Hydrogen peroxide (H₂O₂) is a central reactive oxygen species (ROS) in the global cycle of oxygen (O₂), and is thought to have been a precursor to O₂ via abiotic reactions in early Earth (Slesak et al., 2012). ROS such as H₂O₂ and superoxide (O2•-) may be produced in natural waters via UV (hν)-induced photochemical reactions between dissolved O₂ and organic carbon (Wilson et al., 2000a,b). The absorption of light by appropriate chromophore(s) of DOC results in an excited-state intermediate (DOC^*), which transfers electrons to O₂ to form the superoxide radical O2•- (Equation 1), followed by disproportionation of HO2• to form H₂O₂ (Equation 2, Wilson et al., 2000a,b).

The production of superoxide and hydrogen peroxide from photochemical reactions has important implications for the cycling of dissolved iron (Fe) and other trace elements adsorbed to Fe-oxide solid phases. Several studies have documented diel changes (i.e., changes with light intensity) in the concentrations of dissolved trace elements in stream water containing Fe and DOC (Wilson et al., 2000a,b; Miller et al., 2009). Once hydrogen peroxide is formed, hydroxyl radicals (OH^•) are also produced via the reaction of H₂O₂ and ferrous iron [Fe(II)], (i.e., Fenton reaction, Equation 3), which can lead to the precipitation of highly-insoluble ferric iron [Fe(III)]-oxyhydroxides (Winterbourn, 1995; Wilson et al., 2000a; Kocar and Inskeep, 2003).

Fe(III) can then quickly react with superoxide to form Fe(II), creating a fast ROS turnover (Equation 4, Wilson et al., 2000a; Kocar and Inskeep, 2003). Fenton chemistry can also lead to the chemical decay of hydrogen peroxide and superoxide via the iron-catalyzed Haber-Weiss reaction (Equation 5), producing hydroxyl radicals in the presence of Fe(III) (Kehrer, 2000). In sulfide-rich waters, hydrogen peroxide reacts with hydrogen sulfide (H₂S) to form elemental sulfur at low pH (Equation 6, Hoffmann, 1977).

Free radical reactions, whether abiotic (e.g., Fenton chemistry) or biochemical (e.g., aerobic metabolism), are important sources of internal ROS production in living organisms (Cannio et al., 2000), and these compounds can play important roles in DNA damage, cell signaling, and homeostasis (D'autréaux and Toledano, 2007; Galaris and Pantopoulos, 2008; Cornelis et al., 2011). Organisms of all domains of life (Archaea, Bacteria, and Eukarya) possess defense mechanisms against intra- and extracellular ROS. These detoxification enzymes participate directly in H₂O₂ and Fe cycling (Table 1), and are thought to work in concert to avoid oxidative stress (Pedone et al., 2004; Maaty et al., 2009). Metagenomes of Fe(III)-oxide microbial communities collected in high-temperature acidic geothermal springs of Yellowstone National Park (YNP, Wyoming, USA) have identified chemolithoautotrophic organisms contributing to the early development and production of these microbial mats (Inskeep et al., 2010, 2013; Kozubal et al., 2012; Beam et al., 2016). Curated sequences indicate the presence of numerous oxidative stress response genes (including those listed in Table 1) in two major populations. The rod-shaped bacteria Hydrogenobaculum spp. initiate biofilm attachment, and the iron-oxidizing archaea Metallosphaera yellowstonensis (strain MK1) are responsible for the oxidation of Fe(II) and subsequent accretion of Fe(III)-oxides (Macur et al., 2004; Kozubal et al., 2008; Takacs-Vesbach et al., 2013; Jennings et al., 2014; Beam et al., 2016). Oxygen is a key driver of microbial growth and subsequent mat morphology (Kempes et al., 2014), and its diffusion into the mat is the rate-limiting factor for Fe(III)-oxide mat formation (Beam et al., 2016). Microelectrode measurements of oxygen diffusion into the Fe(III)-oxide mats have revealed steep gradients of O₂ (50-0 μM) from the aqueous-mat interface to ~1 mm depth (Bernstein et al., 2013). The top zone (~1 mm) of the Fe(III)-oxide microbial biofilms, which is in direct contact with the overlaying spring water, thus provides an excellent target microenvironment to study possible changes in ROS stress-related gene expression levels.

The primary goals of this study were to evaluate changes in H₂O₂ concentrations in Fe(III)-oxide mat environments as a function of light intensity (diel cycle) and record the associated microbial response to oxidative stress based on the activity of key genes involved in H₂O₂ cycling. Specific objectives were to (i) measure H₂O₂ concentrations in the outflow channels of acidic geothermal springs as a function of photon flux, (ii) determine the spatiotemporal dynamics of dissolved H₂O₂, Fe(II), Fe(III), O₂, and total sulfide (TS) concentrations occurring within the outflow channels of Fe(III)-oxide springs, and (iii) investigate the expression of key genes known to be involved in oxidative stress response in two of the primary autotrophic populations important in Fe(III)-oxide mat ecosystems. The concentrations of dissolved compounds were measured at different times (diel and seasonal), and at various transect positions within the outflow channels of several Fe(III)-oxide depositing springs in YNP. These geochemical measurements were complemented with in situ expression profiles of oxidative stress response genes in M. yellowstonensis strain MK1 and Hydrogenobaculum spp.

Spatial and temporal changes in H₂O₂ and other dissolved chemical species were assessed over diel cycles in 2014 and 2015, and provide the first insight regarding concentrations of H₂O₂ in acidic Fe(III)-oxide depositing geothermal springs of YNP. Concentrations of H₂O₂ ranged from ~2 to 4 μM across all Fe-oxide sites, and were significantly lower in geothermal source waters and/or in the presence of dissolved sulfide. The similar concentrations of H₂O₂ was likely due to similar geochemistry across the three sites [e.g., Fe(II) ~30–40 μM, pH ~3–3.5, T ~68–80°C]. The observed range in H₂O₂ values in acidic geothermal systems is higher than previously observed in a limited set of thermal systems supporting phototrophic activity (<300 nM; Wilson et al., 2000a), and fall within values observed in rainwater (~6 μM; King et al., 2007), freshwater (~300 nM; King et al., 2007) or coastal waters (50–150 nM; Miller et al., 2005), but are considerably higher than values in the open ocean (~0.7 nM; King et al., 2007). The Fe-oxide mat systems sampled in the current study do not contain phototrophic organisms, although small increases in DO during peak photon flux may in part, be due to inputs from adjacent phototrophic systems.

Dissolved sulfide also plays an important role regulating concentrations of H₂O₂ in geothermal springs. Acidic geothermal springs often contain variable amounts of dissolved sulfide originating from hydrothermal source waters. The majority of dissolved sulfide at low pH is H₂S(aq), which degasses rapidly as the outflow channel equilibrates with atmospheric O₂ (Inskeep et al., 2005; Nordstrom et al., 2005). Dissolved H₂S can react with H₂O₂ (Equation 6; Hoffmann, 1977) so it is expected that high sulfide waters will have lower H₂O₂ levels. For example, the source waters of Beowulf Spring were highly sulfidic (60–100 μM dissolved sulfide), while no sulfide was detected at the discharge of OSP Spring and Echinus Geyser Spring (Table 2). Dissolved sulfide concentrations declined considerably from 64 to 3 μM (May sampling) and from ~100 to 7 μM (July sampling) along the transect points BE_a to BE_c. Hydrogen peroxide levels were consistently lower (~1 μM) in the presence of high sulfide (e.g., BE_a, b), suggesting that high sulfide likely acts to minimize observed aqueous concentrations of H₂O₂ in acid-sulfate-chloride geothermal springs.

High levels of Fe(II) near 30–40 μM in all sites (one order of magnitude higher than observed H₂O₂ concentrations) may also regulate levels of H₂O₂ via abiotic reactions. In Fe-mats from Beowulf spring, diel data suggest that the Fenton reaction (Equation 3) may contribute to small increases in Fe(III) observed during peak photon flux. In many cases, increases in dissolved H₂O₂ were correlated with increases in O₂. For example, dissolved O₂ and H₂O₂ both increase down gradient in thermal outflow channels, and might indicate a common mechanism. Peak O₂ concentrations occurring at peak photon flux also correlated with small increases in H₂O₂, which suggests that photochemical reactions (e.g., Equations 1, 2) may occur in these geothermal waters to generate greater amounts of H₂O₂. Small increases in H₂O₂ were indeed noted in the Beowulf and 100 Springs Plain outflow channels with increasing photon flux, yet in the oxygenated Echinus channel, higher concentrations of H₂O₂ were observed at dawn.

Increases in dissolved O₂ down channel result from ingassing of atmospheric O₂ over the Fe(III)-oxide mats (Bernstein et al., 2013; Beam et al., 2016). However, daytime increases of O₂ (and H₂O₂) at peak photon flux may also suggest possible inputs from adjacent phototrophic systems into the flow channel. Similarly, the increase in O₂ concentrations over the day down the Echinus (ECH_b and ECH_c) and 100 Springs Plain (OSP_b) channels could be explained by oxygen inputs from adjacent phototrophic algal communities (containing members of the Cyanidiales; Castenholz and McDermott, 2010; Reeb and Bhattacharya, 2010), which exist across temperature ranges from ~38 to 56°C, but are not present in sampling positions BE_a-e. Thus, despite a constant input of surface water from the landscape across a diel cycle, these inputs may contain more O₂ during daytime hours as a result of phototrophic metabolism.

Increases in O₂ during the day may in part be responsible for increases in microbial growth of M. yellowstonensis and Hydrogenobaculum, which are both aerobic (Kozubal et al., 2008; Takacs-Vesbach et al., 2013; Jennings et al., 2014). Higher 16S rRNA gene expression levels of Metallosphaera compared to Hydrogenobaculum in both environments (Figure 5) were consistent with higher relative abundances of Metallosphaera observed previously in the upper ~1 mm of mature Fe(III)-oxide mats, which also corresponds to the active O₂-consuming layer (Beam et al., 2016). Given the large changes in ROS that may be expected in natural environments as a function of photon flux, it is highly likely that microorganisms vary the production of enzymes known to react with these highly labile redox-active species across a diel cycle. Reactions involving ROS and Fe can represent an environmental stress for both aerobic and anaerobic organisms, leading to oxidative damage of DNA, RNA, proteins, lipids, and cofactors (Henle and Linn, 1997; Cabiscol et al., 2000; Green and Paget, 2004; Maaty et al., 2009).

Normalized expression levels of genes potentially involved in oxidative stress exhibited very few consistent trends for a given organism across sites or temporal sampling, and/or between the two different autotrophs (Figure 6). For example, in Fe-mat samples from BE-d, Metallosphaera exhibited higher relative expression levels of ROS-related genes, and both organisms showed a significant decrease in expression levels during peak photon flux (~2 p.m.). However, results from Fe-mats sampled at ECH_b (or c) behaved differently where Hydrogenobaculum exhibited higher normalized expression for the majority of ROS-related genes than Metallosphaera (Figure 6). Moreover, no major changes in expression levels for either organism were observed over time in Fe-mats from Echinus Geyser. Small changes in normalized expression occurred for Metallosphaera and Hydrogenobaculum across the three sampling times in ECH_c, but the responses were not consistent between the two autotrophs.

Interestingly, both populations appeared to exhibit less oxidative stress during peak photon flux: normalized levels of all oxidative stress transcripts were lower at the 2 p.m. sampling time, although this was not consistent in Beowulf and Echinus outflow channels. In Beowulf Spring, photochemical reactions and subsequent Fenton chemistry may thus be responsible for H₂O₂ formation and decay during peak photon flux, while microbially mediated H₂O₂ cycling may occur in morning and evenings. More replication and experimentation including nightly sampling will be necessary to develop a mechanistic understanding of the relative roles of abiotic photochemical reactions versus microbial activity in the cycling of ROS in these geothermal environments. Expression analyses of specific ROS genes indicated that M. yellowstonensis exhibits more oxidative stress response than Hydrogenobaculum in Beowulf (Figures 6A,B), but the opposite trend was observed in Fe mats from Echinus Geyser (Figures 6C–F). Although the Metallosphaera population was found to be the most active in ECH_b and c Fe(III)-oxide mats (see 16S transcripts levels in Figures 6B,C), the Hydrogenobaculum population exhibited higher levels of normalized ROS gene expression.

The gene coding for external PRX (Equation 8) was the most expressed in Metallosphaera at all sites and in Hydrogenobaculum in ECH, which suggests that extracellular H₂O₂ is an important source of oxidative stress over a diel cycle in these Fe(III)-oxide mats. Relative expression data show similar transcript levels of SOD (Equation 7) and internal PRX (Equation 8) in both organisms at all sites, which may indicate co-expression of these two genes to coordinate the removal of ROS inside the cell, as previously observed in the archaeon Sulfolobus solfataricus where they form an enzymatic complex (Maaty et al., 2009). Normalized expression of these two genes is higher in M. yellowstonensis than in Hydrogenobaculum in BE_d, ECH_b, and ECH_c Fe(III)-oxide mats, which may reflect a higher level of internal oxidative stress in the bacterial population, and thus a higher cellular response to mitigate ROS. Higher FDX gene transcripts detected in Hydrogenobaculum at all sites also support this hypothesis, as ferredoxin is involved in maintaining intracellular redox potentials (Maaty et al., 2009). In Hydrogenobaculum spp., both rubrerythrin (11) and ferredoxin (Equation 12) enzymes (Weinberg et al., 2004) may participate equally in Fe cycling, as indicated by comparable expression levels at all sites. The very low normalized gene expression of FDX gene (ratio of ca. 10⁻⁶) compared to RBR gene in M. yellowstonensis str. MK1 at all sites (Figure 6), suggests that ferredoxin is not the main enzyme responsible for iron homeostasis in this organism. The very similar expression levels of genes coding for GRX and TRX in both microbial populations at all sites is not surprising, as the TRX/GRX multidomain protein is an ubiquitous thiol redox system in nature (Pedone et al., 2004; Reott et al., 2009).

This is the first study to report on the concentrations of H₂O₂ in acidic geothermal springs of YNP, and only one study has reported on peroxide levels in other high-temperature systems (Wilson et al., 2000a,b). Clearly, additional experimentation will be necessary to develop a mechanistic understanding of factors controlling H₂O₂ levels in geothermal waters of high Fe systems and the implications of spatiotemporal changes in H₂O₂ for other ROS such as the hydroxyl radical and superoxide. This study is also the first to report expression analyses of several oxidative stress response genes in situ, and has documented the expression of specific genes known to be involved in regulation of H₂O₂ and Fe species in two primary chemolithoautotrophs important in the formation of high-temperature Fe(III)-oxide mats. Several of the ROS stress response genes tested have been shown to be upregulated under high H₂O₂ levels (~30 μM) in controlled conditions using a Sulfolobus sp. (Maaty et al., 2009), however, the H₂O₂ values observed in acidic Fe-oxide environments were 10-fold lower, thus difficult to compare. Regarding the widespread genetic machinery across the three domains of life to cope with oxidative stress (Pedone et al., 2004; Maaty et al., 2009), our gene expression results might suggest a more constitutive response to ROS in Fe-mat populations, and more generally during aerobic metabolism.

MM, JB, ZJ, BB, EB, and WI performed field experiments. MM, JB, and BB performed laboratory analyses. MM and WI wrote the manuscript.