Introduction

A wide range of animals are found in symbiosis with microorganisms, from basal metazoans, such as marine sponges (phylum Porifera), to humans (phylum Chordata) (Taylor et al., 2007; Cho and Blaser, 2012). Animal–microbe symbioses encompass a variety of acquisition modes (that is, vertical, horizontal and a combination of both; for example, Cary and Giovannoni (1993); Kikuchi et al. (2007); Schmitt et al. (2008)), and levels of diversity (ranging from monoclonal populations to highly complex assemblages; Martens et al. (2003); Schmitt et al. (2012)). Additionally, a gradient of interactions may occur, ranging from competition to mutualism (Dethlefsen et al., 2007; Coyte et al., 2015; Thomas et al., 2016). To account for this spectrum of patterns, Douglas and Werren (2016) have recently argued that host–microbe symbioses are productively approached as ecological communities. Therefore, to further our understanding of the biology of animal–microbe symbioses, research should contemplate the members of the association, that is, the microbial populations and the host, and how they relate to each other.

Marine sponges are filter-feeders found in a wide range of benthic environments, across gradients of depth and latitude (Wulff, 2012). Sponges perform a broad range of functional roles in benthic habitats (Bell, 2008), including the retention of nutrients and energy within the coral reef through the consumption of dissolved organic matter (DOM) and high cell turnover rates (Yahel et al., 2003; de Goeij et al., 2013). In addition to DOM, sponges feed on detritus and microorganisms of diverse sizes (from 0.5 to 70 μm) (Marta et al., 1999; Hadas et al., 2009). Owing to their basal position in metazoans phylogeny and their association with microbial communities, sponges are emerging as models for the evolutionary study of animal–microbe symbiosis (for example, Fan et al., 2012; Ryu et al., 2016).

Microbial communities can account for up to 40% of the sponge tissue volume and often contain novel and uncharacterized diversity (Taylor et al., 2007; Schmitt et al., 2012). Microorganisms are generally found in the animal’s extracellular matrix (mesohyl), although they are not restricted to this location (Taylor et al., 2007). Owing to the lack of culturable representatives of the microbiome, most of the understanding of the functional roles of microbial symbiont have been acquired by approaches either centered on the community, for example, using meta-omics, 16S rRNA gene analysis and compound-uptake experiments, or on specific genomes, pathways and genes (reviewed by Taylor et al., 2007 and Webster and Thomas 2016). Together, these studies have revealed important metabolic attributes of sponge-associated microorganisms, such as different energy acquisition strategies (for example, heterotrophism and autotrophism), complex element cycling, functional equivalence, production of bioactive compounds and nutrient transfer to the host (Hoffmann et al., 2009; Fan et al., 2012; Freeman et al., 2013; Kamke et al., 2013; Wilson et al., 2014). Nevertheless, the metabolic interactions among the microbiome members and between them and the host cells are rarely explored and thus remain largely unknown (Webster and Thomas, 2016).

The sponge Cymbastela concentrica contains consistently dominant and uncharacterized alphaproteobacterial symbionts, particularly within the family Phyllobacteriaceae (Taylor et al., 2004; Fan et al., 2012; Esteves et al., 2016) as well as community members belonging to the Gammaproteobacteria, the Nitrospira and the archaeal phylum Thaumarchaeota among others (Fan et al., 2012; Esteves et al., 2016). In addition, C. concentrica hosts an abundant population of diatoms (Taylor, 2005). The overall community composition and function of the C. concentrica microbiome have been previously characterized on the community level using metagenomics, metaproteomics and metatranscriptomics (Díez-Vives et al., 2016; Thomas et al., 2010; Liu et al., 2011; Fan et al., 2012; Liu et al., 2012). Here we build on these data sets to unveil nutritional interactions in C. concentrica. Towards this goal, we reconstructed three new prokaryotic genomes and estimated their activity by analyzing metatranscriptomic sequences, while metabolic features of the sponge and diatoms were inferred from an eukaryotic transcriptomic data set produced from C. concentrica.

Materials and methods

Genome assembly, binning and annotation

Metagenomic shotgun sequencing reads for three replicates (BBAY40, 41 and 42) of the microbiome of C. concentrica from a previous study (Fan et al., 2012) were co-assembled using the Newbler assembler (Roche, Penzberg, Germany). Sequence reads for each replicate were then mapped back against assembled contigs >500 bp using the Bowtie aligner (Langmead et al., 2009). Coverage file and contigs were then used to bin and refine genomes using the GroopM tool (Imelfort et al., 2014). Genome bins were checked for completeness and contamination using CheckM (Parks et al., 2015) and then annotated using the IMG system (Markowitz et al., 2014). Genome assembly and annotation are avaiable at IMG (http://img.jgi.doe.gov) under the genome IDs 2608642157, 2608642160 and 2626541593. Phylogenetic analysis of 16S rRNA genes and deduced amino-acid sequences for those bins are described in Supplementary Information.

Transcriptomic analysis

We used the metatranscriptomic reads from the microbiome of three C. concentrica individuals (Díez-Vives et al., 2016) to estimate the gene expression of binned genomes. Briefly, mRNA was previously enriched from three individuals (ind1, ind2, and ind3; biological replicates), which were processed with Dynabeads Oligo (dT)25 (Thermo Fisher Scientific, Waltham, MA, USA) and the Ribo-Zero Bacteria Kit (Epicentre, Madison, WI, USA) in technical replicates (aR and bR). Enriched prokaryotic mRNA was sequenced with HiSeq2000 platform (Illumina, San Diego, CA, USA) with 100 bp paired-end reads. Transcript expression for the coding DNA sequences (CDS) of each genome bin was estimated with RSEM v. 1.2.21 (Li and Dewey, 2011) with default parameters, which implements the Bowtie aligner (Langmead et al., 2009). Transcripts per million (TPM) measures were used as estimations of expression (Li et al., 2010; Wagner et al., 2012). Gene expression estimates in each sponge individual were obtained by averaging TPM values of the technical replicates. Only genes with mapped reads (replicate-average TPM>0) coming from at least two of the three sponge individuals were considered expressed. The global expression estimate for each gene was obtained by averaging TPM values (TPMav) across the three sponge individuals.

The analysis of eukaryotic transcripts is described in Supplementary Information. Briefly, transcripts were taxonomically sorted using the Lowest Common Ancestor algorithm implemented in MEGAN v. 5.11.3 (Huson et al., 2011). Peptides deduced from predicted coding sequences were annotated based on the Kyoto Encyclopedia of Genes and Genomes (KEGG) Orthology (KO). Expression estimates of KEGG functions were calculated by summing TPM values (TPMsum) across coding sequences.

Localization analysis of symbionts

Three individuals of C. concentrica were collected by SCUBA diving from Botany Bay near Bare Island, Sydney, Australia (33.99222° S, 151.23111° E) during October 2015. Specimens were processed, fixed and dehydrated. Probes specific for the obtained genome bins (that is, CcPhy, CcNi and CcThau; see below) were designed and cells were detected using fluorescence in situ hybridization. Sample processing, probe design and fluorescence in situ hybridization conditions are fully described in Supplementary Information.

Results and discussion

Genomes and transcriptomes of the dominant community members in C. concentrica

Three genomes were reconstructed (Table 1) from a previous metagenomic sequence data set of the microbial community of C. concentrica (Fan et al., 2012). Phylogenetic analysis of the 16S rRNA gene sequence recovered from one of the genomes (Ga0068441) placed it with members of the family Phyllobacteriaceae, order Rhizobiales, class Alphaproteobacteria (Supplementary Information; Supplementary Figure S1). Thus the genome will be hereafter referred to as CcPhy (C. concentrica Phyllobacteriaceae). The 16S rRNA gene sequence is only 96% similar to its closest cultured bacterium (Mesorhizobium amorphae CCNWGS0123) and has >99% similarity to sequences that have been previously found in C. concentrica (for example, GenBank: AY942778.1). The cluster represented by these C. concentrica-derived sequences represents a distinct and novel clade within the Phyllobacteriaceae (Supplementary Figure S1). The 16S rRNA gene sequence of the second genome (Ga0068443) was phylogenetically close to other sequences derived from several sponge species (for example, Agelas dilatata, Geodia barretti, Xestospongia testudinaria). These sequences were related yet distinct from sequences of the lineage IV of the Nitrospira genus within the phylum Nitrospirae (Supplementary Information; Supplementary Figure S2). The genome will be hereafter named CcNi. The 16S rRNA gene sequence of the last genome obtained (Ga0078905) clustered within the candidate order Nitrosopumilales (known as group I.1a) (Konneke et al., 2005; Stieglmeier et al., 2014b) within the Thaumarchaeota phylum (Supplementary Information; Supplementary Figure S3). It will be referred as CcThau. The 16S rRNA sequence from CcThau was 99% similar to the sequence from Candidatus Nitrosopumilus sediminis AR2 (GenBank: CP003843.1). According to studies based on OTUs recovered from metagenomic data sets and 16S rRNA gene sequencing, these three organisms are among the most dominant members of the C. concentrica microbiome (Thomas et al., 2010; Fan et al., 2012; Esteves et al., 2016). For example, based on 16S rRNA gene sequencing, the Phyllobacteriaceae OTU may account approximately 17% of all bacteria in the sponge, while the Nitrospira OTU has a relative abundance of ~6% (Thomas et al., 2010). Metagenomic sequencing further indicated that the thaumarchaeal organism has similar relative abundances than the Nitrospira bacterium (Fan et al., 2012).

Table 1 Assembly and annotation statistics of genome bins
Full size table

CcThau has the largest genome assembly with a size of 2.16 mega basepairs (Mb), followed by CcPhy with 2.11 Mb and CcNi with 1.59 Mb (Table 1). Genome completeness estimations based on lineage-specific marker genes were higher in CcThau (97.12%) in comparison to CcPhy (69.17%) and CcNi (66.56%). The estimated contamination of genome bins was low, ranging from 0% to 4.01%. GC content varied from 38.42% for CcThau and 48.03% for CcNi to 60.34% for CcPhy. Metatranscriptomic sequencing reads mapped to these genomes at different extent, where most of the reads mapped to the CcThau CDS (88 303±23 580; mean±s.d.), followed by CcPhy (16 540±6128) and CcNi (7054±3542) (Supplementary Table S1). Expressed genes represented 86.7% of the total CDS in CcPhy (2092 of 2411), 80.5% in CcNi (1353 of 1679) and 64.6% in CcThau (1775 of 2745). High gene expression correlation among biological replicates, as estimated by TPM values, indicated low variability between sponge individuals (Supplementary Information; Supplementary Figures S4). Average TPM values (TPMav) of expressed genes across sponge individuals varied from 1.09 to 27 022.43, with a median of 191.54 in CcPhy, 354.2 in CcNi and 49.18 in CcThau (Supplementary Figure S7).

A heterotrophic, facultative anaerobic metabolism of CcPhy

The genomic and transcriptomic data suggest that the population represented by CcPhy has a heterotrophic metabolism, which is broadly consistent with other members of the family Phyllobacteriaceae (Willems, 2014). The CcPhy expressed an almost complete set of genes for glycolysis (Embden–Meyerhof–Parnas pathway) (TPMav 24–343; lower value−higher value), as well as the anabolic enzyme fructose-1,6-bisphosphatase (TPMav 281±39; mean±s.d.) and tricarboxylic acid (TCA) pathways (TPMav 121–1396), which are linked by the pyruvate dehydrogenase complex (TPMav 88–1316) (Figure 1; Supplementary Table S2). Additionally, CcPhy is able to process sugars via the pentose phosphate pathway. In support of a heterotrophic lifestyle, approximately 13.5% of the protein-coding genes (n=328) were involved in membrane transport (according to Transporter Classification Database (TCDB); Supplementary Table S2). From these, 164 CDS were assigned to the adenosine triphosphate (ATP)-binding cassette (ABC) superfamily (TC: 3.A.1), which were among the most expressed genes in the CcPhy transcriptome (Figure 2). The most expressed genes classified by TCDB encoded for components of different transport systems involved in the uptake of general L-amino acids, glycine betaine/proline, multiple sugars, spermidine/putrescine, monosaccharides, mannitol/chloroaromatic compounds, amino acid/amide and manose/fructose/ribose. Genes encoding for Amt-type ammonium transporters (TPMav 1729±351) and the glutamine synthase-glutamine oxoglutarate aminotransferase pathway (TPMav 168–1157) were expressed, allowing for import and assimilation of extracellular ammonium (Javelle et al., 2005).

Figure 1
figure 1

Predicted model of integrated metabolic processes within Cymbastela concentrica. Selected pathways and gene functions are shown. Transporters with at least partial genomic evidence and expression detected were included. Inferred nutrient fluxes (A–O) are discussed in the section ‘Model of integrated metabolism of the microbial community members in C. concentrica’. Expressed genes are listed in Supplementary Tables S2.

Full size image
Figure 2
figure 2

Most expressed CcPhy genes related to membrane transport. (a) CDS assigned to TCDB were grouped by families and super families. Groups were ordered according to maximum TPMav values (x axis), and only the top 15 are shown. For reference, CDS not classified by TCDB are shown (not in TCDB). The median of all expressed genes is represented by the dashed line. (b) Most expressed transport-related genes. Gray scale refer to gene family or superfamily. Genes are ordered according to TPMav values (x axis). Whiskers represent s.d. of biological replicates (n=3).

Full size image

Genes for the malate synthase and isocitrate lyase (TPMav 211–605) were also expressed. These two enzymes are key for the glyoxylate cycle and their expression is an indication of aerobic growth on fatty acids and acetate (White, 2007). Active uptake of fatty acids could be carried by the product of expressed genes encoding for the proposed fatty acid transporter family (TPMav 80–790). Several genes, whose products are involved in the fatty acid β-oxidation pathway and the downstream propanoyl-CoA degradation pathway were also found expressed (TPMav 88–790) (Supplementary Table S2). Despite existing characterization of the fatty acid contents in sponges and associated bacteria (Gillan et al., 1988; Koopmans et al., 2015), this is to our knowledge the first report of heterotrophic fatty acid-based metabolism of bacteria associated with sponges. Future research focusing on the importance of fatty acids as energy source for sponge microbiome members and the origin of these compounds has the potential to elucidate the ecological relevance of this metabolic feature.

Creatinine degradation via creatine is used by bacteria to assimilate carbon and/or nitrogen (Kim et al., 1986; Shimizu et al., 1986). CcPhy expressed genes related to the degradation of creatine to glycine, that is, creatinase (TPMav 151±25) and the sarcosine oxidase alpha, beta, gamma and delta subunits (TPMav 122–296). A gene encoding for creatinine amidohydrolase was present in the genome but not expressed. Urea is a product of this pathway, which is generated by the hydrolysis of creatine by creatinase (Shimizu et al., 1986). The genome bin CcPhy lacked genes encoding for enzymes involved in the breakdown of urea into ammonium and inorganic carbon (Solomon et al., 2010). Because the genome is not complete and the fact that urease activity has been described in the species of the Phyllobacteriaceae family (for example, Peix et al., 2005), the existence and expression of Phyllobacteriaceae-like urease genes was further investigated for the C. concentrica microbiome. We found transcripts encoding for sequences similar to urease subunits and accessory proteins in the metatranscriptome of C. concentrica (Díez-Vives et al., 2016), but none were related to sequences from Phyllobacteriaceae (Supplementary Information; Supplementary Table S3), further corroborating the absence of urea breakdown in CcPhy.

Members of the Phyllobacteriaceae family are described as having respiratory metabolism with oxygen as the terminal electron acceptor, while anaerobic nitrate respiration and facultative chemolithotrophic metabolism also exist for some members of the family (Willems, 2014). CcPhy has the genetic repertoire to thrive in a variety of oxygenic conditions. The genome encodes for the subunits of two different heme–copper terminal oxidases, the mitochondrial-like aa3-type cytochrome c oxidase and the cbb3-type cytochrome c oxidase. Both these oligomeric cytochrome complexes transfer electrons of the respiratory chain to its final destination, oxygen, but have different O2 affinities and can be differentially regulated according to available oxygen concentrations (Garcia-Horsman et al., 1994; Morris and Schmidt, 2013). For instance, in Bradyrhizobium japonicum (family Bradyrhizobiaceae) the low affinity aa3-type is favored in fully aerobic conditions, while the high affinity cbb3-type is used in microaerobic conditions (Gabel and Maier, 1993; Preisig et al., 1996; Swem et al., 2001).

Genomic data also indicated that, in the absence of O2, CcPhy can perform nitrate (NO3−) respiration, where the membrane-bound nitrate reductase NarGHI complex transfers the electrons to nitrate, reducing it to nitrite (NO2−; Bertero et al., 2003). In this case, it is likely that nitrite is excreted via a nitrate/nitrite transporter NarK as the final product of the CcPhy anaerobic respiration. Nitrite excretion is likely because the genome lacked genes encoding for enzymes capable to further reduce nitrite to nitric oxide (NO), that is, a copper or a cytochrome cd1-containing nitrite reductase (NirK and NirS, respectively), or to ammonium (NH4+), that is, cytochrome c nitrite reductase (NrfA) (Kraft et al., 2011; Zheng et al., 2013). The components and related genes of the cytochrome c oxidases and respiratory nitrate reductase were all expressed and also at similar levels for the aerobic, microaerobic and anaerobic respiratory pathways (Supplementary Figure S8). It is unlikely that any given individual cell within a population performs all three modes of respiration at the same time and therefore our observation indicates that subpopulations within the sponge tissue experience distinct oxygen environments.

Nitrite-dependent chemolithoautotrophy and use of organic nitrogen compounds by CcNi

Nitrospira are known to obtain energy by oxidizing nitrite to nitrate using the key enzyme nitrite oxidoreductase (NXR) (Spieck and Bock, 2005). The NXR of Nitrospira was proposed to be a membrane-bound periplasmic complex composed of at least two subunits, the subunits alpha (nxrA) and beta (nxrB) (Spieck et al., 1998; Lucker et al., 2010). Two adjacent genes encoding for nxrA (gene locus Ga0068443_101117) and nxrB (Ga0068443_101118) were found in CcNi (see Supplementary Information for phylogenetic analysis) representing the fourth and second most abundant transcripts in CcNi transcriptome (TPMav 17 252±9822 and TPMav 26 308±13 520, respectively; Figure 3). Electron transport and ATPase encoding genes were also expressed (Supplementary Table S2), supporting a nitrite-dependent chemolithotrophic electron flow (Lucker et al., 2010). Expressed genes related to other nitrogen transformations were also found and are discussed in Supplementary Information. Taken together, these results indicate an active chemolithotrophic conversion of nitrite to nitrate by the CcNi population (Figure 1).

Figure 3
figure 3

Expression of selected functional genes of CcNi. TPMav values are shown for genes of selected functions. ‘Others’ stands for genes not grouped in the specific functions shown. The median of all expressed genes is represented by the dashed line. Highlighted genes encode for the nitrite oxidoreductase alpha (nxrA), beta (nxrB), putative C (nxrC) subunits (Supplementary Information), cyanate lyase (cya) and pyruvate ferredoxin oxidoreductase (PFOR).

Full size image

Few Nitrospira species can decompose urea using the resulting ammonia for nitrification and biomass production, where urea would be imported from the extracellular milieu by the urea ABC transporter and hydrolyzed by urease (Koch et al., 2015; van Kessel et al., 2015). CcNi possess an putative operon with genes related to the transport and decomposition of urea, including those encoding for subunits of urease, urea ABC transporter, as well as auxiliary proteins (Ga0068443_100614-16, Ga0068443_100619-27) (Supplementary Figure S9). Some of these genes were highly expressed, such as gene for the urtA urea transport system substrate-binding protein (TPMav 2458±357), the urease subunit alpha (TPMav 1332±420) and the urtD and urtE urea transport system ATP-binding proteins (TPMav 978–1255; Figure 3). We did not find genomic evidence for the internal production of urea by CcNi, such as genes encoding for agmatinase or arginase. Because nitrite-oxidizing bacteria encode urea-producing enzymes (Palatinszky et al., 2015), the possibility of Nitrospira-like genes encoding for agmatinase or arginase to be expressed in C. concentrica was investigated. Transcripts encoding for sequences similar to agmatinase or arginase were found in the metatranscriptome of C. concentrica (Díez-Vives et al., 2016), but none of them were related to Nitrospira (Supplementary Information; Supplementary Table S4). This result further supports the absence of pathways for internal production of urea in CcNi.

The gene encoding for cyanate lyase (cya) was abundantly expressed in the transcriptome (TPMav 3712±1501) of CcNi (Figure 3). Cyanate lyase decomposes cyanate to CO2 and ammonia in a reaction dependent of bicarbonate (Johnson and Anderson, 1987). Its activity was described in N. moscoviensis and its proposed biological roles include nitrogen assimilation, cyanate detoxification and energy acquisition (Anderson et al., 1990; Palatinszky et al., 2015). The cya gene was localized downstream from two genes encoding for proteins of the formate/nitrite transporter family (TPMav 513–904) (Supplementary Figure S9). Members of the formate/nitrite transporter family (TC 2.A.44) have broad specificity for small monovalent anions (Lu et al., 2013). It is likely that CcNi putative formate/nitrite transporter is also permeable to cyanate, as similarly inferred for the ammonia-oxidizing archaea Candidatus Nitrososphaera gargensis due to the proximity of genes for transporter and enzymatic degradation (Spang et al., 2012).

Nitrospira species are proposed to fix carbon via the reductive TCA (rTCA) cycle (Starkenburg et al., 2006; Lucker et al., 2010, 2013; Koch et al., 2015). The rTCA cycle shares most of its enzymes with the oxidative TCA (oTCA) cycle, with the exception of three key enzymes, which are characteristic of the reductive cycle: ATP citrate lyase, 2-oxoglutarate:ferredoxin oxidoreductase (also named as α-ketogluterate synthase), and fumarate reductase (Berg, 2011). Alternatively, ATP-dependent cleavage of citrate can be carried by citryl-CoA synthetase and citryl-CoA lyase enzymes (Aoshima et al., 2004a,2004b). CcNi encoded enzymes that are shared between the two forms of TCA cycles. In addition, CcNi encoded a pyruvate ferredoxin oxidoreductase, which can act as pyruvate synthase (EC 1.2.7.1), carboxylating acetyl-CoA to pyruvate (Evans et al., 1966; Furdui and Ragsdale, 2000) (Figure 3). However, the characteristic rTCA key enzymes were not annotated in the genome (Figure 1). The annotation of genes encoding for A, B and C subunits of succinate dehydrogenase (complex II) or fumarate reductase (Ga0068443_100221-23) could indicate fumarate reductase activity and be considered an evidence for the rTCA cycle. These two enzymes are structurally similar and carry opposite reactions, that is, succinate oxidation and fumarate reduction (Hagerhall, 1997). Succinate dehydrogenase/fumarate reductase subunit genes are localized immediately adjacent to citrate synthase gene (Ga0068443_100220) in the CcNi genome. Although the mapping of metatranscriptomic reads to CDS does not allow us to verify whether they are co-expressed as an operon, their adjacent localization and similar expression levels (TPMav 58–169) are suggestive of the participation of succinate dehydrogenase/fumarate reductase in the oTCA cycle as succinate dehydrogenase and not fumarate reductase. In the absence of evidence of other carbon fixation pathways (Supplementary Information), we hypothesize that CcNi fix carbon via rTCA cycle as conducted by other Nitrospira members, despite the lack of the key genes of this pathway in the genome bin.

Ammonia-dependent chemolithoautotrophy and use of organic nitrogen compounds by CcThau

Cultured members of the Thaumarchaeota are able to aerobically oxidase ammonia to nitrite (Stieglmeier et al., 2014a). Genes related to energy acquisition through ammonia oxidation were among the most abundantly transcribed in CcThau (TPMav 3566–31 115, Figures 1 and 4). Ammonia oxidation is carried by ammonia monooxygenase, which in bacteria is composed of three subunits encoded by amoA, amoB and amoC genes (Arp et al., 2007). Additionally, a gene named ‘amoX’ was proposed to be associated with the archaeal ammonia monooxygenase, due to the conserved co-localization with the subunit genes across ammonia-oxidizing archaea (Bartossek et al., 2012). Single copies of genes putatively related to archaeal ammonia monooxygenase were present in pairs, amoA–amoX (Ga0078905_11462, Ga0078905_11463) and amoB–amoC (Ga0078905_103916-17), in the edge of two contigs, possibly forming a single gene array (amoAXCB) as observed in the genomes of Nitrosoarchaeum limnia SFB1 and Nitrosopumilus maritimus SCM1 (Walker et al., 2010; Blainey et al., 2011; Bartossek et al., 2012). The gene encoding for an ammonium transporter (Ga0078905_100724, TPMav 13 184±1736) was estimated to be expressed at an approximate ratio of 1:1 to amoA. In addition, several genes putatively involved in ammonia-dependent chemolithotrophy in Thaumarchaeota were expressed at different levels by CcThau (Kozlowski et al., 2016), including genes encoding for nitrite reductase (NO-forming) (TPMav 107–9653), plastocyanins (TPMav 37–1761), copper-binding proteins of the plastocyanin/azurin family (TPMav 19–200), ubiquinol–cytochrome c reductase cytochrome b subunit (TPMav 304±26), multicopper oxidase (TPMav 18±4) and ATPase subunits (TPMav 26–847). These findings are consistent with the observation of a large transcriptional effort invested in genes related to ammonia-dependent chemolithotrophy by archaeal populations in marine sponges (Radax et al., 2012; Moitinho-Silva et al., 2014).

Figure 4
figure 4

Expression of selected functional genes of CcThau. TPMav values are shown for genes of selected functions. ‘Others’ stands for genes not grouped in the specific functions shown. The median of all expressed genes is represented by the dashed line. Highlighted genes encode for the ammonia monooxygenase subunits alpha (amoA), beta (amoB) and C (amoC) subunits as well as ammonium transporter (Amt).

Full size image

Genomic analysis of several Thaumarchaea species revealed the presence of genes related to transport, production and degradation of urea (Palatinszky et al., 2015). Urea can be used by Thaumarchaeota members as source of ammonia, as shown by biochemical assays of cultivated species (Lehtovirta-Morley et al., 2016), and as carbon source, as inferred from 14C-labeled urea incorporation of prokaryotic communities of Artic waters (Alonso-Saez et al., 2012). The genes encoding for urease subunits, as well as auxiliary proteins, were present in two separate contigs of CcThau but lowly expressed (TPMav 16–77). No encoded urea transporter was found. A set of genes encoding for amino acid/amide ABC transporter of the hydrophobic amino-acid uptake transporter (HAAT) family (TCDB:3.A.1.4), which includes urea transporters (Hosie et al., 2002), were found in the genome. Thus it is conceivable that the encoded products may participate in urea transport. Urea can also be internally produced from the degradation of the amino-acid L-arginine (Shaibe et al., 1985), and the genes encoding for arginine decarboxylase and agmatinase were expressed at low levels (TPMav 10–40).

The expression of a gene encoding for creatinine amidohydrolase, albeit at low levels (TPMav 6±7), further indicates that CcThau degrades creatinine. Creatinine amidohydrolase genes are also present in other members of the candidate order Nitrosopumilales, for example, Nitrosopumilus maritimus SCM1 (Nmar_0222) (Walker et al., 2010). The gene encoding for a creatinase, which would carry the downstream hydrolysis of creatine to sarcosine and urea (Shimizu et al., 1986), was however not found in the CcThau genome bin. The most similar sequence to characterized creatinases from bacteria species was a gene encoding for a Xaa-Pro aminopeptidase (Ga0078905_10021), which is structurally related to creatinase (Supplementary Information; Supplementary Table S5; Bazan et al., 1994). This similarity, however, does not provide enough evidence that CcThau converts creatinine to creatine in order to produce urea and, consequently, ammonia. Therefore, it is unclear in which biochemical context creatinine degradation is carried out by CcThau.

An almost complete set of genes for the 3-hydroxypropionate/4-hydroxybutyrate (HP/HB) pathway was expressed (Figure 1; Supplementary Table S2). The HP/HB pathway is an autotrophic CO2 fixation cycle recently discovered in archaea and present in all Thaumarchaeota species characterized so far (Berg et al., 2007; Bayer et al., 2016). CcThau uses the same set of enzymes as N. maritimus, including 3-hydroxypropionyl-CoA synthetase (ADP-forming) (TPMav 173±53) and 4-hydroxybutyryl-CoA synthetase (ADP-forming) (TPMav 90±31) (Konneke et al., 2014). In one cycle of the HP/HB pathway, two bicarbonate molecules are converted to one acetyl-coA. It is possible that the CcThau HP/HB pathway mainly connects to the central carbon metabolism via succinyl-CoA, which is produced from the generated acetyl-CoA and another pair of bicarbonate molecules by an additional half turn of the cycle (Estelmann et al., 2011). With the exception of the gene for fumarase, for which no orthologous was identified, CcThau showed the expression of the enzymatic repertoire capable to further transform succinyl-coA into precursor metabolites, such as oxaloacetate, phosphoenolpyruvate (PEP), pyruvate and 2-oxoglutarate (TPMav 6–122). This set includes enzymes that are components of the oxidative TCA cycle. In addition, genes encoding for enzymes of the rTCA cycle were also found expressed, such as citrate lyase subunit beta/citryl-CoA lyase and 2-oxoglutarate ferredoxin oxidoreductase (TPMav 32–39). In Thaumarchaeota, oTCA cycle was assumed to be used for anaplerotic reactions and the capacity of Thaumarchaeota to fix carbon via rTCA cycle was not demonstrated (Spang et al., 2012; Zhalnina et al., 2014; Stieglmeier et al., 2014a). Mixotrophic growth has been reported in the Thaumarchaeota phylum (for example, Qin et al., 2014). For CcThau, the possible uptake of urea and amino acids by the amino acid/amide ABC transporter of the HAAT family and the degradation of creatinine by creatinine amidohydrolase were the only evidence to suggest the use of external organic carbon sources (Supplementary Information).

Metabolic features of diatoms and the sponge

From the 184 053 eukaryotic transcripts recovered from C. concentrica (Díez-Vives et al., 2016), the number of transcripts assigned to diatoms (Bacillariophyta) was similar to the combined number of transcripts classified as Porifera or as metazoans with best BLAST hit against sequences from sponges (Figure 5a). Higher expression, estimated from the sum of TPM values, was found for the host compared with the diatom symbiont (Figure 5b). Most transcripts could not be functionally classified for both organisms. Environmental information processing, genetic information processing and metabolism were the pathways with highest overall sums of TPM (Figure 5c). The comprehensive description of the diatom and sponge transcriptomes is beyond the scope of this study and therefore we focused here on key metabolic pathways.

Figure 5
figure 5

Expression of diatom and sponge transcripts in Cymbastela concentrica. (a) Transcripts taxonomically classified as diatoms (Bacillariophyta) or sponge (Porifera or Metazoan with best BLAST hit to sponge sequences) were sorted from other eukaryotic transcripts recovered from C. concentrica. (b) Overall expression of diatom and sponge is indicated by the sum of transcripts’ TPM. (c) Deduced peptide sequences from transcripts’ coding sequences were used for KEGG functional classification. Expression of KEGG functions was inferred from the sum of TPM values across coding sequences.

Full size image

The diatom symbionts invested a large transcriptional effort in functions for energy harvesting via photosynthesis and consumption via glycolysis (Figure 6a; Supplementary Table S6). In addition, functions related to arginine and proline metabolism were found to be abundantly expressed by the diatoms, in particular prolyl 4-hydroxylase (TPMsum 46, Supplementary Table S6), the enzyme that carries out posttranslational modifications of proline (Gorres and Raines, 2010). The expression of carbamoyl-phosphate synthase, argininosuccinate synthase, argininosuccinate lyase and arginase (TPMsum 3–10) are indicative of an active urea cycle in the diatom symbionts. Additional support for this metabolic cycle is provided by the expression of ornithine cyclodeaminase (TPMsum 4) and ornithine decarboxylase (TPMsum 5), which produce proline and putrescine from ornithine, respectively. The urea cycle in diatoms is proposed to redistribute and turnover inorganic carbon and nitrogen coming from catabolism and/or photorespiration (Allen et al., 2011). With respect to sources of nitrogen, the diatoms expressed functions related to nitrate assimilation (nitrate reductase (NADPH), ferredoxin-nitrite reductase, TPMsum 2–5) and production of ammonia via nitrite generated from the oxygenation of alkyl nitronates (nitronate monooxygenase TPMsum 1). The co-expression of these pathways at different intensities has also been recently observed in metatranscriptomes of Skeletonema spp. recovered from surface seawater (Alexander et al., 2015) and indicates the simultaneous use of multiple sources of nitrogen by the diatom symbiont.

Figure 6
figure 6

Expressed metabolic functions in diatoms and the sponge. Sums of TPM are shown for KEGG metabolic pathways found in (a) diatoms (Bacillariophyta) and (b) sponge (Porifera and Metazoan or Metazoan with best BLAST hit to sponge sequences). Legends list the top expressed pathways.

Full size image

Sponge cells abundantly expressed functions of the oxidative phosphorylation (for example, cytochrome c oxidase subunit 6a, TPMsum 418, Supplementary Table S7) and glycolysis (glyceraldehyde 3-phosphate dehydrogenase, TPMsum 418) (Figure 6b). The expression of functions assigned to methane metabolism was largely due to antioxidant enzymes, catalase (TPMsum 468) and peroxiredoxin 6, 1-Cys peroxiredoxin (TPMsum 353), which may protect the sponge cells from reactive oxygen species (Tate et al., 1995; Chen et al., 2000). The most abundant functions related to arginine and proline metabolism were acetylornithine deacetylase (TPMsum 237) and glycine amidinotransferase (TPMsum 94). Both enzymes carry out reactions that produces L-ornithine, the former from N2-acetyl-L-ornithine and H2O and the latter from L-arginine and glycine. The reaction of glycine amidinotransferase is the first of the two-step creatine biosynthesis (Wyss and Kaddurah-Daouk, 2000). Along with creatine, creatinine is expected to be found in sponge cells, as it is a product of spontaneous, non-enzymatic conversion of the former. In addition, the sponge expressed a creatine kinase function (TPMsum 34), which reversibly converts creatine and ATP to phosphocreatine and ADP. This result is in support of previous detection of creatine kinase activity in the tissue of the sponge Tethya aurantia (Ellington, 2000; Sona et al., 2004). In sponges, the creatine kinase/phosphocreatine system was proposed as an intracellular spatial buffer of energy, connecting the ATP produced in the mitochondria with distant cytosolic sites of ATP consumption (Sona et al., 2004; Ellington and Suzuki, 2007).

Model of integrated metabolism of the microbial community members in C. concentrica

Based on the genomic and transcriptomic information of these three prokaryotic symbionts and the expression of eukaryotic transcripts assigned to diatoms and sponges, we developed a metabolic model for the sponge–microbe symbiosis in which nutrient fluxes between cells and the extracellular milieu, that is the sponge mesohyl, were inferred from our data set (see Figure 1).

As shown above, CcPhy is a heterotrophic bacterium with the capacity to uptake a range of organic and inorganic compounds. Such nutrients are likely found across the sponge tissue, which is consistent with a scattered distribution of CcPhy cells throughout the mesohyl (see Figure 7; Supplementary Figure S10). Considering the high relative abundance of Phyllobacteriaceae within the symbiont community (Fan et al., 2012; Esteves et al., 2016), CcPhy may thus be a major contributor to uptake and utilization of DOM by the microbiome. The DOM used by CcPhy may come from the sponge’s filtering activity (Figure 1a; de Goeij et al., 2008, 2013) or from diatoms (Figure 1b) that were found throughout the mesohyl (Figure 7), especially considering that axenic cultures of diatoms have been shown to release between 5% and 21% of their total fixed carbon as dissolved organic carbon (Wetz and Wheeler, 2007). In support of this, photosynthesis functions were abundantly expressed in the diatom transcriptome (Figure 6a). Another source of carbon for the members of the sponge–microbe symbiosis may be from the autotrophic metabolism of CcThau and CcNi (Figures 1c and d, see above).

Figure 7
figure 7

Visualization of microorganisms associated with Cymbastela concentrica by fluorescence in situ hybridization. (a) Rod-shaped cells (white arrows) were observed from hybridization with a Pacific Blue-labeled CcPhy-specific probe. Signals from (b) fluorescein-isothiocyanate-labeled CcNi-specific and (c) Cy3-labeled CcThau-specific probes are shown. (d) Red autofluorescence signals from diatoms were captured (arrows). (e) Clusters of CcNi-specific (green) and CcThau-specific (magenta) probes in the sponge mesohyl are highlighted in the merged color image by circles. CcPhy cells (cyan) did not specifically co-localize with CcNi or CcThau cells. Autofluorescence of the sponge tissue (light blue/cyan) served as a reference for the structure of the sponge mesohyl. Scale bars, 10 μm.

Full size image

Creatine and creatinine are products of the sponge metabolism, as indicated by the expression transcripts encoding for glycine amidinotransferase and creatine kinase. Although it is unknown whether sponge cells exchange creatine between different cells and tissues as described in higher animals (Wyss and Kaddurah-Daouk, 2000), it is conceivable that these compounds are accessible to the microbial community (Figure 1e). The possibility that creatine and creatinine are actively used by the microbiome is supported by the expression of genes able to degrade creatine to glycine by CcPhy and the encoded creatinine amidohydrolase in CcPhy and CcThau genomes. The degradation of these host metabolites by the symbionts is likely widespread, as inferred by the previous observation of related genes in the metagenome of six sponges, including C. concentrica (Fan et al., 2012). Here we could assign this function to be actively used by specific members of the microbiome.

The metabolism of CcPhy, diatom, CcNi and CcThau are linked by the production and consumption of urea. CcPhy internally generates urea from creatine degradation, which can potentially accumulate and diffuse through the bacterial membrane (Figure 1f; Lodish et al., 2000). This assumption is further supported by the fact that heterotrophic marine bacteria have been observed to be net producer of urea even in the presence of hydrolytic activity (Cho et al., 1996; Berg and Jorgensen, 2006). The diatom symbiont is another potential source of urea as suggested by the expression of the urea cycle (Figure 1g). The Nitrospira population represented by CcNi may benefit directly from the urea in the extracellular milieu, as it abundantly expresses the genes related to the transport and degradation of urea. Despite the internal production of urea via L-arginine degradation, CcThau may also uptake urea via the amino acid/amide ABC transporter of the HAAT family. Additionally, urea can spontaneously dissociate to cyanate (Marier and Rose, 1964; Guilloton and Karst, 1985). Similarly, cyanate is formed from carbamoyl phosphate (Allen and Jones, 1964), which is an intermediate of the urea cycle in diatoms. Therefore, urea indirectly links CcPhy and the diatom symbiont to CcNi via cyanate (Figures 1h–j), which could be imported and decomposed to CO2 and ammonia as indicated by the expression of cyanate lyase genes in CcNi.

Urease genes were also found in other sponge-associated microbial genomes and microbiomes (Hallam et al., 2006; Siegl et al., 2011; Bayer et al., 2014; Liu et al., 2016). Together with the phylogenetic diversity found for ureC genes and transcripts found in the sponge Xestospongia testudinaria (Su et al., 2013), these data indicate that urea degradation is a common trait of sponge-associated microorganisms. Although urea can be found in benthic environments (Crandall and Teece, 2012), our data points to the possibility of urea being derived from host metabolites, that is, via bacterial metabolism of creatine and creatinine, and from intracellular metabolism of sponge symbionts, such as diatoms.

Pumping activity and host metabolism have been regarded as sources of ammonia in the sponge tissue (Ribes et al., 2012; Figure 1k). Additionally, urea and cyanate are degraded by CcNi, which is likely a source of ammonia (Figure 1l; Koch et al., 2015). Our analysis indicates that CcPhy takes up ammonium for nitrogen assimilation, while CcThau oxidizes ammonia to generate energy, releasing nitrite (Figure 1m). Nitrite in turn serves as the energy source for CcNi, which expressed the genes necessary for nitrite oxidation. Fluorescence in situ hybridization analysis showed that CcThau and CcNi are often localized in close proximity (Figure 7), supporting such a metabolic interaction. Nitrification, a process in which the oxidation of ammonia is coupled with the oxidation of nitrite, is a common theme in sponge microbiology, being the subject of many studies (Taylor et al., 2007). Based on the use of urea and cyanate as indirect sources of ammonia in nitrification carried out by co-cultures of Nitrospira and ammonia-oxidizing bacteria (Koch et al., 2015; Palatinszky et al., 2015) and the expression profile of CcNi and CcThau, we propose that these compounds should be considered when modeling nitrification in marine sponges.

The end product of the combined aerobic nitrification by CcThau and CcNi is nitrate, which would support anaerobic nitrate respiration by CcPhy (Figure 1n, see above). Such a coupling is possible if the two processes and their associated population are temporally and/or spatially separated by variations in oxygen concentration or redox potential, which have been described to occur in sponges (Schläppy et al., 2010; Lavy et al., 2016). The facultative nature of the CcPhy metabolism supports its temporal separation from the aerobic metabolisms CcThau and CcNi, while the observation that CcPhy rarely co-occured with the CcThau and CcNi cells would indicate also a spatial separation of their metabolisms (see Figure 7). The incomplete denitrification pathway of CcPhy would also yield nitrite (Figure 1o), which again could serve as an energy source for CcNi, thus creating a ‘mini’ N cycle at the interface of aerobic and anaerobic micro-habitats. An alternative path for the N cycle within C. concentrica is the nitrate assimilation by diatoms (Figure 1n).

Conclusion

We described the lifestyle of prominent members of C. concentrica sponge–microbe symbiosis, that is, associated prokaryotes, diatoms and the host, by analyzing their transcriptional activities. Although connected metabolic pathways are expected to occur based on the complexity of nutrient cycles within sponges (Hoffmann et al., 2009), our study is unprecedented in modeling how the physiology of microbiome members are integrated. We thus predicted how and which members can generate, retain, recycle and compete for biochemical resources. In addition, we considered the gene expression of the sponge and diatoms, unveiling the nutritional basis for biological interactions within the sponge–microbe symbiosis. Finally, we demonstrated the power of a holistic approach to further our understanding of the complexity of animal–microbe symbioses.