In apparent accord with HBH predictions, hypoxia reduced TOU in mesocosms even though oxygen was supplied in excess of conventionally defined respiratory demands. The HBH postulates that some classes of chemical compounds will persist in oxygen minimum zones because of the reduced activity of catabolic oxygenase enzymes. In the simplest interpretation of this idea, restoration of a fully oxic state after a period of hypoxia would be predicted to cause rates of oxygen utilization to increase as the activity of oxygenase enzymes accelerated and oxygen-dependent DOM (ODDOM) was more actively catabolized4. TOU data showed that oxygen utilization was lower in the hypoxic treatment but accelerated when it was transitioned to a fully oxic state on day 44, a further indication that hypoxia limited metabolism.
Analyses of community diversity further showed that a few taxa were responsive to oxygen, in some cases strongly, whereas many taxa were unaffected, as expected if their metabolism required respiration but was rate-independent of oxygenase enzymes. Among the microorganisms strongly responding to the difference in oxygen concentrations were lineages of probable annamox bacteria (OM190), chemoautotrophs that catalyze nitrogen loss in OMZs and are known to be inhibited by fully oxic environmental conditions28. Other microbes that were sensitive to the differences in oxygen concentrations were unexpected: SAR11 and marine actinobacteria, common marine heterotrophs, were inhibited by hypoxia, but a collection of bacteria associated with chemolithotrophy, genome minimization and parasitism were more abundant in the hypoxic treatment prior to reoxygenation.
To our surprise, members of the SAR86 clade, a microbial group that has previously been detected in hypoxic seawater, but has not been generally regarded as having adaptions to low oxygen, was abundant in the hypoxic treatment on day 44, whereas members of SAR11 bacteria (Pelagibacterales) were abundant in the oxic treatment14,29,30. The SAR86 ASVs we detected were in the D2472 subclade of SAR86, as described in the Genome Taxonomy Database. SAR86 has been reported from marine oxygen minimum zones in the Tropical Mexican Pacific, and while not significantly abundant within the OMZ core, SAR86 was found in highest abundance in the euphotic zone to the base of the oxycline31. Observations from a seasonally hypoxic estuary have also reported SAR86 being widespread across both high and low DO samples but with notable increase in abundance as DO decreases 14. This suggests these SAR86 ASVs may be specialized members of a ubiquitous group that have undergone niche differentiation. Our findings raise intriguing questions about the ecological impact of SAR86 subclade D2472 in the context of oxygen-depleted marine ecosystems, further emphasizing the incentives for investigations that resolve their functional significance.
To understand the relationship between TOU and TOC we plotted apparent respiratory quotients (ARQ) over the time course of the experiment. ARQ is the ratio of carbon dioxide produced to oxygen consumed (∆TOC:∆O2) by heterotrophic cells, with the assumption that reduced nitrogen, and to a lesser extent sulfur, consumed by chemolithotrophs simultaneously contribute to oxygen consumption32–34. We observed ARQ values in the range of 0.4–0.5 in the first week of the experiment, with higher values consistently in the hypoxic treatment (Fig. 2C). These values are in the range of common ARQ values for natural seawater33. By the third week of the experiment, ARQ had dropped to very low values (< 0.3) rarely observed in natural systems and unlikely to be explained by carbon oxidation to CO2.
A broad range of ARQ values have been reported reflecting the diverse states of dissolved organic molecules, which microbial activity can change significantly 33. Even though fixed values of ARQ are often used, ARQs derived from community respiration are variable, reflecting the changing state of DOM as it is altered by microbial activity. Very low ARQ values indicate highly reduced compounds, such as lipids, that demand a considerable amount of oxygen to fully metabolize to carbon dioxide. Models that allow ARQ to change over time as organic matter is degraded are realistic from the perspective of microbial physiology and biochemistry. Cells often show a preferred order for using molecules, which is canonized in prevalent conceptual models of DOM that empathize a spectrum from the most labile to the most recalcitrant forms. In any case, the very low ARQ values we observed late in the experiment were a clear indication that factors other than TOC oxidation to CO2 were contributing to oxygen uptake after the first 22 days of incubation.
To explain the observation of elevated oxygen uptake relative to TOC decline, particularly after TOC was drawn to relatively constant concentrations, we evaluated the hypothesis that oxidation of reduced nitrogen and sulfur compounds by chemolithoautotrophic microorganisms could consume oxygen in excess of the values calculated from bulk ocean mean stoichiometry. Chemolithoautotrophs use inorganic compounds as electron donors to drive their energy needs and can fix CO2 rather than consume organic carbon during respiration. Supporting this hypothesis, ASV's associated with the family Nitrincolaceae, which degrade nitrogen-rich organic compounds, were always detected in the mesocosms and ammonia-oxidizing archaea (Nitrosopumilaceae) appeared in large numbers between day 44 and day 100 (Fig. 3B). To model the potential for oxygen uptake by ammonium-fueled chemolithotrophy we assumed that all ammonium was oxidized completely to nitrate over the duration of the experiment (Figure S4). We estimate 202.5 µM of oxygen would be required to oxidize 81.2 µM of organic nitrogen (N), estimated from Redfield ratios and [TOC], from ammonium (NH4+) to nitrate (NO3−) in the hypoxic treatment of the experiment. This amount of chemolithotrophic ammonium oxidation, if delayed as indicated by the late rise of N-oxidizing bacteria and archaea in the experiment, could contribute significantly to explaining the low ARQ values we observed late in the experiment. As illustrated in Figure S4, TOU eventually exceeded predictions based on the lowest realistic ARQs that have been reported, and delayed N oxidation, indicating that factors in addition to a temporal offset between TOC and N oxidation might be contributing to the apparent excess of oxygen uptake we observed.
In estimating N chemolithotropy, we did not account for chemoautotrophic CO2 fixation, and the potential for remineralization of the chemoautotrophically produced organic carbon, which would be estimated to increase oxygen respiration from N chemolithotropy by ca. 4%. We also did not model the impact of organic sulfur oxidation by chemolithotrophs, which, according to Redfield stoichiometry, would in principle contribute much less to total respiration than ammonia oxidation.
We made another calculation to evaluate an alternate hypothesis: that unexplained oxygen uptake could be the result of changes in the oxidation state of TOC over the time course of the experiment. TOU models based on changing TOC, shown in Figure S4, do not consider the possibility that of the nominal oxidation state of carbon (NOSC) could change over time due to the recycling of carbon (“the microbial pump”) and/or DOM oxidation by cells, a postulated but unproven process 17. To evaluate the potential of this hypothesis to explain excess oxygen consumption, we assumed that the initial ARQ (∆CO2/-∆O2) of the TOC we detected was 0.30 at T0, but by day 100 was 3.10, which is the extreme of ARQ values reported in the literature33. We estimated that increases in the oxidation state of TOC remaining in the carboys could have consumed 442.3–458.8 µM O2, about 19% of total oxygen use. Deviations in the timing of TOC and DOM oxidation, and changes in the oxidation state of TOC, are not only possible, but indeed likely and both factors would tend to flatten the ARQ curve shown in Fig. 2C by raising ARQ values during the latter phase of the experiment. That said, observed TOU over the duration of the experiment was still higher than expected, suggesting other unexplained factors might be at work35(Figure S4).
Using a relatively simple experimental design we made observations that are in some respects surprising and point to the need for fundamental research into the mechanisms of oxygen loss from marine ecosystems. Microbial community analyses revealed a suite of unusual taxa, including OM190, Woesebacteria, Patescibacteria, and variants of SAR86, responding to low oxygen. These observations underscore the need to classify, cultivate and investigate the taxa shown to respond positively to low oxygen, and identify their roles in the biogeochemistry of hypoxic systems. There is a need for more accurate descriptions of the relationship between oxygen utilization and carbon remineralization by bacterioplankton and how it relates to the NOSC for improved model predictions of the carbon cycle. Research on microbial processes in OMZs present an opportunity to develop microbial and biochemical indicators to identify and potentially predict a specific system’s susceptibility to change. Our observations suggest that the current conceptualization of oceanic oxygen dynamics may be missing key elements, including the intersection of microbial diversity and oxygen and carbon cycling, which could substantially impact the results of global ocean models.