Sponges (phylum Porifera) rank among the oldest extant metazoans and are distributed worldwide across all oceans and major freshwater bodies, displaying various shapes, sizes, and colors, which are possibly influenced by environmental and biotic conditions (Hentschel et al., 2006; Pineda et al., 2015). There are about 8,500 sponge species described to date and likely as many to be described (Van Soest et al., 2012). These sessile, filter-feeding organisms usually shelter dense and complex microbial communities often dominated by diverse, active, and phylogenetically distinct bacteria (Taylor et al., 2007; Kamke et al., 2010; Thomas et al., 2010). Indeed, although sponges intake numerous planktonic microorganisms due to their remarkable filtering activity, their symbiotic communities are taxonomically and functionally different from those found in the water body (Thomas et al., 2010, 2016; Costa et al., 2013). Until now, little experimental evidence exists for the actual participation of sponge symbionts in contributing to host fitness (Webster and Thomas, 2016). However, sponge-associated microorganisms are believed to benefit their hosts through several, eventually interdependent, mechanisms. These include nutrient provision (e.g., through the synthesis of photosynthates and vitamins; Taylor et al., 2007; Siegl et al., 2011); in-host geochemical cycling (e.g., via nitrification; Bayer et al., 2008; Radax et al., 2012), denitrification (Siegl et al., 2011; Fan et al., 2012), or polyphosphate production (Zhang et al., 2015); chemical defense (e.g., via the biosynthesis of polyketides; Piel et al., 2004; Wilson et al., 2014); and removal of metabolic by-products such as ammonia (Webster and Taylor, 2012; Webster and Thomas, 2016) and sulfide (Hoffmann et al., 2005). Particularly, the phylogenetic distinctiveness of the marine sponge microbiome and its vast natural product biosynthesis repertoire have both propelled much research interest in this symbiotic relationship (Taylor et al., 2007; Wilson et al., 2014).

In the last 10 years or so, metagenomics (Handelsman, 2007) and single cell genomics (SCG) (Woyke et al., 2009) approaches coupled to next-generation sequencing (NGS) technologies have become the tools of trend in the inspection of microbial communities thriving in open and host-associated microniches (Handelsman, 2007; Kennedy et al., 2008; Gilbert and Dupont, 2011; Kumar et al., 2015). Functional gene profiling via shotgun NGS revealed that sponge symbiont communities share a suite of common genetic signatures underlying “specific” adaptive strategies such as high frequencies of eukaryotic-like proteins (ELPs), possibly involved in patterns of host–symbiont recognition, and Clustered Regularly Interspaced Short Palindromic Repeats and associated systems (CRISPR-Cas), that may function as a collective anti-viral defense system within the sponge symbiotic consortium (Thomas et al., 2010; Fan et al., 2012; Rua et al., 2015; Horn et al., 2016). However, the frequency and abundance of such genetic elements in other marine microhabitats have not yet been fully examined, making it difficult to diagnose them as exclusive adaptive features of marine sponge symbionts. The linkage between identity and function has been now identified for a number of symbiotic lineages, either via SCG or genome binning from metagenomes (Siegl et al., 2011; Moitinho-Silva et al., 2017; Slaby et al., 2017), greatly increasing our knowledge of the potential physiology of particular sponge-enriched lineages belonging, e.g., to the Cyanobacteria, Proteobacteria, and Poribacteria phyla (Kamke et al., 2013; Gao et al., 2014; Burgsdorf et al., 2015).

In spite of the continued progress enabled by modern cultivation-independent tools, our current understanding of marine sponge microbiome diversity and function mostly derives from comparative studies with the neighboring bacterioplankton (Fan et al., 2012; Rua et al., 2015; Thomas et al., 2016), whereas knowledge of the potential contribution of sediments as sinks and sources of sponge-associated bacteria remains limited. Only recently have studies emerged which investigated sediments in comparative analyses with sponge symbiotic assemblages, using amplicon-based approaches to address the taxonomy and, eventually, in silico functional estimates of the examined communities (Polónia et al., 2014; Thomas et al., 2016). Recent evidence suggests that the density and biochemical composition of particles are major drivers of microbial community structure in aquatic microniches (Zhang et al., 2016). Here, we hypothesize that higher particle/surface availability and cell densities may promote the selection of identifiable traits common to sponge-associated and sediment communities not necessarily favored in planktonic settings. To address this hypothesis, in this study, we tested whether (1) whole taxonomic and functional profiles and (2) abundance distributions of genotypic traits usually regarded as adaptive features of marine sponge symbionts were significantly different across sponge, sediments, and seawater microbial metagenomes.

Spongia officinalis Linnaeus 1759, the first described sponge species, is a canonical bathing sponge (Voultsiadou et al., 2008) displaying widespread occurrence from across the Mediterranean Sea (Dailianis et al., 2011) into the northeastern Atlantic Ocean and beyond. However, signs of population decline as a consequence of human activity, warming temperatures, and bacterial infections have been accumulating in recent years (Webster, 2007; Garrabou et al., 2009). S. officinalis belongs to the chemically rich order Dictyoceratida (Gordaliza, 2010), and as such is the source of diverse biologically active natural products (Gonzalez et al., 1984; Manzo et al., 2011). In spite of the unequivocal economic and societal relevance of S. officinalis, functional information concerning its symbiotic community is scarce. Here, we employ S. officinalis as a model organism to quantitatively address the functional and taxonomic (dis)similarity between sponge, sediment, and seawater microbiomes using shotgun metagenomic sequencing. To reveal the distinctive genomic features of the S. officinalis symbiotic consortium in the context of its natural environment, we used customized pipelines enabling differential abundance analysis of symbionts [i.e., operational taxonomic units (OTUs) set at 97% 16S rRNA gene similarity] and predicted protein families/domains/sites [i.e., InterPro (IPR) entries] across the studied biotopes. Alternative analytical pipelines were used to verify the consistency of the major trends found, and to compare the S. officinalis microbial metagenome with those of other sponge hosts.

The taxonomic distinctiveness of the S. officinalis symbiotic consortium in comparison with those from its neighboring biotopes can be readily observed at the phylum level (Figure 2 and Supplementary Table S2). Consistent with primer-based studies undertaken for other keratose sponges off the Algarve coast (Hardoim and Costa, 2014; Hardoim et al., 2014) and also from the Mediterranean Sea (Erwin et al., 2012a; Pita et al., 2013) and the Great Barrier Reef (Webster et al., 2010), this community is primarily made of a complex mix of so-far uncultivable, sponge-enriched symbiotic bacteria. Owing to our comparative experimental design, we gathered compelling evidence for higher sponge symbiont abundances in sediments than in seawater, revealing an unexpected pattern of distribution of these microorganisms across marine biotopes and extending previous knowledge gained on their occurrence, at very low abundances, in the bacterioplankton (Webster et al., 2010; Webster and Taylor, 2012). Particularly, we identified one possible “generalist par excellence” bacterium in the Rhodobacteraceae clade (OTU 442) which, although clearly being a profuse member of seawater communities, likely performs well both in sediments and sponges. Further, the high prevalence of sponge-enriched Acidobacteria (especially Sva0725 phylotypes), Actinobacteria (Acidimicrobiales phylotypes), and Gammaproteobacteria (several different orders) in sediments adds further layers of complexity to our understanding of sponge symbiont occurrence in the marine realm. Future cultivation-independent, genome-wide studies targeting the adaptive features of these lineages not only hold promise in revealing their likely roles in the sponge endosymbiotic consortium, but may also improve our view of the genetic traits underpinning the persistence of sponge symbionts in the open environment, and consequently of the evolutionary and ecological forces that mediate the dispersal and community assembly of marine sponge symbionts. However, specific studies aiming at uncovering the potential metabolism, linking identity and function, of foundational sponge-associated bacteria are still relatively scarce. SCG and cultivation-independent genome binning from metagenomes have been proven useful in this regard, unveiling, e.g., halogenation capacities within sponge-associated Chloroflexi, Actinobacteria, and Poribacteria spp. (Bayer et al., 2013), non-ribosomal peptide biosynthesis potential within the Chloroflexi (Siegl and Hentschel, 2010) and multiple adaptive features of the keystone sponge-associate cyanobacterium Synechococcus spongiarum (Gao et al., 2014; Burgsdorf et al., 2015). Recently, the ability of several, so-far uncultivable sponge symbionts to utilize carnitine, a quaternary ammonium compound regularly present in the mesohyl matrix of sponges, has been revealed, suggesting parallel adaptation of multiple lineages to a common resource within the in-spongia microniche (Slaby et al., 2017). Our taxon-independent, primer-less sequencing approach revealed a pronounced dominance of one Poribacteria OTU in S. officinalis. It is therefore reasonable to argue that some of the potential metabolic features recently revealed for poribacterial symbionts by means of SCG (Siegl and Hentschel, 2010; Siegl et al., 2011; Kamke et al., 2013) are likely to mediate major bioprocesses and molecular interactions within the S. officinalis endosymbiotic consortium. These features include, among others, polyketide biosynthesis capacities (possibly involved in host’s chemical defense), a vast, specialized carbohydrate degradation repertoire (considered pivotal to host’s nutrient provision), and enrichment of eukaryotic-like repeat proteins (e.g., TRPs, ANKs, LRRs, usually considered to enable symbionts to evade phagocytosis by the host) all of which could be verified, from our community functional profiles, as characteristic of the S. officinalis microbial metagenome. Because these attributes have been commonly verified in diverse sponge symbiont lineages (Slaby et al., 2017), it can be argued that they contribute significantly to the observed difference in taxonomic assembly between sponges, sediments, and seawater observed here.

One important finding in this study was the observation that several of the features identified as genomic signatures of the S. officinalis microbiome displayed higher abundances in sediments than in seawater. Among these traits we highlight IPR entries underlying the coding of an array of ELPs or involved in plasmid assembly, stability and conjugative transfer (e.g., plasmid replication, toxin–antitoxin systems and type IV secretion IPRs), secondary/cytotoxic metabolite biosynthesis (e.g., polyketide and terpene/terpenoid synthases, TolB-like and cytochrome P450 IPRs), remediation of oxidative stress (a luciferase-like domain), organic carbon utilization (e.g., sulfatases) and literally hundreds of other individual IPR entries (Figure 5, Supplementary Figure S2, and Supplementary Table S4). Therefore, some of the features previously regarded as “unique” adaptations of the sponge symbiotic consortium may be well represented in other marine settings. Below, we give emphasis to the above-mentioned functions and discuss their patterns of occurrence across bacterial genomes and the marine biotopes studied here.

Inspection of the IPR entries contributing the most to variation in the functional dataset (Figure 5) revealed the consistent prevalence of ELPs (TRPs, ANKs, LRRs, and WD40) among the most sensitive IPRs differentiating the studied biotopes, all of which were enriched in S. officinalis (Supplementary Figure S2). The abundance of TRPs and ANKs in sponge microbiomes has been well documented (Thomas et al., 2010; Fan et al., 2012), and a role for these ELPs in preventing phagocytosis of bacterial symbionts by the sponge host has been proposed (Nguyen et al., 2014; Reynolds and Thomas, 2016). In the present study, contrary to previous reports addressing other sponge hosts (Thomas et al., 2010; Fan et al., 2012), WD40 repeats were by far the most abundant ELP repeat in the S. officinalis microbiome, with several entries varying markedly in abundance across the surveyed biotopes (Figure 4 and Supplementary Table S4). WD40 repeats are regarded as prevalent in eukaryotes and uncommon in prokaryotes, and act as a protein–protein or protein–DNA platforms to allow for various protein complex assemblies in cellular metabolism (Xu and Min, 2011; Wang et al., 2015). However, evidence from this study and elsewhere (Díez-Vives et al., 2016; Reynolds and Thomas, 2016) is now accumulating for a broad distribution of these motifs among sponge symbiotic bacteria, offering a new angle from which the spread of these macromolecule network hubs can be seen throughout the tree of life. Collectively, the presence of ELPs in prokaryotic genomes has been interpreted as suggestive of lateral host–microbe gene transfer given their presumed eukaryotic origin (Horn et al., 2016). Recently, ELPs were shown to be positively expressed in sponge microbial metatranscriptomes (Díez-Vives et al., 2016), supporting their likely importance in mediating cell–cell interactions within the sponge holobiont. In particular, the expression of WD40 repeats was found to be associated with domains of the Tol-dependent translocation system, which is involved in outer membrane integrity, cell invasion and, eventually, pathogenesis of Gram-negative bacteria, suggesting a pivotal role of these repeats in host–microbe interactions (Díez-Vives et al., 2016). The enriched abundance of both WD-40 repeats and one TolB-like domain (IPR011042) in the S. officinalis microbial metagenome speaks for distinguishing host colonization capacities and/or virulence potential within this symbiotic consortium.

Polyketides have been intensively studied as sponge-derived natural products whose biosynthesis is primarily mediated by bacteria (Piel, 2002; Piel et al., 2004; Wilson et al., 2014), and are thought to play a role in defense of the sponge host against natural enemies, as demonstrated for the bryozoan host Bugula neritina (Lopanik et al., 2004). Terpenes and terpenoids, in their turn, encompass a large class of natural products commonly regarded as of fungal and plant origin whose biosynthesis by bacteria is attracting increasing research interest (Yamada et al., 2015). The most abundant IPR entry related with terpene/terpenoid biosynthesis in S. officinalis (IPR008930) corresponds to a family of terpenoid cyclases/protein prenyltransferases responsible for a wide chemodiversity of terpenoid natural products (Christianson, 2017). Considering the broad distribution of terpene/terpenoid systhase genes across bacterial genomes (Yamada et al., 2015), it is tempting to argue that terpenoid biosynthesis in S. officinalis, and marine sponges in general, could be as well mediated by bacterial symbionts, emerging as a further mechanism possibly conferring host defense against natural enemies or mediating microbe–microbe interactions within the sponge host. Likewise, cytochrome P450 enzymes (IPR001128) are a superfamily of monooxygenases presenting broad substrate spectrum, being widespread in all domains of life. Particularly in bacteria, they are important in the biosynthesis of secondary metabolites such as erythromycin, and bear potential for applications in synthetic biology and the pharmaceutical industry (Girvan and Munro, 2016). Taken together, these observations suggest high microbially driven chemical complexity within the S. officinalis holobiont. Such a vast secondary metabolite repertoire may play pivotal roles in microbiome community assembly, host–symbiont signaling, and host defense. Widely known for their key role in bioluminescence, bacterial luciferases are flavin monooxygenases which incorporate or reduce molecular oxygen in redox reactions, and may have originally evolved as enzymes responsible for reactive oxygen species detoxification (Szpilewska et al., 2003). The sharp enrichment of this trait in the S. officinalis microbiome, followed by sediments, leads us to posit that these enzymes may primarily act as anti-oxidant agents in these particular settings, along with other anti-oxidant enzymes known to be enriched in sponges such as glutathione peroxidases (Thomas et al., 2010), observed here to possess high abundance in both sponges and sediments (Supplementary Table S4, IPR IPR000889). Enrichment in sulfatases/aryl sulfatases have been suggested as a specialization of marine sponge symbionts enabling them to utilize sulfated polysaccharides from the host’s extracellular matrix (Slaby et al., 2017). Sulfatase-encoding genes were abundant not only in the S. officinalis (Supplementary Table S4 and Figure 5) metagenome, but ranked as one major genetic signature of several sponge-associated microbiomes (Supplementary Table S5). Here, we reveal that this trait is equivalently enriched in both sponge and sediment biotopes in comparison with seawater, providing evidence for the common selection of fundamental nutrient acquisition capacities in phylogenetically contrasting microbiomes. In the context of the marine sponge holobiont, sulfatases are supposed to be involved in nutritional exchange between host and microbes, playing a vital role in the cycling of sulfur within the animal.

Altogether, the outcomes delineated above indicate closer resemblance in functional attributes between sponges and sediments than sponges and seawater: a hypothesis corroborated by Bray–Curtis dissimilarity measures calculated for the three biotopes based on the whole array of 10,272 IPR entries uncovered from the data (Table 1). However, the quantitative trend revealed with whole functional profiles must be considered with caution since statistical significance varied depending on data processing methodology and on the reference database employed (Supplementary Appendix S1). Importantly, the level of phylogenetic disparity between all microbiomes was high (Table 1) in spite of our observation for higher residual symbiont abundances in sediments than in seawater (see above). Therefore, it is likely that surface sediments and endosymbiotic sponge communities, although being chiefly composed by different microbial populations (especially regarding their very dominant members) display a certain degree of independent functional convergence. This prompts us to argue that selective pressures common to particle- and host-associated modes of living constitute an important evolutionary force shaping functional assembly in marine biomes. Several factors, ranging from cell–cell interactions to availability (in quality and quantity) of solid surfaces for cell attachment to modes of symbiont acquisition and release by sponges, may contribute to the observed trends. Microbial cell densities alone, known to be about three orders of magnitude higher in coastal sediments (Schmidt et al., 1998) and in Dictyoceratida sponges (Hardoim et al., 2012) than in seawater, may be a key factor promoting genetic exchange and adaptive features likely to prevail in the former biotopes. In highly dense circumstances, gene clusters involved in the biosynthesis of natural products such as polyketides and terpenes—or more specifically terpene-quinones very often enriched in Dictyoceratida species such as S. officinalis (Gordaliza, 2010; Manzo et al., 2011; Li et al., 2017)—are likely to confer selective advantage to its carriers. Similarly, strategies to neutralize cytotoxic effects are likely to elicit a selective advantage in communities where inhibitory compounds abound. ABC transporters (Figure 5) comprise a large family of bacterial trans-membrane proteins mediating the import and export of small and large-sized molecules throughout the cell, and may play a fundamental role as detoxifying agents permitting microbial survival in competitive microniches. Particularly, we found permeases within the ABC transporter category (e.g., IPRs 001851, 0038381, and 025966, Supplementary Table S4) with presumed, manifold detoxifying functions commonly abundant in sponges and sediments. Polyketide synthases, type IV secretion and ABC transporter-encoding genes have all been detected on plasmids from several microorganisms (Stinear et al., 2004; Kadlec and Schwarz, 2009; Bruto et al., 2017). The prevalence of these genes along with the higher incidence of plasmid, transposase and ELP-encoding genes (which by themselves speak for greater genetic exchange potential) in the S. officinalis microbiome, followed by sediments, hints at a possible convergent selection of these traits in phylogenetically divergent microbial communities. Future studies aiming to define the gene content of the community of circular plasmids present in marine sponges will certainly shed new light on the functional features more likely to traffic about in the mobile gene pool within the Porifera.

Physical connectivity between sponges and sediments, although usually given less importance in microbiology studies than seawater intake via filtering, takes place by the capture of particulate organic matter and particles in suspension by the sponge host (Schönberg, 2016). In addition, loss of sponge cells through shedding and sponge-expelled detritus, both found to be significant processes in sponge cell turnover (Alexander et al., 2014), may act as substantial inputs of sponge-associated microorganisms into superficial sediment layers. Thus, marine sediments may serve as both sources and sinks of sponge-associated microorganisms, but the magnitude and relevance of this exchange remains to be addressed. The moderate abundance of a few dominant sponge symbionts in sediments indicates that these bacterial lineages, if not optimal performers, are capable of persisting—for undetermined periods—at considerable densities in this alternative habitat, thereby enhancing their probability of future lateral acquisition by the sponge host. Identifying an active role beyond environmental endurance for these lineages in the complex microbiome of marine sediments is challenging. It is known that several factors such as seawater temperature and the composition and age of biofilm and biofouling communities are decisive for invertebrate larval settlement in benthic ecosystems (Hadfield, 2011; Whalan and Webster, 2014). Therefore, it could be argued that increased inter-domain signaling between host larvae and a seeding community of competent sponge associates on particulate/hard substrate—or microbe–microbe signaling in such circumstances—may contribute to higher larval settlement rates in favorable microniches, promoting the selection, on the sea floor, of sponge symbiont lineages able to persist in the open environment.

Contrasting the trends discussed above, microbial genes involved in motility and chemotaxis were altogether more prevalent in seawater and sediment communities, and much less abundant in the sponge host. The ability to move and orchestrate movement in response to chemical cues and gradients are widely acknowledged as imperative mechanisms dictating the distribution of microorganisms in the oceans (Stocker and Seymour, 2012) and as quintessential features of host-associated bacteria (Wadhams and Armitage, 2004; Rawls et al., 2007). Here we show that the S. officinalis endosymbiotic consortium displays low abundance of genomic features involved in chemotaxis and flagellar, gliding and fimbriae-based motilities when compared to its surrounding environment, supporting the idea that loss of motility may be common among prevalently vertically transmitted symbionts (Bright and Bulgheresi, 2010). Or, alternatively, for symbionts whose mode of acquisition by the host is rather passive from the microbial standpoint. Particularly relevant in distinguishing sediments from sponges and seawater regarding the regulation of virulence and motility were the higher abundances of GGDEF and EAL protein domains and of type II secretion proteins in sediments. The above-mentioned domains modulate the concentrations of cellular c-di-GMP, a signaling molecule involved in the regulation of biofilm formation, virulence, motility and cell surface adhesiveness in Gram-negative bacteria (Argenio and Miller, 2004). Indeed, increased cellular c-di-GMP was found to promote type II secretion activity in Vibrio cholerae (Beyhan et al., 2006). Therefore, signal transduction via c-di-GMP and its modulation appears to be a determining factor in shaping the virulome of marine sediments in a singular fashion. Considering the S. officinalis endosymbiotic consortium, it is likely that this community essentially consists of “sit-and-wait” performers regarding their nutrient foraging strategies, especially if it is assumed that filtering activity alone is responsible for the total import and distribution of organic carbon and energy into the host. Our dedicated sampling of the inner sponge body disregards the profuse and complex community of epibionts known to populate the pinacoderm of keratose sponges, where photosynthetic Cyanobacteria are favorably selected (Erwin et al., 2012b) and motility and chemotaxis traits may be relevant for colonization and biofouling processes. The consistent trend found here for a primarily heterotrophic, less-motile community of endosymbionts highlights the need of approaching distinct microniches within marine sponges for a better understanding of microbiome spatial distributions and dynamics in these hosts (Webster and Thomas, 2016).

Finally, we detected much higher incidence of CRISPR-Cas and restriction endonucleases in the S. officinalis microbiome than in seawater—in accordance with earlier metagenomics surveys (Thomas et al., 2010; Fan et al., 2012; Horn et al., 2016)—and sediments. Therefore, in the context of its two immediate environmental surroundings, the enrichment of both defense mechanisms can be indeed considered a true hallmark of the S. officinalis microbiome, and much likely of marine sponges in general. Much has been discussed on the diversity (Fan et al., 2012; Horn et al., 2016) and role of these genetic elements as an efficient, specific anti-phage defense system permitting bacterial survival within the sponge microbial consortium (Thomas et al., 2010; Fan et al., 2012; Horn et al., 2016). In agreement with this hypothesis, we here observed that the relative abundance of both defense systems and of bacteriophages were inversely correlated in S. officinalis and seawater, where viral particles were 13-fold more frequent than in sponges and CRISPR-Cas were virtually absent. However, low abundances of both defense systems and of viral DNA were detected in sediments, suggesting that viral populations might be regulated by other mechanisms in these settings rather than high abundance of CRISPR-Cas and R-M systems alone. In this regard, it was curious to note that the diversity and assemblage of restriction endonucleases uncovered from sponges and sediments was fairly comparable, with higher abundances in sponges being the primary factor distinguishing these biotopes (Supplementary Table S4). Future efforts are therefore needed to disentangle the relative forces exerted by CRISPR-Cas and restriction-modification systems on the regulation of viral populations within the Porifera and across marine biomes. To this end, better understanding of the structure of phage communities in host- and particle-associated settings will be much required.

In conclusion, the comprehensive comparative metagenomics strategy employed in this study enabled us to critically assess the distribution of genomic features involved in symbiosis across marine habitats, and to address functional convergence vs. divergence in contrasting marine microbial communities more thoroughly. We advocate that such an approach, which in the future shall include the assessment of other invertebrate hosts, is imperative for a holistic understanding of microbial community dynamics and function in marine sponges and benthic ecosystems at large.

This study relied on in situ sampling of microorganisms from marine invertebrates without a nervous system, and as such was exempt from ethical approval procedures according to the current Portuguese legislation (Decreto-Lei n° 113/2013). This study did not occur within privately owned or protected areas. This study did not involve endangered or protected species. The sampling methodology privileged minimally invasive handling procedures, following the guidelines of the European Directive 2010/63/EU.

RC designed the study. EK, JG, JX, and RC performed the experiments. MPR, JG, JX, and RC provided reagents and materials. EK, MR, and RC analyzed the data. EK and RC wrote the main manuscript text and prepared figures and tables. All authors reviewed the manuscript.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.