Enzymes (proteins that catalyze chemical reactions) that enable the reversible oxidation of H₂ to protons (H⁺) and electrons (e⁻; Reaction 1) can be traced back to the last common ancestor of Archaea and Bacteria (Boyd et al., 2014; Weiss et al., 2016), at least 3.8 billion years ago (Battistuzzi et al., 2004; Wolfe and Fournier, 2018). H 2   ↔   2 H + + 2 e −  Reaction  1

Over the past −3.8 Ga, organisms with the ability to metabolize H₂ have diversified to enable H₂ oxidation to be coupled to the reduction of nearly all imaginable and available oxidants, including oxygen (O₂), nitrate (NO₃ ⁻), ferric iron (Fe(III)), sulfate (SO₄ ²⁻), and carbon dioxide (CO₂), among others (Figures 1A,B). In this capacity, H₂ can support the growth of aerobic (respire O₂) or anaerobic (respire an oxidant other than O₂) microorganisms and those that are autotrophic (fix inorganic carbon to biomass) or heterotrophic (use organic carbon as an electron donor and/or for biomass). In the case of heterotrophic microorganisms, H₂ oxidation typically is used to supplement their energy metabolism (Peters et al., 2015; Greening et al., 2016). Alternatively, heterotrophic organisms that are oxidant limited often can ferment, leading to H₂ production (discussed below).

Despite the vast diversity of metabolic backgrounds that feature reversible H₂ oxidation, they are all dependent on one of three classes of enzyme to catalyze its transformation: [FeFe]-hydrogenases, [NiFe]-hydrogenases, and [Fe]-hydrogenases (or Hmd) (Peters et al., 2015; Greening et al., 2016; Greening and Boyd, 2020). Hmd catalyzes the reduction of the substrate methenyltetrahydromethanopterin with H₂ in methanogens (Shima et al., 2008), and won’t be discussed more here given its limited metabolic distribution. The least ambiguous way of examining whether microbial cultures or environmental samples can catalyze reversible H₂ oxidation activity is to measure rates of H₂ consumption in microcosm experiments. However, this is a time-consuming task and is infrequently done (discussed below). Fortunately, like other enzymes/proteins that a cell is dependent on, hydrogenase enzymes are encoded by genes in the genomes of cells. Therefore, organisms with the ability to metabolize H₂ can be readily identified by sequencing the genome of the culture (or the entire community in the case of environmental samples) and subjecting those genomes to comparative bioinformatics analyses. Such studies show that [FeFe]-hydrogenases are widely distributed among Bacteria, anaerobic eukaryotes (ciliates, flagellates, and fungi), and algae (Peters et al., 2015) and were only recently discovered in a limited subset of Archaea (Peters et al., 2015; Greening et al., 2023). In contrast, [NiFe]-hydrogenases are widespread among Archaea and Bacteria but have yet to be detected in Eukarya (Peters et al., 2015). Paralogs (evolutionarily related proteins that have different functions) of [NiFe]-hydrogenases, however, are widespread in Archaea, Bacteria, and Eukarya where they function as entry points for reducing equivalents into electron transport chains (i.e., NADH dehydrogenase, Nuo) (Friedrich and Scheide, 2000; Schut et al., 2013). However, these proteins are derived from [NiFe]-hydrogenases and therefore H₂ metabolism is a trait commonly associated with microbial life rather than higher forms of life.

Common among all characterized [NiFe]- and [FeFe]-hydrogenases, but distinct between these two classes, is an evolutionarily conserved large subunit that harbors the nickel-iron-sulfur or iron-sulfur active site, respectively, where reversible H₂ oxidation takes place (Peters et al., 2015). Variation in how H₂ is metabolized by cells via [NiFe]- and [FeFe]-hydrogenase has occurred through recruitment of genes that encode additional protein subunits that enable diverse redox coupling with intermediate electron carriers in the cell or in the cell membranes such as ferredoxin (Fd), NADH, flavins (e.g., F₄₂₀), quinones (e.g., methanophenozine), and heterodisulfide (e.g., coenzyme M-coenzyme B), among others (Thauer et al., 2010). These intermediary electron carriers pass electrons to electron transport chains specific for reduction of O₂, NO₃ ⁻, Fe(III), SO₄ ²⁻, etc. allowing for energy to be conserved that can be used to power cells. The recruitment of these additional protein encoding genes to [NiFe]- and [FeFe]-hydrogenases was highly advantageous to cells since it allowed for electrons from H₂ (via these intermediary electron carriers) to enter a variety of electron transport chains thereby enabling energy to be conserved from H₂ to power cells. As such, these particular protein recruitments were under strong selective pressure to be maintained, allowing for the diversification and general phylogenetic coherence of large subunit sequences that reflect their mode of metabolic functioning (Figures 1A,B). This feature has been exploited bioinformatically such that is possible to predict the functionality and general directionality of hydrogenase enzymes using genomic data alone (Vignais and Billoud, 2007; Boyd et al., 2014; Peters et al., 2015; Greening et al., 2016; Poudel et al., 2016). This information can be combined with other genomic data to further examine modes by which H₂ oxidation might be coupled with reduction of various oxidants such as O₂, NO₃ ⁻, Fe(III), SO₄ ²⁻, and CO₂. Such information is powerful when trying to understand and predict the potential for microorganisms to influence the flux of H₂ from the deep surface to near surface environments where it can be mined.

The size of a population is generally scalable to the availability of resources to support that population, since communities assemble to maximize utilization of resources until one becomes limiting (i.e., carrying capacity). Using genome sequences from Archaea and Bacteria, it was shown that nearly 30% code for one or more [NiFe]- or [FeFe]-hydrogenase enzyme (Peters et al., 2015; Greening et al., 2016), consistent with a widespread ability to reversible oxidize H₂ among microorganisms. While it is not atypical for an organismal genome to encode multiple copies of hydrogense enzymes, the genomes of several organisms from deep subsurface rock-hosted environments stand out, including that of the bacterium “Candidatus Desulforudis audaxviator” (Chivian et al., 2008). This organismal genome was reconstructed from metagenomic sequence obtained from fracture fluids from a 2.8 km depth in South African gold mines. The genome encodes four copies or homologs of [NiFe]-hydrogenase and six homologs of [FeFe]-hydrogenase, alluding to the central role that H₂ has in the cell’s energy metabolism. More recent work has brought this organism into culture and has demonstrated that it can indeed grow via H₂ oxidation (Karnachuk et al., 2019), as predicted from genomic data (Chivian et al., 2008). Multiple copies of hydrogenases in the genomes of organisms from taxonomic groups (phyla) that characteristically inhabit anoxic habitats is common (Peters et al., 2015), including in the Firmicutes, the group that “Ca. Desulforudis audaxviator” belongs to. Indeed, comparative metagenomic (shotgun sequencing) studies of natural microbial communities from surface environments relative to subsurface environments identified 5- and 2-fold enrichments in the genes encoding [NiFe]- and [FeFe]-hydrogenases in communities from subsurface environments relative to those obtained from surface environments (Colman et al., 2017).

Is the enrichment of hydrogenases in the genomes of organisms from subsurface environments related to the availability of H₂? Hot springs have been proposed as accessible windows into subsurface environments (Colman et al., 2017), and numerous community metagenomes have been generated from these environments (e.g. (Inskeep et al., 2013)). A comparative study of 50 metagenomes from 50 different high temperature hot springs revealed that hydrogenase homologs can be particularly enriched (2X higher than background) in microbial communities from springs that are sourced by reduced volcanic gases and condensed steam or vapor (Figure 1C (Lindsay et al., 2019)). Magmatic degassing of H₂ and high temperature reactions between steam and reduced iron bearing minerals during the ascent of vapor to the surface leads to enrichment of H₂ in this spring type (Lindsay et al., 2019). Indeed, a genomic analysis of one of those springs, termed SmokeJumper3 (SJ3), showed that >70% of the community members encoded at least one hydrogenase homolog (many genomes encoded multiple homologs), the vast majority of which were predicted using bioinformatics approaches to be involved in H₂ oxidation (Lindsay et al., 2019). Further, the total number of cultivatable cells in SJ3 that used H₂ as a reductant coupled to various oxidants (O₂, NO₃ ⁻, Fe(III), SO₄ ²⁻, thiosulfate, and elemental sulfur) was found to be one to two orders of magnitude greater than in two nearby springs (termed SJ1 and SJ2). Geochemical analyses indicate that SJ3 waters harbor the highest concentration of H₂ (2.1 µM) measured in a hot spring in Yellowstone to date, including in SJ1 (1.2 µM) and SJ2 (1.0 µM (Lindsay et al., 2019)). This suggests that organisms in SJ3 are particularly tuned to metabolize H₂ but are likely limited by availability of other nutrients, since H₂ was still measured at high concentration. In other H₂ supported subsurface biospheres such as hydrothermal vents, the number of hydrogenase homologs encoded by resident microbial communities can be even higher, in some cases 40-fold higher than from other surface habitats (Brazelton et al., 2012; Adam and Perner, 2018). Collectively, these observations point to a relationship between the availability of H₂ in subsurface environments and the extent to which microbial inhabitants of those environments depend on H₂ to support their energy metabolism. However, the nature of this relationship is likely to be nuanced by differences in other parameters in these systems, including variation in host rock that controls geochemical compositions, the availability of oxidants, and other nutrients (Adam and Perner, 2018).

The apparent relationship between the prevalence of hydrogenase encoding genes in community metagenomes from subsurface environments and the availability of H₂ in those environments strongly suggests that many of the cells in those environments are dependent on H₂. Yet, not all cells in natural environments are active and it is widely accepted that most (up to 80%) of them are inactive or in a stationary phase of growth (Kolter et al., 1993; Lennon and Jones, 2011), likely due to limitation of energy substrates, nutrients, water, or accumulation of metabolic by-products (waste) that feedback inhibits biochemical processes. Further, the presence of genes allowing for a particular metabolic function does not necessarily mean that function is being carried out. This is because genes must first be converted or transcribed to RNA before the RNA/transcripts can be read and proteins/enzymes be synthesized (i.e., the central dogma of biology). Yet, RNA is easily degraded and is difficult to recover from microbial cells in natural environments. This limitation of RNA, when combined with the added complexity of limited amount of biomass (and therefore RNA) in rock-hosted ecosystems, makes it advantageous to directly measure the activity of enzymes that are involved in H₂ consumption activity. Measurements of H₂ oxidation or consumption generally take one of three forms, each of which can provide different data. One way of measuring H₂ net consumption (consumption-production) is by measuring H₂ concentrations over time following a direct injection of gas into a formation (e.g. (Bagnoud et al., 2016)). This approach is labor-, time-, and cost-intensive but provides estimates that are less prone to “bottle” artifacts (described below). The other common approaches use microcosms, or small bioreactors, that can be sealed and their contents manipulated. A known volume of sediment or ground rock is added to the microcosm along with either filter-sterilized fracture water from the formation or synthetic medium designed to chemically mimic fracture waters. Then, a fixed amount of H₂ is added to the headspace and its net consumption is quantified over time. Alternatively, tritiated hydrogen is added and its conversion to tritiated water is quantified. The difference in these later two approaches is that the former is reflective of net H₂ consumption whereas the latter is measuring H₂ oxidation only. Killed (autoclave or otherwise) controls allow for these activities to be attributed to microbial processes or abiotic processes. Both of the last two approaches, which tend to be carried out in sealed reactors, can lead to under- or over-estimates of activity, since nutrient limitation can develop, microbial communities can be stimulated due to added substrates, or disturbance can affect interactions among populations.

Due to these methodological differences as well as large-scale differences in environment (e.g., acidic pH versus circumneutral pH; impacted by photosynthesis or not), the results of H₂ consumption measurements are often difficult to compare. For example, most measurements of net H₂ oxidation rates in hydrothermal vent communities are reported on a per cell basis (e.g. (Adam and Perner, 2018)). This is done due to the comparative ease by which cells can be enumerated in aqueous samples. However, this assumes that all cells are similarly dependent on H₂ and active, which is not likely to be true (see above). Further, this makes it difficult to compare to rock-matrix associated cells, since quantifying cells in such samples is difficult. Nonetheless, several measurements of net H₂ oxidation have been reported from rock-hosted ecosystems that span a range of temperatures (4°C–70°C) and geologic settings and with circumneutral pH and freshwater salinities, and these were compiled here for comparison. This included a single measurement from a circumneutral hot spring (Lindsay et al., 2018), two measurements from basaltic and carbonate sediments in circumneutral glacial meltwaters (Dunham et al., 2021), and a single measurement from a subsurface clay aquifer/reservoir proposed for use in storing radioactive waste (Bagnoud et al., 2016). These data from rock-hosted, non-photosynthetic ecosystems were then compared to several rates reported from non-rock-hosted environments, including terrestrial soils (Ji et al., 2017; Giguere et al., 2021) and marine and soda lake sediments (Adhikari et al., 2016), both of which are potentially influenced directly or indirectly by photosynthesis. These were all normalized to per unit mass or volume of material that was then converted to mass (soil mass, sediment mass), permitting their comparison.

Average rates of net H₂ oxidation in recently comminuted and fine grained carbonate sediments from Robertson Glacier and basaltic sediments from Kötlujökull Glacier (both −0.1°C), when incubated in microcosms at 4°C, revealed low rates of H₂ oxidation (−1.1 and 7.6 nmol d⁻¹ G dry mass (gdm)⁻¹) (Dunham et al., 2021). In a clay-hosted aquifer (incubated in situ at −17.5°C), a net rate of microbial H₂ oxidation of 1.53 μmol d⁻¹ cm^{3 −1} was reported (Bagnoud et al., 2016). Assuming a density of clay of 1.7 g cm^{3 −1} (Bagnoud et al., 2016), this value equates to 0.9 μmol d⁻¹ gdm^-1. Finally, a rate of net H₂ oxidation of 7.2 μmol d⁻¹ gdm⁻¹ was reported for fine-grained hot spring sediments sampled at the source of a circumneutral pH hot spring (pH, 6.8) when incubated in microcosms at 70°C (Lindsay et al., 2018). Despite the limited size of the dataset, they were plotted as a function of temperature (pseudo-Arrhenius plot), which revealed a linear relationship (R ² = 0.99) that has a slope of −10 and intercept of −5. This relationship yields a −1-2 fold increase in rate per 10°C increase in temperature, which is close to the expectation for a reaction with Arrhenius-like kinetics (2-fold increase in rate per 10°C increase). While this relationship is defined with only 4 datapoints, it is a bit unexpected due to large-scale expected differences in the organisms that inhabit these environments (e.g., archaeal-versus bacterial-dominated, aerobic versus anaerobic), the presumed differences in the type of hydrogenase enzymes they encode and their kinetics, differences in modes of redox coupling, and different amounts of biomass (e.g., 10⁴–10⁵ cells gdm⁻¹ in subglacial sediments to 10⁷–10⁸ cells gdm⁻¹ in hot spring sediments (Lindsay et al., 2019; Dunham et al., 2022)). This is in addition to differences in how the measurements were made (microcosm versus in situ measurements). To further explore the relationship between net H₂ oxidation and temperature, measurements from non-rock-hosted ecosystems were compared to those above.

A net H₂ oxidation rate of 0.84 μmol d⁻¹ gdm⁻¹ was reported for Antarctic soils when incubated at 10°C (Ji et al., 2017), which is close to what would be predicted (14°C) by the linear relationship. A net oxidation rate of 3.36 μmol d⁻¹ gdm⁻¹ was reported for a temperate soil; however, incubation temperature was not reported (Giguere et al., 2021). Applying the linear relationship from above yields a temperature of 40°C, which is ∼15°C higher than most temperate soils. We also note that rates of H₂ oxidation (not net) in marine sediments from a variety of environments range from 0.08 to 17.85 μmol d⁻¹ cm³. Assuming an average density for marine sediments (clay) of 1.7 g cm³ (Tenzer and Gladkikh, 2014), these rates translate to 0.05–10.5 μmol d⁻¹ gdm⁻¹ which are generally higher than for rock-hosted communities. This may be due to these measurements being direct H₂ oxidation measurements (conversion of tritiated H₂ to water), as opposed to net oxidation measurements, and these systems receiving organic carbon from the overlying water column that could further stimulate biomass production. While the relationship between net H₂ oxidation potential and temperature needs to be further refined and evaluated using additional measurements from diverse subsurface systems, the limited data compiled here indicates that microorganisms consume H₂ in rock-hosted ecosystems and that their net H₂ consumption rates are at least partially dependent on temperature.

To begin to examine how microbial H₂ oxidation could impact H₂ recovery in a natural system, we consider a 1 km deep, 1 km diameter cylinder (0.785 km³) of harzburgite/dunite rock in the Samail Ophiolite, Oman. An average rock density for harzburgite/dunite of 2.55 g cm³ (measured in cores BA1B, BA3A, and BA4A (Kelemen et al., 2020)) was used to normalize this volume of rock to mass and a porosity of 3.8% cm³ (average in these same cores (Kelemen et al., 2020)) was used as a crude method to identify the percentage of this rock that could be inhabited by microorganisms. Applying the net H₂ oxidation rate of 3.4 μmol d gdm at 40°C (extrapolated from Figure 2), which is a likely average for a 1 km depth (Nothaft et al., 2021), and assuming a constant H₂ oxidation rate throughout the rock column for simplicity, a net microbial consumption rate is estimated at 52 × 10³ tonnes per year. Assuming complete rock alteration of 1 kg of such rock, −0.3 mol H₂ can be produced (Leong et al., 2023), equivalent to −10¹¹ tonnes for the 0.785 km³ rock volume, which would support microbial hydrogenotrophs for −10⁷ years at this level of activity. Yet, the annual outgassing rate (flux of H₂) from the surface of the Samail Ophiolite is estimated to be 3.6 × 10³ tonnes km³ (Leong et al., 2023), an order of magnitude lower than the estimated net microbial H₂ oxidation potential. This suggests that up to >90% of H₂ produced in the Samail Ophiolite could be lost through microbial consumption. Importantly, the estimates for net H₂ oxidation for rock-hosted microbial communities were derived largely using samples from near surface rock-hosted ecosystems (e.g., subglacial sediments, hot spring sediments) which may differ markedly from those that inhabit subsurface rocks where the flux of oxidants and nutrients could be far lower.

While the focus of this perspective is on H₂ oxidation via hydrogenases, such enzymes are reversible and, under certain conditions, microorganisms can produce H₂. Microbial production of H₂ is most commonly associated with the process of fermentation, whereby organic carbon molecules are broken down in the absence of O₂ or other suitable oxidants. In the absence of other suitable oxidants, microbial cells turn to protons as a source of oxidant to regenerate oxidized intermediate electron carriers (Fd, NAD⁺, etc.) and enable organic carbon oxidation to continue. In the process of oxidizing intermediate electron carriers, protons are reduced to yield H₂ (Reaction 1). Fermentation is a common process in environments with available organic carbon but where other more preferrable oxidants are limiting. Examples of subsurface environments meeting these characteristics are oil and natural gas reservoirs, which are also often replete with H₂ (Zgonnik, 2020). Thus, the biological potential to produce H₂ is highest in subsurface environments that have substantial organic carbon that drives the consumption of available oxidants.

Genomic and metagenomic studies reveal that the ability to reversibly oxidize H₂ is a widespread trait that is integrated into the energy metabolism of nearly all types of microbial cells, ranging from aerobes to anaerobes, autotrophs to heterotrophs, and chemosynthetic to photosynthetic organisms. Metagenomic sequence from rock-hosted subsurface ecosystems reveal enrichment of hydrogenases in subsurface communities relative to those from the surface, alluding to the even greater importance for H₂ cycling in subsurface rocky environments that exclude light as an available energy source. Rates of net microbial H₂ oxidation in several rock-hosted habitats have been measured and a relationship between rate and temperature was established, with rates increasing by a factor of ∼1-2 per 10°C increase in temperature. This suggests that communities may have a significant impact on the flux of H₂ from the subsurface and that this effect might be more pronounced at higher temperature.

Yet, the availability of data to evaluate the potential effect of microorganisms on the flux of H₂ from subsurface environments remains highly limited, despite the recent surge in interest and investment in prospecting for and mining white and orange H₂ as a component of the energy economy (Hand, 2023). This is partly attributable to logistics and costs associated with gaining access to subsurface samples and fluids and ensuring that they are collected in a manner that is conducive for downstream use in geomicrobiological work. We call for closer communication and coordination between prospectors interested in maximizing white and orange H₂ recovery and geomicrobiologists interested in understanding the role of geologic H₂ in sustaining subsurface biospheres. Specifically, coordination and communication between prospectors and geomicrobiologists is needed to ensure that samples can be collected in a manner that is suitable for downstream use in microbiological work. Through such coordination, prospectors and microbiologists will be in a unique position to not only develop a more robust understanding of H₂ cycling activities of microbial communities in rock-hosted ecosystems but to also better guide prospecting efforts and to develop interventive strategies to mitigate H₂ loss due to microbial activities. Further, obtaining such understanding through coordination of efforts could result in new technologies and approaches to engineer subsurface systems to poise microorganisms to enhance the production of H₂ through fermentative or other biological activities.