Dark biological superoxide production as a significant flux and sink of marine dissolved oxygen.

The balance between sources and sinks of molecular oxygen in the oceans has greatly impacted the composition of Earth's atmosphere since the evolution of oxygenic photosynthesis, thereby exerting key influence on Earth's climate and the redox state of (sub)surface Earth. The canonical source and sink terms of the marine oxygen budget include photosynthesis, respiration, photorespiration, the Mehler reaction, and other smaller terms. However, recent advances in understanding cryptic oxygen cycling, namely the ubiquitous one-electron reduction of O2 to superoxide by microorganisms outside the cell, remains unexplored as a potential player in global oxygen dynamics. Here we show that dark extracellular superoxide production by marine microbes represents a previously unconsidered global oxygen flux and sink comparable in magnitude to other key terms. We estimate that extracellular superoxide production represents a gross oxygen sink comprising about a third of marine gross oxygen production, and a net oxygen sink amounting to 15 to 50% of that. We further demonstrate that this total marine dark extracellular superoxide flux is consistent with concentrations of superoxide in marine environments. These findings underscore prolific marine sources of reactive oxygen species and a complex and dynamic oxygen cycle in which oxygen consumption and corresponding carbon oxidation are not necessarily confined to cell membranes or exclusively related to respiration. This revised model of the marine oxygen cycle will ultimately allow for greater reconciliation among estimates of primary production and respiration and a greater mechanistic understanding of redox cycling in the ocean.

Carbon fixation by oxygenic phototrophs produces O2 in marine and terrestrial environments. On short timescales, the concentration of O2 in the atmosphere and dissolved in the global ocean is maintained by an equal and opposite series of oxygen-consuming reactions. The preeminent of these O2 loss terms is cellular respiration, although multiple additional loss processes are operative in oxygenic phototrophs (1–3). These include photorespiration, which occurs when the enzyme RuBisCO (ribulose-1,5-bisphosphate carboxylase/oxygenase) uses O2 as a substrate instead of CO2 in the reaction with ribulose 1,5-bisphosphate (RuBP) as part of the normal functioning of the Calvin–Benson–Bassham cycle. Loss of O2 through photorespiration may be as high as 30% of gross oxygen production (GOP) (4, 5). The Mehler reaction is another significant oxygen loss term in oxygenic phototrophs, occurring when O2 is reduced to superoxide (O2•–) by electrons evolved from reduced ferredoxin in photosystem I in the presence of light (3). The fraction of O2 evolved from photosystem II that is subsequently lost to reduction via the Mehler reaction is estimated at 10% of photosynthetically evolved O2 in higher plants (5). These estimates of photosynthetic oxygen loss are based on model organisms of varying environmental relevance, and a great deal of uncertainty remains concerning how oxygen reduction in the environment is distributed among these many sinks. In algae, for example, O2 photoreduction appears to be largely insensitive to dissolved inorganic carbon concentrations, even when the subject lacked a CO2 concentrating mechanism (2). Studies of Mehler-related oxygen reduction in some cyanobacteria have been shown to exceed 40% of GOP (6, 7). Altogether, these studies demonstrate that Mehler-related oxygen loss in marine cyanobacteria and algae likely represents a larger proportion of total nonrespiratory O2 reduction than is observed in higher plants.

The role of intracellular superoxide production as a significant sink of oxygen has been recognized since the 1950s for its place in the Mehler reaction (8); however, the role of extracellular superoxide production as an oxygen sink in both the presence and absence of light has not been assessed. Yet, dark (light-independent) particle-associated superoxide production accounts for a significant fraction of superoxide measured in natural waters within both the photic and aphotic zone (9, 10). Dark, extracellular superoxide production is in fact prolific among marine heterotrophic bacteria, cyanobacteria, and eukaryotes (9, 11–18). Production of extracellular superoxide proceeds via a one-electron transfer initiated by transmembrane, outer membrane-bound, or soluble extracellular enzymes thought to belong generally to NAD(P)H oxidoreductases (19), and more recently to heme peroxidases and glutathione reductases (9, 20, 21). At circumneutral pH, the superoxide anion (O2•−) dominates over the conjugate acid HO2• (pKa = 4.8) (22). This short-lived, reactive anion is highly impermeable to cell membranes (23). Thus, limited transmembrane diffusion and proportionally low levels of intracellular superoxide are insufficient to account for observed extracellular fluxes (9, 23, 24). Superoxide has a half-life on the order of a few minutes or less (9, 25), with its fate highly dependent on seawater chemistry; superoxide may be reoxidized to O2 or reduced to H2O2, other peroxides, and ultimately water by reactions with redox-active metals including copper, iron, and manganese, organic matter, or via (un)catalyzed dismutation (26, 27).

Extracellular superoxide plays a diverse suite of roles in cell physiology, including cell signaling, growth, defense, and redox homeostasis. In microbial eukaryotes, enzymatic extracellular superoxide production is involved in cell differentiation (19). In plants, superoxide serves as a pretranscription defense in wound repair (16). Recently, extracellular superoxide within marine bacteria was shown to be tightly regulated through the growth cycle with superoxide eliminated upon entering stationary phase (28). Addition of superoxide dismutase (SOD), an enzyme that eliminates superoxide, inhibited growth, suggesting that superoxide is involved in growth and proliferation as previously suggested for pathogenic strains of Escherichia coli and Chattonella marina (29–32). Further, recent evidence indicates that extracellular superoxide production by the diatom Thalassiosira oceanica may play a critical role in maintaining the internal redox conditions in photosynthesizing cells (21). This suite of beneficial physiological processes all result from or result in the reduction of molecular oxygen that is not otherwise considered in biogeochemical cycles of oxygen and related elements.

Our aim in this study is to leverage recent assessments of dark extracellular superoxide production rates by globally important groups of marine microbes to determine the resulting oxygen sink. Our approach consists of two parts: 1) constrain the gross oxygen reduction that results from dark, extracellular superoxide production and 2) estimate the fraction of gross superoxide produced that is ultimately reduced to water. Generally speaking, the most abundant group of organisms in the global ocean is heterotrophic bacteria. Alphaproteobacteria and Gammaproteobacteria comprise >70% of heterotrophic bacteria in the water column, and a subset of Alphaproteobacteria, Pelagibacterales (SAR11 clade), makes up nearly 25% of cells in the ocean (33, 34). Here we compile measured extracellular superoxide production rates of several marine Alphaproteobacteria and Gammaproteobacteria, including SAR11 clade organisms (Fig. 1). Although far outnumbered by marine heterotrophs in the global ocean, oxygenic phototrophs have been shown to produce both light-dependent and -independent extracellular superoxide up to several orders of magnitude higher than average marine heterotrophs on a per-cell basis, meaning a representative accounting of the superoxide flux from phototrophs is necessary to adequately constrain superoxide production (9, 13, 35). In this study we include superoxide production rates of the following organisms: Prochlorococcus and Synechococcus, the two most abundant photosynthetic organisms in the global ocean; Trichodesmium, a dominant N2-fixing cyanobacterium in oligotrophic waters; coccolithophores, the most abundant group of calcifying microalgae present throughout the global ocean; several species of diatoms, a diverse group of silicifying algae found in nutrient rich waters; and Phaeocystis, an alga predominately found in the Southern Ocean. In constructing this estimate of superoxide production, we assign measured dark superoxide production rates to organisms that fall within these groups of organisms and conservatively consider all others to be nonproducing.

Fig. 1.

Extracellular superoxide production rates. Cyanobacteria (green) include Synechococcus, Trichodesmium, and Prochlorococcus; eukaryotic algae (orange) include coccolithophores, diatoms, and Phaeocystis; and heterotrophic bacteria (blue) include Pelagibacterales (SAR11 clade), other Alphaproteobacteria, and Gammaproteobacteria (see for data, sources, and organism details and for diatom rates). Average marine cell oxygen consumption rate is shown in gray ().

Marine heterotrophs belonging to Alphaproteobacteria and Gammaproteobacteria, including two Pelagibacterales members, produce gross extracellular superoxide within a fairly narrow range of 0.1 to 3.7 amol⋅cell−1⋅h−1 (Fig. 1). Oxygenic phototrophs, with the exception of Prochlorococcus, far exceed heterotrophic cell-normalized superoxide production, with net production rates ranging from 4.3 to 13,400 amol⋅cell−1⋅h−1 (17). Prochlorococcus produces significantly less extracellular superoxide across four ecotypes, with average gross rates ranging from 0.007 to 0.091 amol⋅cell−1⋅h−1 (17).

Also shown in Fig. 1 is an estimate for the average marine cell O2 utilization rate assuming a balanced marine oxygen budget (see for calculation), or in other words the total amount of oxygen produced in the global ocean, divided by the number of cells. This estimate demonstrates that dark extracellular superoxide production of several groups of marine organisms exceeds the average oxygen utilization rate. Here we use the more general term “oxygen utilization” instead of “respiration” because other oxygen-consuming biological reactions (e.g., photorespiration and the Mehler reaction) are each responsible for >10% of the marine oxygen sink. To extrapolate these superoxide production values to the global ocean, we provide estimates of total cell numbers in the water column of each organismal group included in this study (Table 1). We assigned cells from each group the mean and standard error (SE) determined from our bootstrapping approach and used a Monte Carlo approach to determine the mean and probability distribution for the whole ocean superoxide flux. A slightly different approach was used for diatoms (). Using these total cell number estimates and modeled dark extracellular superoxide production rates for each organismal group, we calculate that gross dark extracellular superoxide production represents a flux of 3.9 (± 0.5) × 1015 mol O2 per year. For comparison, a central estimate for GOP in all marine environments derived from estimates of marine productivity is 1.09 × 1016 mol O2 y−1 () (36). Thus, gross light-independent extracellular superoxide production by microorganisms represents an O2 loss flux ∼36 (± 5)% of marine GOP. This gross superoxide production estimate is illustrative for demonstrating the size of this reductive flux; however, it is the net reduction of superoxide that ultimately determines the weight of this reductive flux on the global oxygen cycle. We provide constraints on the net reduction of extracellular superoxide below.

Table 1.

Estimate of global superoxide flux: Summary of cell number estimates, cell specific superoxide production rates, and contribution of each marine clade toward the marine superoxide flux

Group	Cell estimate*	Range superoxide production, amol⋅cell−1⋅h−1	Mean model value*, amol⋅cell−1⋅h−1	Superoxide production, mol⋅y−1
Synechococcus	7.0 × 1026	4.3–550	106	6.49 × 1014
Trichodesmium	4.6 × 1023	ND-1,500	465	1.87 × 1012
Prochlorococcus	2.9 × 1027	0.007–0.091	0.06	1.40 × 1012
Coccolithophores	2.6 × 1026	ND-5,300	342	7.78 × 1014
Diatoms	2.6 × 1024	75–13,000	47,560†	1.07 × 1015
Phaeocystis	3.9 × 1025	1,700–4,200	2,925	9.99 × 1014
Pelagibacterales (SAR11)	2.4 × 1028	0.11–0.23	0.15	3.20 × 1013
Alphaproteobacteria (without SAR11)	1.9 × 1028	0.1–3.1	0.9	1.51 × 1014
Gammaproteobacteria	2.7 × 1028	0.1–3.7	1.1	2.57 × 1014
Total				3.94 × 1015

ND = measurement reported below method detection limit.

See for sources, derivations of cell number estimates, and model value assignment.

Calculated using cell surface area normalized rates; see .

While we show that model estimates based on laboratory-based rates yield a dark superoxide flux that is a substantial and previously unrecognized part of the global oxygen budget, a claim that represents such a significant shift in the model of marine oxygen utilization requires some ground truthing with environmental data. In particular, we used a bootstrapping approach to estimate the mean and SE of all extracellular superoxide production rates available in the literature for each group discussed here (). The available data on extracellular superoxide production within axenic cultures do not contain the ideal richness for relying on this numerical approach alone. Thus, we tested our culture-based estimate by calculating expected marine superoxide concentrations based on our estimate of global superoxide production and compared these values to available marine superoxide concentration data. If our estimate for average marine superoxide concentration falls within observations, this provides an independent line of evidence that extracellular superoxide production comprises a significant global oxygen flux, with our reasoning as follows. The mean pseudo-first-order decay rate constant of superoxide in marine environments has been previously characterized at 0.0106 s−1, with a 1σ confidence interval from 0.0050 s−1 to 0.0226 s−1 () (35, 37–39). Using these observed decay rate constants we estimate steady-state superoxide concentrations from biological production to be 152 pM (1σ confidence interval: 71 to 322 pM) in the surface ocean and 0.6 pM (1σ confidence interval: 0.3 to 1.3 pM) below 200 m (). All available superoxide concentration measurements from the marine water column are shown in Fig. 2 along with the calculated expected concentration range based on our model of global superoxide production (gray bar). As indicated, the expected range of superoxide concentrations is largely consistent with measurements collected in the surface ocean, and, in fact, underestimates measured concentrations in all deep ocean measurements. Although not shown in Fig. 2, superoxide concentrations in some coastal systems have been reported as high as 120 nM, four orders of magnitude higher than our estimate, underscoring the conservative nature of our estimate (40).

Fig. 2.

Measured and expected marine superoxide concentration. Compiled marine superoxide measurements from Rose et al. (35) (green diamond), Hansard et al. (38) (yellow triangle), Rusak et al. (39) (red square), and Roe et al. (37) (blue circle). The gray bars indicate the 68% confidence interval for the expected superoxide concentration based on our total superoxide flux estimate. The 68% confidence intervals for the expected superoxide concentrations in the surface (<200 m) and deep (>200 m) ocean are 71 to 322 pM and 0.3 to 1.3 pM, respectively.

Additional Sources of Superoxide

This accounting of the marine superoxide flux only considers extracellular dark superoxide production by dominant organisms with known production rates. In fact, the total O2 flux via all superoxide production pathways in marine environments undoubtedly exceeds our estimates here by a sizable margin and would lead to an even greater contribution of superoxide production on O2 loss. In particular, light-dependent (a)biotic superoxide production is not included in our present estimate. Within sunlit waters, there is significant abiotic photochemical production of superoxide, and extracellular superoxide production rates by marine phototrophs are significantly higher in the light (13, 41, 42). Indeed, the extracellular superoxide production exhibited by multiple species of diatoms more than doubled in the presence of light (42). The same behavior was observed for the coccolithophore Emiliania Huxleyi (18) and certain Trichodesmium ecotypes (13), where light-dependent increases in extracellular superoxide could not be accounted for by abiotic factors. Since phototrophs as a group are responsible for most of the dark extracellular superoxide production, even modest modulation in extracellular superoxide production in the light could produce a substantial increase in our estimate of the gross superoxide flux and the net oxygen sink that results, which we discuss in the next section.

Possibly the best-characterized abiotic source of superoxide in the oceans is the photochemical excitation of chromophoric dissolved organic matter (CDOM) and subsequent reduction of O2 to superoxide in the surface ocean (43). One model using a remote sensing approach to estimate photochemical reactive oxygen species (ROS) production in the surface ocean estimated that midday superoxide concentrations resulting from photochemical production and a range of superoxide sinks ranged from near 10 pM at high latitudes to near 200 pM at low latitudes (41). These results demonstrate that photochemical production of superoxide can exceed dark biological production in sunlit surface waters. Notably, contribution of this photochemical superoxide will vary temporally and decrease with depth upon the attenuation of photoactive wavelengths (290 to 490 nm), with local variations in productivity and the compositional nature of surface ocean CDOM also impacting its reactivity.

The Fate of Superoxide

Both laboratory- and field-based measurements converge on the similar conclusion that dark extracellular superoxide flux is a substantial component of oxygen turnover in the global ocean. To further place this process into the context of the global O2 budget, it is important to distinguish gross dark superoxide production from the net loss of dissolved O2. Superoxide in aqueous systems may decay by oxidation back to O2, with no net effect on marine dissolved oxygen, or via reduction, which may lead to a net loss of oxygen. Superoxide decay is considered the primary source of hydrogen peroxide (H2O2), and thus much of what we know and assume about the fate of superoxide is inferred from studies of hydrogen peroxide concentration and rate measurements. Nevertheless, other secondary and direct sources of hydrogen peroxide have also been identified within microbial cultures (42, 44, 45) and natural waters (10). For the purposes of this estimate, we accept the general premise that superoxide decay is the primary source of hydrogen peroxide and ignore other sources, while recognizing that the foundation of this premise needs further evaluation. It is also often assumed that the primary reduction or disproportionation product of superoxide production is H2O2; however, superoxide decay may also occur through bond-forming redox reactions with dissolved organic carbon, metals, or through unknown sinks (46–50). Such reactions will not produce an H2O2 intermediate. In fact, it has been shown that hydrogen peroxide formation can only account for 45% of net oxygen loss in photochemical oxidation of marine waters (50). While photochemical superoxide production is not a perfect analogy for dark biological superoxide production, the large body of work investigating the fate of photochemically derived superoxide offers the most transferable insight into the likely fate of biologically derived superoxide in seawater. What follows is our attempt to constrain the net reduction of marine dissolved oxygen from the estimate of gross extracellular superoxide production using a combination of measurements from photochemical and biogeochemical literature.

Ultimately, the fate of superoxide and hydrogen peroxide is highly dependent on the availability of dissolved organic matter (and its reactivity), the abundance of redox-active metals (and their redox states), and the expression of enzymes that eliminate ROS such as SOD, catalase, and peroxidases (49, 51, 52). Superoxide can be both oxidized (Eq. ) and reduced (Eq. ), leading to the net consumption of 0 or 1 mol O2 per mole superoxide, respectively (Fig. 3). Superoxide can also undergo uncatalyzed dismutation, a process that results in 50% oxidation to O2 and 50% reduction to H2O2 (Eq. ). Hydrogen peroxide is more stable than superoxide in natural waters, with pseudo-first-order decay rate constants approximately three orders of magnitude lower than that of superoxide. Consequently, H2O2 has a lifetime of ∼1 to 2 d, and typical concentrations are ∼103 higher for H2O2 than for superoxide [from <1 nM in the deep ocean to ∼100 nM in sunlight surface water (53, 54)]. Typical production and decay rates range from 0.8 to 2.4 nM⋅h−1 for dark seawater (53) and 0.9 to 8.3 nM⋅h−1 in sunlit seawater (55). As for the fate of H2O2, previous work has shown that light-independent, biological processes are primarily responsible for its degradation in marine systems, with 65 to 80% of H2O2 degradation resulting from catalase activity (Eq. ) and the remainder resulting from peroxidase activity (Eq. ) (53):Assuming this range of catalase and peroxidase activity holds throughout the water column, between 60 and 67.5% of H2O2 is ultimately reduced to water. Thus, the theoretical net loss of oxygen through the superoxide production pathway can range from 0 (e.g., all superoxide is oxidized back to O2; Eq. ) to 67.5% (e.g., superoxide reduction to H2O2; Eq. ), followed by catalase- and peroxidase-mediated degradation of H2O2 (Eqs. and ) (Fig. 3).

Fig. 3.

Decay pathways of superoxide in biogeochemical systems. Summary of superoxide decay pathways in marine environments. Superoxide, once produced, may be oxidized to O2 and/or reduced to H2O2 (or other peroxides). Peroxides may be oxidized to O2 and/or reduced to H2O. The net dissolved oxygen sink that results from extracellular superoxide production is the product of gross production and that fraction of O2 that is ultimately reduced to water.

There are two ways we can use previous data to estimate the net sink of oxygen that results from the dark biological superoxide flux in the ocean. The first, and more conservative, approach is to multiply the global superoxide production flux by the ratio of hydrogen peroxide production to superoxide production observed in natural waters, PH2O2/PO2−, followed by the oxidation–reduction ratio observed for hydrogen peroxide. This yields the total O2 reduced to water via an H2O2 intermediate and accounts for dismutation (Eq. ), univalent oxidation (Eq. ), and univalent reduction (Eq. ). This method will produce a more conservative estimate of net oxygen loss because it implicitly assumes that all superoxide not reduced to hydrogen peroxide is reoxidized to O2, when in reality a large fraction of the superoxide sink may yield products other than hydrogen peroxide (e.g., organic peroxides or other ROS) (49, 50). The second approach uses the same PH2O2/PO2− ratio to determine the hydrogen peroxide flux that results from the superoxide flux, multiplied by the observed ratio of net oxygen loss to hydrogen peroxide formation in a marine photochemical system where superoxide is implicated as the primary oxidant. This approach may slightly overestimate the net oxygen sink because it ignores the possibility that some oxygen reduction may occur through multielectron transfers (50).

For both of these approaches, we need a reasonable estimate of the production ratio of hydrogen peroxide to that of superoxide (PH2O2/PO2−). In experiments investigating production of superoxide and hydrogen peroxide during DOM irradiation, PH2O2/PO2− ranged from 0.10 to 0.67, with an average value of 0.24 (55). Another similar study on waters collected from the transition between terrestrial and marine environment found that PH2O2/PO2− ranged from ∼0.5 in riverine waters to ∼0.30 in Gulf Stream waters (56). Slightly lower values (0.08 to 0.17) have been observed in photochemical systems with terrestrial organic matter (57), possibly due to the differing nature of terrestrial DOM (aromatic vs. aliphatic). In a brackish and a freshwater pond, PH2O2/PO2− sometimes exceeded the stoichiometry of 0.5 expected for dismutation, suggesting there may be other sources of hydrogen peroxide (10). We use the average value presented by Powers et al. (55) of 0.24 as the most relevant to the marine environment and the best choice at this time for a global marine estimate. Using our first approach and PH2O2/PO2 of 0.24, we estimate a lower bound for the net oxygen reduction resulting from superoxide production at 14.4% of the marine superoxide flux (or 5% of the marine oxygen budget). A significant fraction of this superoxide (52% of the marine superoxide flux) is unaccounted for because it does not produce hydrogen peroxide, meaning the fate of as much as 19% of the marine oxygen budget that cycles through superoxide remains unknown.

Using the second approach, we assume hydrogen peroxide formation and net oxygen loss that result from superoxide formation occur with a fixed ratio in seawater [0.45:1 H2O2:−O2 (50)]. We find that superoxide production will result in a net oxygen loss of 53% of the global superoxide flux (or 19% of the marine oxygen budget).

These two estimation methods suggest dark, biological production of extracellular superoxide is a sink between 5 and 19% in the marine oxygen budget, indicating that the oxygen sink from dark extracellular superoxide production is similar in magnitude to the Mehler reaction and photorespiration. Notably, total superoxide production in photochemical systems has been shown to correlate with CO2 production from DOM (58), suggesting that a significant fraction of the superoxide sink is ultimately through reduction tied to organic carbon oxidation, favoring the higher end of this estimate. Shipboard incubations show that marine superoxide decay is primarily correlated with dissolved organic matter and dissolved manganese, both of which allow for a significant reductive sink of superoxide (26, 51) and subsequent higher O2 loss. Oxygen loss in marine environments has previously been attributed primarily to a combination of respiration, photorespiration, and the Mehler reaction in the surface ocean and respiration alone in the deep ocean. We propose that dark extracellular superoxide production and its net oxygen sink has likely been overlooked and unintentionally incorporated into other sinks because separation of this secondary physiological process presents a methodological challenge. The sinks of marine oxygen resulting from respiration, photorespiration, and the Mehler reaction should therefore be revised downward to accommodate for the reductive sink from extracellular superoxide production.

Further Implications

We anticipate that the identification of an appreciable cryptic sink of oxygen, and consequently organic carbon, may help reconcile spatiotemporal and methodological discrepancies among measurements of marine primary productivity arising as variable contributions of superoxide production to net O2 loss. Adding another layer of complexity, the resulting H2O2 produced in excess during times of increased metabolic activity (e.g., photosynthetic organisms under full light) may store oxidizing equivalents that persist in the environment for hours to days. These processes may lead to heterogeneous O:C reaction stoichiometry as well as possible complexity in corresponding stable isotope dynamics. As measurements of primary productivity collected using multiple methods concurrently can produce primary productivity estimates that vary more than an order of magnitude (59), we propose that cryptic ROS cycling and ROS-related oxygen loss may explain some of these discrepancies (e.g., 14C vs. triple oxygen isotopes). Studies of respiration, respiration stoichiometries (−O2:C), and respiration isotope effects will also be impacted if organism-level extracellular superoxide production rate is an appreciable fraction of respiration rate. The −O2:C ratio of oxygen utilization in the water column has been observed to vary by up to 30% across different locations and ocean depths (60–62). Production of extracellular superoxide (coupled with the oxidation of intracellular reducing equivalents) may lead to differential redox evolution of marine carbon reservoirs that are spatially or functionally separated (e.g., surface vs. deep, particulate vs. dissolved). While superoxide production will not affect the whole ocean −O2:C, it may drastically influence the amount and rate of organic carbon that is directly remineralized to CO2 vs. that which proceeds through partially oxidized dissolved organic compounds. The exact utility of extracellular superoxide for cells and the magnitude of influence superoxide has on the global carbon budget (today and throughout Earth history) both remain important and open questions. Nevertheless, this newly identified gross flux of superoxide in the global ocean underscores the critical role ROS play in the global cycling of O2, carbon, and redox active elements essential to life.

Methods

The dark superoxide production rates compiled in this study, with the exception of coccolithophores, were measured from cells grown to midexponential phase under ideal growth conditions using a flow-injection chemiluminescent approach (9). The study investigating the cell-specific superoxide production rate of coccolithophores measured their production rate throughout their growth curve, which we converted to a time-weighted average using trapezoidal integration of the cell superoxide production rate as a function of time (18). Cell-normalized superoxide production rates presented in the scientific literature are either presented as net or gross superoxide production rates, the latter requiring an exogenous spike of superoxide to determine the proportion of extracellular ROS that is enzymatically degraded by the organisms. Gross superoxide production rates were used here when available; otherwise, net superoxide production rates were used. All compiled rates are presented in Fig. 1. Bootstrap resampling of extracellular superoxide production rates was conducted to estimate the rate distribution of each organism group (), with the exception of diatoms. Diatom production rate was normalized to cell surface area and diatom surface area was determined from a database containing >90,000 georeferenced diatom observations (). Cell abundances for phototroph and heterotroph groups included in this study were compiled or calculated from available literature data and are discussed in greater detail in . We note that when net primary productivity (NPP) estimates were used to estimate cell abundances, each group’s fractional contribution to NPP (as opposed to moles of carbon fixed) was given preference. Fraction contribution to NPP was preferred because it allows for direct comparison between GOP and total superoxide production in a way that minimizes the influence of differences in productivity estimates between studies. Total superoxide flux was estimated using a Monte Carlo approach incorporating superoxide production rate distribution for each organism group. Parameter choices and results are presented in . Superoxide concentrations and uncertainties were estimated using the volume-normalized superoxide flux and a compilation of superoxide decay rates from the literature (). Further experimental methods are described in . All data compiled and generated in this study are provided within the main text and .