Chemosynthesis enables microbial communities to flourish in a marine cave ecosystem

TL;DR

Microbial communities in a marine cave thrive through chemosynthesis, even in the typically light-dependent euphotic zone.

Contribution

The study reveals that chemosynthesis supports diverse microbial life in marine caves within the euphotic zone.

Findings

Interior cave microbes showed higher diversity and chemosynthetic activity compared to entrance communities.

Photosynthetic microbes and genes decreased in the inner cave, while chemosynthetic lineages increased.

Cave communities consume inorganic compounds and fix carbon dioxide, with higher rates in the interior.

Abstract

Chemosynthesis, an ancient metabolism that uses chemical compounds for energy and biomass generation, occurs across the ocean. Although chemosynthesis typically plays a subsidiary role to photosynthesis in the euphotic ocean, it is unclear whether it plays a more important role in aphotic habitats within this zone. Here, we compared the composition, function, and activity of microorganisms colonising the sediment of a marine cave at mesophotic depth, across a transect from the entrance to the interior. Microbes thrived throughout this ecosystem, with interior communities having higher diversity than those at the entrance. Analysis of 132 species-level bacterial, archaeal, and eukaryotic metagenome-assembled genomes revealed niche partitioning of habitat generalists distributed along the cave, alongside specialists enriched across the entrance and interior environments. Photosynthetic microbes and photosystem genes declined in the inner cave, concomitant with enrichment of chemosynthetic lineages capable of using inorganic compounds such as ammonium, sulfide, carbon monoxide, and hydrogen. Biogeochemical assays confirmed that the cave communities consume these compounds and fix carbon dioxide through chemosynthesis, with inner communities mediating higher cellular rates. Together, these findings suggest that the persistent darkness and low hydrodynamic disruption in marine cave sediments create conditions for metabolically diverse communities to thrive, sustained by recycling of inorganic compounds, as well as endogenous and lateral organic matter inputs. Thus, chemosynthesis can sustain rich microbial ecosystems even within the traditionally photosynthetically dominated euphotic zone.

Figures

(A) Map of Port Phillip Bay indicating the location of the study site. (B) Schematic illustration of the surveyed marine cave, showing representative depths, transect across the entrance (0 m), mid (6.5 m), and inner (13 m) cave environments, and typical benthic assemblages, including cnidarians, sponges, and algae within and surrounding the cave. (C) Close-up of the measuring transect highlighting the cave floor substrate. (D) Community richness (Shannon index) of archaeal and bacterial communities in sediment and seawater, based on 16S rRNA gene data. (E) 16S rRNA gene copy number per gram wet weight (g ww) of sediment across cave environments. (F) PCA of beta diversity showing compositional differences in sediment and seawater microbial communities. (G) LEfSe analysis identifying archaeal and bacterial genera enriched in entrance, midsection, and inner cave environments, with linear discriminant analysis (LDA) scores indicating discriminative power. Only top 16 enriched genera per environment are shown, for a full list refer to Supp. Data S1. (H) Chlorophyll a concentration in sediments at the entrance and inner cave, indicating abundance of oxygenic phototrophs.

(A) MAGs maximum-likelihood phylogenetic tree inferred using LG substitution matrix, representing the topography, metabolic repertoires, and niche breadth of 130 archaeal and bacterial MAGs encompassing 13 phyla. Clades are coloured according to phylum-level taxonomy and symbols indicate the potential of each MAG to mediate key energy and carbon acquisition processes. (B) Bar graphs showing genome size (mb) and gene count of generalists, intermediates, and specialists MAGs. Analysis of variance (ANOVA) shows statistical differences among generalists, intermediates, and specialists MAGs genome size and gene count. (C) Heatmap showing the metabolic functions of cave sediment MAGs and eMAGs identified as generalists, intermediates, and specialists.

(A) Relative abundance of genes encoding enzymes for trace gas metabolism, fermentation, sulfur and nitrogen cycling, phototrophy, and carbon fixation across distinct cave environments. Significant differences in gene relative abundance across the cave environments were assessed using categorical models (ANOVA or Kruskal–Wallis test) with post hoc pairwise comparisons (Dunn’s tests) with false discovery rate (Benjamini–Hochberg) correction. We also plotted regression lines across the transect distance (0, 6.5, and 13 m), which highlight the overall spatial trends. Shaded grey areas show 95% confidence interval of the regression lines. (B) Linear regressions showing correlations between the abundance of genes associated with energy metabolism (explanatory variable, x-axis) and those involved in carbon fixation pathways (response variable, y-axis) across the cave environments.

(A) Biogeochemical assays illustrating aerobic metabolic activity of microbial communities from the cave entrance and inner environments, based on the consumption of H2, CO, and NH4+. Heat-killed controls were included to confirm the biotic origin of the observed processes. (B) Oxidation rates of H2, CO, and NH4+ normalized per-cell specialized in each metabolism across entrance and inner cave environments. Shaded grey areas show 95% confidence interval of the regression lines. (C) Aerobic 14CO2 incorporation by sediment community incubated under light (40 μmol m−2 s−1) and in the dark with CO (100 ppm), H2 (100 ppm), NH4+ (1 mM), and S2− (0.8 mM). (D) Oxygen consumption rates of sediment from the entrance and inner cave environments. (A, B) measurements were performed in 120 ml sealed serum vials containing ~10 g of sediment and 50 ml (trace gasses) or 100 ml (O2 consumption) of 0.22 μm-filtered seawater. Trace gas incubations were supplemented with 10 ppm H2 and CO in the vial headspace. Nitrification assays (NOx = NO2− + NO3−) were conducted in uncapped 250 ml Schott bottles containing ~10 g of sediment, 100 ml of 0.22 μm-filtered seawater, and 100 μM NH4+.

Equations

$$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{upgreek} \usepackage{mathrsfs} \setlength{\oddsidemargin}{-69pt} \begin{document} \begin{align*} & Chlorophyll\ a\ (\mu g/g)\\&=[( nm655\ read- nm750\ read)\ x\ 11.4\ x\ 10)]/ weight\ (g) \end{align*}\end{document}$$

Full text

Introduction

Chemosynthesis, the microbial metabolic process that transfers carbon to the biosphere using chemical energy, provides a foundation for marine life [1]. Once thought to be limited to hydrothermal vents and other geologically active environments [2–6], chemosynthesis is now recognized as far more pervasive throughout the ocean [1]. For example, activity-based experiments have shown that surface ocean microbial communities oxidize ammonia (NH_4_^+^), carbon monoxide (CO), and molecular hydrogen (H_2_) for both survival and growth, while co-existing with conventional photoautotrophs and organoheterotrophs [7, 8]. Chemosynthesis is also dominant in certain near-surface waters where either photosynthesis is excluded, for example in waters shielded by Antarctic ice shelves [9], or where chemical inputs are high, for example in mesophotic cold seeps where benthic fauna form mutualistic relationships with chemosynthetic microbial communities [10], underscoring the ecological importance of this process even at relatively shallow depths. In the ocean, chemical energy sources arise from geological seepage and from biotic and abiotic processes that generate compounds capable of fuelling chemosynthesis. For example, microbial processes such as sulfate reduction, nitrogen fixation, and organic carbon degradation generate significant amounts of sulfide (S^2−^), H_2_, NH_4_^+^, and CO that can be used as energy sources for chemosynthetic aerobes and anaerobes [7, 11–16]. Although global fluxes of specific compounds remain unconstrained, largely because marine chemosynthesis knowledge is still fragmentary and these processes are typically assessed only at regional scales [17, 18], estimates suggest that chemoautotrophic carbon fixation could account for ~2.5–22% of total oceanic primary productivity [19]. Yet, the mechanisms that structure and support chemosynthesis at shallow depth remain poorly characterized, particularly the dominant energy sources, the prevailing carbon-fixation pathways, and how they shift across spatial gradients of key marine environments.

The vastness of the ocean and the inaccessibility of many of its environments continue to challenge our understanding of the full breadth of ecological niches and their implications for life on Earth. Among the most enigmatic are marine caves, benthic habitats found along coastal margins worldwide that form unique interfaces between the subsurface and the coastal ocean. Despite their global distribution [20–23], the microbiology of marine caves remains largely unexplored, often obscured by logistical challenges in accessing them. At Cape Palinuro, Italy, fluids enriched in hydrogen sulfide (H_2_S), methane (CH_4_), thiosulfate (S_2_O_3_^2−^), and carbon dioxide (CO_2_) emitted from an active spring in the Grotta Azzura sustain Beggiatoa-like microbial mats [24], reminiscent of those found at abyssal vents [25, 26], highlighting surprising parallels between shallow and deep-sea chemosynthetic ecosystems. Other marine caves lack active vents, but experience severely restricted water exchange, resulting in periodical anoxic conditions accompanied by substantial accumulation of H_2_S throughout the cave water column and sediment [27]. Anoxygenic phototrophs, such as purple sulfur bacteria, can benefit from these conditions and proliferate in dense populations [28]. In these perennially light-limited systems, sulfide availability rather than light can become the primary constraint on biomass production [27]. These foundational studies position marine caves as environments with diverse but distinctive microbiology and biogeochemistry. Yet, no study to date has systematically investigated the composition or function of the microorganisms within marine caves. A mechanistic understanding of the ecological strategies employed by cave-dwelling microbes, the diversity of energy sources employed, and the carbon fixation pathways sustaining primary production remains limited. Equally unresolved are the trophic interactions that structure microbial communities and underpin biogeochemical cycling in these complex systems.

Here we address knowledge gaps by investigating a marine cave at a depth of approximately 30 metres in Port Phillip Bay, Victoria, Australia. This site served as a natural system to explore the spatial transition of microbial metabolisms from the light-exposed, hydrodynamically active sediment at the cave entrance to the dark, stable cave interior. We hypothesized that diminishing light toward the inner cave limits phototroph abundance, whereas the reduced water flow of the interior cave facilitates the formation of stratified microbial communities, where interactions between anaerobic and aerobic metabolisms foster chemosynthetic processes. Bridging gene- and genome-resolved metagenomics with ex situ activity incubations and isotope geochemistry, we clarify the ecological strategies of cave-dwelling microbes, we quantify how inorganic electron donor oxidation and carbon fixation are partitioned across the cave, identifying spatial patterns in the distribution and prominence of metabolic pathways. Our results ultimately provide mechanistic insight into the chemosynthetic processes that enhance the primary productivity of the studied cave ecosystem and more broadly of aphotic environments located in the shallow ocean.

Materials and methods

Sample collection, processing, and physicochemical parameters

The study site was a granitic marine cave located at a maximum depth of ~32 m in Port Phillip Bay, Melbourne (38.293733° S, 144.625867° E; Fig. 1A, B). Sampling during SCUBA diving at this depth for approximately 30 minutes requires strict decompression procedures to ensure divers’ safety, making fieldwork logistically challenging. Additionally, as the cave lies near an active shipping channel, access is restricted to brief windows of less than 1 hour during slack water at the end of an ebb tide and in the absence of vessel traffic. These constraints inherently limited the achievable sample size. We collected triplicate sediment samples over two expeditions in November 2023 and October 2024 along a transect extending from the cave entrance (0 m) through the midsection (6.5 m) to the innermost accessible environment (13 m), and two seawater samples, one from each extremity of the cave (Fig. 1C). We use the collective term interior to refer to the cave midsection and innermost accessible environment. After collection, the samples were transported to Monash University. Samples for molecular, physicochemical, and stable isotope analyses were stored in a −20°C freezer until processing, whereas samples for respiration, trace gas consumption, nitrification, and radioisotope analyses were processed within a day. Sediment physicochemical parameters (pH, electrical conductivity, total carbon, total organic carbon, nitrate-N, ammonium-N, and sulfur) were measured at the Environmental Analysis Laboratory, Southern Cross University, NSW, Australia.

(A) Map of Port Phillip Bay indicating the location of the study site. (B) Schematic illustration of the surveyed marine cave, showing representative depths, transect across the entrance (0 m), mid (6.5 m), and inner (13 m) cave environments, and typical benthic assemblages, including cnidarians, sponges, and algae within and surrounding the cave. (C) Close-up of the measuring transect highlighting the cave floor substrate. (D) Community richness (Shannon index) of archaeal and bacterial communities in sediment and seawater, based on 16S rRNA gene data. (E) 16S rRNA gene copy number per gram wet weight (g ww) of sediment across cave environments. (F) PCA of beta diversity showing compositional differences in sediment and seawater microbial communities. (G) LEfSe analysis identifying archaeal and bacterial genera enriched in entrance, midsection, and inner cave environments, with linear discriminant analysis (LDA) scores indicating discriminative power. Only top 16 enriched genera per environment are shown, for a full list refer to Supp. Data S1. (H) Chlorophyll a concentration in sediments at the entrance and inner cave, indicating abundance of oxygenic phototrophs.

Community DNA extraction and sequencing

Genomic DNA was isolated from each sediment (n = 9) and sea water (n = 2) sample using the FastDNA Spin Kit for Soil (MP Biomedical, California, USA) according to the manufacturer’s protocol, and following guidelines for preventing contamination [29]. For sediment, 0.5 g of material was used per extraction; for seawater, DNA was isolated from biomass collected on 0.22 μm filters after filtering 1.5 L of seawater. A no-template extraction control was processed in parallel. DNA libraries were sent to the Genomics Node of the Monash Genomics & Bioinformatics Platform (Melbourne, Victoria) for library preparation and sequencing on a single lane of a NovaSeqX Plus System (Illumina) using XLEAP-SBS chemistry (2 × 150 bp).

Reads quality control, assembly, and binning

Shotgun metagenomic sequencing of libraries generated from three biological replicates per cave environment yielded an average of 58.8 ± 5.8 million reads for the cave entrance, 54.3 ± 5.1 million reads for the midsection, and 53.9 ± 2.5 million reads for the innermost region. Seawater samples generated 8.4 and 52.7 million reads, while the extraction control yielded a total of 4,308 reads. Raw sequencing data were processed through the Metaphor pipeline [30]. Quality control was carried out using fastp [31], which included the removal of adapters and low-quality sequences, as well as filtering by read length (≥50 bp) and average quality score (≥30), with automatic adapter detection enabled. To enhance genome recovery, particularly of low-abundance taxa, and minimize redundancy, reads from all libraries were co-assembled using MEGAHIT v1.2.9 [32] using default parameters. Assemblies were filtered to retain contigs ≥1000 bp. Genome binning was independently performed with four algorithms: Vamb v4.1.3 [33], MetaBAT v2.12.1 [34], CONCOCT v1.1.0 [35], and SemiBin v2.2.0 [36].

Bacterial and archaeal bin processing

Bins were consolidated using DAS Tool v1.1.6 [37] to generate a refined bin set. These bins were subsequently dereplicated with dRep v3.4.2 [38] at a 95% average nucleotide identity (ANI) threshold corresponding to species level, incorporating CheckM2 [39] metrics. Completeness and contamination were assessed with CheckM2 [39]. In total, we recovered 130 bacterial and archaeal metagenome-assembled genomes (MAGs) meeting the MIMAG criteria [40] for medium- (≥50% completeness; <10% contamination) and high-quality (≥90% completeness; <5% contamination). MAGs were taxonomically classified using Genome Taxonomy Database (GTDB)–Tk v2.3.2 [41] against the GTDB release R220 [42]. The coverage of each MAG across samples was calculated using CoverM v0.7.0 [43] in genome mode with method count. Similarly to previous studies that used abundance and read-mapping thresholds to discriminate habitat-specific taxa [44–46], we defined seawater-exclusive MAGs as those with less than 1,000 reads present in the sediment and/or with a 100-fold increase in the seawater relative to the sediment. The <1,000-read cutoff addresses mapping noise and low-coverage artefacts, whereas the ≥100-fold enrichment ensures robust habitat specificity. A total of five MAGs met these criteria. The maximum-likelihood phylogenetic tree of Archaea and Bacteria MAGs was built through PhyloPhlAn v3.1.1 [47] using the LG substitution matrix with diversity high, then plotted using iTOL v7 [48] and edited in Illustrator v24.0.2.

Metabolic annotation of metagenomic short reads and contigs

Adapter and barcode sequences were removed from raw reads, followed by the elimination of PhiX contaminants and low-quality sequences (minimum Phred score of 20) using the BBDuk utility within BBTools v36.92 (https://sourceforge.net/projects/bbmap). Forward reads that passed quality filtering and were ≥ 130 bp in length were screened for functional genes using DIAMOND blastx [49] against a curated and regularly updated database [50] comprising 62 metabolic marker genes. These genes represent key pathways involved in energy conservation, carbon fixation, phototrophy, and the cycling of hydrogen, CO, methane, sulfur, nitrogen, and iron. Search parameters were optimized for each gene, applying a minimum query coverage of 80% and identity thresholds as follows: 80% for psaA, 75% for hbsT (contigs: 65%), 70% for atpA (contigs: 60%), psbA (contigs: 60%), isoA, ygfK, and aro, 60% for amoA, mmoA, coxL, [FeFe]-hydrogenase, nxrA, rbcL, and nuoF, and 50% for all remaining genes (contigs: rho 30%; rdhA: 45%; cyc2: 35%). To estimate the relative abundance of community members harbouring each gene, read counts were normalized to reads per kilobase million as previously described [51, 52]. Binned contigs were subjected to open reading frame (ORF) prediction using Prodigal v2.6.3 [53]. Resulting ORFs were annotated via DIAMOND blastp [54] using the same reference database and identity thresholds as above, as well as in DRAM v1.5.0 [55].

Eukaryotic bins processing and metabolic annotation

Bins obtained from Vamb v4.1.3 [33], MetaBAT v2.12.1 [34], CONCOCT v1.1.0 [35], and SemiBin v2.2.0 [36] were filtered to a minimum size of 2.5 Mb, then contigs origin within each bin was estimated using Whokaryote v1.1.2 [56] and only bins containing 80% or more eukaryotic contigs were retained. Bin quality was assessed with Busco v5.8.3 [57] and those with 30% or higher completeness were retained. EUKulele v2.0.9 [58] was used to infer contigs taxonomy within each bin. Bins were taxonomically classified when ≥70% of constituent contigs shared the same assignment. FastANI v1.34 [59] was used to dereplicate the bins at species level with 98% ANI according to their taxonomy, which after this step were referred as eukaryotic metagenome-assembled genome (eMAG). After these steps we retained two eMAGs, and calculated their abundance across samples using CoverM v0.7.0 [43] in genome mode with method count. AUGUSTUS v3.5.0 [60] with species “chlamy2011” was used to predict ORFs on eukaryotic contigs and eggNOG-mapper v2.1.13 [61] was used to annotate them against eggNOG v6.0 database [62].

Small subunit rRNAs reconstruction, classification, and analyses

PhyloFlash v3.41 [63] was used to reconstruct small subunit (SSU) rRNA genes from quality filtered metagenomic short-reads generated using the BBDuk utility within BBTools v36.92 (https://sourceforge.net/projects/bbmap/), followed by targeted assembly of SSU rRNAs using SPAdes [64] with standard configuration within phyloFlash. Reconstructed full-length and partial SSU rRNA sequences were taxonomically classified against the SILVA 138.1 SSURef NR99 database formatted for phyloFlash [65]. Read mapping was performed to quantify the relative abundance of Archaea and Bacteria across samples, resulting in a 16S rRNA gene-based abundance table used for downstream community composition analyses.

Community richness (α-diversity) was quantified using the Shannon index, and differences among sediment samples were assessed with analysis of variance (ANOVA) followed by Tukey’s Honest Significant Difference test. Community similarity (β-diversity) across sediment and seawater samples was evaluated from centered log-ratio–transformed Euclidean distance matrices of the 16S rRNA gene abundance table, after removing sequences with ≤10 counts, and visualized by principal component analysis (PCA). To test for differences in community composition among sediment samples and between sediments and seawater samples, we first assessed homogeneity of group dispersions using the “betadisper” function in vegan [66] and then determined statistical significance of between-group differences with permutational multivariate analysis of variance implemented in the “adonis2” function in vegan [66]. Linear discriminant analysis effect size (LEfSe) [67], implemented in microbiomeMarker [68], was applied to identify differentially abundant 16S rRNA gene sequences among entrance, midsection, and inner cave sediments. LEfSe analysis was performed at the genus level on the 16S rRNA gene abundance table after log10 transformation, with genera exhibiting LDA scores >2 considered indicative of enrichment within cave environments.

The multisite metric zeta diversity (ζ) was used to assess incidence-based turnover in community composition (16S rRNA gene sequences) across the cave environments using the zetadiv R package [69]. Zeta decline was calculated with the function “zeta.decline.mc,” performing subsampling set to 1,000 on a Jaccard-normalized presence-absence 16S rRNA gene dataset for ζ orders ζ1 to ζ3. This approach captured community structuring across the cave environments, with increasing ζ values approaching zero. In this context, a power-law model assumes that ζ diversity declines with increasing order (i) according to a scaling relationship, consistent with structured species turnover and the persistence of widespread taxa across multiple sites. In contrast, an exponential model assumes a faster, constant-rate decline, indicating more stochastic or spatially independent community assembly processes. To determine which model best represented the observed pattern, we compared their Akaike Information Criterion (AIC) scores. AIC is a statistical measure that balances model fit and complexity; lower AIC values indicate a model that explains the data well avoiding overfitting. Thus, the model with the lowest AIC was considered the best descriptor of the relationship between ζ diversity and order i.

Ex situ biogeochemical measurements

Incubation assays were carried out to assess aerobic metabolic activity within the cave communities. Each experiment was performed in triplicate, using three independent cave samples per environment (n = 9). Control samples were sterilized via autoclaving at 121°C for 30 minutes, repeated twice. Elemental dynamics observed in live treatments were absent in sterilized controls, indicating that the measured changes were biologically driven.

Trace gas incubations were conducted in 120 ml glass serum vials containing approximately 10 g of sediment in 50 ml of autoclaved and 0.22 μm-filtered seawater, amended with 10 ppm of H_2_ and CO in the vial headspace. Gas sampling commenced immediately following the addition of electron donors, with 2 ml headspace gas extracted at defined time points. Gas concentrations were quantified via gas chromatography equipped with a pulsed discharge helium ionization detector (TGA-6791-W-4 U-2, Valco Instruments Co. Inc.), using certified calibration standards (0, 10, and 100 ppm in N_2_; BOC Australia).

Nitrification incubations were conducted in uncapped 250 ml Schott bottles containing approximately 10 g of sediment in 100 ml autoclaved and 0.22 μm-filtered seawater, and supplemented with 100 μM NH_4_Cl. At each timepoint, 10 ml aliquots were filtered through 0.45 μm pore-size membranes and stored at −20°C for subsequent analysis. Filtered samples were analysed for combined nitrite and nitrate (NO_x_^−^) concentrations using a Lachat QuikChem 8000 Flow Injection Analyzer (FIA), following standard APHA methods [70].

For measurements of oxygen consumption, incubations were prepared in sealed 120 ml glass serum vials containing approximately 10 g of cave sediment in 100 ml of autoclaved and 0.22 μm-filtered seawater, and maintained in the dark while shaking. Dissolved O_2_ concentrations were measured at 16 h 25 m, 42 h 35 m, 69 h 55 m, and 91 h 25 m using a FireSting oxygen probe (PyroScience).

Cell-specific oxidation rates were calculated for the samples used in the trace gas and nitrification incubations, by dividing bulk trace gas oxidation rates by the estimated abundance of organisms capable of each metabolism. Cell numbers were estimated from 16S rRNA gene copy numbers quantified by qPCR using a plasmid standard containing the 16S rRNA gene V4 region, adjusted by the average copy number of relevant metabolic marker genes per organism as estimated from short-read metagenomic data across the same samples, following a previously described protocol [52].

Radioisotope incorporation analysis

Cave sediment samples (0.5 g) with 1 ml of autoclaved and 0.22 μm-filtered seawater and an approximate concentration of 0.1 μM of radiolabeled sodium bicarbonate solution (NaH^14^CO_3_, Perkin Elmer, 53.1 mCi nmol^−1^) were prepared in 7 ml scintillation vials with ambient air headspaces. Six replicates of both entrance and inner cave sediments were prepared for five experimental conditions: light (40 μmol m^−2^ s^−1^), dark + H_2_ (100 ppm), dark + CO (100 ppm), dark + S^2−^ (800 μM), and dark + NH_4_^+^ (1 mM). Experiments were incubated for 3 days at room temperature, shaking at a constant rate (50–100 rpm). Following incubation, concentrated HCl was added dropwise to each vial and allowed to react for 24 h with intermittent shaking to ensure complete acidification and release of unbound dissolved inorganic carbon as ^14^CO_2_. Acid was added uniformly across all samples until effervescence ceased before 350 μl of 10 M NaOH was added to neutralize. Finally, 5.2 ml of scintillation cocktail (EcoLume, MP Biomedicals) was added and ^14^C activity was quantified using an automated liquid scintillation counter (Tri-Carb 2810 TR, PerkinElmer). Disintegration per minute values were then normalized against heat-killed controls, averaged among treatments and converted to ^14^C fixation rates (^14^C pmol day^−1^ g_ww_^−1^).

Stable carbon isotope measurements

Stable isotope analysis was conducted to investigate pathways contributing to organic matter formation in the cave environment. Approximately 1 g of sediment was dried at 60°C for 24 h, homogenized using a clean mortar and pestle, and further dried overnight at 60°C. Samples were weighed into 8 × 5 mm silver capsules and decarbonated by repeated treatment with 10% HCl, followed by drying at 60°C until effervescence ceased. Isotopic analyses were performed using an ANCA GSL2 elemental analyser coupled to a Hydra 20–22 continuous-flow isotope ratio mass spectrometer (CF-IRMS; Sercon Ltd., UK). Analytical precision for δ^13^C measurements was ±0.2‰. Internal standards (sucrose, gelatin, and bream) calibrated against international reference materials (USGS40, USGS41, IAEA C-6) were analysed alongside samples to ensure accuracy.

Chlorophyll a measurements

Chlorophyll was extracted by adding 9 ml of acetone to 5 g of cave sediment (four samples from the cave entrance and four from the inner cave), followed by vortex mixing and sonication in a high-power ultrasonic water bath for 5 minutes. Samples were then stored at 4°C overnight to facilitate pigment release. The following day, 1 ml of ultrapure water was added to each tube prior to centrifugation at 2,000 rpm for 10 minutes. From the supernatant, 3 ml of the clarified extract was transferred into a cuvette, and absorbance was recorded at 665 and 750 nm using an Eppendorf BioSpectrometer. Chlorophyll a concentrations were subsequently calculated using the following equation:

\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{upgreek} \usepackage{mathrsfs} \setlength{\oddsidemargin}{-69pt} \begin{document} \begin{align*} & Chlorophyll\ a\ (\mu g/g)\\&=[( nm655\ read- nm750\ read)\ x\ 11.4\ x\ 10)]/ weight\ (g) \end{align*}\end{document}

Statistics and visualization

Downstream statistical analyses were performed in RStudio (version 1.2.5033) using R packages ggplot2 [71] for drawing plots and charts then Illustrator v24.0.2 was used for figure editing, phyloseq [72], microbiomeMarker [68], and vegan [66] for community analysis, and ecolutils [73] for Levins index [74].

Results and discussion

Physicochemistry and microbial structure change across a marine cave transect

Cave physicochemical conditions were profiled across a transect extending from the entrance to the innermost accessible environment. As we moved deeper into the cave, light levels dropped sharply towards total darkness (Fig. 1C). In parallel, sediment grain size became finer (Fig. S1), suggesting lower hydrodynamic energy and transport from underwater currents. However, the cave water column remained fully oxygenated throughout the transect approaching 100% on air saturation (Fig. S2). Though we did not measure in situ concentrations of other gases, seawater flowing into the cave is most likely enriched in H_2_ and CO, which in Port Phillip Bay are supersaturated relative to the atmosphere with concentrations of 1.8 ± 0.26 nM and 8.5 ± 1.7 nM, respectively [7]. Sediment physicochemical analysis indicates a transition across the transect to a less alkaline (pH: 8.52 entrance; 8.41 inner) with lower organic carbon content (TOC: entrance 4.8%; inner 3.0%; Fig. S3) environment. Compared to other surficial marine sediments collected in Australian waters (TOC: 0.02 to 2%) [75], the sediments of the marine cave were organic carbon rich, likely reflecting a combination of endogenous chemosynthetic primary production and lateral transport of photosynthetically-derived organic matter, balanced by consumption by organoheterotrophic microorganisms.

We tested whether archaeal and bacterial communities in the sediment varied across different cave environments and in relation to the overlying seawater. The innermost cave had the highest microbial richness despite exhibiting the lowest 16S rRNA gene copy numbers (Fig. 1D, E). Microbial communities in the sediment differed significantly across the three cave environments and from the seawater (Fig. 1F). These observations (Fig. 1D–F), together with the clear distinction in physicochemical properties across the entrance, midsection, and inner cave (Fig. S3), suggest that deterministic processes may have been the leading force shaping the structure of the microbial communities colonising these environments. Consistently, *z-*diversity analyses [73] showed microbial turnover across the cave environments and community assemblages were best described by a power law regression (AIC: −13.43; Fig. S4), indicating that they are likely structured by environmental gradients and exhibit habitat specialisation.

Our inferences on the cave microbial community structure are reinforced by analysis of the relative abundance of microbial genera inhabiting the three cave environments (Fig. 1G; Supp. Data S1). Specifically, many taxa abundant in seawater were most abundant at the illuminated cave entrance, including diverse Cyanobacteria (Fig. 1G; Supp. Data S1), a result consistent with chlorophyll a quantification (Fig. 1H), and the highly abundant SAR11 clade (Fig. 1G; Supp. Data S1). Along the transect, the cave midsection showed the features of a transition zone with enrichment of taxa adapted to a broad range of redox conditions, from ammonia-oxidising bacteria such as Nitrosococcaceae [76], through facultative anaerobes including Reichenbachiella [77], to anammox bacteria “Ca. Scalindua” [78] (Fig. 1G; Supp. Data S1). Moving deeper into the darker cave, enriched taxa comprise genome-streamlined putative fermenters such as Woesearchaeales and “Ca. Magasanikbacteria” [79], as well as sulfide-cycling microorganisms such as Desulfoconvexum [80] and Sulfurimonas [81] (Fig. 1G; Supp. Data S1), which likely exploit localized anoxic or microoxic niches formed by micro-stratification of the inner cave sediments resulting from reduced hydrodynamic flow. Furthermore, the inner cave hosted taxa such as Deferrisoma, Calditrichaceae, ABY1 (Patescibacteria), and AT-s2–59 (Pseudomonadota), typically associated with deep-sea habitats [82–85] (Fig. 1G; Supp. Data S1).

Functional traits distinguish cave generalists, entrance specialists, and interior specialists

To characterize the diversity, niche breadth, and functional potential of cave-dwelling microbes, we employed genome-resolved metagenomics, recovering 132 genomes spanning 14 phyla (Fig. 2A; Supp. Data S2). Most of the MAGs affiliated with bacteria, especially the Pseudomonadota (59% MAGs; Fig. 2A), with one archaeal MAG (ammonia-oxidizing Nitrosopumilus; Fig. 2A) and two eMAGs (photosynthetic Chlorophyta) also retrieved (Supp. Data S2). We used Levins index [74] to categorize each taxa into specialist, intermediate, or generalist based on their habitat distributions across the cave environments (Fig. 2A). Of the MAGs recovered from the sediment, 50% were classified as generalists found ubiquitously across the cave systems, whereas 28.8% of the specialist population colonized the lit cave entrance, and 10.4% the interior (Supp. Data S2). The habitat generalists may have flexible strategies to adapt to the varying physicochemical gradients within the cave ecosystem. Aligning with this concept, habitat generalists had larger genomes than specialists after normalising for genome completeness (4.6 ± 1.5 Mb vs 3.7 ± 2.2 Mb; Fig. 2B) and substantially more genes (4,744 ± 1,466 vs 3,858 ± 2,360; Fig. 2B), likely reflecting a broader repertoire of metabolic pathways and regulatory mechanisms. When examined separately, specialists from the cave entrance displayed reduced genome sizes (3.2 ± 2.2 Mb vs 5.1 ± 1.7 Mb; Fig. 2C) and gene repertoires relative to those inhabiting the cave interior (3,254 ± 2,245 vs 5,530 ± 1,861; Fig. 2C). This pattern is consistent with genome streamlining, a strategy associated with oligotrophic, light-exposed marine environments [86], such as the cave entrance, where selective pressures favour genomes that retain only the core metabolic functions.

(A) MAGs maximum-likelihood phylogenetic tree inferred using LG substitution matrix, representing the topography, metabolic repertoires, and niche breadth of 130 archaeal and bacterial MAGs encompassing 13 phyla. Clades are coloured according to phylum-level taxonomy and symbols indicate the potential of each MAG to mediate key energy and carbon acquisition processes. (B) Bar graphs showing genome size (mb) and gene count of generalists, intermediates, and specialists MAGs. Analysis of variance (ANOVA) shows statistical differences among generalists, intermediates, and specialists MAGs genome size and gene count. (C) Heatmap showing the metabolic functions of cave sediment MAGs and eMAGs identified as generalists, intermediates, and specialists.

The cave habitat generalists were predicted to be metabolically flexible microbes capable of using a wide range of energy sources, carbon sources, and electron acceptors (Supp. Data S2). Organoheterotrophy was pervasive, as highlighted by the diverse repertoire of carbohydrate-active enzymes (CAZymes), especially those involved in the breakdown of complex polysaccharides such as cellulose, chitin, and pectin, originating from plant, algal, and fungal sources (Supp. Data S2). This indicates that laterally transported organic carbon plays an important role in sustaining habitat generalists within the cave. In addition to using polysaccharides, some generalists were predicted to use one-carbon compounds, as highlighted by the recovery of a putative methanotrophic Rhizobiaceae genome (Fig. 2A; Supp. Data S2). This Rhizobiaceae genome encodes an XmoA-type copper-containing monooxygenase, but forms a novel cluster distinct from characterized methane, ammonia, and hydrocarbon monooxygenases and thus its substrate is unknown (Fig. S5). Autotrophy emerged as a subsidiary trait among generalists and present in only 4.7% of this group, encompassing both photosynthetic and chemosynthetic microbes. The habitat generalist alga Pycnococcus was abundant throughout the cave, likely owing to recurrent deposition from seawater inflow where it is also common (Supp. Data S2), as well as its capacity to persist under low-light conditions, potentially supported by mixotrophic strategies such as heterotrophic carbohydrate scavenging (Supp. Data S2). Equally abundant were two chemoautotrophs classified as Methyloligellaceae and Woeseiaceae carrying genes for CO_2_ fixation via the Calvin–Benson–Bassham (CBB) cycle, supported by electrons harvested from the oxidation of S^2−^ / S_2_O_3_^2−^, and H_2_ respectively (Supp. Data S2). The capacity for lithoheterotrophy was also common, with more than 64% of generalist MAGs carrying genes for the oxidation of CO, H_2_, S^2−^, S_2_O_3_^2−^, and arsenite (AsO_3_^−3^; Fig. 2A; Supp. Data S2), with the capacity for H_2_ and S^2−^ oxidation 2.4- and 2.2-fold enriched in generalists relative to specialists (Fig. 2D; Supp. Data S2). Most generalists were predicted to be capable of both aerobic respiration (93% MAGs) and anaerobic respiration (80% MAGs), suggesting most are facultative anaerobes adapted to the spatiotemporal variations in oxygen availability in cave sediments. Particularly widespread was the capacity to use nitrate, nitrite, nitric oxide, nitrous oxide, arsenate, and organohalides as electron acceptors (Supp. Data S2). Fermentative metabolism was also ubiquitous, especially for H_2_ and acetate production (Fig. 2D; Supp. Data S2). Collectively, these findings indicate that cave habitat generalists can use a broad range of resources to thrive across the cave system despite sharp redox and resource gradients.

The habitat specialists inhabiting the lit cave entrance differed in their metabolic capabilities compared with those adapted to the cave interior. For instance, specialist facultative phototrophs including members of the Rhodobacteraceae, Porticoccaceae, Pirellulaceae, Halieaceae, RS24 (Alphaproteobacteria), and TMED25 (Alphaproteobacteria), as well as the alga Picochlorum, were most abundant at the entrance (Fig. 2D; Supp. Data S2), likely selected by increased light availability. The lit cave entrance also harboured specialists capable of CO, H_2_, H_2_S, and S_2_O_3_^2−^ oxidation, including Geminicoccaceae, Rhodobacteraceae, Actinomarinaceae, Endozoicomonadaceae, and the gammaproteobacterial AqS2 (Fig. 2D; Supp. Data S2). However, these metabolisms were considerably more enriched in specialists inhabiting the cave interior, such as the Cyclobacteriaceae, Aestuariivirgaceae, Hyphomicrobiaceae, Flavobacteriaceae and other unresolved taxa within the Pseudomonadota (Fig. 2D; Supp. Data S2). Oxidation of inorganic compounds can also sustain autotrophic carbon fixation. Consistent with this, members of the Desulfocapsaceae were abundant in the inner cave, where genomic evidence suggests the capacity to couple sulfur and gas derived energy metabolism with Wood–Ljungdahl pathway, enabling CO_2_ fixation while simultaneously maintaining redox balance under energy-limited conditions (Supp. Data S2). The cave interior likely provides favorable environments for lithotrophic activity, where reduced hydrodynamic disturbance limits sediment mixing, promotes redox stratification, and fosters metabolic interactions in which anaerobes supply byproducts that fuel lithotrophs. The latter is reflected by the higher prevalence of genes associated with anaerobic metabolisms in interior specialists compared to entrance specialists; for instance sulfate reduction was exclusively encoded by this group, whereas denitrification, arsenate reduction, and organohalide reduction pathways were considerably enriched relative to the cave entrance (Fig. 2D; Supp. Data S2). Altogether, these results indicate that specific functional guilds may exhibit niche specialisation around spatially restricted energy sources.

Inner cave environment enriched in chemosynthetic communities

To test our hypothesis that chemosynthesis underpins cave microbial communities, we used gene-centric metagenomics to resolve how microbial metabolism shifted across the transect. Consistent with the physicochemical profiles and 16S rRNA gene analyses, the relative abundance of genes mediating CO_2_ fixation, trace gas metabolism, phototrophy, fermentation, nitrogen and sulfur cycling significantly varied across the cave environments (Fig. 3A). Genes for photosystem- and rhodopsin-based light harvesting were present throughout the cave but decreased substantially in the inner cave, with photosystem II genes (psbA) declining by 76-fold (Fig. 3A; Supp. Data S3). In contrast, the inner cave environment hosted higher abundance of genes for the oxidation of inorganic compounds (Fig. 3A). [NiFe]-hydrogenases were particularly enriched (entrance: 25.0 ± 12.1%, inner: 85.2 ± 16.6%; Fig. 3A), suggesting H_2_ oxidation is a key process. Global surface oceans are supersaturated with H_2_ [16, 87, 88], providing a potential source to support hydrogenotrophic metabolisms in the cave ecosystem. However, a portion of the H_2_ oxidized within the cave likely originates from in situ biotic production via fermentative pathways and nitrogen fixation, with the genes encoding fermentative [FeFe]-hydrogenases, group 3 [NiFe]-hydrogenases, and nitrogenases also enriched in the inner cave (Fig. 3A). CO oxidation also emerged as a key metabolism, with anaerobic CO dehydrogenases (cooS) increasing 20.8-fold in the inner cave, whereas aerobic CO dehydrogenases (coxL) remaining relatively stable (Fig. 3A). Consistent with the likely presence of redox gradients across the cave, the inner cave was enriched with genes mediating anaerobic respiration, including key denitrification steps (narG, napA), dissimilatory nitrate reduction to ammonium (nrfA), and dissimilatory sulfate reduction (dsrA). Byproducts of these pathways, such as NO_2_^−^, NH_4_^+^, and S^2−^ from hypoxic niches, likely fuel chemosynthetic activity. Accordingly, genes for NO_2_^−^ (nxrA), and S_2_O_3_^2−^ (soxB) oxidation were detected across the whole cave (Fig. 3A), whereas S^2−^ (sqr) and NH_4_^+^ (amoA) oxidation genes were enriched in the inner cave (2.6- and 2.0-fold; Fig. 3A).

(A) Relative abundance of genes encoding enzymes for trace gas metabolism, fermentation, sulfur and nitrogen cycling, phototrophy, and carbon fixation across distinct cave environments. Significant differences in gene relative abundance across the cave environments were assessed using categorical models (ANOVA or Kruskal–Wallis test) with post hoc pairwise comparisons (Dunn’s tests) with false discovery rate (Benjamini–Hochberg) correction. We also plotted regression lines across the transect distance (0, 6.5, and 13 m), which highlight the overall spatial trends. Shaded grey areas show 95% confidence interval of the regression lines. (B) Linear regressions showing correlations between the abundance of genes associated with energy metabolism (explanatory variable, x-axis) and those involved in carbon fixation pathways (response variable, y-axis) across the cave environments.

Potential for inorganic carbon incorporation through autotrophic pathways was high across the whole cave, but especially in the inner environment. Marker genes for all CO_2_ fixation pathways except the rTCA cycle were enriched in the inner cave (Supp. Data S3). The CBB (rbcL 9.8 ± 2.6% of the community), Wood-Ljungdahl pathway (acsB 3.7 ± 5.5%), and 4-hydroxybutyrate pathway (hbsT 2.9 ± 1.2%) were the dominant CO_2_ fixation processes and exhibited on average a 1.7-, 16.1-, and 2.4-fold increase between the entrance and the inner cave (Fig. 3A; Supp. Data S3), respectively. Enrichment of carbon fixation pathways in the inner cave likely reflects that cave microbes are more limited by lateral organic carbon inputs, but also potentially have an elevated supply of inorganic energy sources, in the likely redox-stratified sediments. These data also highlight parallels between the inner cave and the deep ocean, where autotrophic archaea are similarly enriched [89]. Linear correlation analyses demonstrated that CO_2_ fixation genes were most strongly associated with those for inorganic compound oxidation rather than light capture, i.e. chemosynthesis not photosynthesis (Fig. 3B). Specifically, H_2_ and to a lesser extent S^2−^ oxidation genes were most closely associated with CBB cycle genes (Fig. 3B), whereas CO and NH_4_^+^ oxidation genes were most associated with the Wood-Ljungdahl pathway and 4-hydroxybutyrate pathway (Fig. 3B), respectively. These inferences are consistent with the MAG-based predictions (Fig. 2A) and wider culture-based studies on chemosynthetic microorganisms [90–92].

Chemosynthetic metabolisms are active across cave environments

To assess the activities of cave microbes, we measured rates of inorganic substrate consumption, photosynthetic and chemosynthetic carbon fixation (Fig. 4A–C), and aerobic respiration (Fig. 4D) through ex situ assays using sediment collected from the cave entrance and inner environments. In line with the genomic predictions, microbial communities from both locations consumed oxygen, with the inner cave mediating 1.4-fold higher bulk aerobic respiration rates (entrance: 217 O_2_ nmol L^−1^ h^−1^; inner: 302 O_2_ nmol L^−1^ h^−1^; Fig. 4D; Supp. Data S4). Aerobic respiration is likely primarily driven by a combination of organic carbon derived from both in situ chemosynthetic primary production and laterally transported allochthonous substrates. Cave communities also consumed the inorganic energy sources H_2_ and CO aerobically, consistent with the genomic predictions several bacteria including within phyla Pseudomonadota and Actinomycetota can aerobically use these gases (Fig. 4A). NH_4_^+^ oxidation also occurred at both extremities of the cave, likely reflecting activity of the archaeal Nitrosopumilus (Fig. 4A). Although bulk consumption rates were similar between the two environments, normalization against cell abundance revealed the cave interior communities mediated significantly higher cellular rates of H_2_ and CO oxidation (2 and 1.8-fold higher than the entrance, respectively), whereas cellular nitrification rates were slightly lower than at the entrance (1.4-fold; Fig. 4B; Supp. Data S4). This pattern suggests that active and diverse chemosynthetic communities reside in the inner cave.

(A) Biogeochemical assays illustrating aerobic metabolic activity of microbial communities from the cave entrance and inner environments, based on the consumption of H2, CO, and NH4+. Heat-killed controls were included to confirm the biotic origin of the observed processes. (B) Oxidation rates of H2, CO, and NH4+ normalized per-cell specialized in each metabolism across entrance and inner cave environments. Shaded grey areas show 95% confidence interval of the regression lines. (C) Aerobic 14CO2 incorporation by sediment community incubated under light (40 μmol m−2 s−1) and in the dark with CO (100 ppm), H2 (100 ppm), NH4+ (1 mM), and S2− (0.8 mM). (D) Oxygen consumption rates of sediment from the entrance and inner cave environments. (A, B) measurements were performed in 120 ml sealed serum vials containing ~10 g of sediment and 50 ml (trace gasses) or 100 ml (O2 consumption) of 0.22 μm-filtered seawater. Trace gas incubations were supplemented with 10 ppm H2 and CO in the vial headspace. Nitrification assays (NOx = NO2− + NO3−) were conducted in uncapped 250 ml Schott bottles containing ~10 g of sediment, 100 ml of 0.22 μm-filtered seawater, and 100 μM NH4+.

We complemented consumption assays with measurements of chemosynthetic and photosynthetic carbon fixation rates using radioisotope incorporation (^14^CO_2_). Incorporation of ^14^CO_2_ was consistently detected in all incubations, but as expected for sandy sediment collected at mesophotic depths [93], photoautotrophy was minimal (Fig. 4C). At the cave entrance, chemosynthetic ^14^CO_2_ incorporation rates were low overall (14.1 ± 6.7 pmol^−1^ d^−1^ g_ww_^−1^) with modest stimulation by S^2−^ (19.2 ± 5.4 pmol^−1^ d^−1^ g_ww_^−1^) relative to other treatments (Fig. 4C; Supp. Data S3). In contrast, inorganic energy sources enhanced ^14^CO_2_ incorporation in the inner cave (Fig. 4C; Supp. Data S3). For instance, CO-supplemented incubations supported the highest ^14^CO_2_ incorporation with average rates 2-fold higher relative to the cave entrance (Fig. 4C). Likewise, H_2_-supplemented incubations incorporated on average 2.6-fold more ^14^CO_2_ in the inner cave compared to the entrance, consistent with the higher proportion of oxidative groups 1 and 2 [NiFe]-hydrogenases (Supp. Data S3). The substantial stimulation by H_2_ in the inner cave aligns with observations from other subsurface and oligotrophic environments showing H_2_ is a key electron donor sustaining communities under nutrient- and light-limited conditions [94, 95]. Conversely, NH_4_^+^ or S^2−^ did not stimulate ^14^CO_2_ incorporation, indicating that these compounds are either minimally used as electron donors by the communities or were already available in sufficient concentrations within the sampled sediment porewater (Fig. 4C). We note that our experiments likely underestimated the gross carbon fixation potential of the community due to the presence of abundant dissolved inorganic carbon in seawater and carbonates in the cave samples (Fig. S1), and the use of ^14^C-labelled CO_2_, which is discriminated against by microbial enzymes in favour of lighter, naturally occurring ^12^C.

The isotopic composition of bulk organic carbon (δ^13^C) was similar between the entrance (−22.5‰) and inner cave (−22.1‰) samples, though there is a slight enrichment in the inner cave (Supp. Data S4). These δ^13^C values suggest that organic matter in the cave entrance and inner environments was predominantly derived from CBB-based carbon fixation [51, 96–99]. Since CBB cycle is the most abundant carbon fixation pathway in both cave photoautotrophs and chemolithoautotrophs, the exact contributions of photosynthesis and chemosynthesis to organic carbon supply in caves remain to be resolved, as does the relative contribution of allochthonous and autochthonous sources.

Conclusions

Our findings demonstrate that even a relatively small marine cave harbours pronounced physicochemical gradients that structure microbial communities adapted to local conditions. Metabolically flexible generalists co-exist throughout the cave alongside specialists occupying well-defined niches shaped by resource availability; entrance communities rely on light-derived energy, either directly via phototrophy or indirectly through degradation of organic carbon (including phytodetritus), whereas interior specialists are increasingly dependent on inorganic compounds (either aerobically or anaerobically) for energy conservation and carbon fixation. The inner cave’s constant darkness, limited organic carbon inputs, reduced water flow, and redox gradients create unique pressures and a wide variety of niches for microorganisms. These conditions are ideal for chemosynthetic microbes that use inorganic compounds supplied to the cave or produced autochthonously by other microbes (e.g. fermenters and sulfate reducers). Concomitant with decreased illumination, the progressively attenuated hydrodynamic flow along the cave transect likely promotes localized biogeochemical cycling that supports diverse metabolic processes. In particular, the reduced turbulence of the inner cave appears to facilitate the formation of stratified microbial communities, where chemosynthetic processes are fostered by compounds recycling between anaerobic and aerobic metabolisms. These patterns likely reflect ecological dynamics that have shaped the cave environments over time.

Our findings highlight that chemosynthetically-supported ecosystems are not confined to deep or extreme environments, but can form in aphotic habitats within the euphotic zone. Comparable gradients of light and hydrodynamics occur in many other marine settings, including ledges and crevices, but also artificial structures such as shipwrecks, solar farms, and piers, where chemosynthesis is also likely to play an important role in structuring ecological interactions and biogeochemical processes. Altogether, these results support emerging paradigms that chemosynthesis underpins a much larger fraction of oceanic productivity than was previously thought and reveals substantial hidden productivity in cryptic and littoral habitats.

Supplementary Material

wraf286_Supplemental_Files