Ammonia oxidation and recalcitrant carbon degradation fuel mixotrophic growth in the symbiont community of a deep-sea sponge

Abstract

Background

Sponges are important members of shallow-water, benthic ecosystems, where they often rely on their microbial symbionts to acquire organic or inorganic carbon. Sponges are also found in the deep sea, however, how they metabolically interact there with their symbionts remains underexplored. Here, we combined metagenomic, metatranscriptomic and stable-isotope labelling approaches to investigate the metabolic activities of the microbial community of the deep-sea sponge Calyx sp.

Results

Approximately 84% of the total estimated microbial abundance was composed of nine heterotrophic phyla, whilst the remaining 16% consisted of two autotrophic ammonia-oxidising archaea. Metatranscriptomic analysis revealed the high expression of genes involved in the degradation of recalcitrant polysaccharides of algal origin, suggesting that an undegraded fraction of marine snow plays a role in the nutrition of this deep-sea holobiont. Additionally, we detected active ammonia oxidation and carbon fixation pathways in the autotrophic community members and, through ex situ incubations with labelled carbonate show a potential to fix 13.67 mg CO₂ per g dry weight in a year.

Conclusions

This study highlights the mixotrophic lifestyle of a deep-sea sponge microbiome, expanding our knowledge of the sponge–microbe symbiosis in the oligotrophic environment of the deep ocean.

Download video file ^{(45.6MB, mp4)}

Video Abstract

Graphical Abstract

Supplementary Information

The online version contains supplementary material available at 10.1186/s40168-026-02411-y.

Introduction

Sponges (Porifera) are basal metazoans [1] that play a foundational role in marine benthic ecosystems due to their capacity for converting dissolved organic matter (DOM) into particulate organic matter (POM), which in turn can be consumed by other trophic levels in oligotrophic environments [2–4]. Sponges also provide an array of microhabitats within their bodies that allow diverse communities of microbial symbionts to thrive [5, 6]. These relationships can be beneficial to the sponge host, as the microbial symbionts can aid, for example, with nutrient acquisition and defence against pathogens [7, 8].

Deep-sea sponge grounds represent one of the largest biogenic habitats on the planet, with some aggregations spanning thousands of square kilometres and contributing significantly to benthic biodiversity and ecosystem functioning [9, 10], and they have been identified as Vulnerable Marine Ecosystems by the United Nations due to their ecological importance and sensitivity to anthropogenic disturbances. Importantly, deep-sea sponges can process vast quantities of seawater, with individual sponges filtering up to several thousand litres per day, thereby playing a crucial role in benthic–pelagic coupling and organic matter cycling in the deep ocean [11].

Shallow-water sponges largely rely on filter feeding to obtain their organic carbon and nutrients, whilst microbial symbionts contribute to the uptake of DOM and provide essential cofactors and vitamins [12]. Shallow-water sponges may also host autotrophic symbionts, such as cyanobacteria or diatoms, which can contribute to their overall biomass production by providing organic compounds through photosynthesis [13, 14]. In contrast, deep-sea sponges must adapt to life in a dark, oligotrophic environment, where organic compounds are often recalcitrant and limited [15, 16]. To partially compensate for the lack of photosynthesis, some deep-sea sponges appear to have formed successful symbioses with autotrophic ammonia-oxidising archaea (AOA) of the family Nitrosopumilaceae [17, 18], which can generate energy and reductive power for carbon fixation by oxidising ammonia to nitric oxide or nitrite [19, 20], and can be dominated by a single AOA phylotype that can constitute up to 95% of the symbiont community [18, 21]. However, other deep-sea sponges contain more complex microbiomes that comprise a diverse array of heterotrophic symbionts, including members of the phyla Pseudomonadota, Chloroflexota, Acidobacteriota, Desulfobacterota and Spirochaetota [22–25].

Omics-based analyses of metabolic interactions between symbiont and shallow-water sponges have revealed a number of redox pathways for energy and biomass generation in the holobiont, including the capacity to degrade a range of sponge-derived and environmental carbon sources, including taurine, creatine and coral- and macroalgae-derived DOM [26–29]. However, information on such interactions in deep-sea sponges is still scarce, with some work suggesting methane oxidation [30], nitrogen cycling [31] and ammonia-based chemolithoautotrophy [19]. While the genomic potential for carbon fixation has been detected in the microbiomes of several deep-sea sponges [17–19, 21, 22], the mechanistic basis of autotrophy in these systems remains poorly understood. Prior work [17, 19, 21] has largely been limited to the identification of marker genes or pathway inventories in single-symbiont-dominated systems, without quantitative assessment of the autotrophic contribution to holobiont biomass or resolution of which carbon fixation pathways are active and in which taxa. A mechanistic, multi-pathway characterisation of autotrophy in a complex deep-sea sponge microbiome has therefore not been reported.

To address this, we characterised the microbial community of the deep-sea sponge Calyx sp., for which information so far has been scarce and which represents an interesting system for studying deep-sea sponge–microbe interactions. Unlike many other deep-sea sponges that are dominated by a single AOA phylotype, Calyx sp. maintains a complex community of both autotrophic and heterotrophic symbionts, which allowed us to analyse the contribution of organic and inorganic carbon to the holobiont’s nutritional requirements. Here, we investigated the degradation of carbohydrates and hydrocarbons as well as carbon fixation pathways through metagenome sequencing, genome reconstruction and metatranscriptomics. Through ex situ isotope labelling experiments, we further determined the quantitative contribution that autotrophy made to biomass production in the Calyx sp. holobiont.

Materials and methods

Sponge sampling and identification

Three Calyx sp. individuals were collected during a deep-sea expedition at the Campos Basin, near Rio de Janeiro, Brazil, in September 2021, according to previously described methodology [18, 32]. Briefly, sampling was performed at approximately 827 m depth using a remotely operated underwater vehicle (ROV) equipped with individual sampling containers to minimize cross-contaminations of samples, which were sealed but allowed for pressure equilibration during ascent. Upon surfacing, subsamples of each sponge were cut with sterile equipment and washed three times with sterile Ca- and Mg-free seawater to remove loosely associated microorganisms. Samples were then immediately frozen in liquid nitrogen.

Sponges were identified as previously described using phylogenetic analysis (Supplementary Fig. S1) [18]. One of the specimens was subsampled at its base (hereafter S38), middle (hereafter S39) and close to the osculum (hereafter S40). Another individual was subsampled close to the base (hereafter S10) and close to the osculum (hereafter S92). Finally, the third individual was only subsampled in the middle (hereafter S43) and subsequently maintained in aquaria installed on board the vessel and used for the isotope labelling experiment described below. Subsamples were immediately snap-frozen on board and stored at − 80 °C. All six subsamples were subjected to 16S rRNA gene sequencing analysis. S38 and S40 were selected for metagenomic shotgun sequencing due to the quality of their extracted DNA and as they covered all the taxa found in the other samples in the amplicon sequencing results. The metagenomic reads were pooled prior to assembly (see metagenomic analysis section below). All samples were subjected to metatranscriptomic analysis, except for S92, which failed our quality check (Supplementary Table S1).

Water nutrient analysis

Water samples were collected during ROV dives and transferred to containers. Amber glass containers were used to store samples for hydrocarbon analysis. For metal analysis, samples were acidified to pH < 2 with HNO₃, while samples were kept at 4 °C and stabilised with 20 mM Zn-acetate/10 mM NaOH for sulphur analysis. Total hydrocarbons were extracted with n-hexane (as per EPA SW-846 8015D) and quantified with a 5977B Series Gas Chromatographer (Agilent, Santa Clara, CA, USA) with flame ionisation. Total Cu, Fe and Mg were measured by inductively coupled plasma optical emission spectroscopy (SMEWW Method 3120 B). Nitrate, sulphate and sulphite concentrations were determined using 0.45 µm-filtered aliquots: nitrate using cadmium-reduction/Griess with 540 nm spectrophotometry (EPA-NERL 353.3—nitrite corrected), sulphate turbidimetrically via BaSO₄ at 420 nm (EPA-NERL 375.4) and sulphite with the pararosaniline–formaldehyde method at 548 nm (OAR 340–202-0070).

DNA extraction, 16S rRNA gene sequencing and analysis

DNA was extracted from aliquots of the frozen sponge samples using the DNeasy PowerSoil Pro Kit (Qiagen, Hilden, Germany) according to the manufacturer’s instructions. The V4 hypervariable region of the 16S rRNA gene was amplified using the primer pairs 515 F (Parada) and 806R (Apprill) [33, 34] and sequenced according to the manufacturer’s instructions and previous procedure [18] on a MiSeq platform (Illumina) at the Ramaciotti Centre for Genomics (University of New South Wales, Sydney, Australia).

Raw sequences were quality-filtered using Trimmomatic (version 0.38) in a sliding window of 4 bp and reads with a quality score below 30 or shorter than 100 bp were discarded [35, 36]. Paired-end reads were merged using USEARCH (version 11.0.667) [37] with a minimum 8 bp overlap and sequences shorter than 250 bp or longer than 300 bp were removed. The UNOISE3 algorithm [37] was used to cluster, denoise and generate sequences for amplicon sequence variants (ASVs). Chimeras were removed using the UCHIME3 and the SILVA database version 138.1 in the high-confidence mode [37]. Taxonomy was assigned to the ASV sequences using Bayesian Lowest Common Ancestor (BLCA) [38] (identity and coverage intervals set to 95–100%, respectively) against a filtered version of the Genome Taxonomy Database (GTDB version 214), from which all sequences shorter than 1000 bp were removed [39].

All non-bacterial and non-archaeal ASV sequences were removed, as well as samples with less than 5000 reads (Supplementary Table S2). Reads from the samples kept in the aquarium (see below) were subsampled using SeqKit [40] (seqkit sample -p 0.15, seed 11) and normalised for read counts using the DESeq2 package [41]. To identify which bacterial ASVs were affected by incubation in the aquarium (see below), we used a one-factor design with ‘incubation time’ as a fixed factor. Sample sizes were as follows: field samples (n = 6), 48 h acclimatation (n = 4) and 3-months acclimatation + 48 h (n = 3). A generalised linear model (GLM) was fitted separately to each ASV using the mvabund package [42]. Negative binomial regression was specified (family = ‘negative.binomial’) with ‘composition = true’ to account for the compositional nature of the data. All assumptions for the model were verified and met. An adjusted (Benjamin–Hochberg correction) p-value < 0.05 was considered significant.

Microbial cell enrichment, shotgun DNA sequencing and metagenomic analysis

Microbial cells were enriched from frozen sponge tissue using a modification of a previous protocol [43]. Briefly, approximately 20 g of Calyx sp. tissue was homogenised in 20 mL of Ca- and Mg-free seawater (CMFSW) using a bench blender (250 mL Triton® cup, Hamilton Beach, Glen Allen, VA, USA). The resulting homogenate was filtered through a 100 µm sterile filter (MF-Millipore, Carrigtwohill, Ireland) and centrifuged twice at 300 × g for 15 min at 4 °C to remove sponge tissue. The supernatant was sequentially filtered through 20 µm and 3 µm filters (MF-Millipore) to remove remaining sponge cells and then centrifuged at 15,000 × g for 15 min at 4 °C.

DNA was extracted from the resulting cell pellet using the DNeasy PowerSoil Pro Kit (Qiagen), following the manufacturer’s protocol with a modification in the FastPrep step (MP Biomedicals, Irvine, CA, USA), where an intensity of 4 for 30 s was applied. DNA yield was quantified using a Qubit 3 fluorometer (Thermo Fisher Scientific, Waltham, MA, USA), and purity was evaluated with a Nanodrop 1000 spectrophotometer (Thermo Fisher Scientific). The DNA quality was further analysed using a TapeStation (Agilent).

Library preparation and DNA sequencing were performed at the Ramaciotti Centre for Genomics on a Nextseq 500 platform with 2 × 150 bp chemistry (Illumina) according to the manufacturer’s instructions. Reads were checked for adaptors using Trimmomatic v. 0.39 [35]. S38 and S40 were initially assembled using SPAdes [44] (v. 3.5.15, –meta parameters: -k 21,33,55,77; -k 21,33,55,77,99; -k 21,33,55,77,99,127). For each assembly, metagenome-assembled genomes (MAGs) were binned using Metabat2 v2.12.1 [45] with default settings, Maxbin v2.2.7 [46] with default settings and Semibin v1.4.0 [47] with the “single_easy_bin –training-type self” setting. Metawrap v1.3.2 [48] was used to select MAGs with completeness > 50% and contamination < 10%, and bins from the two assemblies were dereplicated using dRep v3.4.3 [49]. Bins were then improved using iterative read mapping and reassembly, and MarkerMAG v1.1.28 [50] was used to identify unbinned 16S rRNA genes belonging to the MAGs for incorporation to the final bins.

Taxonomy was assigned to the MAGs using GTDK-Tk v2.3.0 [51] based on the GTDB v214. MAG completeness and contamination were assessed using CheckM [52]. The relative abundance of each MAG was calculated based on the number of mapped reads as a proportion of the total number of binned reads. Functional annotation of the MAGs was performed by running similarity searches with Diamond v2.1.4 [53] against the Cluster of Orthologous Groups of proteins (COG) [54], the Kyoto Encyclopedia of Genes and Genomes (KEGG) [55] and the CAZY database (release July 2024) [56]. BLAST v2.0.5 [57] was used for searches against the UniProt database (release 032023) [58]. For carbon fixation pathway analysis, we used hidden Markov models to find key enzymes of each pathway and then inferred the completeness of the pathway by using gapseq [59, 60]. The reverse tricarboxylic acid (rTCA) cycle, the reductive acetyl-CoA pathway (Wood–Ljungdahl, WLP) and the reductive glycine pathway were not considered in our present analysis as they cannot be inferred solely by bioinformatics analyses [61, 62].

RNA extraction and gene expression analysis

Total RNA was extracted from 1.0 g of snap-frozen sponge tissue using Trizol and bead-beating and then purified with a PureLink RNA Mini kit (Thermo Fisher Scientific), which includes a DNase treatment. PCR amplification of the 16S rRNA gene using the universal primers 515F/806R with DNase-treated RNA as a template (without reverse transcription) was performed to ensure no genomic DNA was left. rRNA was depleted using the Ribopools kit with a mix of eukaryote and prokaryote probes (SiTools Biotech GmbH, Planegg, Germany) according to the manufacturer’s instructions. Samples were followed by Illumina NextSeq library construction. RNA extracted from the samples collected in the field or used in the carbon fixation experiment was sequenced on an Illumina NovaSeq 6000 system or NovaSeq X Plus system, respectively, with the same 2 × 150 bp chemistry (Illumina) at the Ramaciotti Centre for Genomics. Sample S92 was excluded from metatranscriptomic analysis due to insufficient RNA quality after extraction, which did not meet the thresholds required for rRNA depletion and library preparation.

Raw reads were quality-filtered using Trimmomatic v.0.38 [35] and merged using PEAR v.0.9.11 [63]. SortMeRNA v.4.3.3 [64] was used to remove rRNA reads and the remaining reads were mapped back against the obtained MAGs using Bowtie2 v.2.4.5 [65]. Expression of the coding sequences for each MAG was estimated with HTSeq v.2.0.2 [66] and gene counts were normalised based on gene length. Data was normalised using DeSeq2 [41] and expressed as normalised counts. To account for possible batch effects due to the different sequencing runs, we normalised the total counts by the total number of reads for each sample and divided the result by 10⁶ to obtain counts per million of transcripts (CPM). For differential gene analysis using CPM, we performed a Mann–Whitney U test, and p-values were adjusted using the false discovery rate method (FDR). Adjusted p-values < 0.05 were considered significant.

Sponge maintenance and stable-isotope incubations

Calyx sp. was maintained in a custom-built tank system at the AquaRio public aquarium facility (Rio de Janeiro, Brazil). Filtered seawater from the aquarium’s 170,000 L reservoir was processed through UV sterilization and ozonation and then allowed to equilibrate in a chemostat system housed in a 6 °C cold room. The treated seawater was then supplied via a gravity-fed system from an insulated header tank into 35 L tanks at a constant flow rate of 50 mL min⁻¹. Tanks were equipped with a circulation pump (SunSun JVP130–Sensen Group Co., Ltd, Zhoushan City, China) to maintain water movement and kept in a dark, 4 °C cold room. The Calxy sp. individual from the expedition was sectioned into explants ~ 1 cm thick. These clones were acclimatized for 3 months and then sectioned further into 32 explants with an approximate volume of 1 cm³ before the experiment. Each sponge explant was then transferred, without air exposure, into individual 500 mL flasks (acid-treated and air-tight) containing 450 mL of deep-sea water collected during the expedition and filtered through 0.22 µm filters (MF-Millipore). Half of the flasks were supplemented with 450 µL of a 2 mM ¹³C-labelled bicarbonate solution (NaH¹³CO₃, Thermo Fisher Scientific), whilst the other half was supplemented with 450 µL of a 2 mM unlabelled bicarbonate solution (Thermo Fisher Scientific). All incubations were conducted in the dark, at 4 °C and with shaking at 60 rpm. Sponge explants were sampled after 6, 12, 24 and 48 h, and every time point and both treatments had four independent replicates. Sponge samples were immediately rinsed in ultrapure water and cut into three subsamples, which were snap-frozen. Subsamples were used for 16S rRNA gene sequencing, metatranscriptomics (see above) and stable-isotope analysis. These fragments are named Explant Rep. 1–Explant Rep. 4 for the acclimatized samples prior to stable-isotope incubation and Explant Rep. 5–Explant Rep. 8 after the incubation.

Stable isotope analysis

Frozen sponge samples from the incubation experiments were lyophilised for 72 h and then homogenised using a mortar and pestle. The carbon content and the ¹³C/¹²C ratio were measured with an Elemental Analyser Flash 2000 (Thermo Fisher Scientific) coupled with an isotope ratio mass spectrometer Delta V Advantage (Thermo Fisher Scientific). A three-point calibration curve was used to correct the isotopic data. The ∂13C values were measured against caffeine (IAEA-600, ∂13C = − 27.85‰), casein (∂13C = − 26.79‰) and calcite (IAEA-603, ∂13C = − 10.03‰), which acted as internal stable-isotope standards. The ¹³C/¹²C ratios were calculated as previously described [67, 68]. Isotope data from the ¹³C incubations are corrected for the background ∂13C values of sponges that were not incubated with NaH¹³CO₃ and normalised against the total organic carbon of the sponge explant and the incubation time. Assimilation rates are expressed as µmol ¹³C/mmol C_org⁻¹ d⁻¹.

Results and discussion

Metagenome-assembled genomes from deep-sea Calyx sp. samples

Metagenomic analysis of Calyx sp. retrieved 142 MAGs (Supplementary Table S3), recruiting 69.81% of all obtained sequencing reads (S38 and S40 combined), which also likely contained some remaining sponge reads. Ninety-one MAGs had estimated completenesses higher than 80% and contaminations below 5% (Supplementary Table S3) and were used for subsequent analyses. Those 91 MAGs make up approximately 74.7% of the microbial community abundance based on read coverage of all binned reads (S38 and S40 combined) and span 14 different phyla (Fig. 1). Thirty-eight MAGs were assigned to the bacterial phylum Chloroflexota, 18 to Acidobacteriota, eight to Pseudomonadota and seven to Actinomycetota (Fig. 1). The phylum Poribacteria had five MAGs assigned to it, four MAGs were placed into the phylum Latescibacterota and two MAGs each were assigned to the Bacteroidota, Gemmatimonadota and the archaeal phylum Thermoproteota. Deinococcota, Desulfobacterota_B, Desulfobacterota_D, the placeholder phylum JAAXHH01, and Nitrospirota all had one MAG each assigned to them. Similarly to other shallow- and deep-sea Demonspongiae [5, 18, 69], Calyx sp. thus hosts a diverse microbiome, suggesting different microhabitats and diverse metabolic activities within the holobiont. Several of the prevalent taxa, such as Chloroflexota and Actinomycetota, are commonly enriched in high microbial abundance (HMA) sponges [70], which is consistent with another member of the sponge genus (Calyx podatypa) being classified as HMA [71].

Fig. 1.

Phylogenetic tree of 91 MAGs of the current study. The outer ring is coloured by phylum, whilst the inner ring is coloured by class. Grey bars on the outside represent the relative read abundance for each MAG

One MAG of the phylum Thermoproteota was further assigned to the family Nitrosopumilaceae and genus Nitrosopumilus, and was the dominant microorganism in the community, recruiting 15.7% of all binned reads. Combined with a second bin assigned to the genus Nitrosopumilus, the Nitrosopumilaceae family makes up approximately 16% of the microbial community. This contrasts with our previous observation on the same Calyx sp. samples, which showed a relative abundance of approximately 50% based on ASVs assigned to the family Nitrosopumilaceae [18], which we suspect could be caused by a primer bias.

The most abundant phylum in the community based on read coverage is the Chloroflexota (recruiting 22.75% of all binned reads) across its 38 MAGs. While there are autotrophic organisms within the Chloroflexota [72], the dominant class found here (20 bins, 18.61% of all the binned reads) is Dehaloccocoidia, which are known for their potential to degrade organohalides, rather than being autotrophic [73]. The third and fourth most abundant phyla are the metabolically diverse and mostly heterotrophic phyla Actinomycetota [74] (18 MAGs) and Acidobacteriota [75] (7 MAGs), which recruited 10.12% and 10.19% of all the binned reads, respectively. This large diversity of heterotrophic organisms in the community led us to analyse the possible niche partitions with respect to energy sources and carbon-degradation profile.

Carbohydrate metabolism of the microbial community of Calyx sp.

To define how the microbial community might be contributing to the heterotrophic capacity of the Calyx sp. holobiont, we investigated the expression of genes involved in carbohydrate degradation (Fig. 2) using the database for carbohydrate-active enzymes (CAZYmes, Supplementary Table S4) [56]. We did not consider glycosyltransferases in our analysis, as they are in general related to the biosynthesis of carbohydrates [76].

Fig. 2.

Log₁₀-transformed averaged mRNA counts per million of transcripts for carbohydrate degradation genes based on CAZYdb annotation clustered according to the Euclidean distance. CBM = carbohydrate-binding modules, CE = carbohydrate esterases, GH = glycoside hydrolases, PL = polysaccharide lyases. A list with the detailed functional description of each CAZY family is available in Supplementary Table 4. Black fields in the heatmap indicate no observed expression

Analysis of polysaccharide lyase (PL) transcripts revealed expression of enzymes involved in the degradation of recalcitrant, algae-originated polymers [77–80], including alginate (PL15, PL17, PL39), ulvan (PL25, PL40) and pectin (PL1_2, PL2, PL10_1, PL10_3). PL expression was concentrated in specific taxa rather than broadly distributed across the community. MAGs assigned to the classes UBA2968 (phylum Latescibacterota) and WGA-4E (candidate phylum Poribacteria) accounted for the majority of PL transcripts, suggesting these organisms are the primary degraders of algal-derived polysaccharides in the Calyx sp. holobiont. In contrast, glycoside hydrolases (GH) and carbohydrate esterases (CE) showed broader expression across multiple phyla (Fig. 2), indicating more widespread involvement in general carbohydrate metabolism. Although we did not find a strong phylogenetic signal for the expression or distribution of CAZYmes, MAGs assigned to the bacterial placeholder classes UBA2968 and WGA-4E had the largest number and variety of PL transcripts among all community members.

Glycoside hydrolases (GH) are the most diverse class of CAZymes, and transcripts for enzyme families related to the degradation of algal compounds, such as cellulose and hemicellulose (cellulases, GH5, GH5_12, GH5_13, GH5_18, GH5_26, GH5_39, GH5_48), and their degradation products, such as mannan (80), are prevalent in the Latescibacterota class UBA2968. Transcripts related to the degradation of xylans, such as alpha- and beta-xylanases (GH3, GH10, GH39, GH51, GH67, GH116 and GH141), xyloglucan:xyloglucosyltransferases (GH16_3 and GH16_21) and ß-arabinofuranosidases (GH127, GH142 and GH146) were prevalently detected in the classes WGA-4E and Gammaproteobacteria. Transcripts encoding carbohydrate esterases (CE) involved in xylan hydrolysis (CE 1, 7 and 9) were also detected, suggesting again the potential degradation of dissolved organic matter from algal origin [81]. Transcripts for peptidoglycan lytic transglycosylases (GH23, GH102 and GH103) were predominantly detected in the Gammaproteobacteria and may be related to the turnover of murein, a component of the bacterial cell wall [82]. Expression of alpha-amylases for the degradation of starch (GH13, GH13_3, GH13_8, GH13_9, GH13_10, GH13_11, GH13_16, GH13_18, GH13_20, GH13_23, GH13_26) was scarce compared to the GHs involved in the degradation of more complex compounds, such as xylan and pectin.

Expressed genes related to the oxidation of animal-related compounds encoded chitosanase (GH46), chitinase (GH18), sialidases (GH33 and GH156) and galactosaminidases (GH109, GH129, GH163). MAGs assigned to the class Dehalococcoidia showed the highest expression levels for genes encoding sialidases, glucoamylases and beta-galactosidases [83, 84]. This suggests their primary nutritional strategy may involve degradation of host-derived compounds rather than environmental or algal substrates [85].

Genomic evidence for the degradation of algae- and sponge-derived compounds has also previously been observed in metagenomes of the shallow-water sponges Aplysina aerophoba, Cliona orientalis, Stylissa flabelliformis, Rhopaloeides odorabile, Coscinoderma matthewsi and Carteriospongia foliascens [8, 86]. The genomic potential for carbohydrate degradation between the Calyx sp. holobiont and these sponges was similar, as we also find members of the Chloroflexota and candidate Poribacteria among the symbionts with the highest potential for carbohydrate degradation. However, a study reporting on CAZymes transcription patterns in the microbiome of Ircinia ramosa, Ircinia microconulosa and Phyllospongia foliascens showed that genes involved in the degradation of simple carbohydrates, such as starch and fructose, are more active in these shallow-water sponges [87] when compared to the deep-sea Calyx sp. This is likely because simpler, more bioavailable carbohydrates are quickly degraded across the water column, whilst more complex and less bioavailable molecules, such as hemicellulosic compounds (e.g. xylans, pectin), tend to sink to the deeper parts of the ocean [88]. Whilst there seem to be differences between the expression levels of the GHs involved in the degradation of algae-derived compounds between microbiomes of shallow-water sponges and Calyx sp., those involved in sponge-derived carbohydrates are similar. However, more deep-sea sponges need to be analysed to generalise this trend.

Hydrocarbon metabolism of the microbial community of Calyx sp.

Due to the substantial oil exploration in the Campos Basin and the presence of aliphatic hydrocarbons in waters from our sampling site (Supplementary Table S5), we also assessed the metabolic potential of hydrocarbon degradation in the Calyx sp. microbiome (Fig. 3 and Supplementary Table S6). We did not find a strong phylogenetic correlation between expressed hydrocarbon degradation pathways or functions and MAG taxonomy. Some members of the class Dehalococcoidia (bins 92, 31, 86, 74, 4, 13, 36, 39 and 6) express genes for haloalkane dehydrogenases, which hydrolyse a haloalkane into a halide and primary alcohol, with the latter being further oxidised via an alcohol dehydrogenase and an aldehyde dehydrogenase to a carboxylic acid, which then can undergo ß-oxidation [89]. Besides a role in energy generation, haloalkane dehydrogenases may also contribute here to detoxification, as previous work has indicated that sponge symbionts have the potential to degrade halogenated compounds that are secreted by the host as a chemical defence [90].

Fig. 3.

Log₁₀-transformed averaged mRNA counts per million of transcripts for genes involved in hydrocarbon degradation according to the KEGG database. A list with all the KO functions analysed for hydrocarbon degradation is available in Supplementary Table 6. Black fields in the heatmap indicate no observed expression

Bins 92, 42, 48, 49, 74, 97, 39, 11 and 6, assigned to the Dehalococcoidia, also express alkanesulphonate monooxygenases that catalyse a flavin-dependent oxidation of alkanesulphonate to an aldehyde and sulphite [91]. However, for this to occur, the alkanesulphonates need to be imported into the cell, and none of the MAGs were found to encode canonical genes for the sulphonate transport system (ssuACB) [92]. Whilst this could be due to genome incompleteness, not even MAGs with completeness > 95% had these genes present, and it is unlikely that all MAGs of this taxon would have by chance missed these genes by incompleteness. This suggests either an unknown alkanesulphonate transport system or that the alkanesulphonate monooxygenases found are promiscuous enzymes that catalyse the desulphonation of organosulphur compounds (i.e., taurine or methanesulphonate) [93, 94] that do not rely on a ssuACB-dependent importer [95].

Some Dehalococcoidia MAGs found in the Calyx sp. microbiome also have the genetic capacity for the degradation of phthalate, with bins 96 and 128 expressing both phthalate 4,5-cis-dihydrodiol dehydrogenase (K18067) and 4,5-dihydroxyphthalate decarboxylase (K04102) and bins 11, 20, 6, 19, 138, 39 and 49 expressing only the 4,5-dihydroxyphthalate decarboxylase. No MAG seems to encode or express canonical forms of the phthalate 4,5-dioxygenase, but the former two enzymes are among the most expressed hydrocarbon degradation enzymes in the dataset. Hence, we propose that the MAGs might encode a different enzyme that catalyses the deoxygenation of the aromatic ring or that the enzymes might be absent due to MAG incompleteness. Bin 138 also expresses a 2,6-dioxo-6-phenylhexa-3-enoate hydrolase (K10222), which catalyses the last step in the cleavage of biphenyl. However, the MAG does not encode any other enzymes related to this degradation pathway. Bin 85 is the only MAG that expresses all subunits for the phenol/toluene 2-monooxygenase that is also involved in the degradation of benzoate. Finally, we found multiple MAGs from different classes encoding partial catechol degradation pathways.

Most of the hydrocarbon degradation pathways were found to be incomplete in any given MAG, and/or have secondary functions, such as sulphur assimilation [92] or detoxification of haloalkanes [96]. The catabolism of catechol, a common intermediate in the degradation of most aromatic hydrocarbons [97], is also incomplete in most organisms investigated here, but different MAGs encode different, complementary steps that could lead to the full degradation. This may suggest a syntrophic degradation, where different community members would obtain energy from different oxidation steps. Whilst this concept is yet to be proven for sponge microbiomes, there are several instances for syntrophic degradation of hydrocarbons and other recalcitrant compounds in other microbial systems, such as marine sediments [98–103].

Autotrophic metabolism in the microbial community of Calyx sp.

Among the 91 analysed MAGs, ten have the genetic potential for carbon fixation (Fig. 4, Supplementary Table S3). These include the two MAGs (the abundant bin 1 and bin 2) assigned to the genus Nitrosopumilus, which have the full gene complement for the 3-hydroxypropionate/4-hydroxybutyrate cycle (3HP/4HB). This pathway was initially characterised in Nitrosopumilus maritimus, where it was shown to operate as the primary CO2 assimilation route [104, 105].

Fig. 4.

Log₁₀-transformed averaged counts per million of transcripts involved in carbon fixation based on the MetaCyc database functions. Black fields in the heatmap indicate no observed expression

The use of the reverse tricarboxylic acid (rTCA) cycle for autotrophic carbon fixation is a conserved trait across the order Nitrospirales. It was first described in Candidatus Nitrospira defluvii based on genomic and isotopic evidence [106] and has since been confirmed as a universal feature of the order through comparative genomics [107]. Bin 27 (phylum Nitrospirota, order Nitrospirales, placeholder genus UBA8639) contributes to 0.81% of the community abundance and encodes seven out of the nine genes required for carbon fixation through the rTCA cycle, missing only the genes for fumarase and 2-oxoglutarate synthase. As the MAG is incomplete (91%) yet encodes for the essential key enzymes citrate lyase (acl) subunits A and B and pyruvate–ferridoxin reductase [108] for the rTCA that all characterised members of the Nitrospirales use [109], we expect the corresponding organism to be autotrophic.

Bins 85 (Class Gammaproteobacteria, placeholder family UBA4486 and genus JAAXHV01) with a completeness of 86.5% and a community abundance of 0.36% as well as bin 113 (Class Alphaproteobacteria, placeholder family CAIVPW01 and genus JABHVE01) with a completeness of 87.3% and a community abundance of 0.24% encode ten and 11 out of the 13 reactions of the Calvin-Benson-Bassham (CBB) cycle [110], respectively. Bin 113 encodes both key enzymes of the CBB (phosphoribulokinase (PRK) and RuBisCO), and bin 85 has a gene for the PRK, making both corresponding organisms likely also autotrophic/mixotrophic. The CBB has been shown to be the predominant carbon fixation pathway among chemolithotrophic Pseudomonadota and is well documented in sulphur-oxidising Gamma- and Alphaproteobacteira [111]. Further supporting this is the genetic potential of both MAGs for chemolithotrophy using sulphite oxidation, carbon disulfide oxidation (bin 113 only), and carbon monoxide oxidation (bin 85 only). Interestingly, bin 85 also encodes twelve out of the 18 unique enzymes required for the 3-hydroxypropionate bicycle (3HP) [112], including a propionyl-CoA carboxylase and an acetyl-CoA carboxyltransferase, which might support carbon fixation through anaplerotic reactions [60].

Finally, bins 13, 31, 7, 86 (all assigned to Actinomycetota), bin 77 (Chloroflexota) and bin 85 (Pseudomonadota) have the key enzyme methylmalonyl-CoA mutase for the 3HP, but the completeness of this pathway is low (Supplementary Table S3). Nevertheless, some MAGs have carboxylases (bin 86 has both a propionyl-CoA carboxylase and an acetyl-CoA carboxyltransferase, whilst bins 7 and 13 have a propionyl-CoA carboxylase) that could contribute to intermediate anaplerotic reactions of the 3HP. Previous work has suggested that carbon monoxide oxidation may generate the reductive power to fuel anaplerotic reactions in the microbiome of the shallow-water sponge Petrosia ficiformis [113]. However, these reactions were suggested to be more likely involved in the energetic balance of the holobiont, rather than the accumulation of biomass. Furthermore, anaplerotic reactions have been found to contribute to 0.5–1.2% of the total biomass production in free-living deep-sea Alphaproteobacteria [114]. Further studies are necessary to quantify the contribution of the genetically indicated anaplerosis to the biomass accumulation in sponges.

The Calyx sp. microbiome during aquarium acclimation

In order to generate samples for ex situ carbon fixation experiments, we acclimatised explants of Calyx sp. for 3 months in aquaria, where they appeared to be healthy and developed new oscula. Sequencing of the V4 hypervariable region of the 16S rRNA gene was used to profile microbial community composition of field and acclimatised samples. After pooling spatial subsamples of field individuals, alpha diversity was moderate across the three Calyx sp. specimens (Shannon H' 2.61–3.21, mean 2.94 ± 0.31; Pielou’s evenness J 0.45–0.56, mean 0.52 ± 0.06), indicating that no single taxon dominates the community. Microbial community profiling revealed a shift in community structure after acclimation (Fig. 5, Supplementary Table S7), similarly to what is normally observed in sponges kept in captivity [115]. Acclimatised samples were then incubated for 48 h with ¹³C-labelled bicarbonate (see “Materials and methods” section for details), and there was no statistical support for any further changes in community structure during this incubation (mvabund overall statistical test: LRT = 1122.017; p-value = 0.432; Fig. 5).

Fig. 5.

Taxonomic profiles at the class level of bacterial and archaeal communities in Calyx sp. Samples collected from the deep sea (S10, S38, S39, S40, S43 and S92), after the 3-months acclimation (Explant Rep. 1 – Explant Rep. 4) and after the 3-months acclimation + 48 h incubation with H¹³CO₃^- (Explant Rep. 6 – Explant Rep. 8). Samples derived from three individuals: one subsampled at the base (S38), middle (S39) and osculum (S40); a second subsampled at the base (S10) and osculum (S92); and a third subsampled at the middle only (S43). The profiles are presented as relative read abundances only showing taxa with an average higher than 1%

A statistical analysis between samples from the field and those after 3-months acclimation + the 48 h incubation revealed 452 ASVs changes (Supplementary Table S8), which were grouped at the class level (Supplementary Table S9). Acclimation led to a significant reduction of the relative read abundance (RRA) of Nitrososphaeria and Nitrospiria (2.3 and 6.43-fold and p-value < 0.001 and p-value = 0.008, respectively). Overall, this reduction suggests a relatively lower genetic potential for ammonia- and nitrite-dependent carbon fixation in the aquarium samples. There was also a significant reduction of Dehalococcoidia (p-value < 0.001) in the acclimatized sponge samples.

We observed the appearance of the classes Bacteroidia and Clostridia among the microbial community. Whilst Clostridia are generally heterotrophs (except for some members of the genus Clostridium, which can grow autotrophically via H₂ oxidation and the WLP [116]), Bacteroidia have the potential to fix carbon, mainly in the group of green sulfur bacteria [117]. A more granular analysis revealed that the Bacteroidia ASVs are assigned to the families Crocinitomicaceae and Flavobacteriaceae. Whilst members of the Crocinitomicaceae have the potential for autotrophic growth via H₂ oxidation, we likely did not have the necessary high concentrations of dissolved hydrogen in our aquaria for H₂ oxidation to occur [118, 119]. As for the Flavobacteriaceae, to the best of our knowledge, they are strictly heterotrophic organisms. We also noticed a significant increase in the population of Alphaproteobacteria (p-value = 0.01), with 54.8% of its assigned reads (or 7.93% of all binned reads) belonging to Rhodobacterales. Within this order, the genera Aliiroseovarius and Halocynthiibacter dominated with more than 1% of all binned reads, but they are both described to be heterotrophic [120–123]. We also found an increase in the RRA of the class UBA9042 of the phylum Myxococcota, but no key enzyme for any carbon fixation pathway was found in the genomes of this class using previously developed hidden-Markov models [60]. Finally, we also observed the appearance of members of the classes Gracilibacteria and Planctomycetia, which are also heterotrophic [124, 125].

Despite these changes, several points support the validity of our carbon fixation measurements detailed below: Nitrososphaeria remained detectable and metabolically active in acclimatized samples, as evidenced by the expression of ammonia oxidation (amoABC) and 3HP/4HB cycle genes (Fig. 6). No exogenous autotrophs with substantial carbon fixation potential were introduced during acclimation, and the enriched taxa (Bacteroidia, Clostridia, Alphaproteobacteria) are predominantly heterotrophic. Finally, the reduced AOA abundance in acclimatised samples suggests that our measured carbon fixation rates below likely represent a conservative estimate of the in situ rates, as the field population would have a higher proportion of active carbon-fixing archaea.

Fig. 6.

Log₁₀-transformed counts per million of transcripts of genes involved in ammonia oxidation and the 3-hydroxypropionate/4-hydroxybutyrate carbon fixation cycle

Ex situ carbon fixation in the Calyx sp. holobiont

Isotope labelling with sodium bicarbonate (NaH¹³CO₃) was used to quantify the carbon fixation in the aquarium-held Calyx sp. explants (see Material and Methods for details). After 6 h of incubation, we detected an incorporation of 0.32 ± 0.01 µmol ¹³C/mmol C_org, or 11.07 ± 1.98 µg C/g dry weight sponge per day (Supplementary Table S10). These rates are somewhat lower than the 34 µg C/g dry weight sponge per day reported for the deep-sea sponge Hymedesmia coriaceae [126], and much higher than what has been observed for the deep-sea sponges Higginsia thielei and Nodastrella nodastrella (5 ng and 36 ng C/g dry weight sponge per day, respectively) [127]. However, these lower rates in the latter two sponges could be due to trawling and box-corers sampling that could have damaged or stressed the sponges, in contrast to the controlled sampling method we used (see methods). Similarly, we also observed that carbon fixation rates reduced with longer incubation times (Supplementary Table S10), likely due to the sponge experiencing stress in the closed incubation system.

Aiming to further define the chemolithotrophic processes that support the observed autotrophic activities, we performed RNA shotgun sequencing of the aquarium samples after incubation. Only bin 1 (Nitrosopumilus sp.) recruited enough reads to cover at least 80% of its genes across three replicates, and its expression was compared to those of field samples (Fig. 6). In both the field and the aquarium samples, the genes involved in ammonia oxidation (amoABC) are among the most highly expressed. The gene amoC is expressed more in the aquarium samples than in the field samples (p-value = 0.035), whilst amoB is expressed less in the aquarium samples (p-value = 0.035). The gene amoC encodes a regulatory/stabiliser subunit, and the selective retention of its transcripts under low oxygen concentrations or host-related food depletion, as previously found [128], is consistent with its higher transcript levels after the incubation when food and/or oxygen may be limited. Previous work on the ammonia-oxidising bacterium Nitrosomonas europaea has also suggested that AmoC has a chaperone-like function that stabilizes ammonia oxidation during periods of inactivity or energy stress [128, 129].

The gene amoB is proposed to encode the catalytic subunit of the ammonia monooxygenase due to its structural similarity with particulate methane monooxygenases [130] and its transcripts are promptly degraded under low levels of ammonia. Low levels of amoB transcripts might thus indicate an impaired nitrification, which leads to a shortage of ATP and NADPH, which are required for the 3HP/4HB cycle. Consistent with this, the carbon fixation genes for the acetyl-CoA carboxylase subunit B, acetyl-CoA carboxylase subunit A, malonic semialdehyde reductase, 3-hydroxybutyryl CoA dehydratase, and succinic semialdehyde reductase are also expressed less in the aquarium samples than in the field samples.

Overall, the reduction in expression levels of amoB and the genes for the enzymes of the 3HP/4HB indicate that carbon fixation in the acclimatised samples occurs at a lower level than when compared to the field and that the ex situ bicarbonate incorporation rates measured above might represent lower boundaries of what the Calyx sp. holobiont can achieve in situ. Further in situ studies across more individuals and under different environmental conditions are, however, required to fully understand and quantify the natural carbon fixation rate of this sponge (and deep-sea sponges in general).

Ecological implication of findings

The deep sea is a diverse environment, with some locations being rich in nutrients, but most being characterised as oligotrophic [131], mainly because easily bioavailable particulate and dissolved organic carbon that sinks from the surface water is quickly consumed by planktonic organisms in the water column [132]. Hence, most of the organic compounds that are found in the deeper ocean would be considered recalcitrant [133]. Previous work has shown that microbial symbionts of shallow-water sponges actively uptake DOM through heterotrophic carbon metabolism and transfer the incorporated carbon to the host [68, 134]. Similarly, bulk isotope labelling experiments with particulate or dissolved microalgal material have revealed that deep-sea sponge holobionts rely on DOM to meet metabolic demands, with its uptake accounting for the majority of carbon required by many sponge species [135, 136]. As the symbionts in Calyx sp. have the expressed genomic potential for a complete degradation of a plethora of complex carbohydrates and the partial degradation of hydrocarbons, our results expand on this previous work by supporting the notion that DOM uptake and utilisation in deep-sea sponge holobionts is adapted towards recalcitrant matter. However, in the Calyx sp. holobiont, heterotrophic metabolism appears to be not sufficient to meet carbon demands as we also observed autotrophic activities of AOA.

AOA are also found in shallow-water sponges [137–139] and in the tropical Ianthella basta [140], which has been shown to fix between 0.18 and 0.61 mg C per gram wet weight sponge per year [132], thus only making a minor contribution to carbon acquisition in the holobiont. In the deep-sea Calyx sp. holobiont, the AOA, which appears to be responsible for most of the carbon fixation, has rates of at least 13.67 mg CO₂ per gram of sponge dry weight per year (Supplementary Table S10). In recent work, photoautotrophic organisms were shown to provide nutrition that satisfies up to 52% of the respiratory demand of the tropical shallow-water sponge Chondrilla caribensis [10]. Our work thus indicates that the AOA in deep-sea sponges can act as primary producers in a similar trophic function but on a different scale to the photoautotrophs in shallow-water sponges.

Overall, our findings highlight the diverse strategies that the Calyx sp. holobiont applies in generating biomass from organic and inorganic carbon in the deep sea. The presence of both autotrophic and heterotrophic microorganisms indicates a mixotrophic lifestyle relying on recalcitrant substrates from marine snow, while simultaneously obtaining organic carbon from inorganic carbon fixation. This shows that sponges in the deep sea can have dual functions in influencing the carbon cycle, and future work should quantify how this influences carbon fluxes in their environment.

Authors’ contributions

ANG, LH, RP and TT conceptualised the study. ANG analysed and curated data, prepared all figures and tables, and wrote the original draft; ANG, LH, JN, MAA, MEM, BR, LM, TD,, GD, and CV contributed to methodology and experiments. All authors reviewed the manuscript. ANG and TT edited the manuscript.

Funding

This work was supported by a project registered as ANP 21005–4, “PROBIO-DEEP—Survey of potential impacts caused by oil and gas exploration on deep-sea marine holobionts and selection of potential bioindicators and bioremediation processes for these ecosystems” (UFRJ/Shell Brasil/ANP), sponsored by Shell Brasil under the ANP R&D levy as “Compromisso de Investimentos com Pesquisa e Desenvolvimento”. R.S.P. was also supported through KAUST grant number BAS/1/1095–01-01.

Data availability

All data are available in the main text or as supplementary materials. Metagenomic and metatranscriptomic reads are available in the Sequence Read Archive (SRA) accessions SRR30902886 - SRR30902887 and SRR30902953 - SRR30902965, respectively. MAGs are available as BioSamples SAMN42181007 - SAMN42181148. All the information has been submitted under BioProject accession PRJNA1130053.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.