The JMVFs are located at the southern part of the Mohns Ridge in the Norwegian-Greenland Sea (71.2°N, 5.5°W) (Figure 1A), and consist of the Troll Wall Vent Field and the Soria Moria Vent Field. Sampling (Figures 1B–E) was carried out at both vent fields, in which 3–5 cm-thick microbial mats on top of three iron mounds (Mat1-3) (Vander Roost et al., 2017) and samples from the interior of four 30 cm-high iron mounds (IM1-4) were collected. In addition three environments showing no signs of hydrothermal activity were sampled. A sample of basaltic glass with many rusty-colored cracks (Bas) was sampled several hundreds of meters away from any active sites of the Troll Wall Vent Field. In the rift valley of the Troll Wall two sediment cores were obtained: one (Sed1) on the edge of an iron mat and the other (Sed2) from background sediments. The temperature of bottom seawater at the JMVFs is consistently -0.7°C, whereas the temperature at 30 cm below the seafloor (cbsf) in close vicinity of the Sed2 sample was 1.0–1.3°C, indicating very diffuse venting. In the vicinity of the Sed1 sample, the temperature was 56–86°C at 20 cbsf and 1–2°C at ∼1 cbsf, indicating significant hydrothermal activity. The Sed1 sample was 12 cm long and was divided into two subsamples: one at 2 cbsf and one at 20 cbsf. However, no PCR product was obtained from the latter subsample and it was not considered further. The Sed2 sample was 60 cm long and was divided into five subsamples covering a depth-profile going from 0 to 60 cbsf. Lastly, bottom seawater samples were obtained: two replicate samples (SW1-2) from the open water masses 50 m above the seafloor at the Troll Wall Vent Field and one from the seafloor at the Troll Wall Vent Field (SW3). These samples were collected using a hydraulic suction pump connected to a remotely operated vehicle. DNA was extracted and analyzed in duplicate or triplicate per sample, except for the SW3 sample (Supplementary Table 1). In this way, replicates encompass both technical variability and variation between subsamples.

Once shipboard, pH and alkalinity of the fluids from the hydraulic suction device were measured by a portable pH-meter and an autotitrator (Titrando 888, Metrohm), respectively. Ammonium, nitrate+nitrite, and phosphate were measured photospectrometrically by a Quaatro continuous flow analyzer (Seal). 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 HNO₃ and stored in acid-cleaned HDPE bottles. Both aliquots for ICP-OES and aliquots for anion analysis by Ion Chromatography (Metrohm) were stored at 4°C until on-shore analyses.

DNA was extracted from around 0.5 g material using the FastDNA^®Spin Kit for soil (MP Biomedicals, Solon, OH, United States) following the manufacturer’s protocol. Cells were lysed at a speed setting of 6.0 for 40 s with a FastPrep instrument (MP Biomedicals, Santa Ana, CA, United States). 16S rRNA gene amplification was targeted with universal primers for Bacteria and Archaea: Uni787F (5′-ATTAGATACCCNGGTAG-3′) (Roesch et al., 2007) and Uni1391R (5′-ACGGGCGGTGWGTRC-3′) (Turner et al., 1999). The PCR mix consisted of 12.5 μL 2× HotStar Taq master mix (Qiagen), 1 μM of primer, 1 μL DNA template and ddH₂O up to a total volume of 25 μL. The PCR program initiated with (5′ 95°C), followed by 30 cycles of (45″ 95°C), (45″ 53°C), and (1′ 72°C) and with a final elongation step (7′ 72°C). The size of 16S rRNA gene amplicons was evaluated on a 1.5% agarose gel. Subsequently, PCR products were purified with the MinElute PCR purification kit (Qiagen) following manufacturer’s protocol and eluted with 15 μL ddH₂O.

Amplicons were tagged in a second PCR, with PCR-mixes containing 10 μL 2× HotStar Taq master mix (Qiagen), 1 μM of MID-barcoded primers, 0.8 ng of template and ddH₂O up to a total volume of 20 μL. Ultimately, purification of the 16S rRNA-tagged amplicons was achieved by AMPure XP Bead Purification (Agencourt) following the manufacturer’s protocol. DNA quantification was done with a Bioanalyzer (Agilent Biosystems) and equivalent DNA amounts were pooled prior Multiplex GS FLX+/Titanium 454 pyrosequencing (Roche) at the Norwegian High-Throughput Sequencing Centre in Oslo (NHS), Norway and at Microsynth in Balgach, Switzerland.

In order to compare 16S rRNA gene amplicons of the current study with previous analyses of iron mats in the Troll Wall Vent Field (Vander Roost et al., 2017), we reanalyzed unfiltered 16S rRNA amplicon sequences (raw sff files available from accession numbers ERS903624, ERS903625, ERS903628, ERS9036233-ERS9036235, ERS903639) together with the sequence data obtained in the current study. Sequence files were processed using the 454 SOP MOTHUR protocol¹ (Schloss et al., 2011). For sequence filtering, AMPLICONNOISE (Quince et al., 2011) was used as implemented with the “shhh-flows” MOTHUR command. Sequences containing more than six homopolymers were removed. Subsequently, sequences were aligned to the SILVA reference alignment database (silva.nr_v11.align). A screening step required sequences to start before position 2530, and selecting an end position such that 98% of the sequences ended after that position. Chimera-removal was applied by the UCHIME program, implemented in the “chimera.uchime” MOTHUR command. Statistics from the sequence processing are given in Supplementary Table 2. Richness and diversity indices were calculated by the “summary.single” MOTHUR command.

A total of 284,011 high-quality sequences were subjected to taxonomic classification and OTU analysis was carried out on 97% sequence identity level. This resulted in 4957 OTUs after removal of singletons. All sequences assigned to Zetaproteobacteria were extracted from the dataset and compared to predefined “ZetaOtus” using ZetaHunter (McAllister et al., 2011). ZetaHunter was also used for the identification and assignments of ZetaOtus from a previously published dataset of a microbial JMVF mat dominated by Epsilonproteobacteria growing on top of sediments (Lanzén et al., 2012; Urich et al., 2014) (NCBI Sequence Read Archive accession code SRP004929).

Cluster analyses and Non-metric multidimensional scaling (NMDS) analyses were performed using the R package VEGAN (Schloss et al., 2011). For cluster analyses we used the ward.D2 algorithm (Murtagh, 2014) on Bray-Curtis distances. Bubble plots were generated in ggplot2 version 2.2.1 (Wickham, 2009) using the ggplot command.

For phylogenetic reconstruction of Betaproteobacteria, near full-length 16S rRNA gene sequences (>1200 bp) of Sideroxydans, Leptothrix, and Gallionella were obtained from the SILVA database². An alignment of these sequences together with OTUs (62 bp) from axial Seamount (Meyer et al., 2013) and OTUs generated in the current study (238 bp) was generated using the online SINA aligner³. Based on this alignment a phylogenetic tree was generated in ARB (Ludwig et al., 2004) in two steps: first a backbone tree of the near full-length sequences was generated using the RAxML maximum likelihood algorithm as implemented in ARB. Then, the shorter sequences were added on the tree using the quick-add option in ARB.

For phylogenetic analyses of Zeta- and Betaproteobacterial OTUs used in bubble plot analyses, a maximum likelihood phylogenetic reconstruction was performed in MEGA version 5.2.2 (Tamura et al., 2011), using the Tamura–Nei model (Tamura and Nei, 1993) with default settings.

Raw sequence data generated through this study have been submitted to the European Nucleotide Archive (ENA) under the accession numbers ERS1737039–ERS1737065.

In order to explore the habitat range and distribution of iron oxidizers within the JMVFs, we analyzed samples covering a wide range of habitats, including iron mounds (4 samples), sediments (2 samples), basalts (1 sample), and deep-sea water (2 samples). Results of taxonomic profiling, based on 16S rRNA gene sequencing, were compared to previous analyses of iron mats from the same region (Vander Roost et al., 2017) (Supplementary Tables 1–3 and Figure 2). Overall, the dataset included 284011 reads after filtering (Supplementary Table 2), which were clustered into 4957 OTUs. Putative iron-oxidizing bacteria, i.e., Zetaproteobacteria (FeOζ, 18 OTUs) and members of the Betaproteobacterial genera Gallionella, Sideroxydans, and Leptothrix (FeOβ, 6 OTUs) were detected in most samples. Apart from the iron mats, relative abundances of these iron oxidizers were low (<3%), with the exception of iron mound 1 (IM1a, b), with relative abundances averaging 10.14 and 0.29% of Zeta- and Betaproteobacteria, respectively (Table 1 and Figure 2).

Thaumarchaeota were widely distributed (Figure 2). However, while dominating in the open water masses (SW1-3) and upper parts of iron mounds and sediments, the relative abundance of this group rapidly decreased at increasing sediment depths (Sed2e-Sed2a). Most iron mound samples (IM2-4) and the basaltic glass sample (Bas1) had a class level microbial community composition similar to the open water masses (SW1-3), but with higher relative abundances of Chloroflexi, Planctomycetia, and in most cases, Alphaproteobacteria (Figure 2). In one of the iron mounds (IM1) we observed a microbial community more similar to iron mats (Mat1-3) with relative abundances of Zetaproteobacteria reaching up to 17.25%. No characteristic microbial mat structures similar to the Mat1-3 samples were observed by visible inspection of the IM1 sample. However, it cannot be excluded that an elevated flow of hydrothermal fluids occurred at the IM1 sampling site and that an iron mat was under formation. Iron mat samples and deep-sea water samples formed distinct clusters in cluster and NMDS analyses (Figure 3). Most iron mound samples clustered together with basalt samples and sediment samples. Samples from different depths at the same sampling site [0–10 cm below the seafloor (cmbs) for iron mound samples, 0–20 cmbs for sediment samples] generally clustered more closely together than samples from different sites. Hence, there seems to be larger differences in community structure between different mound and sediment sites than there is between different depths in the upper tens of centimeters below the seafloor. Nineteen FeOζ OTUs (classified as “ZetaOtus” by ZetaHunter) and six FeOβ OTUs were identified. Most of these OTUs were found in several locations (Figures 3C, 4). Notable exceptions are FeOζ23, only observed in deep-sea water, and FeOζ4, only observed in sediments. In an attempt to identify FeOζ in hydrothermally more active areas of the JMVFs than the samples considered in Figures 2–4, we investigated a previously published 16S rRNA gene amplicon dataset from a nearby, microbial mat dominated by sulfide-oxidizing Epsilonproteobacteria, growing at the base of an active venting chimney (Lanzén et al., 2012; Urich et al., 2014). Using ZetaHunter we identified four Zetaproteobacterial sequences in this dataset (corresponding to a relative abundance of 0.003%), which all belonged to ZetaOtu18, one of the overall most-dominating FeOζ at the JMVFs (Figure 4). Apart from FeOβ OTU336, all FeOβ observed in iron mats were also identified in iron mounds, basalts or sediment samples (Figure 4).

Given the notoriously low abundance of FeOβ in marine samples (0–2.72%), it is difficult to exclude the possibility that this group represents contaminants introduced during the extraction of DNA. However, based on an observed significant positive correlation between FeOβ and FeOζ in JMVF iron mats, we have previously argued that this possibility is unlikely (Vander Roost et al., 2017). A similar positive correlation was found when data from the iron mounds, sediments, basalt, and deep-sea water were added (Supplementary Figure 1A). Moreover, a significant positive correlation between FeOβ and FeOζ was also observed when iron mat samples were not included (Supplementary Figure 1B).

Phylogenetic analysis revealed that two of the FeOβ observed in this study (BetaOTUs 436 and 1670) are close-relatives of organisms previously reported in marine environments (Figure 5). Both are close-relatives (>97% sequence similarity) of the marine Gallionellaceae clones YS18Uc25 (AB329936) from the Southern Mariana Trough (Kato et al., 2009b) and ELSC-TVG13-B96 (GU220769) from the Lau Basin hydrothermal vent (Li et al., 2012). Three other BetaOTUs (118, 130 and 4464) are more closely related to freshwater strains within Gallionella and Sideroxydans (Figure 5). A sixth detected FeOβ (BetaOTU336) classified as Leptothrix (order Burkholderiales) and was identical to a marine FeOβ Illumina-sequenced read (62 bp) from the Axial Seamount, Juan de Fuca Ridge (Meyer et al., 2013).

Almost all FeOζ and FeOβ detected in mounds, sediments, basalts, and seawater were also observed within iron mats. Some notable exceptions are ZetaOtu23 and ZetaOtu4, which were exclusively observed in seawater and sediment, respectively. However, given the limited number of samples analyzed in this study, we cannot conclude that ZetaOtus 4 and 23 are truly unique to these sites. In fact, ZetaOtu4 has previously been reported in various Fe(II)-rich sites around the world, particularly at Loïhi (Scott et al., 2017). Similarly, with the isolation of a member of ZetOtu23, at the Great Salt Bay (GSB2 strain) from mild steel corrosion experiments (McBeth et al., 2011), this might be a generalist representative, showing a cosmopolitan spread and exhibiting opportunistic growth when conditions become favorable. Nearly all detected FeOζ were assigned to predefined ZetaOtus observed outside the JMVFs, indicating that they are not endemic to this region. Given that the detected FeOζ are known to colonize a wide range of environmental settings, this would suggest that most, if not all, of these iron oxidizers should be considered generalists, in the sense that they are able to colonize in a wide range of environmental settings. Interestingly, Scott et al. (2015) did not find any evidence of any specific associations between ZetaOtus and different hydrothermal systems existing under different geological settings (basalt versus ultramafic rock types). This is in line with the detection of cosmopolitan Zetaproteobacteria in three hydrothermal systems on the Mid-Atlantic Ridge (Rainbow, TAG, and Snake Pit). Hence, it appears that a limited number of ZetaOtus reoccur both within and between hydrothermal systems, whereas relative abundance may be determined by environmental factors such as energy availability from Fe(II) oxidation.

With the notable exception of Ghiorsea bivora (Mori et al., 2017), members of the Zetaproteobacteria are generally considered obligate Fe(II) oxidizers (Scott et al., 2015). However, based on the global distribution of this OTU, G. bivora appears to be absent from H₂ environments whilst seemingly present under Fe(II) poor conditions. This would suggest that G. bivora uses H₂ only in addition to Fe(II) in natural environments (Mori et al., 2017). Nonetheless, the discovery of G. bivora illustrates that Zetaproteobacteria have some metabolic versatility, and we therefore cannot entirely exclude the possibility that at least some of the detected FeOζ may utilize alternative electron donors. Nevertheless, assuming that all FeOζ are obligate iron oxidizers, the observed relative abundance of FeOζ within the JMVFs are consistent with the hypothesis that the relative abundances of different functional groups in hydrothermal systems can largely be explained by the relative densities of energy sources at a given site (Dahle et al., 2015, 2018). We have previously shown that Fe(II) concentrations and Fe(II)/H₂S ratios are particularly high in the iron mats, presumably due to shallow circulation of hydrothermal fluids in the JMVF rift valley (Johannessen et al., 2016; Vander Roost et al., 2017). Hence, densities of energy available for iron oxidation can be expected to be considerably higher in the iron mats than in surrounding sediments (Vander Roost et al., 2017; Supplementary Table 1). A variable and moderate abundance of FeOζ in more or less active iron mounds is arguably connected to corresponding shifting densities of Fe(II) oxidation energy. These variations stem from the hidden hydrothermal flow paths within iron mounds that shape mound accretion (Johannessen et al., 2016). The lowest fractions of FeOζ were observed in the environments with presumably the lowest densities of potential energy from iron oxidation, i.e., low-temperature deep-sea water and high-temperature sediments with high concentrations of H₂S.

The detection of Zetaproteobacteria within the water column has previously been reported from a stratified estuary, but to the best of our knowledge, the current study is the first to report the presence of Zetaproteobacteria in deep-sea water. It is plausible that the detected Zetaproteobacteria are not active in this environment, but merely transported from benthic environments to the oxygenated water column through the flow of hydrothermal fluids. Alternatively, and perhaps more in line with the observation of a unique ZetaOtu for this environment (ZetaOtu23), pelagic iron oxidizers may grow in flocs, partly consisting of solid FeS particles which act as a source of Fe(II) and allow the formation of microniches with sharp oxygen gradients. This type of mechanism was recently proposed as an explanation for the presence of pelagic FeOζ in Chesapeake Bay (Chiu et al., 2017).

The first evidence for the capability of Fe(II) in basaltic glass to serve as an electron donor for FeOζ, was demonstrated recently by the growth of ZetaOtu9 (growing up to nearly 40% of the microbial community) on freshly synthesized basalt chips incubated on the abyssal plain of the Atlantic Ocean (Henri et al., 2015; Mori et al., 2017). In more altered, naturally occurring basalts, however, microbial communities are highly diverse, consisting predominantly of Thaumarchaeota, a feature that has been related to organic matter contents (Durbin and Teske, 2012). Intriguingly, however, iron-oxidizing microorganisms occur close to or below the detection limit in such environments (Thorseth et al., 2001; Lysnes et al., 2004; Edwards et al., 2011). The current study confirms the presence of Zetaproteobacteria in the basalt samples. Nonetheless, further investigations are needed to reveal how Fe(II) becomes available to the Zetaproteobacterial cell and at what rates they oxidize iron.

A significant correlation was recently identified between the relative abundance of Zetaproteobacteria and Gallionellaceae. This strongly suggests that Gallionellaceae represent an intrinsic part of the microbial communities in iron mats of the JMVFs (Vander Roost et al., 2017). In the current study we provide evidence that FeOβ representatives of the genus Leptothrix (family Burkholderiaceae) may also co-occur with Zetaproteobacteria.

The detection of FeOβ in our samples is somewhat puzzling, given that all isolated members within this group are freshwater organisms. Remarkably, however, three of the detected FeOβ (Leptothrix OTU336 as well as Gallionella OTU436 and OTU1670) (Figure 4) are close relatives of sequence types previously detected in marine environments. This provides evidence that they belong to distinct marine clades. However, until we manage to obtain pure cultures of marine FeOβ, we can only speculate with regard to their salt tolerance. Moreover, the occurrence of FeOβ at relatively low abundance is also intriguing and will require both more sequencing and cultivation efforts to fully comprehend.