Introduction

Marine sponges (phylum Porifera) are the most ancient extant metazoans with fossil records, indicating their emergence more than 600 million years ago (Love et al., 2009). These animals are sessile filter feeders with an enormous filtering capacity that is known to affect nutrient concentrations in the surrounding environment (Gili and Coma, 1998; Maldonado et al., 2005, 2012). In addition to their evolutionary and ecological significance, sponges have attracted recent scientific attention owing to their specific and unique microbiology (Hentschel et al., 2012). The microbial biomass in sponges is located in the extracellular matrix, the so-called ‘mesohyl’, and can make up 35% of the sponge body mass (Vacelet, 1975). Collectively, representatives of more than 30 bacterial phyla and both archaeal lineages have so far been found in sponges from various geographic locations (Webster et al., 2010; Schmitt et al., 2012; Simister et al., 2012). The microbial diversity of marine sponges is well investigated (Taylor et al., 2007), and the collective repertoire of ‘omics’ approaches has been instrumental to shed light on the functional genomic traits of the collective sponge microbiome (Thomas et al., 2010; Fan et al., 2012; Liu et al., 2012; Radax et al., 2012). However, community-wide approaches do not provide sufficient information about functions of specific symbiont clades. Providing a thorough understanding of symbiont function is further complicated by the fact that many sponge symbiont lineages remain uncultivated, such as for the many candidate phyla found in these animals (Schmitt et al., 2012).

One such candidate phylum, termed Candidate phylum Poribacteria, was originally discovered and described in marine sponges (Fieseler et al., 2004). Poribacteria are widely distributed and highly abundant in sponge species around the world (Fieseler et al., 2004; Lafi et al., 2009; Schmitt et al., 2011, 2012), and also occur freely in seawater, albeit at very low abundances (Pham et al., 2008; Webster et al., 2010; Taylor et al., 2013). They were shown to be affiliated with the Planctomycetes, Verrucomicrobia, Chlamydiae (PVC) superphylum (Wagner and Horn, 2006). Poribacteria are vertically transmitted via reproductive stages (Schmitt et al., 2008, Webster et al., 2010). Owing to their abundance and diversity within sponges, Poribacteria can be regarded as a model sponge symbiont.

Single-cell genomics has become the most useful tool to investigate the genomic repertoire of distinct uncultivated microbial symbionts (Kamke et al., 2012) and other microorganisms (Woyke et al., 2009; Yoon et al., 2011; Stepanauskas, 2012). It also has previously been successfully applied to a poribacterial cell (Siegl et al., 2011). Unlike other ‘omics’ approaches, this method can connect phylogenetic identity with the functional potential of uncultivated microbial organisms, even from high diversity environments. Here we used single-cell genomics to analyze five poribacterial cells from the Mediterranean sponge Aplysina aerophoba, expanding the existing data set from one (Siegl et al., 2011) to a total of six poribacterial single-amplified genomes (SAGs). We provide an in-depth genomic analysis of one of the main symbiont lineages in the complex microbiota of marine sponges. The property of carbohydrate degradation emerged as the most common feature among the analyzed genomes. We therefore focused on carbohydrate degradation potential of Poribacteria in this study and discussed the results in context of a nutritional basis of the sponge–microbe symbiosis.

Materials and methods

Sample collection and processing

Samples of the marine sponge A. aerophoba were collected in September 2009 by scuba diving to a depth of 5–12 m at the Coast of Rovinj, Croatia (45°08′N, 13°64′E). The animals were transported to the University of Wuerzburg (Wuerzburg, Germany) and kept in seawater aquaria until further processing within 1 week of collection. Fresh sponge samples were used for extraction of sponge-associated prokaryotes using an established protocol of tissue disruption, density centrifugation and filtration by Fieseler et al. (2004).

Single-cell sorting, whole genome amplification and PCR screening

Single-cell isolations were conducted with freshly extracted and purified sponge-associated prokaryotes using the fluorescence-activated cell sorting Vantage SE flow cytometer with FACSDiVa option (Becton Dickinson, Heidelberg, Germany) as described previously (Siegl and Hentschel, 2010). For cell lysis and whole genome amplification (WGA) by multiple displacement amplification, we followed the same procedure as Siegl et al. (2011). To identify phylogenetically WGA products and check for possible contamination, we screened the WGA products obtained by polymerase chain reaction (PCR) using 16S or 18S rRNA gene primers targeting Eubacteria, Archaea, Poribacteria and Eukaryotes, as described previously (Siegl and Hentschel, 2010). Poribacteria-positive WGA products were additionally screened with the degenerated PCR primer pair 27f-B and 1492r-B (Cho and Giovannoni, 2004) that covers Eubacteria more broadly and enables to obtain longer 16S rRNA gene sequences from Poribacteria. Subsequent cloning of PCR products, restriction fragment length polymorphism analysis and Sanger sequencing was conducted to confirm the presence of a single cell using the same procedures as Siegl and Hentschel (2010).

WGA products that originated from single poribacterial cells were then subjected to another round of multiple displacement amplification under the same conditions as stated above. Before genome sequencing, we conducted a S1 nuclease treatment and DNA purification as described by Siegl et al. (2011). Five SAGs were selected in the PCR screening process and subjected to whole genome sequencing: Candidatus Poribacteria WGA-3G, WGA-4C, WGA-4CII, WGA-4E and WGA-4G (hereafter referred as 3G, 4C, 4CII, 4E and 4G, respectively). These were complemented by one poribacterial SAG sequence from an earlier study by Siegl et al. (2011), Candidatus Poribacteria WGA-A3 (hereafter referred as 3A). The existing assembly for this SAG was used for annotation and further analyses as described below.

Genome sequencing, assembly and annotation

A detailed description of all steps of genome sequencing, assembly, annotation and quality checks can be found in Supplementary Text S1. Briefly, a combination of Illumina and 454 pyrosequencing was conducted for double displacement amplification products of SAGs 4C, 4E and 4G at LGC Genomics (Berlin, Germany) and the DOE Joint Genome Institute (JGI, Walnut Creek, CA, USA). SAGs 3G and 4CII were also sequenced at JGI using Illumina HiSeq2000 technology only. Illumina sequences were normalized using DUK, a filtering program developed at JGI, and used for assembly including 454 reads (if available). For Illumina/454 hybrid assemblies, a combination of Velvet (Zerbino and Birney, 2008), Allpaths-LG (Zerbino and Birney, 2008) and the 454 Newbler assembler (Roche/454 Life Sciences, Branford, CT, USA) was used. For Illumina assemblies, we used the programs Velvet and Allpaths-LG. All assemblies were submitted to the IMG/ER annotation pipeline (Markowitz et al., 2009) for gene prediction and automatic functional annotation. All data sets were quality checked and screened for possible contamination (Supplementary Table S1). Manual curation and functional analyses were conducted within the IMG/MER system (Markowitz et al., 2012a), unless stated otherwise.

The Whole Genome Shotgun projects were deposited at DDBJ/EMBL/GenBank under the accession nos. ASZN00000000 (3G), APGO00000000 (4C), ASZM00000000 (4CII), AQTV00000000 (4E) and AQPC00000000 (4G). The versions described in this paper are versions ASZN01000000 (3G), APGO01000000 (4C), ASZM01000000 (4CII), AQTV01000000 (4E) and AQPC01000000 (4G). Raw data, genome assemblies and annotations can also be accessed under the IMG software system (http://img.jgi.doe.gov) under genome IDs 2 265 129 006–2 265 129 011. Additionally included 16S rRNA gene sequences for phylogenetic analysis from PCR screenings were submitted to GenBank under accession numbers KC713965–KC713966.

ANI and tetranucleotide frequency analysis

Average nucleotide identities based on BLAST (ANIb) and tetranucleotide frequencies were estimated using the JSpecies software (v.1.2.1; http://www.imedea.uib.es/jspecies/about.html) with default parameters (Richter and Rosselló-Móra, 2009).

Genome completeness estimation

Genome size and completeness were estimated using two conserved single copy gene sets that have been determined from all bacterial (n=1516) and all archaeal (n=111) finished genome sequences in the IMG database (Markowitz et al., 2012b). The sets consist of 138 bacterial and 162 archaeal conserved single copy genes that occurred only once in at least 90% of all genomes by analysis of an abundance matrix based on hits to the protein family database (Punta et al., 2012). HMMs of the identified protein families were used to search both, all SAG assemblies and all combined assemblies by means of the HMMER3 software (Finn et al., 2011; http://hmmer.janelia.org/help). Resulting best hits above precalculated cutoffs were counted and the completeness was estimated as the ratio of conserved single-copy gene to total conserved single-copy genes in the set after normalization to 90%. Thereafter, the estimated complete genome size was calculated by division of the estimated genome coverage by the total assembly size.

Phylogenetic analysis

Poribacterial 16S rRNA gene sequences were aligned using the SINA aligner (Pruesse et al., 2012) and manually checked in the ARB software package (v.5.3; http://www.arb-home.de/) (Ludwig et al., 2004). The SILVA 16S rRNA database version 111 (Pruesse et al., 2012) was used for selection of reference sequences plus additional poribacterial 16S rRNA gene sequences in the SILVA database in January 2013. Only sequences ⩾1100 bp were used to construct a maximum-likelihood bootstrap tree with 1000 resamplings with the RAxML software (v.7.2.8; http://www.exelixis-lab.org/) (Stamatakis, 2006). The resulting tree was reimported into ARB and short poribacterial sequences (⩽1099 bp) were added without changing tree topology using the parsimony interactive tool in ARB.

Screening for carbohydrate-active enzymes

Protein sequences of poribacterial genomes were screened against the HMM profile-based database of carbohydrate-active enzymes obtained from dbCAN (Yin et al., 2012) in December 2012 using hmmsearch in the HMMER software package (v.3.0; http://hmmer.janelia.org/help) (Finn et al., 2011). Results were filtered using an e-value cutoff <10−5. In addition, all returned hits were manually evaluated based on their functional annotation in IMG/MER and excluded in case of conflicting results. Comparison between glycoside hydrolase (GH)-encoding genes (E.C.: 3.2.1.x) between Poribacteria and all free-living planktonic organisms available in the IMG software system in May 2013 was conducted additionally using functional annotation tools in IMG.

Results and discussion

General genomic features

SAG sequencing

Final genome assembly sizes for the poribacterial cells ranged from 0.19 to 5.44 Mbp (Table 1). For genomes 3G, 4C and 4E, genome recovery was large enough to estimate genome coverage of 98.54%, 38.36% and 58.20%, respectively, whereas the largely fragmented assemblies of 3A, 4CII and 4G did not permit for genome size estimation. The estimated poribacterial genome sizes ranged from 4.25 to 6.27 Mb (Table 1) and do not suggest genome size reduction. The guanine–cytosine content ranged from 47% to 50%, with the exception of genome 4C (41%) (Table 1). Protein coding genes accounted for 95.5–99.6% of the retrieved genomes (Table 1). Approximately 30% of these could not be functionally assigned (50% for genome 3A). The investigated genomes encoded for only one copy of the 16S rRNA gene, which is consistent with previous reports (Fieseler et al., 2006; Siegl et al., 2011).

Table 1 Summary of assembly and genome statistics of poribacterial genomes
Full size table

Definition of phylotypes

Phylogenetic analysis of nearly full-length 16S rRNA gene sequences of Poribacteria showed that three of the six analyzed SAGs (3A, 3G and 4CII) clustered closely together, with genome 4G also in close proximity (>97% sequence similarity), whereas the other two genomes (4C and 4E) each fell separate from this group (Figure 1). Average nucleotide identity and tetranucleotide frequency analysis confirmed a closer relationship between SAGs 3A, 3G, 4CII and 4G than to the other two SAGs (Supplementary Table S2 and Supplementary Figure S1), and they were therefore defined as one phylotype named group I (Figure1). The other two SAGs each represent a separate phylotype. Group I represents, however, a ‘composite phylotype’ as the values for tetranucleotide frequency and average nucleotide identity analysis are under the defined thresholds of 0.99 for tetranucleotide frequency and 95–96% ID for average nucleotide identity (Richter and Rosselló-Móra, 2009).

Figure 1
figure 1

Phylogenetic maximum-likelihood tree based on 16S rRNA genes of the candidate phylum Poribacteria. Sequences from poribacterial SAGs are shown in bold and gray shading. The tree was constructed based on long sequences (⩾1100 nucleotides), shorter sequences were added without changing tree topology and are indicated by dashed lines. Bootstrap support (1000 resamplings) of ⩾90% is shown by filled, and ⩾75% by open circles. The outgroup consisted of several Spirochaetes sequences. Scale bar represents 4% sequence divergence.

Full size image

GHs and other CAZymes

The ability of Poribacteria to degrade and transform complex carbohydrates was assessed by screening genome data against the dbCAN (Yin et al., 2012) and classified according to the carbohydrate-active enzymes (CAZy) database (Cantarel et al., 2009). Most poribacterial hits matched GHs and glycosyl transferase, whereas carbohydrate binding modules, carbohydrate esterases and polysaccharide lyases made up for a smaller proportion (Supplementary Table S3, Supplementary Figure S2). GH and glycosyl transferase also showed the highest diversity with up to 18 different GHs and 18 glycosyl transferases as detected in SAG 3G. We detected 17, 76, 21, 14 and 63 hits to GH families in SAGs 3A, 3G, 4C, 4CII and 4E, respectively (fasta file available in Supplementary Information). No GHs were found in 4G, which is however largely incomplete. GH frequencies of 1.6%, 1.3% and 1.9% genes/genome were estimated for 3G, 4C, and 4E, respectively. In a recent study by Martinez-Garcia et al. (2012), marine Verrucomicrobia genomes contained approximately 0.9–1.2% genes encoding for GHs as compared with the average of 0.2% in other bacteria. Considering slight differences in screening methods between our study and that by Martinez-Garcia et al. (2012), the poribacterial GH frequency was found to be similar to that of Verrucomicrobia. Poribacterial genomes also show an almost linear correlation between the number of GHs and genome coverage with R2=0.9525 (Supplementary Figure S3). Similar to Verrucomicrobia, these findings could indicate a specialization of Poribacteria towards carbohydrate degradation.

We additionally compared the poribacterial GH frequencies with those of all available finished genomes of marine planktonic bacteria in IMG (n=102, May 2013) and also the nearly closed Verrucomicrobia single-cell genome AAA168-F10 by Martinez-Garcia et al. (2012) based on annotation of EC numbers 3.2.1.x (see CAZY database). GH frequencies based on this analysis were on average 0.32% for planktonic bacteria, 0.24% for Verrucomicrobia AAA168-F10 and 0.08–0.42% for poribacterial genomes. A Poribacteria-specific set of enzymes was however not identified, which may be because of the fact that the most dominant GH families in Poribacteria are not accessible in the IMG system because of the lack of EC number annotations available for these families.

Overall, 22 different GH families were identified in all poribacterial genomes (Table 2 and Supplementary Figure S4). Glycoside hydrolase family 109 (GH109) was the most abundant family of GHs. Most of the poribacterial GH109 proteins are annotated as predicted dehydrogenases and related proteins. An α-N-acetylgalactosaminidase activity is described for GH109, although more functions may be assigned to this family (Henrissat ‘GH109’ in CAZypedia; URL: http://www.cazypedia.org, accessed December 2012). α-N-acetylgalactosaminidases cleave N-acetylgalactosamine residues from glycoproteins and glycolipids. The two most apparent sources are the sponge glycoconjugates or glycoproteins of cell walls, and organic matter (both dissolved and particulate) from seawater. The dominance of GH109 is in agreement with the poribacterial ability to use uronic acids as a carbon source (see below), as both point towards degradation of glycoproteins in sponge matrix components.

Table 2 Overview of glycoside hydrolase families in Poribacteria
Full size table

The family with the second largest amount of poribacterial hits, GH74, contains many xyloglucan-hydrolyzing enzymes with diverse functions (Yaoi and Ishida ‘GH74’ in CAZypedia; URL: http://www.cazypedia.org, accessed December 2012). Generally, these enzymes act on the β-1,4-linkages of glucans and might degrade several different substrates including poly- and oligosaccharides of all kinds of organisms. Some GH74 hits from Poribacteria genomes encode for proteins with the potential for chitin deacetylation. Chitin is a polysaccharide of N-acetyl-D-glucosamine entities and represents a major component of the skeleton of A. aerophoba (Ehrlich et al., 2010). Glucans also occur in other glycoconjugates, some of which are responsible for self–non-self recognition in sponges (reviewed by Fernandez-Busquets and Burger, 2003). Xyloglucan-degrading activity may also be inferred from the poribacterial GH74 proteins. Xyloglucans are a major component of plant cell walls and also occur in some green algae (Del Bem and Vincentz, 2010).

GH family GH23 also appears to be widely present in Poribacteria. Proteins of bacterial origin in this family are peptidoglycan lyases that cleave the β-1,4-linkage between N-acetylmuramyl and N-acetylglucosaminyl residues in peptidoglycan (Clarke ‘GH23’ in CAZypedia, URL: http://www.cazypedia.org, accessed December 2012). Poribacterial proteins with hits to this family showed similarity to lytic transglycosylases of other bacteria and contained protein family domain Pfam01464-transglycosylase SLT domain, making it likely that these proteins indeed act as peptidoglycan-lytic transglycosylases.

Of further interest are families GH32 and GH33, which are represented by several hits in genome 4E. Most hits to GH32 were annotated as β-fructofuranosidase (EC: 3.2.1.26), a central enzyme for sucrose degradation. Most hits on poribacterial genomes to GH family GH33 were annotated as neuraminidase/sialidase-like enzyme, exo-α-sialidase (EC: 3.2.1.18), or have high similarities to genes encoding for proteins with this function. This enzyme cleaves terminal sialic acid residues from oligosaccharides, glycoproteins and glycolipids. Sialic acids are N-or O-substituted derivates of neuraminic acid, most often N-acetylneuraminic acid.

Carbon degradation pathways

Single-cell genomic analyses identified Poribacteria as aerobic, heterotrophic organisms. Genes for the major central pathways, such as glycolysis, pentose phosphate pathway, tricarboxylic acid cycle, the non-phosphorylative or semiphosphorylative Entner–Doudoroff pathway and oxidative phosphorylation were largely present (Supplementary Table S4 and Figure 4). Group I might utilize glucose completely over the ED pathway, as none of the genomes belonging to this phylotype encoded for phosphofructokinase (EC: 2.7.1.11) or fructose bisphophate aldolase (EC: 4.1.2.13). This suggests that glycolysis is missing in this phylotype. Instead, group I genome 3G encoded for enzymes enabling transfer of glucose into glucuronate by glucose 1-dehydrogenase (EC: 1.1.1.47) and gluconolactonase (EC: 3.1.1.17) to then be degraded by the ED pathway. The phylogenetically more distant genome 4E encoded for both complete glycolysis and ED pathway. Apart from glucose, Poribacteria seem to be able to use a variety of additional carbon sources (Supplementary Table S5).

Galactoside, fructoside, xyloside and rhamnoside degradation

Poribacterial genomes showed the potential for degradation of galactoside polymers such as melibiose and lactose. Oxidative degradation of galactosides was supported as well as parts of the Lelior pathway (Supplementary Figure S5 and Supplementary Table S5). The potential for degradation of the fructoside polymers levan and sucrose was also encoded in poribacterial genomes (Supplementary Figure S6) and additionally genes coding for enzymes involved in degradation of D-xylose over the xylose isomerase pathway (Supplementary Figure S6). The genomic potential for rhamnoside degradation was shown by genes relevant for the L-rhamnose isomerase pathway and oxidative L-rhamnose degradation, which was strongly supported by multiple gene copies encoding for enzymes involved in this pathway (Supplementary Figure S7 and Supplementary Table S5). For a detailed description of these pathways in Poribacteria, please refer to Supplementary Text S2.

The substrates of these degradation pathways can generally be found in oligo- and polysaccharides of glycoconjugates or biopolymers of various organisms, and especially in the cell walls of many plants and bacteria (Sutherland, 1985; Rehm, 2010; Ray et al., 2011; Visnapuu et al., 2011; Singh et al., 2012). Therefore, they might be freely available in the sponge mesohyl as a result of sponge feeding on bacteria and microalgae.

Inositol degradation

A nearly complete inositol dehydrogenase pathway was present in group I genomes, as well as in genome 4E, and partially in 4C. This pathway degrades myoinositol to glyceraldehyde-3-phosphate, which is further used in the central metabolism (Figure 2). Inositol phosphates are cell wall compounds in all eukaryotes and archaea, and are rarely found in bacteria (Michell, 2011). In addition, phosphorylated inositol is a precursor for several important lipid molecules including sphingolipids, ceramides and glycosylphosphatidylinositol anchors (Reynolds, 2009), as well as many stress-protective solutes of eukaryotes (Michell, 2011). Inositol phosphate is part of the signal transduction in sponges, as shown in Geodia cydonium, where production of inositol triphosphate increases after sponge cell aggregation (Müller et al., 1987). It seems therefore likely that either the sponge itself or eukaryotic microorganisms could be a source of inositol for Poribacteria. Inositol degradation has been reported from a variety of other bacteria (Fry et al., 2001; Yoshida et al., 2008, 2012; Kohler et al., 2010) including Sinorhizobium symbionts of soybean where this pathway was shown to provide a competitive advantage in the plant rhizosphere (Galbraith et al., 1998). We hypothesize that Poribacteria not only use myoinositol as a carbon source but also as an agent for regulating metabolic functions involved in sponge–microbe symbiosis.

Figure 2
figure 2

Schematic reconstruction of inositol degradation as encoded on poribacterial SAGs. Numbers within circles represent the number of genomes encoding for the corresponding enzyme.

Full size image

Uronic acid degradation

Analysis of the poribacterial genomes further revealed the presence of several genes connected to uronic acid degradation (Figure 3 and Supplementary Table S5). The degradation of polymers such as pectin or pectin-like glycoconjugates likely occurs in all investigated phylotypes as indicated by the presence of genes encoding for polygalacturonase (EC: 3.2.1.15) in genome 4E and pectate lyase (EC: 4.2.2.2) in group I genome 3G and genome 4C. This was further supported by the presence of genes encoding for enzymes participating in 5-dehydro-4-deoxy-D-glucuronate degradation such as oligogalacturonide lyase (EC: 4.2.2.6), 2-deoxy-D-glucconate 3-dehydrogense (EC: 1.1.1.125) and 2-dehydro-3-desoxy-D-glucokinase (EC: 2.7.1.45) in SAG group I genomes and partially in genome 4C. Furthermore, glucuronoside polymers appear to be degradable by Poribacteria as indicated by the presence of α- and β-D-glucuronosidases (EC: 3.2.1.139; EC: 3.2.1.31) in group I genomes 3G and 4CII, respectively. The occurrence of genes encoding for galacturonate isomerase (EC: 5.3.1.12), alteronate hydrolase (EC: 4.2.1.7) and mannonate dehydratase (EC: 4.2.1.8) in high copy number on several poribacterial genomes of the phylotypes represented by group I and 4E points towards the possibility of galacturonate and glucuronate catabolism. The products of these degradation steps could then enter the ED pathway via 2-dehydro-3-desoxyphophogluconate aldolase (EC: 4.1.2.14; 4.1.3.16), which was found in group I genomes and in genome 4E.

Figure 3
figure 3

Schematic reconstruction of uronic acid degradation as encoded on poribacterial SAGs. Numbers within circles represent the genomes encoding for the corresponding enzyme. Dashed circles represent manually annotated genes.

Full size image

Uronic acids are sugar acids that can be found in various biopolymers of plant, animal or bacterial origin (Sutherland, 1985; Rehm, 2010). The occurrence of uronic acids in glycosaminoglycans (GAGs) is especially worth noting, because the extracellular matrix of sponges is largely constructed by these polymers (Fernandez-Busquets and Burger, 2003). Therefore, Poribacteria should literally be submerged in uronic acid-containing substances. It is known that GAGs of sponges are different from those of higher animals (Misevic and Burger, 1993), with the main components being fucose, glucuronic acid, mannose, galactose, N-acetylglucosamine and sulfate in sponge GAGs (Misevic and Burger, 1986, 1993; Misevic et al., 1987).

Transporters

The poribacterial genomes code for a range of transporters representing different families (Table 3 and Figure 4), and here we concentrate on those involved in carbohydrate metabolism. Typical sugar transport systems/phosphotransferase systems were missing from all poribacterial genomes. Genes coding for proteins of the tripartite ATP-independent periplasmatic transporter family were found on all three phylotypes, which are often involved in organic acid transport such as C4-dicarboxylates, keto-acids and sugar acids (N-acetyl neuraminic acids, sialic acid). The most dominant transporter family in all poribacterial genomes was the ATP-binding cassette superfamily. ATP-binding cassette transporters were found for a variety of broader substrate categories such as amino acids, di- and oligopeptides, carbohydrates, lipoproteins, metal ions and systems involved in cell protection and competition with other organisms (Supplementary Table S6). Most of the detected carbohydrate ATP-binding cassette transporters were simple or multiple sugar transporters and did not show any further specification. However, we detected a system of D-xylose transport on group I genome 3G and transporters for maltose, malto-oligosaccharides, arabinose and lactose on several other genomes besides 3G.

Table 3 Transporters classes on poribacterial genomes
Full size table
Figure 4
figure 4

Schematic overview of a poribacterial cell in the sponge extracellular matrix illustrating pathways of carbohydrate metabolism and glycosaminoglycan degradation by poribacterial enzymes. The dashed arrow represents glycolysis that is not supported by the dominant poribacterial phylotype group I.

Full size image

Sulfatases

Genome analysis of the six poribacterial genomes revealed a total of 103 genes coding for sulfatases (Table 4). Individually, group I genomes coded for 57, genome 4C for 25 and genome 4E for 21 sulfatase genes. Genes coding for choline sulfatase (betC; EC: 3.1.6.6) were identified in group I genomes, and 4E. This enzyme transforms choline sulfate to choline and then to glycine betaine, which is often used as an osmoprotectant (Le Rudulier et al., 1984) or can be degraded via glycine and serine to pyruvate. This degradative pathway is largely present on genome 4E, whereas glycine degradation was encoded in all three phylotypes. As choline-O-sulfate is synthesized by different microorganisms (Spencer and Harada, 1960; Fitzgerald and Luschinski, 1977; Rivoal and Hanson, 1994), it is likely available as carbon and sulfur substrate for Poribacteria.

Table 4 Sulfatase genes on poribacterial genomes
Full size table

Two genes coding for arylsulfatase (EC: 3.1.6.1) were detected in group I genomes 3G and 4E. General substrates for this type of enzyme are phenol sulfates, but the exact kind of phenol sulfate is difficult to determine because of high similarities between different types of substrates. Genome 3G codes also for a cerebroside-sulfatase (EC: 3.1.6.8), which also hydrolyzes phenol sulfates, ascorbate 2-sulfate and galactose-3-sulfate residues in lipids. These enzymes might provide Poribacteria with an organic sulfur source from hydrolyzed sulfate esters in the absence of inorganic sulfate. Poribacteria have the genomic potential for assimilatory sulfate reduction over adenosine-5′-phosphosulfate and 3′-phosphoadenosin-5′-phosphosulfate and subsequent cysteine synthesis (data not shown). The use of organic sulfur compounds under sulfur-limiting conditions has been shown in various bacteria (Kertesz, 2000). The presence of sulfated lipids and polysaccharides in the sponge extracellular matrix is well documented (Zierer and Mourao, 2000; Vilanova et al., 2009), and it may thus serve as a possible carbon and sulfur source.

Group I and genome 4C both encode for N-acetylgalactosamine-4-sulfatase (EC: 3.1.6.12), which lyses the sulfate groups of N-acetylgalactosamine-4-sulfate from chondroitin and dermatan sulfate. Chondroitin is known to be part of sponge GAG chains, and N-acetylgalactosamine was found in GAG chains and glycoproteins of sponges (Fernandez-Busquets and Burger, 2003). In addition, genome 4C codes for iduronate-2-sulfatase (EC: 3.1.6.13) and SAG group I genome 3G for N-sulfoglucosamine sulfohydrolase (EC: 3.10.1.1); however, their substrates are yet to be identified in sponges.

Symbiotic heterotrophy in sponges

Symbioses frequently have a nutritional basis, such as nitrogen fixation in Rhizobium–legume symbioses (Lodwig et al., 2003), supplementation of amino acids in Buchnera symbionts of aphids (Ramsey et al., 2010; Hansen and Moran, 2011), chemoautotrophy in marine mussels and worms (Woyke et al., 2006; Petersen et al., 2011) or photosynthesis in symbionts of corals or ascidians (Weis and Allemand, 2009; Schnitzler and Weis, 2010; Donia et al., 2011). Here we present bacterial heterotrophy, that is, the ability to utilize diverse carbon sources, as a potential functional basis for the interaction of microbial symbionts with sponges. The role of heterotrophy in symbiosis in general has so far been underestimated. Only recently, studies of gut microbiomes of human, ruminants or termites, and other terrestrial hosts (Warnecke et al., 2007; Hess et al., 2011; Zhu et al., 2011; Schloissnig et al., 2013) have received major attention. Here we show that symbiotic heterotrophy could also be of relevance in a marine host, using poribacterial symbionts of marine sponges as an example. Whether the host gains benefit from the heterotrophic metabolism of its symbionts is an interesting question for future investigations.

The carbohydrate degradation potential of Poribacteria opens up various nutritional sources that are taken up by the host’s extensive filtration activities and transported into the sponge interior. Sponges feed on dissolved and particulate organic matter (Yahel et al., 2003; De Goeij et al., 2008) as well as heterotrophic bacteria and eukaryotes (Maldonado et al., 2012; Perea-Blázquez et al., 2012). It has further been shown that sponge feeding on dissolved and particulate organic matter from bacterial and algal sources can be mediated by bacterial symbionts (De Goeij et al., 2008). Here we provide the genomic background and detailed carbon degradation pathways behind this ecological observation. We could show that Poribacteria have the potential to catabolize various substrates that can be found in, for example, cell wall components, glycoproteins and polysaccharides from marine algae and prokaryotes (Rehm, 2010; Jiao et al., 2011).

The carbon degradation repertoire of Poribacteria would further be consistent with degradation of compounds from the extracellular matrix of the sponge host (Figure 4). Our hypothesis is supported by (i) the high abundance of GH families with acetylgalactosaminidase or sialidase activities (and several GHs consistent with this hypothesis), (ii) the ability to degrade uronic acids, frequent components of glycoconjugates and especially GAGs, as well as (iii) the presence of sulfatases with GAGs as the specific substrate. Remarkably, Poribacteria are significantly enriched in high microbial abundance over low microbial abundance of sponges or seawater (Schmitt et al., 2012; Taylor et al., 2013) One major difference between high microbial abundance and low microbial abundance sponges (apart from the amount and diversity of microorganisms itself) is that high microbial abundance sponges contain a notably expanded mesohyl matrix as compared with their low microbial abundance counterparts (Hentschel et al., 2003; Weisz et al., 2008). The extracellular matrix may thus provide both habitat and nutrient source to Poribacteria. As sponges are known to continuously remodel matrix components (Bond, 1992), it is likely that polymers that become available in this process, serve as nutrient substrates for Poribacteria. This process is unlikely harmful to the host as no signs of tissue destruction in healthy A. aerophoba sponges can be observed neither in the natural environment nor by transmission electron microscopy of the mesohyl matrix (Vacelet, 1975; Friedrich et al., 2001)

The bacteria–sponge interaction may go beyond nutrition and may have also a mechanistic basis. In this context, the so-called ‘sponge aggregation factor’ is of relevance, which shows a proteoglycan-like structure and has a major role in cell-specific aggregation cell–matrix connections and adhesion processes (Müller and Zahn, 1973; Misevic and Burger, 1993; Fernandez-Busquets and Burger, 2003). It might be conceivable that Poribacteria influence the function of the sponge aggregation factor through glycosylase, glyocotransferase activities, or by modification of sulfate groups on the mucopolysaccharide chains. The poribacterial potential to influence sialylation by GH family 33 enzymes might also affect this process. It has been shown that sialic acid residues are an important component of glycoproteins and mucopolysaccharides on sponge cell surfaces (Garrone et al., 1971) and that these also have a major role in sponge cell aggregation (Müller et al., 1977). It might thus be possible that poribacterial enzymes interfere with mechanistic processes related to adhesion, aggregation and self–non-self recognition.