High compositional and functional similarity in the microbiome of deep-sea sponges

Abstract Sponges largely depend on their symbiotic microbes for their nutrition, health, and survival. This is especially true in high microbial abundance (HMA) sponges, where filtration is usually deprecated in favor of a larger association with prokaryotic symbionts. Sponge-microbiome association is substantially less understood for deep-sea sponges than for shallow water species. This is most unfortunate, since HMA sponges can form massive sponge grounds in the deep sea, where they dominate the ecosystems, driving their biogeochemical cycles. Here, we assess the microbial transcriptional profile of three different deep-sea HMA sponges in four locations of the Cantabrian Sea and compared them to shallow water HMA and LMA (low microbial abundance) sponge species. Our results reveal that the sponge microbiome has converged in a fundamental metabolic role for deep-sea sponges, independent of taxonomic relationships or geographic location, which is shared in broad terms with shallow HMA species. We also observed a large number of redundant microbial members performing the same functions, likely providing stability to the sponge inner ecosystem. A comparison between the community composition of our deep-sea sponges and another 39 species of HMA sponges from deep-sea and shallow habitats, belonging to the same taxonomic orders, suggested strong homogeneity in microbial composition (i.e. weak species-specificity) in deep sea species, which contrasts with that observed in shallow water counterparts. This convergence in microbiome composition and functionality underscores the adaptation to an extremely restrictive environment with the aim of exploiting the available resources.


LOW COMPOSITIONAL DIVERGENCE
To further investigate similarities in the microbiome between species or locations, we analyzed the microbial members and functions that were shared between the sample groups (the aggregation of the 3 replicates from the same species and location), and the ones that differed.
Interestingly Gpac_DR15 shared about 72% of ASVs with Povi_DR15, which is as similar as any two sample groups of Povi from different locations.Gbar_DR15-Gpac_DR15 shared 68%, and Gbar_DR15-Povi_DR15 shared 67%.All samples together had 44% of ASVs in common.
We then investigated what differentiated our samples by checking the differentially abundant (DA) ASVs between species and locations.As expected, few ASVs were DA between the species: 9 DA-ASV between Gpac_DR15-Povi_DR15, 14 between Gbar_DR15-Gpac_DR15, and 23 between Gbar_DR15-Povi_DR15 (out of 9,751 ASV considered >0.01%RA and >= 2 samples, Table S10, Fig. S1A).Most of the DA ASVs represented low abundant taxa, covering several genera with no clear group being representative of any species.Only one DA-ASVs assigned to Woesearchaeia (Nanoarchaeota) reached up to 2.7% of aRA and was more abundant in one Gpac sample compared to the other species.Among locations there were a maximum of three DA-ASVs detected (Table S10, Fig. S1B), and the most abundant one was an ASV assigned to Nitrososphaeria (Thaumarchaeota).

LOW FUNCIONAL DIVERGENCE
In terms of functions, pair comparisons shared between 96 to 87% of their abundant transcripts (>1 tpm, Fig. 3).Here as before, Povi_DR15 and Povi_DR9 shared the greatest number of transcripts (93%), and Gpac_DR15 shared with Povi_DR15 up to 92% of transcripts, but Gbar_DR15 and Gpac_DR15 shared 90%.In this case, locations shared more transcripts among them than among species.All Povi samples shared 89% of transcripts, and the 3 species from DR15 shared 85% of transcripts.
We also checked differentially expressed (DE) transcripts.We selected transcripts with more than 10 cpm and in more than 2 samples, leaving pairwise datasets of 18,000 -19,000 transcripts.Relatively small differences in gene expression were also found among species, with Gpac-Povi showing the smallest differences (240 DE-trans), 686 DE-transcripts between Gbar-Gpac, and 1,132 DE-transcripts between Gbar-Povi (Table S11), this represented from 1.2 to 6% of the transcripts.These DEtranscripts included one or more enzymes that were related with a number of different pathways (Table S11), although around 70% did not have any KO annotation.From the annotated ones, the Gbar-Povi comparison showed that amino acid metabolism and biosynthesis of secondary metabolites were differentially upregulated in Povi (Fig. S2A).Quorum sensing pathways included 26 DE genes upregulated in Gbar and 51 DE genes in Povi, similar to ABC transporters (23 and 49, respectively).
Nitrogen metabolism had 19 genes upregulated in Povi, and 2 in Gbar, while Sulfur metabolism had 7 genes upregulated in Gbar.Xenobiotics biodegradation and metabolisms was notably upregulated in Povi, as well as cell adhesion molecules and ECM-receptor interaction (within signaling molecules).
The comparison between Gbar-Gpac pointed to similar pathways being DE expressed, albeit to a smaller extent.And the lower 240 DE-transcripts detected between Gpac-Povi were centered notable in sensing and interacting with environment and other cells such as the quorum sensing (10 DE-genes upregulated in Povi, and 1 in Gpac), ABC transporters (9 and 8), notch signaling, two component system, and ECM-receptors.Only few genes were related with differences in nitrogen metabolism, oxidative phosphorylation and porphyrin and chlorophyll metabolism.
DE-transcripts between locations for Povi were lower, ranging from 11 DE-trans (Povi_DR9 vs Povi_DR15) to 123 DE-transcripts (Povi_DR10 vs. Povi_DR15), this is 0.05 to 0.6% of the transcripts considered (Table S11).In the latter, Povi_DR10 always presented upregulated genes in many different pathways (Fig. S2B), being noteworthy porphyrin and chlorophyl metabolism, phosphonate and phosphinate metabolism, lipid and amino acid metabolism.However, when Povi_DR10 was compared to Povi_DR4 upregulated genes were mostly associated with pathways of carbohydrate, energy, and cofactors and vitamins metabolisms (Fig. S2B), while Povi_DR4 had genes upregulated in many other functions like prodigiosin biosynthesis, quorum sensing, or lipid metabolism.The comparison Povi_DR10-Povi_DR9 featured 5 pathways notably upregulated in Povi_DR9 and none in Povi_DR10.
Between Povi_DR4-Povi_DR9 (114 DE-transcripts), genes were usually upregulated in Povi_DR9 covering many different pathways, the largest differences were in butanoate metabolism, carbon fixation, and benzoate degradation among others.
Generally, these results show the relatively small differences (i.e., low percentage of differentially expressed transcripts between groups) in the microbiome of different sponge species at functional level.
Any differences were centered in microbial genes related with the sensing of and exchange with their environment, and few metabolic pathways.Interestingly, P. ovisternata showed greater upregulation in a variety of metabolic pathways, many of them among xenobiotic biodegradation.Unfortunately, poor taxonomic classification of the associated ASV sequences at the species level did not allow us to confidently identify possible xenobiotic-degrading microorganisms.

LIPID METABOLISM
Fatty acids biosynthesis pathway accounted for 2,834 aTPM (Table S17), and modules inside were complete (Table S14).Sponges are rich in short chain fatty acids (FA), mid-chain branched fatty acids (MBFAs) and the sponge specific demospongic acids (long chain acids, LCFAs).Sponges seem to rely on short-chain fatty acids from bacterial origin as precursors for demospongic acids, since they have incomplete biosynthesis pathways for short-chain fatty acids [11,12].Genes involved in the short-chain fatty acid biosynthesis (type II FAS, FabDHFGILZ) were high expressed in our metatranscritpomes.
MBFAs are often abundant compounds in HMA sponges, produced by bacterial-type polyketides synthetases (PKS), and Poribacteria was proposed as the potential producer of these compounds [13].
In Geodia barretti, and other North-Atlantic deep-sea demosponges, methyl-branched fatty acids were the dominant bacterial FAs, and were used as precursors for branched LCFAs [14].LCFAs constitute a major part of sponge membrane phospholipids (PLs) and probably serve a structural and functional role [15].

AMINO ACID METABOLISM
Modules leading to amino acid biosynthesis were mostly complete, and highly expressed (Table S14).
Metagenomic studies showed the genetic potential of different SAGs to synthesize different amino acids [16].The collective sponge microbial community seems to be able to synthetize most of amino acids, with highest expression in pathways such as valine, leucine and isoleucine biosynthesis (6, For example, biosynthetic pathways of several amino acids were present in the prokaryotic metatranscritpome of Xentospongia mutua, but host-derived transcripts only included catabolic reactions for these amino acids [6].
Apart from glucose, symbionts seem to be able to use a variety of additional carbon sources, based on the expression of degradation pathways for galactoside (1,297 aTPM), fructoside (225 aTPM), xyloside (213 aTPM), rhamnoside (199 aTPM), inositol (617 aTPM) or uronic acids (885 aTPM; pathway descriptions are based in key enzymes described in [17]).Some intermediate genes were not expressed at detection levels and their absence has been identified in Poribacterial and genomes Chloroflexi genomes, however intermediate products could be converted by other enzymes [3,17].
Degradation of complex carbohydrates appears to be a dominant feature in sponge symbionts [17,18].
Symbiont ability to utilize diverse carbon sources can be of relevance for the host, as the sponge feeding on dissolved and particulate organic matter from the extensive filtration activity can be mediated by the bacterial symbionts [19][20][21].Moreover, symbionts can utilize compounds from the extracellular matrix of the sponge as nutrient substrates [17,18,22].With the lack of sun light for photosynthesis, the symbiotic heterotrophy can perform a more important role in these deep-sea communities.

Carbon fixation
With respect to autotrophic carbon fixation, the cumulative aTPM values showed that the reductive citrate cycle (also called Arnon-Buchanan cycle or reductive tricarboxylic acid) presented the genes with highest expression (6,319 aTPM), followed by dicarboxylate-hydroxybutyrate cycle (3,485 aTPM),

Nitrogen metabolism
All modules associated with nitrogen metabolism were complete: denitrification (1,176 aTPM), nitrification (1,130 aTPM), dissimilatory nitrate reduction of ammonium (DRNA, 443 aTPM) and assimilatory nitrate reduction (15 aTPM) (Table S14, Fig. 6).During nitrification, ammonia is oxidized to nitrite by the sequential actions of ammonia monooxygenase and hydroxylamine dehydrogenase, providing reductant for the cell [24].Ammonium transporter (Amt family, K03320) was expressed at high rates (1,535 aTPM), being the most expressed type of transporter, indicating the great use of this molecule.The three genes conforming the ammonia monooxigenase operon (K10944-46, pmo-amoCAB) were co-transcribed as a single transcript, with expression levels from 99 to 337 aTPM.
Baterial CuMMO (ammonia oxidizing bacteria, AOB) genes belonged to Actinobacterias, Methylococcales (Gammaproteobacteria) and Methylacidiphilae (Verrucomicrobia), although these clustered with the closely related particulate methane and butane monooxygenase in a phylogenetic tree (pMMO and pBMO; data not shown), and therefore were probably not involved in ammonia oxidation.To prevent overestimation of nitrification, only expression values from Thaumarchaeota were considered (Fig. 6 and 7).The next step in nitrification, hydroxylamine oxidation (hao) to nitrite, expressed by Gammaproteobacteria, probably serves in the detoxification of hydroxylamine, instead to contributing to nitrification, as previously suggested [8], since the ammonia oxidation gene was not confirmed for this group.Nitrite oxidizing bacteria (NOB) perform the second step of the nitrification, the In denitrification, nitric oxide reductase (norB) was expressed by Bacteroidetes, Deltraproteobacteria, Planctomycetes and Firmicutes, while norC was expressed by Betaproteobacteria and Spirochaetes.
The last enzyme of the denitrification (nosZ) was expressed by Alphaproteobacteria and Chloroflexi (Fig. 7).Complete denitrification has been found in sponge-associated Alphaproteobacteria [44].
Inside the symbiotic cells, ammonia can be used as source of energy (nitrification) but also for biosynthesis of amino acids (ammonia assimilation).Enzymes for ammonia assimilation were also highly expressed, mostly via glutamine synthetase (K01915, glnA, GLUL; 732 aTPM), and in a lesser extend through glutamate dehydrogenase (K00261 GLUD1_2, gdhA; 55 aTPM).Ammonia can be derived from host metabolism (ammonia excretion of the sponge cells), but can also be generated by the dissimilatory nitrate reduction and by coupled denitrification and nitrogen fixation.
The enzymes characteristic of dissimilatory nitrate reduction to ammonium, DNRA (cytochrome c dependent nitrite reductases (nrfAH, K03385-K15876), and NADH dependent nitritre reductase (nirBD, K00362-63) are specific to the reduction of nitrite to ammonia and were expressed in low values (0.1 to 11 aTPM).Enzymes for the assimilatory nitrate reduction were also detected but with low expression (max.value of the ferredoxin-nitrite reductase, nirA, K00366, was 13 aTPM).Enzymes for the anaerobic anammox or the nitrogen fixation (NF) were not detected.In the metagenomes of Vazella portalessi some of the enzymes for the nitrogen pathways were missing but alternative routes were proposed to link the processes [43].
In addition, urease activity (by detection of ureC gene) has been identified in the sponge Xestospongia testudinaria [45], and urease-encoding gene clusters and urea transporters are also reported from sponge metagenomes and metatranscriptomes [46][47][48].We detected expression of the three urease subunits (ureABC, with accumulated 45.9 aTPM), several accessory proteinc (ureD-J, accumulated 53.2 aTPM), and the urea transport system (urtA-E with accumulated 103.2 aTPM), proving urea utilization in our deep-sea sponges.Urease genes were expressed by 14 different phyla, with dominance of Actinobacteria, and Firmicutes.Urea is one of the dominant organic nitrogenous compounds in oligotrophic oceans [49] and likely serves as an alternative nitrogen source to ammonia, nitrate, and nitrite within the sponge holobiont.
Formaldehide can either be dissimilated to CO2 for energy production by the formaldehyde dehydrogenase (fdhA, 94.3 aTPM) and formate dehydrogenase (FDH at 43.3 aTPM), or assimilated for biomass production by the xylulose monophosphate, ribulose monophosphate and serine pathways (Fig. 6).Only the ribulose monophosphate pathway was complete with an accumulated expression of 97 aTPM (Fig. S4, Table S14).
Methanogenesis was incomplete, with key enzymes such as the methyl-coenzyme M reductase (Mcr) complex, and methylenetetrahydromethanopterin dehydrogenase (Mtd) not being expressed (Fig. 6).

Sulphur metabolism
Sulphur is an essential nutrient for microbial synthesis of certain amino acids, vitamins and enzymes.
A source of sulphur is the sponge derived taurine.Transporters for taurine acquisition (tauACB, accumulated 31.9 aTPM) and degradation to sulfite (taurine dioxygensases, 54.5 aTPM) were expressed (Table S14, Fig. 7).Sulfite can be reduced to sulfide or oxidized to sulfate.Sulfate can also be produced through thiosulfate oxidation, and thiosulfate may be produced by incomplete oxidation of sulfides [8].Modules of assimilatory sulfate reduction (ASR, 830 aTPM), dissimilatory sulfate reduction (DSR, 311 aTPM) and thiosulfate oxidation by SOX system (149 aTPM) were complete (Table S14, Fig. 7), suggesting that these processes are important in the sponge microbiome.

Carbon monoxide oxidation
Another potential energy source is the CO oxidation via carbon monoxide dehydrogenase (CODH)mediated pathway.CO-oxidizing bacteria are lithoheterotrophs common in sponge microbiomes [57,58], highly overrepresented compared to seawater [4].The presence of this CODH complex in 65% of the sponge symbionts analysed by Burgsdorf et al. [59] suggested that CO oxidation is the most abundant process related to a litoheterothrophic lifestyle in sponge symbionts.Our results support this observation since CODH subunits are expressed by many different phyla at very high values in our metatranscriptomes (coxS at 1,668 aTPM, coxM at 1,922 aTPM, and coxL at 1,583 aTPM (Table S14, Fig. 6).The function of coxL is variable, and it has been related to the oxidation of CO but also oxidation of different organic substrates in sponge symbionts [59].CO dehydrogenase was expressed by many microbial phyla (Fig. 7), most of them were previously confirmed in other sponge species [58,59].

Phosphonate utilization
Microorganisms can utilize phosphonates (reduced organiphosphorous compounds) as potential source for growth in conditions of limiting inorganic phosphorous (Pi) concentrations.Microorganisms can metabolize these compounds via cleavage of the C-P bond through at least three mechanisms: hydrolytic (2AEP, 2-aminoethylphosphate degradation pathway via PhnWX), radical (C-P lyase), and oxidative (PhnYZ) [60].Genes for bacterial phosphate management are included in the so-called Phosphate regulon (Pho regulon), whose control is performed by a two-component regulatory system (TCRS).The involvement of the microbiome in Pi metabolism was investigated in several pathways.
Gene products of the Pst system (phospahte-specific transport) encoding an ABC transporter of inorganic phosphate (pstSCAB-phoU operon) were expressed at 45 to 134 aTPM.During Pi starvation, TCRS activates the expression of proteins for the transport and use of phosphonates which are encoded in the C-P lyase complex and consist of 14 genes (phnCDEFGHIJKLMNOP).Three gene products constitute the ABC transporters for phosphonates (phnCDE) and were expressed from 25 to 73 aTPM.However, the seven gene products involved in the catalysis of degradation of phosphonates were expressed in low values (< 0.2 aTPM).Accessory protein phnP had an expression of 52.7 aTPM.
Other enzymes of phosphonate catabolism such as phosponoacetate catabolism performed by phnA were expressed at 92.5 aTPM, and 2-AEP (2-aminoethylphosphonate) catabolism (phnW,phnX) at 4.9 and 5.5 aTPM respectively.In a global oceanic survey, bacterial C-P lyase abundance declined in the mesopleagic zone relative to the epipelagic zone, while other phosphonate degradation pathways remained well represented [61].The proportion of phosphonate producers in free-living bacterioplankton were reported to steadily increase with depth to nearly 30-40% at 200 m [60].
Phosphate biosynthesis in our metatranscriptome was expressed, pepM (K01841) at 12.8 aTPM, and ppd (K09459) at 260.8 aTPM.It seems that phosphate metabolism in depth is focused on the biosynthesis of phosphonates.In sponges, phosphate is important for regulatory functions within the holobiont [62], genes involved in phosphorous cycle were detected in V. pourtalessi microbiome [43] and SAUL genome [22].Sponge microbiome can also produce and store polyphosphate granules, which can comprise 25 to 40% of total phosphate present in sponge tissue [63], and it is conceivable that it serves as P-storage reservoir [64].However, Acker et al. [60] found that phosphonate biosynthesis and catabolism pathways were mutually exclusive, therefore pointing to other proposed functions such as incorporation of phosphonates into membrane lipids or capsular polysaccharides to protect cells against enzymatic attack or phage attachment.

OTHER FUNCTIONS RELATED TO THE SYMBIOTIC LIFESTYLE
ATP binding cassette (ABC) transporters pathways were highly expressed (ko02010 at 9,568 aTPM, Table S17).The diverse and abundant range of membrane transporters encoded by the sponge microbiome (12,380 aTPM, Table S16) provides mechanisms to facilitate putative metabolic exchanges [5], and indicates the strong reliance on nutrient uptake from the sponge microhabitat.Among these transporters we found high expression for amino acids, carbohydrates, lipoproteins, metals, and lipoproteins transport (Table S12).Quorum sensing pathway, includes a large family of 282 genes for sensing and transferring compounds inside the cell, which were highly expressed (ko02024 at 21,428 aTPM, Table S17), and can be used for microbiome population control [65].
Many symbiotic microorganisms are known to interact with their host via microbial secretion systems (ko03070, combined expression of 2,767 aTPM).Most genes for secretion systems Type II (TS2) were expressed, as well as Sec-RCP and Twin arginine targeting (Tat) for transporting proteins across the inner membrane into the periplasm (Table S17).This system excretes effectors into the milieu and processes mediated by T2S include suppression of innate immunity, adherence to host surfaces, biofilm formation, invasion into and growth within host cells, nutrient assimilation, and alterations in host ion flux [66].The syringe-like types III, IV and VI were also expressed in our metatranscriptomes, but few genes were not detected.These systems can transport proteins across an additional host cell membrane, delivering secreted proteins directly to the cytosol of a target cell, to eliminate microbial competitors and to translocate toxic effector proteins.Genes for the secretion systems were also identified in metagenomic analysis of Ircinia ramosa [8].Robbins et al. [39] identified that only few lineages encoded the necessary genes to form secretion systems, suggesting that eukaryotic-like repeat proteins are unlikely to be introduced in the sponge via these traditional pathways.
In agreement with the results from Robbins et al. [39], Thaumarchaeta presented substantially lower frequency of ELPs, with what the authors explained as that the microorganism could be utilizing alternative mechanisms to maintain associations with the host.
In addition to these proteins, cell-cell adhesion molecules may be required for the attachment to the host tissue.Cadherin domains and Fibronectin III domains mediate cell adhesion and biofilm formation in eukaryotes and have also been found to serve the same function in bacteria [76][77][78].These genes were enriched in the sponge-associated MAGs and identified in most bacterial lineages [39].Genes containing cadherins and fibronectins were also high expressed in our metatranscriptomes (1,426 and 799 aTPM respectively), and taxonomically widespread to 30 phyla (Table S19).Most frequent type were cadherins expressed by Proteobacteria, followed by fibronectin in Actinobaceria and Thaumarchaeota.In general, the high expression and wide distribution of ELPs and adhesion proteins in the sponge-associated members suggests that these features are critical for successful establishment in the sponge niche.
Production of biologically active secondary metabolites is an important defence mechanism utilised by sponges for protection against predators or epibionts [79].Many secondary metabolites are produced by polyketide synthases (PKS), mainly Type I PKS, and non-ribosomal peptide synthetases (NRPS) [80,81].Some of these compounds are produced by the sponge and others by associated microorganisms [9,[82][83][84][85].Despite being found in metagenome bins of sponge symbionts [22,86,87], their expression was low in our metatranscriptome.We found a total of 85 genes annotated as COG3321 (Acyl transferase domain in polyketide synthase enzymes) or COG3319 (Thioesterase domain of type I polyketide synthase) with expression of 9.6 aTPM.
Mobile genetic elements such as transposons, plasmids and prophages can facilitate adaptation to either specific niches or to changes in environmental conditions in sponge-associated microbes [5,88].
Transposable insertion elements (selected by the category X of the COG classification: mobilome, prophages, transposons) were detected in 22,635 genes and their accumulated expression was 7,567 aTPM, across 53 different phyla, dominating Proteobacteria and Cyanobacteria (Table S20).Highest expression values were detected in transposases (COG0675, COG3293, COG3415) and a retron-type reverse transcriptase (COG3344/PF00078) previously detected in members of the 'sponge-associated unclassified lineage' (SAUL) that are frequently recorded from sponges [22].
Restriction-modification systems (R-M) are considered bacterial defence systems against incoming, foreign DNA, that are also involved in adaptation to changes in the environmental conditions, and in host colonization time protect against DNA exchange with non-symbiont and/or pathogen microorganisms [89][90][91][92].R-M systems have been described in sponge symbionts [18,22,93] and may facilitate horizontal DNA exchange between sponge symbionts.These elements were annotated in 10,865 transcripts with 1,806 aTPM (highest COG0863, DNA modification methylase) (Table S21).
Members of sponge microbial communities have also incorporated into their genomes systems to effectively protect themselves and minimise the introduction of foreign DNA into their chromosomes [5,89].In this context, clustered regularly interspaced short palindromic repeats (CRISPRs) and their associated proteins (Cas) [94] are commonly enriched in sponge-associated microbial communities compared to seawater [4,5,22,75,89,95,96], with only one gene, a CRISP-nuclease (COG3513), present both in sponges and seawater [89].The expression of genes annotated as CRISPR and their associated proteins (Cas) were not particularly high expressed in our metatranscriptome (the accumulated expression of 3,728 CRISPR-related transcripts accounted for 361 aTPM, Table S22), suggesting low pressure from potential invading DNA in microbes living within deep-sea sponges.

MOST ABUNDANT INDIVIDUAL KO TERMS IN DEEP-SEA SPONGES
Among the fifteen most abundant individual terms (KO) for all species (Fig. S6), we found three proteins related with a peptide/nickel system: a substrate binding protein (K02035 at 8,934 aTPM), and two permease proteins (K02033 and K02034, with 3,698 aTPM and 2,243 aTPM respectively).Proteins for the transport of peptides/nickel were also the most abundant genes in the metagenomes of several tropical Ircinia spp.[97] and in a metaproteomic study of Aplysina aerophoba.These were followed by proteins for the transport of sugars (K02027) and amino acids (K01999) [98], that were also among our top 30 most expressed genes (Table S12).The metal nickel is a fundamental cofactor of many enzymatic reactions of prokaryotes and eukaryotes.Nickel-containing enzymes are involved in at least five metabolic processes, including the production and consumption of molecular hydrogen, hydrolysis of urea, reversible oxidation of carbon monoxide under anoxic conditions, methanogenesis, and detoxification of superoxide anion radicals [99].The peptide/nickel transporter system has been hypothesized as a response to the presence of antibiotic chemicals, conferring resistance to vancomycin [100].
Hypoxanthine phosphoribosyltransferase (K00760, 6,160aTPM), the second most expressed protein in this study, is involved in the purine salvage pathway (nucleotide metabolism), catalysing the conversion of hypoxanthine and guanine to their respective mononucleotides and which is essential for life processes.The chaperonin GroEL (K04077, 3,810 aTPM), also referred as heat-shock protein family 60, and other molecular chaperons help to compensate protein stability due to heat stress [101] or other stressors such as salt or ethanol stress, osmotic pressure, presence of reactive oxygen species and toxic compounds [102,103].However, expression in obligate symbionts is usually high even under nonstress conditions (i.e., constitutive) [104], being commonly the most highly expressed protein in symbionts of insects [105][106][107][108]. GroEL has being proposed for several roles in microorganism-insect interaction [104,109,110].For instance, in primary symbionts with reduced genomes, the chaperonin could assist in the folding of conformationally damaged proteins created by the negative effects of deleterious mutations occurring due to genome erosion [109,111], but also as target of antimicrobial peptides (AMPs) for endosymbiont control, limiting cell division of endosymbionts [110].High abundance or expression of GroEL genes have been reported for several sponges [67,98,112], although its function within sponges is not completely understood.Ribosomal proteins (L27, K02899, and L21, K02888) help in stabilizing the specific structures of rRNAs in mature subunits and facilitating the accurate folding of rRNAs during ribosome assembly [113].These proteins perform a fundamental role in cell physiology, and therefore their high abundance was not surprising (2,612 aTPM and 2,574 aTPM, respectively).

FUNCTIONAL CONVERGENCE
We investigated the functional convergence of the sponge microbiome in our target species by focusing on the community members preforming the same metabolic functions in the metatranscriptome.Each annotated gene (KO) was expressed by up to 38 different phyla (Table S12) and half of them were common to more than 12 phyla.For example, nitric oxide reductase (norB) was expressed by Planctomycetes (2.9 aTPM), Bacteroidetes (2.8 aTPM), Deltraproteobacteria (1.12 aTPM), and Betaproteobacteria (1 aTPM).NorC subunit was expressed by Betaproteobacteria (0.43 aTPM) and Spirochaetes (0.34 aTPM).The last enzyme of the denitrification (nosZ) was detected in Alphaproteobacteria (0.80 aTPM) and Chloroflexi (0.12 aTPM) (Fig. 7A).The adenosine-5′phosphosulfate reductase alpha subunit (aprA), a key enzyme in microbial sulfate reduction and sulfur oxidation, was expressed mostly by Euryarchaeota, but also by Alpha-, Beta-, Delta-and Gammaproteobacteria, and Fervidibacteria (Fig. 7A).Another metabolic module with high functional redundancy was CO-oxidation, with many bacterial and archaeal taxa showing high expression values of the carbon monoxide dehydrogenase (CODH) genes coxL, coxM, and coxS (Fig 7A).Several aerobic bacteria genera were reported previously to utilize CO as energy supplement when organic substrates are limiting [58,114,115].Carbon fixation reactions showed wide distribution among many microbial phyla (Fig 7B).
These modules had a larger percentage of completion in this shallow HMA sponge than in the deepsea HMA sponges (Fig. S4).In carbon fixation modules, Poribacteria also had a higher participation in several reactions than in the deep-sea water species (Fig. S8C).In H. caerulea, there were many reactions and enzymes absent, but among the ones present, the main microbial taxa were shared with the HMA species, except for Poribacteria (Fig. S7D and S8D).
Functional redundancy has also been identified among these processes as the use of different enzymes performing equivalent reactions.In denitrification and ammonia oxidation, Fan et al. [5] noticed a preference of copper-containing nirS genes in the symbiotic community of Cymbastella concentrica, while cytochrome cd-1 dependent nirK was more frequent in other species performing the same reaction.In our deep-sea sponges, nirS was expressed at low values (from 0 to 0.2 tpm), but nirK was highly expressed (from 573 to 950 tpm), even though previous work could only amplify nirS in G. barretti [34].These values were equivalent in all three species.We also found a dominance of membranebound narG (208 aTPM) expression compared to periplasmic napA (1.3 aTPM) in all our species (Fig 7A), contrary to earlier reports of different frequencies of these genes in different sponge species [5,95].
There is extensive functional convergence of the microbiome in sponges [5,30], and therefore, sponge microbiomes may share a set of core functional genes rather than a common set of taxa [116][117][118].In effect, this is what was termed "guilds" [119], a group of species that exploit the same class of environmental resources in a similar way, regardless of taxonomic position.A guild with many members will contribute to the stability of the ecosystem [120,121], because as environmental conditions shift, different members of the guild will become dominant, but the function will be carried out anyway.
However, members of a guild are supposed to interact mostly by competing with each other, since they use the same resources in a similar way [122].It is likely that the remarkable similarity of the microbiome across deep-sea sponge species, prevented us from detecting variations in the expression of different genes between the sponge species, but the number of members expressing the same gene within a sponge indicates that the sponge environment is a stable ecosystem with large functional redundancy.
lower expression values of 4.7 aTPM.The last reaction of nitrification, done by the nitrite oxidoreductase alpha and beta subunits (the enzymes K00370-71, narG, narZ, nxrA and narH, narY, nxrB) showed again high expression values (208 and 207.5 aTPM).The key enzyme codified by the ammonia monooxigenase (amo) gene in the nitrogen fixation pathway belongs to the same copperdependent membrane monooxygenase family (CuMM) as the particulate methane monooxygenase (pMMO, K01944-46) and probably other CuMM containing members, and cannot be distinguished by homology alone[25].Similarly, the nitrite oxidoreductase enzymes (K00370-71) are also used in the inverse reaction, nitrate reduction, present in denitrification and DNRA, confounding the actual expression values of each function.
[5,20,[41][42][43]rarely detected, suggesting incomplete pathways, which would result in the release of nitrous oxide into the surrounding seawater, or the presence of alternative routes producing final N2 and O2 bridging major pathways[5,20,[41][42][43].Corroborating those studies, the last steps of denitrification were expressed in low values (9.6 and 0.82 aTPM, respectively), indicating that the cycle can be closed, but the lower expression could produce the observed accumulation of nitrous oxide.
oxidation of nitrite to nitrate.NOB representatives among our transcripts included mostly phylum Nitrospirae and some Alphaproteobacteria. metabolism, while the copper-containing nitrite reductase (K15864, nirS) was expressed in low values (0.07 aTPM).Previous studies have shown that only enzymes responsible for the first steps of denitrification (nitrate reductase and nitrite reductase) are enriched in sponge symbionts, while enzymes for the final two steps, nitric oxide reductase (norBC, K04561-K02305), and nitrous-oxide reductase (nosZ,