Evidence of dark oxygen production at the abyssal seafloor (2024)

A benthic chamber lander was deployed in the NORI-D license area six times in May–June 2021 (5D cruise), five times in November–December 2021 (5E cruise) and five times in August–September 2022 (7A cruise) (Extended Data Fig. 1 and Extended Data Table 1). The lander comprised three independent, autonomous, square benthic chambers (484 cm²) separated by approximately <0.5 m. After arriving at the seafloor, the lander waited for 0.07–1.34 d before the chambers were pushed into the sediment to create an enclosed microcosm of the seafloor. Ten minutes into the incubation period, the enclosed chambers were injected with 50 ml of one of three solutions: (1) 0.45-µm-filtered, cold surface seawater containing 79.2 mg of freeze-dried Phaeodactylum tricornutum algae, (2) 32 µM Na₂HCO₃ and 40 µM NH₄Cl dissolved in cold artificial seawater (salinity 35) and (3) 0.45-µm-filtered, cold surface seawater. On some occasions, the injection mechanism failed allowing the response to control (no injection) conditions to be measured. The seafloor in the study area had a temperature of 1.6 °C ± 0.006 °C (SE, n = 28) and a pH of 7.41 ± 0.05 (SE, n = 17). Immediately after the injection, the overlying water was mixed with a submersible stirrer at 60 rpm for 1 min before the stirrer was turned off that allowed any particulate substrates to settle for 1 h. After 1 h, the stirrer was then turned on again for the remainder of the experiments. During the 5E expedition, the stirrers were programmed to continually stir the overlying water even immediately after injection.

The syringe samplers removed approximately 50 ml of seawater from the water phase of each chamber at 0.1 or 0.03, 1, 3, 9, 28, 38 and 47 h into the incubation experiment. Oxygen optodes (CONTROS HydroFlash O₂ manufactured by Kongsberg Maritime Contros GmbH) mounted in the lid of each chamber logged O₂ concentrations in the chamber every 10 seconds throughout each experiment. Two days before the first lander deployment of each cruise, the optodes underwent a two-point, multi-temperature calibration using 0 and 100% O₂ calibration solutions at 1.2, 7, 18 and 30 °C following the recommendations of Bittig et al. (ref. ¹⁸). On the 5D cruise, we also calibrated the sensors 2 d after the last lander experiment so we could estimate optode drift, which was negligible (0.27 µmol l⁻¹ d⁻¹) over the course of the six-week cruise. The 0% and 100% O₂ saturation solutions were created by bubbling 0.45-µm-filtered surface seawater in a bottle sitting in a water-chilling/heating unit with N₂ gas (0%) or an aquarium air bubbling unit (100%) for 30 min. The O₂ concentration of the calibration solutions was confirmed in triplicate by Winkler titration. After incubating seafloor sediments for 47 h, the lander chambers were closed by a shutter door at the base of the chambers, and the chambers were then pulled slowly out of the sediment, which took 1 h. The lander was then recalled from the seafloor. In eight instances, the lander programme did not finish and the doors did not shut, preventing the sampling of sediment and determination of the volume of the water phase in the chambers (Extended Data Table 2). Once the lander was back and secured on deck, the chambers were opened and the water above the sediment removed via syphoning into a bucket. The distance from the top of the sediment to the base of the chamber lid was then measured in four places to get an accurate water depth for water volume estimates. Whenever possible, a photograph was then taken of the chamber sediment and nodules from directly above the opening of the chamber. All syringes containing water samples were removed and taken to the shipboard lab for immediate processing or stored in a cold lab (4 °C) before processing. The optodes were removed and their onboard data downloaded to a computer. Finally, the nodules were removed from the chambers and washed of attached organic debris with cold (4 °C), 0.45-µm-filtered surface seawater and placed in sterile Whirlpak bags to be weighed in the laboratory later. The number of polymetallic nodules at the seafloor determined from chamber counts was 1170 ± 97 m⁻².

Unfiltered syringe sample seawater was carefully transferred from each 50 ml syringe to a 12 ml exetainer via a 10 cm tube attached to the syringe nozzle, ensuring no air bubbles were introduced and immediately fixed for microWinkler titration. The sample was then mixed thoroughly using a glass bead placed in the exetainer and placed in the dark in a 4 °C refrigerator for 30–45 min to allow the precipitate to settle. Once the precipitate had sedimented, the exetainers were shaken again and left for 2–3 h before Winkler titrations were performed. All titrations were completed within 12 h after sampling to determine dissolved O₂ concentrations. Each Winkler sample (approximately 5 ml) was titrated twice, and duplicate measurements showed minor differences in O₂ concentration (5D cruise error: 3.5 ± 0.3 μmol l⁻¹, n = 71; 5E cruise error: 1.3 ± 0.2 μmol l⁻¹, n = 69; 7A cruise error: 2.8 ± 0.4 μmol l⁻¹, n = 84). Winkler O₂ concentration data were averaged for each syringe sample. The O₂ concentrations estimated by Winkler analysis were 22 ± 1% (n = 42, SE, 5D cruise), 8 ± 4% (n = 39, SE, 5E cruise) and 24 ± 2% (n = 40, SE, 7A cruise) lower than the concentrations measured by the optodes at the same time point in the same incubations most likely due to out gassing of supersaturated O₂ caused by depressurization and warming of the externally mounted syringes (whose samples were used for Winkler analyses) during the lander recovery to the surface.

Back on shore, the final O₂ concentration values were calculated following Bittig et al. (ref. ¹⁸) from the optode, calibration and in situ pressure data that was derived from the depth where each lander deployment was made. Time stamps in the optode data were compared to the lander computer programme times so the optode readings could be aligned to the schedule of the chamber experiment. The total change in O₂ concentration in each chamber was then calculated from the volume of the water phase above the sediment and the difference in O₂ concentration from when the chambers started to seal off the sediment to the point when the maximum O₂ concentration was reached.

Benthic O₂ microprofiling

Benthic O₂ microprofiles were made during lander deployments AKS313, AKS316, AKS318 and AKS321 during the 5E cruise using a UNISENSE deep-sea microprofiling unit mounted <0.5 m from the benthic chambers. The microprofiles were made using 20 cm O₂ microsensors that penetrated the sediment in 0.05 mm steps. The microsensors were calibrated 2 h before the lander deployments at in situ temperature (1.6 °C) at 0% and 100% O₂ saturation (above). At each sampling depth, the microsensor stopped for 5 s before each measurement was made. The sensor then recorded five individual O₂ concentration measurements. The average of these five measurements was taken for each depth point. The sediment surface was determined manually based on the turning point in the slope of O₂ concentration with depth where O₂ started to become depleted. SCOC was determined from Fick’s first law of diffusion.

Microbiology sampling

Nodule and sediment samples for microbial community analyses were collected from the 5D experimental chambers. Approximately 30 g of sediment from each of the 0–2 cm and 2–5 cm horizons and 50 g of intact nodules were placed in separate sterile Whirlpak bags with a pre-sterilized spatula and then transferred to a −80 °C freezer. DNA from approximately 10 g of nodules and 250 mg of sediment were extracted using the Qiagen PowerMax soil and PowerSoil extraction kits, respectively. Extracted DNA was then shipped on dry ice to Laragen Inc. and sequenced using a proprietary in-house method. The V4 region of 16 S rRNA genes were amplified using the Earth Microbiome Project protocol¹⁹ with the 515 F (5′‐GTGYCAGCMGCCGCGGTAA²⁰) and 806 R (5′-GGACTACNVGGGTWTCTAAT²¹) primers. Raw fastq files were processed using a custom pipeline (https://github.com/Boston-University-Microbiome-Initiative/BU16s) built with QIIME 2020.2 (https://www.nature.com/articles/s41587-019-0209-9). Adaptor sequences were removed using cutadapt (https://doi.org/10.14806/ej.17.1.200), read truncation positions were determined by mineer (more below), amplicon sequence variants (ASVs) were generated using dada2 (trunc-len-r²⁰) (https://doi.org/10.1038/nmeth.3869) and ASVs were clustered to 99% identity with the SILVA 132 database (https://academic.oup.com/nar/article/42/D1/D643/1061236) using the vsearch cluster-features-closed-reference (https://doi.org/10.7717/peerj.2584). Due to drops in sequencing quality, all reverse reads were truncated by 49 bases (from a length of 301 to 252) as determined by minERR, an algorithm for determining optimal sequence length based on sequence quality scores (https://github.com/michaelsilverstein/mineer). Family- and genus-level abundance was computed by summing the relative abundance of all ASVs with the same family/genus classification within each sample. Spearman correlations were then computed between family- and genus-level abundance and observed optode-derived total O₂ changes. Sequences have been archived at National Centre for Biotechnology Information GenBank under the Bioproject ID PRJNA1117483.

Polymetallic nodule surface area measurements

Photographs of the surface sediment and nodules in the chambers were imported into Image J. The outline of each nodule in each chamber photograph was then traced and the surface area of the nodule automatically calculated in Image J (assuming each surface nodule was flat in shape) and logged as an Image J file before being exported and saved as an Excel file.

Radiolysis O₂ production estimates

To estimate the potential radiolytic O₂ production, published concentrations of ²³⁸U, ²³⁵U, ²³²Th, ⁴⁰K (refs. ^{22,23,24,25,26}) in seawater were used (Supplementary Table 1). For nodules, ²³⁸U, ²³⁵U and ²³²Th isotopes of three nodules from chamber experiments from the 5D cruise were measured by Multicollector-Inductively Coupled Plasma Mass Spectrometer using previously described methods^27,28,29 and averaged; ⁴⁰K values were derived from the literature¹². Nodule and seawater contributions were calculated using a kinetic model developed by ref. ⁹ that incorporates 32 reactions (equation (1) in ref. ³⁰). The nodule boundary layer was assumed to be fully integrated with the seawater, surpassing the respective ~23 to ~452 μm stopping power distance of alpha and beta particles used to model geologic materials³¹. Sediment radiolytic O₂ was calculated as half of the previously quantified H₂ production rates in equatorial Pacific subsurface sediment³², given the stoichiometry of water’s radiolytic decomposition (an equivalency that probably offers an overestimate of derived O₂). Contributions from these three components (nodules, sediment and seawater) were scaled by the benthic chamber’s size and contents to produce an estimate of 0.18 μmol l⁻¹ of O₂ generated over 48 h according to the following expression.

Here (O₂)_t is the mass (kg) of O₂ produced over a given time t (yr), Q_iz is the mass (g) of the isotope, E_a is the average energy (eV) released from the decay of one atom; G(O₂) is the radiation chemical yield of molecules per 100 eV of the radiation energy; M_O2 is the O₂ molecular mass (g), A_iz is the isotope atomic mass (g) and λ is the isotope-specific decay constant (y⁻¹). The overall (O₂)_t value summed the contributions from ²³⁸U, ²³⁵U, ²³²Th and ⁴⁰K across water, nodule and sediment sources.

Electrochemistry measurements

Voltage potentials were measured using a Keithley DMM6500 digital multimeter on nodules previously collected by coring in the UK1, NORI-D and BGR license areas. Nodules were initially immersed for seven days in Instant Ocean artificial seawater (salinity 35). To measure the potentials, two electrodes (platinum wire, 99.9% purity) were first washed in perchloric acid, rinsed in Milli-Q water and dried before being attached to alligator clamps attached to the multimeter. The platinum wires were then immersed in Instant Ocean artificial seawater in a glass petri dish to measure background voltages (0.003 ± 0.001 V, SE, n = 17) until stable. Once stable, a nodule was placed in the petri dish and the platinum probes placed on the nodule at random locations, ensuring contact in one of two ways. We either carefully drilled a hole into some nodules so one platinum wire could be fixed inside it while the second platinum wire was firmly pressed against the nodule surface using a clamp. Alternatively, the platinum wires were pressed firmly against two different spots on the nodule surface and held in place using a clamp. Voltages were then recorded for 1–2 min until the signal was stable. This procedure was repeated up to 20 times in different randomly selected regions of the nodules depending on their size. Measurements were undertaken on 12 nodules at 21 °C (n = 153) and a single control rock composed of metamorphosed carbonate (n = 10). Two nodules from UK1 were also retested after being cooled to 5 °C (n = 18) by placing them in Instant Ocean water in a refrigerator overnight. Voltage potentials (n = 20) between two nodules were measured using four nodules collected from UK1. Potentials measured during each measurement were averaged and corrected for the background seawater voltage measured using only Instant Ocean seawater in the absence of a nodule. Measured resistances inside some of the nodules that were broken up were in the kΩ to 100s of kΩ range, though it is unclear if these resistivities change at the nano- or microscale requiring further investigation.

Geochemistry modelling

The chemical stability and solubility of manganese (IV) oxide (birnessite) to dissolved Mn²⁺ as a function of pH and O₂ activity was modelled using the Geochemist Workbench Professional (version 12) software, with the in-built and internally consistent THERMO database. The conditions used for generating the phase diagram (Extended Data Fig. 5) represent bottom seawater as measured in the eastern CCZ with a temperature of 1.6 °C and chlorine and manganese concentrations of 0.55 M Cl and 2e⁻¹⁰M Mn, respectively.

Ex situ core incubations

Opportunistic ex situ experiments were undertaken during the 5D cruise using sediment cores retrieved by a multi-corer from the CTA area (Extended Data Fig. 1). Immediately after the multi-corer arrived back at the surface, cores were removed and transferred to a cold lab held at in situ temperature. The cores were then exposed to the following five treatments (administered using a 60 ml syringe), which included (1) Na₂HCO₃ (0.3 μM final concentration, n = 3), (2) NH₄Cl (10 μM final concentration, n = 3) and (3) NH₄Cl (50 μM final concentration, n = 3), (4) 0.3 μM Na₂HCO₃ + 10 μM NH₄Cl (final concentration, n = 3) and (5) HgCl₂ (1.1 μM final concentration, n = 3). No-injection controls (n = 3) were also performed and separate core experiments in which four nodules were incubated for 48 h by themselves with no additions. After addition, the water phase of each core was stirred and a 50-ml sample of top water was taken for microWinkler analysis (as above). Stoppers were then placed on the top of the cores, ensuring no air bubbles were present. The stoppers were secured tightly and the cores fully submerged in a large bucket containing 0.45-µm-filtered, cold, surface seawater (salinity 35). The bucket was covered with five black plastic bags and secured in the cold room with the lights turned off. After 48 h, the cores were removed from the bucket, and the cores were inspected for the presence of air bubbles. Only one core, a HgCl₂ treatment, had a gas bubble beneath the bung, which was rejected from further analysis, leaving n = 2 for this treatment. The other cores were then re-sampled for dissolved O₂ and analysed as before. Core-specific water volume measurements were used together with the change in O₂ concentration to calculate the total net O₂ change per core.

To determine if our ex situ DOP detection was affected by intrusion of O₂ from the atmosphere into the core tube, two controls were performed: a shipboard test with an O₂ microprofiler and a lab-based test using the Winkler method. Shipboard, a clean core tube was filled with Milli-Q water and sparged with N₂ for 10 min before beginning the test. A Metrohm 8663 Multimeter was inserted through a predrilled hole in the rubber stopper, allowing for O₂ concentration to be recorded every 5 s. An increase from 39 to 69 µmol l⁻¹ was observed over ~5 h, corresponding to a rate of 0.14 mmol m⁻² d⁻¹ or 4% of the 3.5 mmol m⁻² d⁻¹ mean net DOP measured in the ex situ experiments. Back in the home laboratory, three of the original core tubes were filled with 4 °C, 0.2-µm-filtered artificial seawater (salinity 35) and sparged with N₂ for 8 min through a filtered pipette tip to achieve an initial dissolved O₂ concentration of ~100 µmol l⁻¹ (for example, the approximate starting O₂ concentrations for the shipboard experiments). The tubes were sealed with rubber stoppers and electrical tape, being careful to avoid bubble formation. They were then submerged in a 32-gallon plastic garbage can of unfiltered seawater (O₂ concentration: 228.12 µmol l⁻¹) in a dark cold room (8 °C) for 48 h. After 48 h, the tubes were quickly unsealed and analysed one at a time to prevent additional O₂ dissolution from the air. A 50-ml sterile syringe was used to slowly collect 10 ml of seawater from the centre of the core tube, being sure to avoid bubble entrainment into the syringe. The sample was carefully expelled into a 10-ml reaction vial and fixed using the adjusted values for a 10-ml sample according to a volume-scaled Winkler titration protocol³³ and the reagents from the LaMotte Dissolved Oxygen Test Kit. The fixation of each collected sample was done in less than 2 min in a fume hood. Dissolved O₂ increased by 0.11 mmol m⁻² d⁻¹ during the 48 h, which corresponds to between 3.2% of the mean net DOP rate observed in the ex situ experiments (3.5 mmol m⁻² d⁻¹). Both of our control experiments provide high confidence that the diffusion of external O₂ into the core tubes did not cause the O₂ production measured in the ex situ core incubations.

Calculations to quantify intrusion of O₂ from the polyoxymethylene chambers and lids

Oxygen intrusion was estimated from Stephens³⁴ who calculated that 20.66 µmol l⁻¹ of O₂ could diffuse out of 428 cm² of polyoxymethylene plastic when immersed for 48 h in hypoxic water (O₂ diffusion rate: 0.02 µmol O₂ cm⁻² d⁻¹). To determine the total area of plastic that would be available for diffusion (869–1,584 cm²), we added the surface area of the lid to the surface area of the four walls that would be exposed at the seafloor (based on the depth of the water phase—above). The minimum and maximum areas available for diffusion were multiplied by 0.02 µmol O₂ cm⁻² d⁻¹ to estimate that 41.9–76.5 µmol O₂ l⁻¹ would diffuse out of the polyoxymethylene chamber walls and lid in 48 h under hypoxic conditions. Thus, we are highly confident that O₂ leakage from the plastic chambers could not replicate the high O₂ concentration seen in some of our oxygenated experiments (Fig. 1).