the Southwest Indian Ridge [11,37]. Using ZetaHunter, 22 de novo OTUs were constructed from all TWVF samples. However, individual relative abundances of these Zetaproteobacterial OTUs were low (<1.7%). Hence, none of the most dominating ZetaOtus were found to be unique for TWVF. The four most abundant OTUs assigned to Zetaproteobacteria in the ZetaHunter-independent OTU clustering, corresponded to ZetaOtus 18, 14, 15 and 6 (Fig 4).

Functional assignments
Based on taxonomic information and extensive BLAST searches, we inferred functions of OTUs closely related to cultured organisms (S3 Table). From this, we constructed functional- , derived from a mine water treatment plant (Freiberg, Germany). In order to evaluate whether the detected Gallionellaceae represented contaminants introduced during DNA extraction or PCR, we compared relative abundances of Gallionellaceae with estimated cell numbers. We reasoned that if the Gallionellaceae represented contaminants, they would have a tendency to occur with higher abundances in samples with low DNA yields (i.e. samples with low estimated cell numbers). However, no such correlation was observed. Instead, we observed a correlation between relative abundances of Gallionellaceae and Zetaproteobacteria (R 2 = 0.7845,  [14]), as well as information from [11,35].

Discussion
Using 16S rRNA gene amplicon sequencing data we have investigated shifts in seafloor microbial mats from the Troll Wall Vent Field in environments variable in terms of hydrothermal activity and distance from high-temperature venting smokers. With cell numbers 1-2 orders of magnitude lower in the background seawater than in the iron mats, the influence of iron mats by background seawater can be expected to be minor. Moreover, the high similarity between replicates indicates that the observed differences between iron mats are not only an effect of biases in DNA extraction and PCR amplification, but reflect real variation. Zetaproteobacteria were detected in all iron mat samples, indicating that this class is widespread at TWVF. However, relative abundances of iron oxidizers and their co-occurring taxa seem to vary largely among sites. Below we compare these findings with other iron mats and discuss a potential metabolic interaction between members of Zetaproteobacteria and members of Nitrosopumilus. Based on geochemical data from TWVF fluids, we discuss how the observed transition from microbial mats dominated by sulphur oxidizers near hydrothermal chimneys, to microbial mats dominated by iron oxidizers in the rift valley, might be linked to shifting fluidflow patterns and, ultimately, shifting energy landscapes. This is summarized in a conceptual model of the geobiological interactions taking place at the TWVF (Fig 6).

Comparison with iron mats from other locations
On the taxonomic level, the microbial iron mats analysed in this study share some of the characteristics of other iron mats around the world-i.e. the presence of close relatives of sulphate reducers within Deltaproteobacteria [39][40][41], the presence of Epsilonproteobacteria [42,43], a high abundance of close relatives of ammonium-oxidizing Nitrosopumilus among Archaea [40,41,43], and the presence of Chloroflexi [39,41]. Yet, to our knowledge, this is the first report indicating that methanogens may make out a large fraction of microorganisms in iron mats. Members of methanogens and sulphate reducers are strict anaerobes, hence their presence is likely to reflect steep oxygen gradients at the sampling sites. Presumably, obligate anaerobes grow in the lower parts of the iron mat or in the transition from iron mat to the underlying sediments, and arguably contribute to the production of methane and hydrogen sulphide, consumed by aerobic methanotrophs and sulphide oxidizers in upper layers.
With respect to marine iron oxidizers, insights into the biogeographical patterns of Zetaproteobacteria have been documented in several recent papers [11,[44][45][46][47]. On a global level, the distribution of Zetaproteobacteria OTUs has been found to be more strongly correlated with geographic occurrence than with environmental parameters such as temperature, pH, or total Fe concentration [11,48]. Our results are in agreement with this in the sense that many of the same OTUs of Zetaproteobacteria seem to dominate in both rift margin as well as rift valley samples. Hence, shifting geological settings between the iron mats (see below) seem to have a large effect on relative abundances of Zetaproteobacteria and what taxa they co-occur with, but without causing large shifts in what OTUs of Zetaproteobacteria the different mats host. Moreover, our ZetaHunter analyses (Fig 5) did not reveal any new dominating ZetaOtus specific to TWVF. Similar observations were made from the Snake Pit, Rainbow and TAG site of the Mid-Atlantic Ridge [14], suggesting that Zetaproteobacteria from the Atlantic region are highly similar to their relatives from around the world and probably do not constitute any distinct phylogenetic clades. The detection of ZetaOtus 1, 2 and 9 supports earlier findings that these ZetaOtus might be cosmopolitan [11,35]. Recently, it was suggested that ZetaOtu1 may not be truly cosmopolitan, due to its low abundance at MAR sites [35]. However, ZetaOtu1 was identified as one of the dominating ZetaOtus at TWVF (Fig 5) demonstrating that this ZetaOtu may also be found in high abundances at MAR. High abundances of ZetaOtus 6 and 14 at White microbial mats are abundant on and around the active chimneys and have relative abundances of sulphide oxidizers and methane oxidizers that are consistent with communities predicted from energy models considering mixing of hightemperature fluids with seawater [1]. (2) Rare-earth element analyses suggest that hydrothermal fluids flowing through iron mats in the rift margin, have the same source as the high-temperature fluids flowing through the chimneys (not published). However, subseafloor precipitation of sulphides lead to conditions where Fe(II), leaching out from minerals or produced by iron reducers, is more stable and forms the basis for the development of iron deposits inhabited by low abundances of FeOB on the seafloor. Yet, high abundances of methanotrophs and methanogens are indicative of a microbial degradation of organic matter with organics as the main source of energy under this setting. Accumulation of organic carbon in the rift margin may partly be a result of transport of biomass from the most active hydrothermal areas down towards the rift valley. (3) As suggested by rare-earth element analysis, fluids circulating through the rift valley belong to a shallow circulation system that is separated from the high-temperature fluids [17]. Sulphate is therefore not reduced to sulphide abiotically, and lower amounts of Fe(II) precipitate in the crust. This forms the basis for conditions on the seafloor with relatively high densities of potential energy from iron oxidation, as reflected by communities with high relative abundances of FeOB. SOB = sulphur-oxidizing bacteria, MA = methanogenic archaea, MOB = methanotrophic bacteria, FeOB = iron-oxidizing bacteria. https://doi.org/10.1371/journal.pone.0185008.g006 TWVF are also inconsistent with the recent suggestion that these ZetaOtus could be considered as geographic biomarkers for the Loihi Seamount. The comparison between TWVF and other MAR sites demonstrates that the distribution of ZetaOtus can differ significantly between geographically distant MAR sites (Fig 5). For example, ZetaOtu18 is the second most abundant ZetaOTU at TWVF, but barely present elsewhere on MAR. Taken together, these findings illustrate that our knowledge about the global distribution of ZetaOtus, including those from MAR sites, is still too limited to draw strong conclusions on biogeographical patterns. It is also too early to exclude the possibility of geochemical setting being the main driver for shaping communities of Zetaproteobacteria worldwide.
The presence of Gallionellaceae in all iron mat samples is noteworthy. This group is traditionally considered to consist exclusively of freshwater iron oxidizers. Although some recent studies report on the presence of this group in marine environments [42,[49][50][51][52][53], their relative abundance has been so low (<0.1%) that it has remained unclear if they represent contaminants or not. Low abundances of Gallionellaceae were also observed in the present study. However, if they represented contaminants of some sort, one would expect a negative correlation between DNA content in the samples and relative abundances of Gallionellaceae and no relationship between relative abundances of this family and other taxa. Our results indicate that the opposite is true: we found no correlation between estimated cell numbers and relative abundances of Gallionellaceae as opposed to a highly significant correlation between relative abundances of Zetaproteobacteria and Gallionellaceae, providing strong evidence for Gallionellaceae being an intrinsic part of the microbial communities in the iron mats.
Numerous hydrothermal systems hosting iron mats containing biogenic stalks have been reported along mid-ocean spreading ridges [54][55][56][57][58][59][60]. Nevertheless, extensive studies on community level have been limited [14]. Given the large variability in the overall community structure observed so far in only a few of these mats (this study, [14,17]), we are most likely only beginning to grasp the full diversity of iron mats in these settings, as well as how they form and what roles they play in elemental cycling.

Interactions between Zetaproteobacteria and Nitrosopumilus
In general, little is known about the interactions between iron oxidizers and co-occurring taxa in iron mats, particularly when it comes to Archaea. In the iron mats of TWVF, Zetaproteobacteria consistently co-occurred with close relatives of ammonium-oxidizing Thaumarchaeota within the Nitrosopumilus genus (Figs 2 and 4). Genomic analyses have revealed that Zetaproteobacteria frequently have genes involved in assimilatory or dissimilatory reduction of nitrate or nitrite, such as nitrate reductase and nitrite reductase [61]. Based on the wide distribution of nirK genes, encoding nitrite reductase, in iron mats of the Loihi Seamount, it was recently proposed that the habitat range of Zetaproteobacteria includes anoxic habitats containing nitrite [46,62]. From an analysis of a single iron mat from TWVF, we have already proposed that there might be a link between ammonium-oxidizing Thaumarchaeota within the Nitrosopumilus genus and iron-oxidizing Zetaproteobacteria, whereby nitrite produced by Nitrosopumilus, can be used by Zetaproteobacteria in anaerobic respiration [17]. In the present study we show that a co-existence between members of Zetaproteobacteria and Nitrosopumilus is a common feature of iron mats in TWVF, in both the rift valley as well as the rift margin. This opens for the possibility that Nitrosopumilus assist in the formation of iron mats and sustain Zetaproteobacteria communities in the entire area. Alternatively, members of Nitrosopumilus can get entrapped in the iron mats during water circulation. Such entrapment has previously been suggested for explaining the presence of planktonic psychrophiles in a hydrothermal vent ecosystem at Loihi Seamount [63]. In order to understand the relationship between Nitrosopumilus and Zetaproteobacteria in more detail, it would be interesting to measure in situ metabolic rates in addition to doing more comprehensive genomic studies involving the reconstruction of near full-length genomes.

Shifts in microbial communities and energy densities from the rift margin to the rift valley
Zetaproteobacteria seem to depend on Fe(II) as an electron donor [14]. However, although Fe (II) concentrations in a given habitat may be sufficient to support a population of Zetaproteobacteria, a number of other factors may control their relative abundance and what organisms they co-occur with. Thermodynamic modelling provides evidence that, on a functional level, shifts in community structures within and between hydrothermal systems can be explained by shifting chemical energy landscapes [1]. With this in mind, it is interesting to note how observed shifts in community composition (Figs 2 and 4, S1 Fig) are co-occurring with shifts in geochemical setting. Specifically, hydrothermal fluids flowing through sediments at the base of the hydrothermal chimneys can be expected to have the same source as the fluids emitted through the chimneys, which have much higher concentrations of H 2 S and methane, than Fe (II) (S1 Table) [1,64]. At the seafloor the fluids are highly diluted by seawater, and community composition modelling, based on thermodynamic considerations, predict that the communities growing on these fluids will be dominated by sulphide oxidizers and, to a lesser extent, methane oxidizers [1]. This is largely consistent with observed community structures in white mats growing in the vicinity of high-temperature venting chimneys (S1 Fig). Furthermore, analyses of rare-earth elements indicate that the fluid flow in the central rift valley is disconnected from the high-temperature fluids venting through the chimneys, and belong to a separate, more shallow circulation system [17]. As a result, sulphate from seawater is not reduced to sulphide, H 2 S/Fe(II) ratios are low, and energy densities are more dominated by iron oxidation (Fig 6, S1 Table). The composition of rare-earth elements in the diffuse venting fluids of the rift margin indicate that these fluids have the same source as the high-temperature fluids (unpublished data). Unfortunately, we lack data on H 2 S concentrations for these samples. However, the temperature gradients in this region are less steep than at the base of the chimneys (S1 Table), indicating lower fluxes of hydrothermal fluids. Most of the sulphide may therefore have had time to precipitate within the subseafloor as metal sulphides, diminishing the substrate availability for sulphide oxidizers in the iron mats without removing all of the dissolved Fe(II). Yet, energy landscapes seem to be mostly influenced by degradation of organic matter where methanogens consume hydrogen or acetate, presumably produced by fermentative organisms, to produce methane, which in turn is consumed by aerobic methanotrophs (Fig 6). Taken together, microbial mats of the rift valley and the rift margin have largely different community structures (Figs 3 and 4), which seem to reflect that they reside under very different energetic settings. Arguably, obtaining spatially high-resolution data on geochemical settings and distribution of Fe(II) within hydrothermal systems, is key to our understanding of why Zetaproteobacteria are distributed the way they are and what taxa and functional groups they co-occur with.

Conclusion
This study demonstrates that members of Zetaproteobacteria are widespread throughout the TWVF, comprising all ZetaOtus considered as cosmopolitan (i.e. ZetaOtus 1,2 and 9). We found no evidence for the presence of abundant Zetaproteobacterial OTUs, specific for TWVF. Relative abundances of Zetaproteobacteria and their co-occurring taxa vary considerably between different iron mats. This also seems to include members of iron-oxidizing Gallionellaceae. Microbial communities in iron mats from TWVF appear to show a large beta-diversity, which may well be revealed to be even larger if more samples are analysed in the future. Hence, the TWVF seems to be an excellent natural laboratory for investigating the general factors that control the structure and diversity of iron mat microbial communities. Future studies of iron mats from TWVF, involving hydrothermal flow rate measurements, cultivation, shotgun metagenomics studies, and more extensive chemical analyses, seem to be promising approaches in this respect. Nevertheless, building on previous work on linkages between chemical energy and community structure [1], this study provides further evidence that variability in energy density is a key factor shaping microbial communities, including those in iron mats, within and between hydrothermal systems.