Widespread Production of Extracellular Superoxide by Heterotrophic Bacteria

See allHide authors and affiliations

Science  07 Jun 2013:
Vol. 340, Issue 6137, pp. 1223-1226
DOI: 10.1126/science.1237331

Sending Out an ROS

The global imprint of biological activity in aquatic environments is often considered a consequence of enzyme-mediated redox reactions that support metabolic activity, such as reducing oxygen during respiration. But some organisms also release redox-active reactive oxygen species (ROS) into the environment—to acquire trace metals or to prevent viral infections—which can influence global processes like nutrient availability and contaminant transport. Photosynthetic organisms like phytoplankton are thought to be the primary biological source of ROS in freshwater and marine environments. However, Diaz et al. (p. 1223, published online 2 May; see the Perspective by Shaked and Rose) now show that a broad range of ecologically and phylogenetically diverse heterotrophic bacteria also produce large quantities of superoxide. Production rates vary widely across 30 common bacterial isolates but in some cases were greater than production rates of phytoplankton. Because these bacteria do not require light to grow, they may be the dominant source of ROS in dark environments like the deep ocean, terrestrial soils, or lake sediments.


Superoxide and other reactive oxygen species (ROS) originate from several natural sources and profoundly influence numerous elemental cycles, including carbon and trace metals. In the deep ocean, the permanent absence of light precludes currently known ROS sources, yet ROS production mysteriously occurs. Here, we show that taxonomically and ecologically diverse heterotrophic bacteria from aquatic and terrestrial environments are a vast, unrecognized, and light-independent source of superoxide, and perhaps other ROS derived from superoxide. Superoxide production by a model bacterium within the ubiquitous Roseobacter clade involves an extracellular oxidoreductase that is stimulated by the reduced form of nicotinamide adenine dinucleotide (NADH), suggesting a surprising homology with eukaryotic organisms. The consequences of ROS cycling in immense aphotic zones representing key sites of nutrient regeneration and carbon export must now be considered, including potential control of carbon remineralization and metal bioavailability.

Heterotrophic bacteria are ubiquitous and abundant components of natural ecosystems. These metabolically versatile organisms alter the global environment by mediating the redox transformation of many elements, including carbon, iron, and mercury. Biologically produced reactive oxygen species (ROS) may play an unrecognized role in these pathways because of the susceptibility of these elements to redox reactions with ROS (14), but the ability of common heterotrophic bacteria to produce extracellular ROS has remained largely unexplored (2). Nevertheless, the recent discovery of extracellular superoxide (O2) production by a manganese-oxidizing marine bacterium belonging to the prolific Roseobacter clade (5) suggests that this capability may be present in environmentally relevant heterotrophs.

ROS production in natural waters (fig. S1) has long been linked to abiotic photooxidation of organic compounds (6). Yet, biological ROS production has been historically recognized (7), and recent field evidence indicates that biological production is likely the dominant ROS source in many marine and freshwater systems (810). Although all aerobic organisms produce intracellular ROS as a metabolic byproduct (11), ROS are generally toxic to living cells. Intracellular concentrations are therefore kept very low by enzymes that scavenge ROS, such as superoxide dismutase and catalase. This strict regulation of intracellular ROS and the inability of charged ROS to pass the lipid bilayer (2) point to biologically directed processes rather than adventitious cell rupture or leakage as the primary source of biogenic ROS in the environment. In fact, a number of cultivated eukaryotic phytoplankton and cyanobacteria produce extracellular superoxide under laboratory conditions (8, 1214) for reasons equivocally related to metal nutrient acquisition (13, 15) and virulence (12). These photosynthetic organisms have been inferred as the source of total biological superoxide production in the surface ocean based only on indirect evidence, such as agreement between culture- and field-based fluxes. The ultimate source of biogenic superoxide in the environment therefore remains unclear.

We analyzed extracellular superoxide production by a broad range of ecologically and phylogenetically diverse heterotrophic bacteria (16). Using a high-sensitivity flow-through chemiluminescence approach (fig. S2) (13, 17), we detected superoxide production by 27 of 30 environmentally common isolates, including members of three Proteobacteria clades, Bacteroidetes, and two Gram-positive genera, each representing Actinobacteria and Firmicutes. The bacterial strains selected were originally isolated from a wide variety of aquatic and terrestrial environments, including lakes, soils, hydrothermal vents, pelagic marine sediments, estuaries, and the surface and deep (>1 km) ocean. The organisms ranged from psychrophiles to thermophiles and exhibited benthic, planktonic, and symbiotic lifestyles. The three organisms that tested negative for superoxide production were Shewanella oneidensis MR-1 and two betaproteobacteria, Thiobacillus thioparus (ATCC 8158) and Leptothrix discophora SS-1. As a group, these organisms exhibited no distinctively common traits to distinguish them from those that did produce extracellular superoxide. Superoxide production is thus widespread among heterotrophic bacteria, exhibiting no clear patterns with respect to ecology or phylogeny (Fig. 1).

Fig. 1 Superoxide production across a broad phylogenetic and ecological diversity of bacteria.

Relative superoxide production within this 16S rRNA gene tree is indicated in the outer ring and is scaled to the highest observed rate (Ruegeria sp. TM1040; black); white represents values below detection. The middle ring identifies psychrophiles (yellow), mesophiles (orange), and thermophiles (red). The inner ring represents freshwater (light blue), estuarine (medium blue), and marine (dark blue) habitats. Font color indicates whether the original isolate was benthic (black), planktonic (red), or isolated from soil (blue). Organisms isolated from surface environments (<1 km) are indicated by normal font weight, whereas isolates obtained from greater than 1 km appear in bold font. Asterisks indicate that the bacterium was originally isolated from another organism or biological consortium. Leaf colors indicate phylogenetic group, including two Gram-positive isolates, one each from Actinobacteria (blue) and Firmicutes (orange), as well as Gram-negative bacteria from Alphaproteobacteria (purple), Betaproteobacteria (green), Gammaproteobacteria (yellow), and Bacteroidetes (pink).

Rates of superoxide production normalized to the proportion of metabolically active cells varied by a few orders of magnitude, between 0.02 ± 0.02 amol cell−1 hour−1 (mean ± SEM) and 19.4 ± 5.2 amol cell−1 hour−1 (Fig. 2A and table S2). Experiments involving standard additions of superoxide to these live cultures also revealed concomitant superoxide degradation. For example, the recovery of standard superoxide additions ranged from 1 to 100%, with an average of 52 ± 3% (table S2). A clear relationship between superoxide production and standard recovery was not observed (fig. S3), indicating that enhanced production is not simply a consequence of a depressed ability to degrade superoxide.

Fig. 2 Gross rates of extracellular superoxide production by heterotrophic bacteria.

(A)When normalized to abundance of metabolically active cells, rates of superoxide production by heterotrophic bacteria representing Firmicutes (orange), Actinobacteria (blue), Bacteroidetes (pink), Gammaproteobacteria (yellow), and Alphaproteobacteria (purple) are lower than previous estimates for phytoplankton (green). (B) However, when these active-cell rates are normalized to cell surface area, superoxide production by heterotrophs is comparable with, and in several cases greater than, previous estimates for phytoplankton. Measurements for heterotrophs represent the mean ± SEM for biological replicates analyzed during stationary phase (indicated by asterisks) or mid-exponential growth. Phytoplankton data are from (8) and are based on the assumption that superoxide recovery is 100%. Comparing net superoxide production by heterotrophic bacteria (table S2) with these phytoplankton rates yields similar results.

Although cell-normalized rates of superoxide production by heterotrophic bacteria are lower than those measured previously for marine phytoplankton (Fig. 2A), rates normalized to cell surface area are comparable with, and in several cases greater than, those of phytoplankton. For instance, the most prolific heterotrophic superoxide producer, Ruegeria sp. TM1040, exhibited rates over two times higher than that of the cyanobacterium Synechococcus sp. WH7803 (Fig. 2B and table S3). This heterotrophic bacterium belongs to the highly common Roseobacter clade, which can account for up to 30% of marine bacterioplankton communities (18).

If we assume that bacteria producing superoxide at the rate of Ruegeria sp. TM1040 represent 30% of the total bacterioplankton population (18) consisting of 106 cells mL−1, the extrapolated rate (2 × 103 fmol L−1 s−1) can account for 90 to >100% of biological superoxide production reported previously in the Gulf of Alaska (9) and >100% of biological superoxide production within the photic zone of the Costa Rica Dome upwelling region (table S4) (8). Similar results were obtained from an extrapolation based on the median extracellular superoxide production rate observed for heterotrophic bacteria (1.2 amol cell−1 hour−1) (table S4). Although extrapolated rates of superoxide production by Synechococcus are similar to the potential contribution of heterotrophs in the Costa Rica Dome (8), biological superoxide production in this region does not correlate with Synechococcus abundance (8), further suggesting a role for heterotrophic bacteria.

Superoxide signals were stable (~2% coefficient of variation) over time periods of 1 to 40 min, timeframes that are comparable with or much longer than the observed half-life of superoxide (~2 min). This stability is consistent with steady-state production and would not be expected in the case of sudden superoxide release via cell lysis. In fact, for those organisms tested during mid-log and stationary phases (with the exception of Pseudomonas putida GB-1), superoxide production was comparable or substantially lower during stationary phase, when cell lysis would be greatest (Fig. 2). Moreover, calculations based on the superoxide content of a typical bacterial cell revealed that release of intracellular superoxide is negligible compared with any superoxide production rate observed (table S5).

The biochemical processes responsible for superoxide production may be similar between heterotrophic bacteria and eukaryotic organisms. Enzymes within the broad class of nicotinamide adenine dinucleotide (NAD+) and nicotinamide adenine dinucleotide phosphate (NADP+) [NAD(P)H] oxidoreductases, including NADPH (the reduced form of NADP+) oxidase and peroxidase, are thought to be involved in extracellular superoxide production by raphidophycean algae (19), diatoms (13), fungi (20), and plants (21). These conclusions are based in part on the observation that exogenous NAD(P)H stimulates extracellular superoxide production among model eukaryotic organisms, whereas diphenyleneiodonium (DPI), an inhibitor of NAD(P)H oxidoreductases and other NAD(P)H oxidizing enzymes, strongly attenuates it (13, 19, 20). For example, in the toxic eukaryotic raphidophycean flagellate Chattonella marina, extracellular superoxide production occurs through a loosely surface-associated putative NAD(P)H oxidase (19) that readily detaches from the cell (19). Similarly, NADH (the reduced form of NAD+) oxidases have been implicated in superoxide production by bacterial pathogens and Escherichia coli (2224). Consistent with this previous work, addition of exogenous NAD(P)H to Roseobacter sp. AzwK-3b significantly increased extracellular superoxide production via an enzymatic response (Fig. 3, fig. S4, and table S6). In fact, similar to the case of C. marina, we detected NADH-stimulated superoxide production by protein (or proteins) collected from the cell-free filtrate of Roseobacter sp. AzwK-3b (Fig. 3). We thus infer that exogenous NADH is able to boost extracellular superoxide production by stimulating this extracellular protein without the need to enter the cell, as in the case of C. marina (19). Under natural physiological conditions, however, endogenous NADH is likely to be the primary electron donor. The physiological architecture responsible for electron flow from intracellular NADH to the surface-associated NADH oxidoreductase in Roseobacter sp. AzwK-3b and C. marina is not yet known but may involve quinone-like electron carriers cycled within the membrane, as observed for E. coli (23).

Fig. 3 Extracellular superoxide production by Roseobacter sp. AzwK-3b.

(A and B) Addition of NADH increased superoxide production by Roseobacter sp. AzwK-3b in (A) whole cells and (B) (left) cell-free filtrates and (right) protein extracts, as measured with two separate methods. NAD+ had no significant effect (A), whereas DPI, a broad inhibitor of NAD(P)H oxidoreductases and other NAD(P)H-oxidizing enzymes, decreased superoxide production (A). Similarly, NADH-stimulated superoxide production in filtrates and protein extracts substantially decreased when protein activity was reduced (proteinase-K) or eliminated (boiling before incubation with NADH) (B). Asterisks indicate treatments that significantly differ from the control condition (A) or the NADH condition (B) (P < 0.05) (table S6). Measurements are means ± SEM. (C) Proteins from Roseobacter sp. AzwK-3b were visible in whole-cell extracts (Coomassie gel, lane b) and also in cell-free filtrates (Coomassie gel, lane d). Reaction of the protein gel with the superoxide probe nitroblue tetrazolium (NBT) under control conditions revealed superoxide production at a single location in the separated whole-cell protein extract (Control gel, lane b). Addition of NADH stimulated superoxide production by this and other soluble whole-cell proteins (NADH gel, lane b) and also stimulated a single protein band present in the cell-free filtrate (NADH gel, lane d). Addition of DPI quenched NADH-stimulated superoxide production in whole-cell protein extracts (NADH+DPI gel, lane b) and completely inhibited it in the filtrate protein band (NADH+DPI gel, lane d). Protein ladders (lanes a) were prestained; appearance of the ladders therefore does not reflect superoxide production. Lane description for all gels (labeled in Coomassie gel) are (a) protein ladder, (b) soluble whole cell proteins, (d) cell-free filtrate proteins, (f) boiled (denatured) whole-cell soluble proteins, and (c and e) loading buffer only.

Also consistent with previous results from the analysis of eukaryotic microorganisms (13, 19, 20), DPI significantly decreased enzymatic production of extracellular superoxide by Roseobacter sp. AzwK-3b and eliminated production by the superoxide-generating protein (or proteins) within the cell-free filtrate (Fig. 3, fig. S5, and table S6). Although NADPH oxidases have not been found in prokaryotes (20), a number enzymes within the broad class of NADH oxidoreductases are encoded by the genome of Roseobacter sp. AzwK-3b. These findings reveal a widespread ability for extracellular superoxide production within prokaryotic and eukaryotic organisms, including bacteria, phytoplankton, fungi, and plants (13, 1924), with the potential for a surprising homology in the mechanisms of formation between these disparate groups.

Given the magnitude of fluxes observed in culture (Fig. 2 and tables S2 and S3) and comparison with field-based rates (table S4), superoxide production by heterotrophic bacteria must be considered in areas conventionally regarded as important sites of ROS production by phytoplankton and abiotic photooxidation pathways, such as the surface ocean. Furthermore, superoxide production among cosmopolitan heterotrophic bacteria, whose life cycles and metabolisms do not directly depend on light, may explain ROS production in aphotic ocean regions (8, 25, 26) and provide a source of ROS to other aphotic environments, such as pelagic sediments and terrestrial soils. Because the aphotic ocean represents the largest mass of water on Earth and the largest aqueous habitat for life (27), bacterially mediated ROS cycling will undoubtedly affect global biogeochemistry. For example, given the ability of ROS to enhance carbon remineralization directly (1, 28) or indirectly (5), variations in the efficiency of the biological pump (29) may be partly attributable to wide variations in ROS production by heterotrophic bacteria. ROS cycling by heterotrophic bacteria may furthermore influence numerous metal cycles, including mercury (Hg), a bioaccumulating toxin (30). Dark redox reactions in aphotic waters, likely of biological origin, control the content of Hg in the ocean by regulating the balance between Hg(II) and Hg(0) (the latter of which is volatile and outgasses to the atmosphere) (3, 4). Because inorganic mercury cycling is canonically shaped by reactions involving ROS (3, 4), the production of ROS by abundant and ubiquitous communities of heterotrophic bacteria in the deep ocean may drive this enigmatic dark cycling of Hg and, ultimately, the content of Hg in the ocean.

Supplementary Materials

Materials and Methods

Figs. S1 to S6

Tables S1 to S8

References (3151)

References and Notes

  1. Materials and methods are available as supplementary materials on Science Online.
  2. Acknowledgments: This study was supported by the National Science Foundation under grants OCE-1129594, OCE-1131734, EAR-1024817, and EAR-1025077 and by the Radcliffe Institute for Advanced Study at Harvard University and the Ford Foundation Fellowship Program. We thank E. D. Ingall for manuscript comments, D. W. King for technical assistance, and the thoughtful comments of four anonymous reviewers. We also thank the following individuals for contributing bacterial isolates for analysis in this study: H. Ehrlich, D. Emerson, S. Fendorf, Y. Masue-Slowey, T. Mincer, M. A. Moran, B. Tebo, R. Belas, and A. Templeton. The GenBank accession numbers for 16S ribosomal RNA (rRNA) gene sequences generated in this study are as follows: Bacillus sp. AzsLept-1c, JX515653; Tencibaculum sp. UAzPsLept-5, JX515654; Marinobacter sp. AzsJAc-4c, JX515655; Marinobacter sp. AzsJAc-4, JX515656; Halomonas sp. VMMm1-3c, JX515657; Microbulbifer sp. HLm1-1, JX515658; Paracoccus sp. HIJAc-3c, JX515659; Alteromonas sp. BIII82, JX524853; Alteromonas sp. BIII45, JX524854; Pseudoalteromonas sp. SSW22, JX524855; Marinobacter sp. LCB-12, JX524856; Halomonas sp. LOB-22, JX524857; and Citreicella sp. LOB-21, JX524858.
View Abstract

Navigate This Article