Microbial ecology, the study of microbial communities and their interactions with the natural environment, relies on analyzing community composition, diversity, and abundance. Ideally, these analyses should be coupled with what is known about the metabolisms and physiological activities of the dominant taxonomic groups, to better understand the relationships of the microorganisms with each other and with the environment. Enormous progress has been made with molecular and -omics techniques to characterize microbial communities, especially of the non-cultivated fraction. However, the information these methods can give about the actual geochemical processes happening in the environment is rather indirect. Detected species, families or higher taxonomic groups might not be active, metabolisms inferred from DNA sequences require translation steps and even the presence of certain enzymes requires knowledge of their kinetics as well as data on the substrate availability to asses if a process really happens. These methods should therefore be complemented by more direct process analyses. Approaches to link microbiota to geochemical processes include to combine microscopy and measured physiology of enrichments (Schulz et al., 1996; Schulz and Schulz, 2005; Boetius et al., 2000; Otte et al., 1999), combine hyperspectral analysis and microsensing (Klatt et al., 2015a, 2015; Merz et al., 2021) or to compare, in well stabilized sediments or microbial mats, the communities and main biogeochemical conversions with high spatial resolution (Ramsing et al., 1993; Teske et al., 1998; Santegoeds et al., 1998; Schramm et al., 1996). These studies were all done in habitats with high activities and unique geochemistry.

The sea floor of the Gulf of Aqaba in the Red Sea may offer a natural laboratory. This habitat is characterized geochemically by only a few studies. These Mn- and Fe-rich sediments showed deep oxygen penetration and low sulfate reduction rates (Blonder et al., 2017; Boyko et al., 2018). The Red Sea is highly oligotrophic and its sea bottom has little organic input, hence chemoautotrophy may be relatively important (Middelburg, 2011). From deeper sediments NH₄⁺, Mn²⁺ and Fe²⁺ diffuse upwards, fueling chemoautotrophic processes (Boyko et al., 2019, 2018; Blonder et al., 2017; Steiner et al., 2019). The sediments are aeolian and have a low sedimentation rate of 0.4-0.7 mm per year (Blonder et al., 2017). Bioturbation, inferred from isotope distributions (Steiner et al., 2016), is limited. The bioturbation coefficients at the surface were estimated to be in the order of 10^-12 (m²s^-1), 3 orders of magnitude lower than the diffusion coefficients of the solutes in the porewater, and rapidly decreasing with depth (Steiner et al., 2016). Therefore, the sediments are considered vertically stratified. Advection in these very fine-grained sediments can be ignored, allowing rates to be calculated from porewater profiles using diffusion-reaction modeling. These physico-chemical characteristics make these sediments a useful system to assess the reflection of the biogeochemical profiles on the community structure.

In this contribution, we provide a detailed description of the geochemistry and microbial community composition of sediments from a deep site in the Gulf of Aqaba. We hypothesize that the microbial community is as stratified as the biogeochemical profile, which would allow us to infer which taxa are responsible for the main geochemical conversions: oxidation of Mn²⁺, Fe²⁺ and NH₄⁺. We used a combination of centimeter resolution 16S rRNA gene community analysis, geochemical depth profile analyses and flux calculations to comprehensively describe these sediments.

NH₄⁺ oxidation is relatively well confined to described microbial taxa (Beman et al., 2011; De Boer and Kowalchuk, 2001; Martinez-Garcia et al., 2008). Mn²⁺ and Fe²⁺ oxidation is much less defined. Chemical Fe²⁺ oxidation by O₂ occurs under pH neutral conditions very fast, thus most biological processing is anaerobic, driven by light (Widdel et al., 1993), Mn-oxides (Jørgensen and Nelson, 2004) or NO₃^- (Wang et al., 2020). Mn²⁺ oxidation is thought to be mainly biological by a wide group of microorganisms (Hansel, 2017). Mn-oxide reduction can be both biological and chemical driven by many reduced species such as Fe²⁺, sulfide and NH₄⁺ (Johnson et al., 2016). The active oxidation and reduction processes can drive rapid Mn cycling in sediments. An important complication of Fe and Mn oxidation is their potential to generate reactive oxygen species (ROS) (Hansel, 2017; Luther, 2010; Dixon and Stockwell, 2014; Koppenol, 1993). These develop upon single electron transfer leading to reduction of O₂ to superoxide (O₂^-.) (Luther, 2010). This radical can be degraded by superoxide dismutase to form the less reactive peroxide H₂O₂, that subsequently is converted by catalase to O₂ and water. ROS species can be highly destructive for life (e.g. the hydroxyl radical OH^.) or less toxic (e.g. O₂^- and H₂O₂). Although ROS are short-living, their presence in sediments was demonstrated and shown to significantly influence microbial degradation rates (Van Erk et al., 2023). Consequently, Mn and Fe oxidation could have lethal side-effects for microbes without effective ROS defense. The ROS presence was investigated by H₂O₂ microsensors and porewater analysis by a chemiluminescent method.

In summary, the aim of this study was to link the various spatially separated geochemical processes, inferred from vertical porewater profiles and solid phase chemistry, to the microbial community analyzed by 16S rRNA amplicon sequencing on horizontally sliced sediment cores.

Sampling was conducted in March 2022 and June 2023 in the Gulf of Aqaba, northern Red Sea, aboard the R/V Sam Rothberg. The sampling site was located at 29°28’28.8”N, 34°55’20.2”E, at a water depth of approximately 700 m (Figure 1). A detailed overview of the site and its sediments is described previously (Boyko et al., 2022). Sediments were collected using a multicorer from the ship (MC-400 Hedrick/Marrs Multi-Corer, Ocean Instruments), the coreliners were scrubbed and rinsed with freshwater before each use. Bottom water temperature was ~21 °C—comparable to ambient conditions in the on-site laboratory where all sample processing was conducted. NH₄⁺, dissolved Mn, solid-phase Mn, TOC, O₂, pH and H₂O₂ profiles were determined on sediments sampled in March 2022. Dissolved Fe²⁺ and microbial analyses were conducted on sediments sampled in June 2023. Because the fixed microbial samples from 2022 had leaked during transport to Bremen, the DNA quality was unsatisfactory and sampling for microbiology was repeated in 2023.

Sediment cores were collected using liners pre-drilled with holes for porewater extraction using Rhizons (Rhizosphere Research Products, The Netherlands). These were used for microsensor-based and chemiluminescent H₂O₂ analyses, as well as for Fe²⁺, Mn²⁺, NO₃⁻, NH₄⁺, dissolved inorganic carbon (DIC), total organic carbon (TOC). For pore water extraction Rhizons were used at a depth resolution of 1–2 cm, and another core was sliced and centrifuged for solid phase and pore water analysis at depth resolution of 0.5 cm. Porewater was filtered (0.22 µm) and immediately collected into prepared tubes for the respective analyses. O₂, and pH profiles were measured using microsensors. O₂ and H₂O₂ microprofiles were obtained using home-built microsensors as described previously (Van Erk et al., 2023; Revsbech, 1989). pH microprofiles were measured using a glass microelectrode (MI-413, Microelectrodes Inc., USA) on the same setup.

For dissolved and solid-phase Mn analyses, three sediment cores were sliced at 0.5 cm intervals from 0 to 5 cm depth. The sediment slices were transferred into 50 mL centrifuge tubes, flushed with N₂ (99.999%) and centrifuged for 10 minutes at 9000 rpm (Soto et al., 2023). The resulting supernatant was filtered through a 0.22 µm syringe filter. For dissolved Mn analysis, 100 µL of the supernatant was mixed with 5 mL of 0.1 N HNO₃. Following centrifugation, sediments were stored at -20°C.

For solid-phase Mn, sediment was dried at 80°C, ground, combusted at 450°C, and digested in 6 M HCl (100x dilution). Pore water and solid phase Mn concentrations were measured by microwave plasma atomic emission spectrometry (MP-AES, Agilent).

For two of the cores, supernatant for NH₄⁺ analysis and NO₃^- analyses was stored at -20°C and later transported to the Max Planck Institute in Bremen, where analyses were carried out following previously described protocols (Billerbeck et al., 2006a).

For Fe²⁺ measurements, porewater was directly extracted through Rhizons (Rhizosphere Research Products) and mixed directly with an equal volume of Ferrozine reagent (0.29 g Ferrozine in 2 mL Milli-Q water) that was pre-filled in the syringes. Samples were stored at 4°C until spectrophotometric measurement at a wavelength of 562 nm.

A separate core was used for analyzing methane (CH₄) profiles with 1 cm resolution in porewater as described previously (Sivan et al., 2014). The limit of detection was 4 nM.

For chemoluminescent H₂O₂ analyses, porewater was extracted through pre-drilled holes in eight sediment cores using Rhizons (Rhizosphere Research Products, The Netherlands). Directly after collection, 100 µL porewater was transferred to a 2 mL tube pre-filled with 100 µL Ferrozine solution (0.29 g Ferrozine in 2 mL mQ water). Subsequently, chemoluminescent H₂O₂ detection was performed with a home-build system, as described earlier (Van Erk et al., 2023). Daily calibrations of the chemiluminescent H₂O₂ detection method were performed using a standard range, and catalase was added to selected subsamples to confirm signal specificity to H₂O₂.

For total organic carbon (TOC), two additional sediment cores were sliced at 1 cm intervals (0–10 cm). The 0–1 cm and 5–10 cm slices were stored frozen and analyzed at MPI Bremen following previously established protocols (Billerbeck et al., 2006b). DIC samples were fixed (5 µL saturated HgCl₂ per ml), brought to the MPI Bremen in glass vials with septum (Exetainer, Labco, UK) and analyzed as described previously (Billerbeck et al., 2006b).

The chemicals for these analyses were analytical grade and obtained from Sigma-Aldrich, H₂O₂ (3% Beyvers GmbH, Berlin) was obtained from a pharmacy.

H₂O₂ was added experimentally to high concentrations (100 µM - 1 mM) in the water column, from where it diffused into the sediment, and O₂ and H₂O₂ concentrations were measured using microsensors. This was done on (1) intact sediments to study the effects on sediments above the brown band, (2) with the oxic sediment removed to expose the brown band, and (3) with also the brown band removed to study the effects of H₂O₂ additions on the anoxic sediment below the brown band.

Local fluxes were calculated from the porewater profiles using Fick’s law of diffusion: J = D_s × dC/dx with D_s the diffusion coefficient of the solutes corrected for salinity, temperature and porosity (Van Stroe-Biezen et al., 1993; Li and Gregory, 1974; De Beer et al., 2006), C the concentration and x the depth. The values for D_s used were (in m²s^-1): Fe²⁺ 2.63 ×10^-9, Mn²⁺ 2.6 × 10^-9, NO₃^- 7.3 × 10^-9, O₂ 9.5 × 10^-10, NH₄⁺ 7.6 × 10^-9, H₂O₂ 5.9 × 10^-10. Conversion rates at specific sediment depths were calculated from the profiles as previously described (Gieseke and De Beer, 2004).

Three sediment cores were sliced at 2 cm intervals, and each sediment layer was preserved in 50% ethanol. To detach bacteria from sediment grains, samples were sonicated on ice with a Sonopuls GM Mini20 probe equipped with a microtip MS 2.5 (Bandelin, Berlin, Germany). Samples were sonicated for 30 s, at an amplitude of 86% and pulse of 0.2 s, sediment was removed using a table centrifuge with 2 ml eppendorf tubes and the supernatant was subsequently diluted 1:100. After dilution, cells were filtered on polycarbonate filters pre-stained with Irgalan Black. Filters were stained with 2 mL of acridine orange, rinsed with sterile-filtered citric acid buffer, and mounted onto glass slides with Cargille immersion oil. Additional staining attempts using DAPI, SYBR Green, and SYBR Gold were unsuccessful due to excessive background fluorescence, which prevented cell identification. Cells were counted using an epifluorescence microscope.

Three sediment cores were collected and sectioned at 1 cm intervals from the surface downward. Each sediment slice (3 mL of sediment) was preserved in 12 mL of LifeGuard^® solution to stabilize nucleic acids. One DNA extraction per sample was done using the PowerSoil^® Kit (QIAGEN, Hilden, Germany), following the manufacturer’s protocol. PCR amplification was performed for 34 cycles using the primers Bac-341F (CCTACGGGNGGCWGCAG) and Bac-805R (GACTACHVGGGTATCTAATCC) (Herlemann et al., 2011), and purified using AMPure XP beads (Beckman Coulter, California, USA). Sequencing was conducted on an Illumina NextSeq 2000 platform, generating 2 × 300 bp paired-end reads at the Max Planck Genome Center Cologne. All samples were sequenced in the same run.

Raw sequences were demultiplexed with Cutadapt v1.15 and processed using DADA2 v1.3 (Callahan et al., 2016) for quality filtering, denoising, and amplicon sequence variant (ASV) inference. Raw sequences were filtered based on length (reads shorter than <200 base pairs were removed) and maximum number of expected errors (2 for forward reads, and 3 for reverse reads). Taxonomic assignment was done using the SILVA database (silva_nr99_v138.1_train_set.fa.gz). Before any analysis, sequences classified as chloroplasts and mitochondria were removed, and only bacterial reads classified at the phylum level were kept. ASVs were filtered by length (>396 and <430 bp), and those that were observed in only one sample were removed, regardless of their abundance. We suspect those sequences to be largely PCR artifacts and preferred to have a rather conservative analysis when it comes to community composition and diversity. The proportion of ASV retained after each step is shown in Supplementary Table 1. Once the dataset was filtered, diversity analyses were done at the ASV level using the iNEXT package (Hsieh et al., 2016), standardizing for 99% sample coverage, which was the achieved coverage for most samples (for rarefaction curves, see Supplementary Figure 6S). Community composition was remarkably similar between cores with most sequences not being classified at the genus level. Therefore, the community composition analysis was done at the family level over the average community between the three cores (community composition at the genus level for each core is shown in Supplementary Figure 7S). Since our objectives regarding community analysis were mostly descriptive and exploratory, no statistical tests were applied. Instead, data is presented to show trends of the microbial community across depth.

The sediments consisted of very fine light colored clay with a porosity of 0.7 and a grain size of 45 µm (Rasheed et al., 2003). No faunal activity or worm burrows were observed in any of the cores. Due to the fine grainsize the transport of porewater solutes was purely diffusional (Wilson et al., 2008; Shepherd, 1989). In all sampled cores a characteristic brown band was observed at 2 cm below the surface (Figure 1). This coincides with the Mn solid phase maximum (Figure 2).

Porewater depth profiles (Figure 2) and fluxes derived from these porewater profiles (Table 1) show occurrence of various aerobic and anaerobic microbial processes. O₂ diffuses from the water column downwards. From larger depths NH₄⁺, Mn²⁺ and Fe²⁺ diffuse upwards and are oxidized at slightly different depths. NH₄⁺ diffusing upwards from the anoxic zone penetrates 1 cm into the oxic zone. Mn²⁺ disappears in the brown band where also the maximum in solid phase Mn is found. Fe²⁺ diffusing upwards disappears 1–2 cm below the brown band, in the anoxic zone.

The brown band was colored by Mn-oxides, as exactly at that depth the solid Mn content showed a maximum (Figures 2A; Supplementary Figure 2S). O₂ diffusing downwards meets Mn²⁺ diffusing upwards (Figures 2A; Supplementary Figure 2S) in the brown band, driving aerobic Mn²⁺ oxidation. Porewater solute fluxes were calculated using Fick’s law of diffusion (Table 1). The O₂ flux into the band was half of the Mn²⁺ influx into this zone, close to the stoichiometry of aerobic Mn²⁺ oxidation (Mn²⁺ + 0.5O₂→MnO₂ + 2H⁺). Thus, the brown band is the location where all upwards diffusing Mn²⁺ is oxidized aerobically.

The curvature of the oxygen profile indicates a homogeneously distributed oxygen consumption in the oxic zone above the Mn²⁺ containing zone, by both NH₄⁺ diffusing upwards from the anoxic zone (Figure 2B) and organic matter deposited from the water column. The O₂ flux from the sediment surface downward is approximately 30x higher than the upward NH₄⁺ flux into the oxic zone (Table 1). This is far above the stoichiometry of nitrification (NH₄⁺ + 2O₂→ NO₃^- + H₂O + 2H⁺). Hence all NH₄⁺ diffusing into the oxic zone can be oxidized by nitrification. Nitrification in the oxic zone results in elevated NO₃^-, with a maximum just below the sediment surface. From its maximum NO₃^- diffuses partially downwards into the anoxic zone below the brown band, and partially upwards into the water column. Besides NH₄⁺, additional e-donor is needed to explain the O₂ consumption in the oxic zone, which is probably organic matter. Indeed TOC (Supplementary Figure 1S) decreases linearly with depth, indicating gradual degradation. The DIC profile mirrors the TOC profile, except at the highest point at 1 cm below the seafloor (bsf), likely due to out-diffusion into the water column.

The Fe²⁺ diffusing upwards disappears in the anoxic zone just below the Mn-enriched brown band. This could be due to anaerobic oxidation with NO₃^- as e-acceptor. The flux of Fe²⁺ is approximately 5x the flux of NO₃^- into their overlapping zone (Table 1), close to the stoichiometry of Fe oxidation via NO₃^- (5Fe²⁺ + NO₃^- + 12 H₂O → 5Fe(OH)₃ + 0.5N₂ + 9H⁺). The slight pH minimum at 4–5 cm bsf supports the possibility of this process.

CH₄ concentrations showed no trend with depth in the upper 25 cm. They were on average 2.3 nM, all values were below the detection limit (data not shown).

No other processes can be inferred from our geochemical data. Although NO₃^- and NH₄⁺ overlap, the profiles do not indicate the presence of ANAMMOX, the anaerobic oxidation of NH₄⁺ by NO₂^-, as the NH₄⁺ profile does not show depletion or a curvature in this zone. No net NH₄⁺ oxidation occurs in the anoxic zone, as there the profile is linear.

The oxidation of metals, like Mn²⁺ and Fe²⁺, can result in the formation of reactive oxygen species (ROS) (Luther, 2010; Jomova et al., 2012). ROS was shown to significantly decrease organic matter mineralization rates (Van Erk et al., 2023). Therefore, we expanded the study to H₂O₂ and its cycling. H₂O₂, measured both by chemiluminescence analysis on extracted porewater and by microsensors, was low but significantly present in the porewater, also in the anoxic zone below 2 cm (Supplementary Figure 3S). The concentrations measured by both methods were approximately 1 µM. The H₂O₂ microsensors became erroneous below the brown band, possibly by an interference. Porewater H₂O₂ measurements using chemiluminescence resulted in similar concentrations of H₂O₂ throughout most of the sediment (Supplementary Figure 3S). Addition of catalase to a selection of extracted porewater samples reduced the signal to zero, confirming that measured signals were indeed H₂O₂. Porewater samples were directly mixed with a Ferrozine solution (an Fe²⁺ scavenger) during porewater extraction, inside the syringe used for extraction. Thus Fe²⁺ interferences could be excluded.

H₂O₂ addition experiments showed H₂O₂ degradation under simultaneous O₂ evolution, indicating a catalase-like process: 2H₂O₂→2H₂O + O₂. This process could be observed in sediments above the brown band and in the brown band, but not in the sediments below the brown band (Supplementary Figure 4S).

From the profiles the kinetic parameters of H₂O₂ degradation were estimated using diffusion-reaction modeling (Gieseke and De Beer, 2004). Each profile represents a range of concentrations and local degradation rates. The rates were linearly proportional to the concentration, as occurs under first order kinetics (Supplementary Figure 5S). The first order reaction rate constant C was 0.00206 (s^-1) above the brown band and 0.00288 (s^-1) within the band. The rates in the zone above the bands showed larger variations than in the band due to variable MnO₂ content.

The cell numbers decreased from 2 × 10⁹ cells mL⁻¹ at the sediment surface to approximately half of this value at 8–9 cm depth (Figure 3A). Notably, a local minimum was observed at the brown band in all 3 cores, below which cell numbers increased again.

Estimated richness, defined as the number of ASVs at 99% community coverage, was between 1264-3566 (Figure 3B), and exhibited a slight increase below the brown band. Evenness, estimated by Pielou’s evenness index, was exceptionally high and constant across depth with a mean value of 0.91 and a standard deviation of 0.014, with a slight decrease towards the deepest layers. It indicates a balanced community distribution and no dominant single taxonomic group at any depth.

All three cores were remarkably similar, and the community composition showed clear changes with depth (Figure 3D, see Supplementary Figure 8S for MDS plot on Jensen-Shannon distances between communities). Together, 13 families and 10 groups not classified at the family level showed more than 2% in at least one sample, and in total, they accounted for about 50% of the community at any given depth (Figure 3E). All of these groups were present at all depths, with some of them showing clear changes in their relative abundances with depth, while others remained more constant (Supplementary Figure 9S). Woeseiaceae, Nitrospiraceae, Nitrosococcaceae, and Cyclobacteriaceae, together with unclassified groups from Ca. Actinomarinales, ‘Subgroup 21’, ‘AT-s2-59’, and ‘SAR202 clade’ were more abundant above the brown band than below, while Spirochaetaceae, Desulfatiglandaceae, Desulfosarcinaceae, Methylomirabilaceae, and Anaerolineaceae, together with unclassified groups from Aminicenantales, Thermodesulfovibrionia, and Syntrophobacterales, were more abundant below the band than above. Sequences belonging to archaeal Nitrosopumilus were also found above the brown band, but were not included in the analysis (see Methods). Ca. Scalanduaceae was the only group that peaked at the brown band. Kiloniellaceae, Omnitrophaceae, Thermoanaerobaculaceae, together with unclassified groups from “NB1-j”, “BD2–11 terrestrial group”, and “bacteriap25” showed fairly stable relative abundances, with little variation across depth.