Validating the Cyc2 Neutrophilic Iron Oxidation Pathway Using Meta-omics of Zetaproteobacteria Iron Mats at Marine Hydrothermal Vents

Iron oxides are important components of our soil, water supplies, and ecosystems, as they sequester nutrients, carbon, and metals. Microorganisms can form iron oxides, but it is unclear whether this is a significant mechanism in the environment. Unlike other major microbial energy metabolisms, there is no marker gene for iron oxidation, hindering our ability to track these microbes. Here, we investigate a promising possible iron oxidation gene, cyc2, in iron-rich hydrothermal vents, where iron-oxidizing microbes dominate. We pieced together diverse Zetaproteobacteria genomes, compared these genomes, and analyzed expression of cyc2 and other hypothetical iron oxidation genes. We show that cyc2 is widespread among iron oxidizers and is highly expressed and potentially regulated, making it a good marker for the capacity for iron oxidation and potentially a marker for activity. These findings will help us understand and potentially quantify the impacts of neutrophilic iron oxidizers in a wide variety of marine and terrestrial environments.

KEYWORDS Cyc2 Fe oxidation pathway, biogeochemistry, environmental microbiology, hydrothermal vent, iron cycling, iron oxidizers, metagenomics, metatranscriptomics, microbial ecology, Zetaproteobacteria N eutrophilic Fe-oxidizing microbes are common in marine and terrestrial environments (1), precipitating reactive Fe oxyhydroxides that sequester organic carbon, phosphate, arsenic, and many other metals (2)(3)(4). However, it has been difficult to study the effects of neutrophilic Fe oxidation in natural systems due to myriad challenges that have slowed the discovery of genetic markers of neutrophilic Fe oxidation. These Fe oxidizers are difficult to culture, so only recently have we obtained enough isolate genomes to deduce hypothetical neutrophilic Fe-oxidizing pathways. Comparative genomics has led to multiple proposed pathways, each involving an outer membrane cytochrome (5)(6)(7). However, only one pathway is present in all well-established neutrophilic Fe-oxidizing isolates (Zetaproteobacteria and Gallionellaceae), centering on a fused cytochrome-porin, Cyc2 (7)(8)(9)(10). Yet, beyond comparative genomics, we lack evidence of the Cyc2 pathway function in neutrophilic Fe oxidizers, particularly the uncultured Fe oxidizers that dominate natural Fe systems.
The Cyc2 pathway in the neutrophilic Fe oxidizers is modeled after the Fe oxidation pathways found in acidophilic Fe oxidizers Acidithiobacillus ferrooxidans and Leptospirillum sp., where Cyc2 Fe oxidase function has been verified (11,12). Weak homologs to cyc2 from these organisms were first found in the genomes of the Gallionellaceae Sideroxydans lithotrophicus ES-1 and Gallionella capsiferriformans ES-2 (13). The genome of Zetaproteobacteria type strain Mariprofundus ferrooxydans PV-1, on the other hand, lacked homologs to known Fe oxidation genes until a proteome study discovered that cyc2 was in fact carried by PV-1 but missing from the draft genome (14). Subsequently, cyc2 homologs were found within the few Zetaproteobacteria lineages with genomic representation (9,15). Despite the identification of cyc2-like genes, low amino acid sequence homology (20% sequence identity between PV-1 and A. ferrooxidans Cyc2) suggests that their function is speculative and needs to be validated. Without a means of testing this function biochemically or genetically, we focus on more comprehensive comparative genomics and expression in Fe-oxidizing environments.
To this end, we turned to Zetaproteobacteria in natural Fe microbial mats. The Zetaproteobacteria discovered to date are all considered to be Fe oxidizers, since every isolate grows by Fe oxidation and uncultured Zetaproteobacteria are typically found in Fe-oxidizing environments (10,(16)(17)(18)(19)(20)(21). The Zetaproteobacteria are often the dominant organisms in marine hydrothermal Fe mats (22)(23)(24)(25), where they play a key role in mat formation (26). This abundance and ubiquity in Fe-oxidizing mats make Zetaproteobacteria ideal for study through metagenomic and metatranscriptomic approaches. Furthermore, their taxonomic diversity allows for a robust comparative genomics study. We sampled paired metagenomes and metatranscriptomes from three hydrothermal venting regions: Loihi Seamount, the Mid-Atlantic Ridge (MAR) (Rainbow, TAG, and Snake Pit vents), and the Mariana Backarc (Urashima vent field). Recovery of highquality metagenome-assembled genomes (MAGs) allowed us to improve the limited genomic representation of the Zetaproteobacteria (see reference 10 for a summary of genomic representation prior to this study). Using the MAGs for reference mapping, we explored in situ environmental expression of the Zetaproteobacteria within undisturbed natural Fe mats. In addition, we examined expression responses to Fe(II) using shipboard Fe(II) amendment experiments. With these results, we assess and update the model neutrophilic Fe oxidation pathway expressed in natural environments.
This article was submitted to an online preprint archive (27).

RESULTS
Microbial Fe mat sampling and geochemistry. Over three expeditions, we sampled a wide diversity of Fe microbial mats (see Table S1 in the supplemental material). Sampled mats varied in their physical setting, with meter-scale Loihi mats found in direct and indirect flow from vent fissures, mat mounds on the scale of tens of centimeters at the MAR at the diffuse-venting periphery of black smoker fields, and the mats at the Mariana Backarc Urashima vent fields covering a 7-m-tall Fe chimney ( Fig. S1 and Text S1). Temperatures ranged from 10 to 63°C, while geochemical conditions also varied widely, notably concentrations of Fe(II) from 1.3 to 190 M and O 2 from Ͻ3 to 123 M within the mats (Table 1). Mariana mats had shallow O 2 gradients, while at Loihi, O 2 was undetectable (Ͻ3 M) at 1 cm below the mat surface. These Fe(II) and O 2 conditions favor biotic Fe oxidation (10). At Mariana, total dissolved Fe was depleted by 49% to 74% in our low-temperature mats relative to the conservative mixing of the local high-temperature zero-Mg endmember (Fig. S2), which suggests that a substantial amount of Fe is being oxidized and precipitated within these mats.
Zetaproteobacteria abundance and diversity. We initially assessed Zetaproteobacteria abundance and diversity using a 16S rRNA gene survey (Fig. S3). Fe mat communities at all sites hosted abundant Zetaproteobacteria populations, from 16.4 to 95.9% of the total bacterial community at Loihi, 10.7 to 31.3% at the Mid-Atlantic Ridge, and 37.1 to 79.9% at Mariana. Many samples were dominated by the Zetaproteobacteria, notably sample S1 (96% Zetaproteobacteria), a centimeter-scale sample of actively growing Fe mat surface. In addition to Zetaproteobacteria, the mats hosted variable flanking microbial communities that differed between the three sites ( Fig. S3 and S4) but were similar to previous studies (9,22,23,28,29). Overall, the relatively simple, Zetaproteobacteria-rich composition of these marine Fe mats makes them good systems for studying neutrophilic Fe oxidation mechanisms.
We used the 16S rRNA gene community profiling results to choose metagenomic (MG) and metatranscriptomic (MT) samples, aiming to recover high-abundance and   diverse Zetaproteobacteria to produce high-quality genomes with sufficient MT read depth (Fig. S3). We recovered 126 total high-quality MAGs from our samples (Ͼ70% complete, Ͻ10% redundant) (see Tables S4 and S5 at Table S2), which were compared to a collection of published high-quality genomes (Text S1). MAGs from this study improve the representation of nine different Zetaproteobacteria operational taxonomic units (ZOTUs) spanning the Zetaproteobacteria phylogenetic tree by providing 2 to 13 additional high-quality MAGs for each of these ZOTUs ( Fig. 1; Table S2). Many of these ZOTUs previously had poor genome representation (labeled ZOTUs in Fig. 1B). These diverse ZOTUs were abundant and active within our Fe mats (abundance by 16S rRNA gene and MG; activity by MT) (Fig. 2). MAG relative abundance generally matched relative activity, with the exception of MAG S6_Zeta1 (ZOTU6), which had higher activity than expected, likely in response to the shipboard incubation conditions. By substantially improving Zetaproteobacteria genome representation and pairing this with metatranscriptomes, we are poised to investigate genetic commonalities and diversity across the Zetaproteobacteria, particularly of the Fe oxidation mechanism.
Phylogeny of the putative Fe oxidase Cyc2. The key component of the proposed neutrophilic Fe oxidation pathway is Cyc2, which has been shown to oxidize Fe(II) in acidophiles. Our preliminary analyses showed that some Zetaproteobacteria genomes have multiple cyc2 copies that were not closely related. To investigate these, we developed a comprehensive Cyc2 phylogeny. This phylogeny includes sequences from terrestrial to marine and circumneutral to acidic environments, as well as both known Fe oxidizers and organisms not known to oxidize Fe ( Fig. 3; see Fig. S8 for detailed tree with sequence names at https://doi.org/10.6084/m9.figshare.c.4646336). Cyc2 sequences form three clusters, but the Fe-oxidizing function has been verified only for the Cluster 2 Acidithiobacillus ferrooxidans Cyc2 (11) and the Cluster 3 Cyc2 homolog Cyt 572 of Leptospirillum sp. (12). However, most of the neutrophilic Fe oxidizers fall within Cluster 1, a well-supported group (93% bootstrap) that is largely comprised of the Zetaproteobacteria, Gallionellaceae, and Chlorobium ferrooxidans. This strongly suggests that Cluster 1 Cyc2s share a function.
Putative Fe oxidation pathway distribution revealed by comparative genomics of the Zetaproteobacteria. Our Cyc2 phylogeny shows that Zetaproteobacteria possess Cyc2 from both Clusters 1 and 3. All ZOTUs possess a Cluster 1 cyc2 gene; with our new genomes, this includes four additional ZOTUs that are now known to possess cyc2 (ZOTUs 1, 7, 13, and 14; Fig. 4). In contrast, fewer ZOTUs have the Cluster 3 cyc2 gene, and 65% of genomes with Cluster 3 cyc2 (n ϭ 15) also have Cluster 1 cyc2. This suggests that both Cyc2 types have a use in the Zetaproteobacteria, though it is unknown how Cluster 1 and 3 Cyc2s may differ in function. ZOTU2 is unusual in that only 3 of the 10 genomes appear to have cyc2, though this may be due to assembly issues specific to ZOTU2 (Text S1). In any case, the presence of cyc2 in all Zetaproteobacteria OTUs suggests its centrality to these neutrophilic Fe oxidizers.
In addition to cyc2, other proposed genes for the Fe oxidation pathway were also widely distributed in Zetaproteobacteria genomes. Homologs of cyc1 were present in the genomes of all ZOTUs except ZOTU9 (Fig. 4); cyc1 encodes a diheme c-type cytochrome thought to be a periplasmic electron carrier in the A. ferrooxidans Fe oxidation pathway (11). ZOTU9, which includes the Fe-and H 2 -oxidizing Ghiorsea spp. (30), must use another periplasmic electron carrier. Indeed, many other putative periplasmic cytochromes can be found in Zetaproteobacteria genomes (see below). Cyc1 or another electron carrier likely passes electrons to a terminal oxidase or to complex I via complex III (reverse electron transport). Genes for the bc 1 complex were found in all ZOTUs, whereas we found alternative complex III (ACIII) genes in only a few Zetaproteobacteria, primarily in ZOTU11 and Family 2 (ZOTUs 4, 10, and 13). We found three types of aerobic terminal oxidases: (i) cbb 3 -type cytochrome c oxidase, (ii) aa 3 -type cytochrome c oxidase, and (iii) cytochrome bd-I ubiquinol oxidase. Further, two distinct forms of the cbb 3 -type cytochrome c oxidase were found, clustering in the proximal and distal cbb 3 subtrees defined by Ducluzeau et al. (31) (Fig. S5). All ZOTUs possess genes for one or more of these terminal oxidases, suggesting that all Zetaproteobacteria are aerobic Fe oxidizers (Fig. 4). Taken together, these findings allow us to update the neutrophilic Fe oxidation pathway model (Fig. 5).
In situ expression of the putative Fe oxidation pathway. Our next step was to determine whether the putative Fe oxidation pathway genes are expressed in the environment; high expression would lend support for the model. In situ expression from six unique ZOTUs in 10 different samples (total of 21 observations) shows that cyc2 genes from Cluster 1 are highly expressed in all Zetaproteobacteria and samples, ranging from 3.0ϫ to 555ϫ baseline constitutive gene expression (Table 2). Cluster 1 cyc2 was frequently the highest-expressed gene in the genome, particularly in the Mariprofundaceae (Family 1). Interestingly, cyc2 expression levels differed between Zetaproteobacteria Families 1 and 2, though expression was still high in all Zetaproteobacteria. The cyc1 and cbb 3 -type terminal oxidase genes are expressed up to 17.3ϫ and 56.6ϫ constitutive gene expression, respectively. On average, this places the cyc1 and terminal oxidase genes in the 73.4 and 75.4 percentile range in Zetaproteobacteria metatranscriptomic expression, respectively (Fig. S6). This gene expression is consistent with protein expression levels of M. ferrooxydans PV-1, where the corresponding proteins were expressed at or above the 87th percentile (14). In combination, our data suggest that genes in the core model Fe oxidation pathway are highly expressed in situ by diverse Zetaproteobacteria under various environmental conditions. Variation in gene expression may help us resolve which modules are most commonly used during Fe oxidation in the environment. For example, expression of complex III module genes further supports the importance of bc 1 over ACIII for reverse electron transport. Average expression of bc 1 was 6.6ϫ constitutive expression, while ACIII expression was much lower at 0.6ϫ. In ZOTUs with both complexes, bc 1 genes were expressed 1.5ϫ to 18.6ϫ higher than the ACIII complex. The limited distribution of ACIII in only a few Zetaproteobacteria lineages, despite our deep sampling with near-complete genomes, combined with its low expression suggests that ACIII is not required for Fe oxidation under the sampled conditions.  Validating the Cyc2 Neutrophilic Fe Oxidation Pathway Comparison of relative in situ expression may also help identify genes that may be involved as intermediate electron carriers, particularly in ZOTU9, which lacks cyc1. We identified at least 14 different putative periplasmic c-type cytochrome genes (PCs) with high expression (Ͼ90th percentile) in one or more Zetaproteobacteria genomes (see Table S6A at https://doi.org/10.6084/m9.figshare.c.4646336). Some of these genes (cyc1, PC12, PC61, PC16, and PC38) were more highly expressed in some genomes than cyc2 in the Mariprofundaceae, and all were found in at least one genome where they were more highly expressed than cyc1. Interestingly, these putative periplasmic cytochromes were found and expressed at different levels in different Zetaproteobacteria lineages, with some unique to a single ZOTU (e.g., PC73 in ZOTU2) and most found in several ZOTUs. The cytochromes c previously found on a conserved cassette identified in Zetaproteobacteria isolates and single amplified genomes (SAGs) (including cyc1, PC2, and PC3) (15) are most highly expressed in genomes from ZOTUs 1 and 2. These results help us narrow the list of potential electron carriers in the Zetaproteobacteria.
Expression of Fe oxidation pathway genes in Fe-amended mat incubations. To link gene expression more specifically with Fe oxidation, we added Fe(II) to mat  samples and analyzed metatranscriptomic responses over time. We performed shipboard incubations at Loihi Seamount and Mariana, using fresh Fe mats, live and killed, while monitoring Fe oxidation. Microbes within the mat were actively oxidizing Fe(II) faster than abiotic processes, with the pseudo-first-order Fe oxidation rate constants 3.7ϫ (Loihi) and 5.3ϫ (Mariana) higher in live samples than azide-killed ones (Fig. S7). These results show that we stimulated biotic Fe oxidation, which should lead to increased expression of Fe oxidation genes.
As with the in situ samples, cyc2 was highly expressed in the Zetaproteobacteria during the time series experiments, reaching a maximum of 97.1st to 100th percentile in the four most active Zetaproteobacteria lineages. After Fe(II) was added, there was an increase in total cyc2 expression (sum of all cyc2 genes), as well as cyc2 expression by each individual MAG ( Fig. 6A and B). Expression increased at different rates for each ZOTU, with some peaking earlier and others peaking at the end of the experiment. Expression of cyc2 increased less for the most abundant Zetaproteobacteria (e.g., S6_Zeta1), which already had high expression of cyc2 prior to Fe(II) amendment. Less-abundant Zetaproteobacteria (S6_Zeta11/S6_Zeta23) were also expressing cyc2 prior to Fe(II) amendment but had a much larger change in expression (5.7ϫ to 6.5ϫ), reached their maximum quickly, and maintained a higher expression over the course of the experiment. Overall, while the degree and timing of response differed, all Zetaproteobacteria increased their cyc2 expression in response to Fe(II) amendment.
Other genes in the Fe oxidation pathway also generally followed this trend, with cyc1 expression in 4 of 5 genomes and cbb 3 -type terminal oxidase genes in 2 of 4 genomes also increasing after Fe(II) addition. However, low read recruitment depth led to substantial noise. To correct for this noise, we normalized expression using six constitutively expressed genes (Text S1; also see Table S7 at https://doi.org/10.6084/ m9.figshare.c.4646336) and focused on expression patterns in S6_Zeta1, which had high read depth (Fig. 6C). Constitutive normalized expression of S6_Zeta1 shows a similar pattern of cyc2 expression change over the time series compared to expression patterns prior to constitutive normalization, with a maximum increase of 3.6ϫ after Fe(II) amendment. The genes encoding Cyc1 and the cbb 3 -type terminal oxidase also increased after Fe(II) addition, reaching a maximum fold change of 1.5ϫ to 3ϫ. This trend was also observed for 7 of 8 putative periplasmic cytochromes more highly expressed than cyc1, increasing 2.1ϫ to 4.5ϫ over the course of the experiment (see Table S6B at https://doi.org/10.6084/m9.figshare.c.4646336). These results suggest that the Fe(II) amendment increased the expression of many genes thought to be in the Fe oxidation pathway. If the Zetaproteobacteria represented by the S6_Zeta1 genome is an autotrophic Fe oxidizer, Fe(II) amendment should stimulate genes for carbon fixation, central metabolism, and growth. Like cyc2 expression, genes for central metabolic pathways, including the tricarboxylic acid (TCA) cycle, increased in expression 2.0ϫ to 3.4ϫ over 1 h in S6_Zeta1 from ZOTU6 (see Fig. S9 at https://doi.org/10.6084/m9.figshare.c.4646336). Similarly, expression of genes for glycogen synthesis increased 1.8ϫ in the first 2 min after Fe(II) addition. Carbon fixation genes increased 1.9ϫ in the first 12 min. Some of the highest fold changes after Fe(II) amendment were seen in genes related to proper protein folding (molecular chaperones groEL, groES, and dnaK) and membrane protein quality control (htpX-type protease) (32). For example, groEL increased 119ϫ after Fe(II) addition. Though these gene responses may correspond with shock to the cell after Fe(II) amendment, these genes were also highly expressed under in situ conditions in S1_Zeta1, which suggests that they may be necessary for promoting active Fe oxidation in the environment. Together, these data suggest that Fe(II) amendment stimulated genes for both Fe oxidation and growth.

DISCUSSION
The main objective of this study was to assess the Cyc2-based Fe oxidation pathway in neutrophilic Fe-oxidizing bacteria (FeOB), using environmental metagenomics and metatranscriptomics of marine Fe mats. This work contributes 53 new nearly complete Zetaproteobacteria genomes paired with expression profiles, both in situ and from incubations, along with a comprehensive Cyc2 phylogeny. Using these, we have characterized the distribution and usage of the model Fe oxidation pathway across the full range of known Zetaproteobacteria in Fe mats at three geographically distinct venting regions. The emerging pattern shows that the pathway as a whole is highly expressed, with increased expression in all Fe pathway genes following Fe(II) amendment. The cyc2 gene is among the highest expressed and is the only gene in the pathway shared by all Zetaproteobacteria, suggesting that it plays a key role in Fe oxidation.
Assessing the Fe oxidation pathway model through Zetaproteobacteria comparative genomics. In this study, we assessed the current model for Fe oxidation by comparing genomes representing the full diversity of the Zetaproteobacteria (Fig. 1B). Since the Zetaproteobacteria are thought to be an entire class of Fe-oxidizing bacteria, all genomes should have an Fe-oxidizing pathway, which may be conserved. Our results show that overall, the basic model of neutrophilic Fe oxidation in the Zetaproteobacteria holds (Fig. 4). All Zetaproteobacteria lineages, including novel ones presented in this study, possess genes encoding the putative Fe oxidase Cyc2, an intermediate electron carrier, and a terminal oxidase. However, our survey shows that each of these components can have multiple versions, suggesting that the pathway contains interchangeable modules, as depicted in our updated model (Fig. 5). Across all Zetaproteobacteria, there are two types of Cyc2, multiple potential periplasmic cytochrome electron carriers, and four terminal oxidases. The variations are likely linked with specific adaptations related to niche, with some genomes possessing multiple versions of certain components, perhaps to span multiple niches (33). Within each ZOTU, individual genomes possessed pathway gene variations consistent with the ZOTU as a whole, even though some genomes were missing genes that a majority of others in the ZOTU possessed. These false negatives could result from incomplete MAGs, which is why we focused on ZOTUs. Accounting for modularity and genome variability within ZOTUs, comparative genomics has confirmed that this Fe oxidation pathway is common to all Zetaproteobacteria lineages. Support for cyc2 as an Fe oxidation gene in neutrophilic Fe oxidizers. Cyc2 homologs in neutrophilic FeOB are commonly referred to as "putative Fe oxidases" based on homology to the functionally characterized A. ferrooxidans Cyc2, though sequence similarity is low. Indeed, our phylogenetic analysis shows that Cyc2 sequences are highly diverse, and most of the neutrophilic FeOB Cyc2 homologs fall into Cluster 1 (Fig. 3). This cluster forms a distinct group from the clusters containing biochemically characterized Fe oxidases: Cyc2 from A. ferrooxidans (Cluster 2) and Cyt 572 from Leptospirillum sp. (Cluster 3). Although none of the Cluster 1 Cyc2s have been biochemically characterized, the high bootstrap support suggests a common function, and the prevalence of sequences from Fe-oxidizing isolates strongly suggests involvement in Fe oxidation.
Another clue to function lies in the expression of cyc2 in Fe mat environments, where Fe oxidation by the Zetaproteobacteria is required for carbon fixation and growth. Genes central to fitness are often highly expressed (34,35). We measured cyc2 expression in five different active Fe mats from three different hydrothermal vent fields and confirmed that cyc2 is frequently the highest-expressed gene. This is true in diverse Zetaproteobacteria lineages ( Table 2), suggesting that the pathway is important to Zetaproteobacteria fitness in many different environments. The level of cyc2 expression in the Zetaproteobacteria is consistent with the high expression in Fe-oxidizing isolates of Acidithiobacillus sp. (often above microarray detection limits) and Ferrovum sp. (8ϫ above average expression) (36)(37)(38). In the environment, the neutrophilic Gallionellaceae have been shown to express cyc2 highly, up to the 100th percentile in an Fe-rich aquifer (39). Together with the expression of cyc1 and terminal oxidase genes, this shows that the putative Fe oxidation pathway is consistently expressed under Fe-oxidizing conditions. The especially high expression of cyc2 across various Fe-oxidizing taxa and Fe-oxidizing environments supports its role in Fe oxidation in both acidophiles and neutrophiles.
To link cyc2 to Fe oxidation, we followed its expression when Fe(II) was added to Fe mat samples. In separate experiments at Loihi and Mariana, Fe(II) amendment resulted in both active biotic Fe oxidation (see Fig. S7 in the supplemental material) and increased cyc2 gene expression. This increase in expression was found not only for the whole sample but also in every Zetaproteobacteria genome detected within these samples (Fig. 6). Although there was variation in the timing and magnitude of the response, which may be lineage specific, the fact that expression increased in all Zetaproteobacteria suggests that cyc2 expression is stimulated by the presence of Fe(II). The Fe(II) amendment also resulted in increases in carbon fixation and central metabolism gene expression, suggesting a link between cyc2 expression, neutrophilic Fe oxidation, and growth (see Fig. S9 at https://doi.org/10.6084/m9.figshare.c.4646336).
Can the cyc2 gene be used as a marker for Fe oxidation? Unlike many other energy metabolisms, neutrophilic Fe oxidation is challenging to track in the environment due to the lack of an isotopic signature and difficulties distinguishing biotic from abiotic Fe oxides. Until recently, there have not been any candidates for a widely applicable Fe oxidation genetic marker; instead, it seemed that there were many different potential Fe oxidases, with various levels of functional verification (e.g., references 6, 7, and 40). Our work adds to the mounting evidence that Cyc2 is an Fe oxidase. The cyc2 gene is widely distributed across many Fe-oxidizing lineages, with homologs in acidophiles and neutrophiles. Specifically for neutrophiles, cyc2 is common across the well-studied neutrophilic chemolithotrophs Gallionellaceae and Zetaproteobacteria. As we have sequenced more of these neutrophilic FeOB genomes, this association has held true (9,13,15,21,41). However, our Cyc2 phylogeny has identified a substantial number of cultured and uncultured organisms that have not yet been shown or tested to be capable of Fe oxidation, work that could bolster confidence. In all, the cyc2 gene is a promising genomic marker of the capacity for Fe oxidation across many different Fe-oxidizing lineages, including neutrophiles.
Not only is cyc2 common to all well-established neutrophilic Fe oxidizers, it is also highly expressed in environments where neutrophilic Fe oxidizers predominate (this study and reference 39). This opens the possibility of cyc2 expression levels as an indicator of microbial Fe oxidation activity. Indeed, when we stimulated Zetaproteobacteria Fe oxidation in incubations, expression of cyc2 increased along with an increase in carbon fixation and central metabolism genes. This is consistent with Fe-oxidation-fueled chemolithoautotrophic growth, and so relative cyc2 expression levels can correspond to increases in Fe oxidation activity. However, our results suggest that cyc2 expression levels may not be easily related to Fe oxidation activity in the environment. All Zetaproteobacteria in our samples were expressing cyc2 prior to Fe(II) amendment, when there was no detectable dissolved Fe(II). This could represent baseline expression by obligate Fe oxidizers, which always need to be prepared for Fe oxidation. In this case, relative changes in cyc2 expression would remain more informative for activity. Alternatively, cyc2 expression before Fe(II) amendment could result from cryptic cycling of Fe between Fe oxidizers and reducers (42). Such cryptic cycling would make developing a genetic marker for activity even more important for tracking Fe oxidation activity. Because of these potential complications, further transcriptomics experiments should focus on isolates or microcosms without Fe reducers. In combination with our results, such experiments will help us understand how to use cyc2 expression levels to interpret Fe oxidation activity in the environment.
Conclusions. Using paired metagenomes and metatranscriptomes from the Zetaproteobacteria, we have been able to demonstrate that the Cyc2-based neutrophilic Fe oxidation pathway is widespread and highly expressed in the environment, validating the environmental importance of the pathway. We have shown that the Cluster 1 cyc2 gene, conserved in the Zetaproteobacteria and other neutrophilic Fe oxidizers, is highly expressed in multiple Fe mat environments and is stimulated by Fe(II) addition, suggesting it may be regulated. This makes cyc2 an excellent marker of Fe oxidation capability and may allow us to detect and monitor the activity of Fe oxidizers in the environment. However, to correlate expression with activity, further efforts should focus on testing the regulation of cyc2 in diverse organisms and simple communities. The phylogeny of Cyc2 shows at least three distinct clusters, with some neutrophilic Fe oxidizers possessing multiple copies (e.g., Clusters 1 and 3 in the Zetaproteobacteria). This may be akin to an example in Pseudomonas aeruginosa, which has multiple cbb 3 -type cytochrome c oxidases that are optimized for high and low O 2 concentrations and for resistance to respiratory inhibitors cyanide and nitrite and thus allow growth under different conditions (43). If Cyc2 variants are similarly optimized, they may enable Fe oxidation under various conditions, a hypothesis that could be tested by independently monitoring cyc2 from different clusters in diverse habitats and growth conditions. Without a marker of activity, the roles of neutrophilic Fe oxidizers have been virtually invisible outside model Fe-oxidizing environments, like Fe microbial mats. By applying our findings to other environments, we can start to reveal how Fe-oxidizing microbes drive key biogeochemical cycles in the varied marine and freshwater habitats where they thrive.  Table S1 in the supplemental material). To preserve in situ expression, 18 samples were collected using devices half-filled with 2ϫ RNAlater (Invitrogen, Carlsbad, CA, USA). Samples collected using a syringe sampler device (44) provided ϳ10 to 30 ml of mat material representing a discrete microbial population, as opposed to 2 liters (scoop) or Ͼ5 liters (suction sample) of bulk sample. After settling for a few hours at 4°C, the overlying supernatant was removed and samples were stored at Ϫ80°C.  (45,46). At Mariana, geochemistry samples were collected using the hydrothermal fluid and particle sampler (HFPS) (47) or the microbial mat sampler (44). The HFPS pulls fluid through a titanium inlet nozzle at 1 to 4 liters/min. The fluid flows through a continuously flushed titanium and Teflon manifold and is diverted into sample containers or to a SeaBird (Bellevue, WA) SBE 63 oxygen sensor and an AMT (Rostock, Germany) deep-sea glass pH electrode. In extremely low-outflow vent environments, seawater will be entrained in the HFPS and dilute the in situ fluid. We collected temperature, pH, and O 2 concentrations for ambient water, at the surface of the chimney, and with the nozzle inserted into the microbial mat. HFPS chemistry results represent the fluid composition at the measured temperature, including any entrained seawater that occurs during sampling. The microbial mat sampler has a much lower intake rate (Ͻ0.2 liter min Ϫ1 ) and is better able to capture chemical microenvironments. The microbial mat sampler was equipped with 0.22-m inline filtering and a check valve for chemical analysis. Fe(II) and total Fe concentrations were assayed using the ferrozine method with 40 mM sulfamic acid to stabilize Fe(II) (48,49); the detection limit was estimated at 0.12 M Fe(II). Samples recovered with the HFPS were processed as described previously (47) and analyzed shipboard for pH by glass electrode and on shore for total dissolved iron by atomic absorption at National Oceanic and Atmospheric Administration/Pacific Marine Environmental Laboratory (NOAA/PMEL) and by inductively coupled plasma mass spectrometry (ICPMS) at the University of Washington Department of Civil Engineering.

MATERIALS AND METHODS
Fe(II) amendment experiments. Shipboard Fe(II) amendment experiments were conducted on bulk mat samples from Loihi (J2-677-SSyellow) and Mariana (J2-801-SC8). Samples were transported to the ship after 2 h (Loihi) and 11 h (Mariana) of remotely operated vehicle (ROV) operations and allowed to settle at 4°C for 1 h prior to removing the majority of the supernatant and starting the experiment. One sample was taken immediately prior to Fe(II) amendment [pre-Fe(II) addition]. Water bath temperature and initial Fe(II) amendment concentration were set to mimic environmental conditions. Fe oxidation pseudo-first-order rate constants (k 1 ) were calculated using a log-linear fitted trend line.
At Loihi Seamount, Fe mat floc was added to two 250-ml vessels; one remained alive while the other was killed using 1 mM azide, which has been shown to interact with Fe(II), though not at this concentration and time interval (50). Both vessels were shaken by hand several times a minute in a 35°C water bath. To initiate the experiment, 100 M FeCl 2 was added. After this, starting after 2 min and subsequently at 10-min intervals, samples from each vessel were removed for Fe(II) and total Fe Validating the Cyc2 Neutrophilic Fe Oxidation Pathway January/February 2020 Volume 5 Issue 1 e00553-19 msystems.asm.org 13 measurements by the ferrozine method (48). Concurrently, 30 ml from the living vessel was mixed 1:1 with 2ϫ RNAlater (Invitrogen). This mixture was held at 4°C for a few hours prior to freezing at Ϫ80°C. At Mariana, Fe mat material was sparged with a 2% O 2 gas mix (pH 5.9/6.2 before/after sparge). Each time point (n ϭ 5) and treatment (duplicate living and 3 mM azide killed) had its own 125-ml reaction vessel with 30 ml mat material. In addition to a pre-Fe(II) addition sample, one sample was taken at the end which did not experience any Fe(II) addition. Both of these nonamended samples had low Fe(II) concentrations (below detection [BD] and 0.3 M, respectively). Each reaction vessel was amended with 333 M FeCl 2 and suspended in a 28°C water bath with frequent mixing by hand. Starting after 4 to 6 min and subsequently at 10-min intervals, one vessel was sacrificed at each time point, for sampling for Fe(II), total Fe, and pH and mixing of 25 ml of material 1:1 with 2ϫ RNAlater.
DNA and RNA extraction. DNA samples were extracted using the FastDNA Spin kit for soil (MP Biomedicals, Santa Ana, CA, USA) according to the manufacturer's instructions, except that 250 l of a 0.5 mM sodium citrate solution (pH 5.8) was added prior to lysis. RNA samples were extracted using the NucleoSpin RNA kit (Macherey-Nagel, Bethlehem, PA, USA), with modifications detailed in Text S1. Prior to library preparation, an internal in vitro-transcribed RNA standard, pTXB1, was added (Text S1; see also 16S rRNA, metagenome, and metatranscriptome sequencing. Microbial community composition was first estimated using a PacBio-based 16S rRNA gene survey, with SILVAngs used for taxonomic classification (see Text S1) (75). Zetaproteobacteria operational taxonomic units (ZOTUs) were classified from these 16S rRNA gene sequences using ZetaHunter (51). Samples were chosen for metagenomic (MG) and metatranscriptomic (MT) sequencing based on the microbial community composition and Zetaproteobacteria diversity. MG and MT libraries were prepared and sequenced at the University of Delaware Sequencing and Genotyping Center. Sequencing details are provided in Text S1.
Metagenome assembly, binning, and annotation. Raw sequence reads were trimmed to remove adaptors, poor-quality regions, and short sequences (Trimmomatic) (52), and paired reads were merged if overlapping (Flash) (53). Metagenome libraries were assembled from quality-controlled (QC'ed) reads using metaSPAdes v3.10, with read error correction disabled to improve recovery of real community genomic variation (54). Only contigs mapping Ն1ϫ read coverage over 90% of their length were utilized in downstream analysis (ϳ92% remained).
RNA read recruitment and expression estimates. Raw total RNA reads were quality controlled (see above) using Trimmomatic (average 99% of reads passed) (52). rRNA reads were removed using SortMeRNA (v2.1b) (70). The resulting non-rRNA reads, primarily mRNA, were used for subsequent recruitment for expression estimates. MT reads were recruited to the MG from the same sample, with the following exceptions: 665-MMA4 was recruited to 665-MMA12; S7_B5, S8_B2, S8_B3, S9, and S24 were recruited to S7_B4 MG. Reads were mapped using Bowtie 2, with default parameters (71).
To determine gene read recruitment, we used BEDTools to extract the read count from each gene coordinate region (72). We used three normalization methods for estimating gene expression: (i) transcripts per million (TPM), normalizing for sequencing effort and gene and read lengths (73); (ii) TPM values further normalized to the average expression of six constitutively expressed genes (adk, gyrA, recA, rpoB, rpoC, and secA) (74) to correct for changes in organism relative abundance (constitutive normalized expression); and (iii) TPM values normalized to the maximum expression for the time series for visual representation.
Data accessibility. High-quality full-length reads (20-pass minimum) from the PacBio 16S rRNA gene survey were submitted to GenBank (MK048478 to MK048944). Raw metagenome and metatranscriptome reads, as well as 5-pass-filtered PacBio 16S rRNA gene reads, were submitted to the NCBI SRA under BioProject accession PRJNA555820. Metagenome assemblies from this study and reassembled metagenome assemblies from the work of Fullerton et al. (9) were submitted to the JGI IMG database (sequence project IDs Gp0295814 to Gp0295821 and Gp0295823 [this study]; analysis project IDs Ga0256915 and Ga0257019 to Ga0257023 [Fullerton et al. {9}]). Zetaproteobacteria MAGs were also submitted to the JGI IMG database (see sequence project IDs listed above). Specific accession numbers per sample are shown in Table S4

SUPPLEMENTAL MATERIAL
Supplemental material is available online only. TEXT S1, PDF file, 0.8 MB. The authors declare no conflicts of interest.