We first aimed to investigate the NRPS gene diversity in twelve marine sponge species from the Bahamas and the Mediterranean. For this purpose NRPS A domain gene fragments were amplified as previously described [25] using degenerate primers A3 and A7R. Altogether 62 A domain DNA sequences (ca. 750 bp) were amplified from the metagenomes of all 12 sponge species, from the ascidian Ecteinascidia turbinata and from seawater. No PCR product was obtained from the sediment sample. Additionally, a NRPS A domain from the bacterium Streptomyces sp. Aer003, which had previously been isolated from Mediterranean Aplysina aerophoba (Acc# JN830622) was amplified and sequenced. Sequences from each sponge metagenome that exhibited ≥98% sequence identities were presumed to have been amplified from identical genes, thus taking into account the PCR-induced errors that may arise from using degenerate primers [20]. Using this criterion, a total of 24 sequences were considered different (Acc# JN815085–JN815111). A neighbor-joining tree was constructed including related sequences from BLAST analysis (Figure 1).

The majority of these sequences (22/24) formed one large distinct cluster with 97–99% in-cluster amino acid identity. The cluster also included the NRPS A domain sequences from the isolate Streptomyces sp. Aer003 and from seawater. The closest relatives were all Actinobacteria with the A domain sequences from Streptomyces roseosporus (ZP_04696845) and S. fungicidus (ABD65957) being the closest relatives (64, 65% sequence identity). The substrate for the NRPS adenylation domains appears to be hydoxy-phenyl-glycine (hpg) as predicted by NRPSpredictor2 [26]. Notably, Xestospongia muta clone 8 and Aplysina cauliformis clone 19 fell outside of this large cluster with the A domain sequences of Stenotrophomonas maltophila (YP_002028658) and Pseudomonas putida (YP_001750394) being their closest phylogenetic neighbors, respectively and with valine being the likely substrate. Overall, the NRPS gene diversity discovered in this study was very low. However, a bias cannot be ruled out as primers targeting specifically NRPS systems from actinomycetes were used [25]. If primer sets targeting different groups of microorganisms were employed, a higher diversity of NRPS genes would be expected. The fact that a closely related NRPS sequence clade was found in all sponge samples independent of their geographic location, in Caribbean seawater and in a Mediterranean streptomycete isolate suggests a wide geographic distribution of this NRPS-bearing bacterium.

PCR screening using degenerate primers [25] targeting two independently constructed metagenomic libraries harboring altogether ca. 2.4 Gb Aplysina aerophoba microbial community DNA [22] resulted in the identification of 14 NRPS-positive library pools, that is, at least 14 NRPS-containing cosmid clones. Sequence analysis of the NRPS PCR products from AApAY1 [22] NRPS-positive, metagenomic cosmid clones revealed nearly identical DNA sequences (data not shown). Two overlapping cosmid clones were sequenced and assembled into a 57031 bp long DNA fragment (AANRPS; Acc# HQ456128) with an overall GC content of 67%. Remarkably, the entire DNA fragment did not exhibit significant similarity on the DNA level to any known sequence in the database, except for NRPS identified previously in the Chloroflexi symbiont of A. aerophoba [21]. Phylogenetic analysis of the translated NRPS A domains confirmed the NRPS gene of a Chloroflexi symbiont of A. aerophoba (“uncultured sponge symbiont cosmid clone ln22” (ACX49739)) as the nearest phylogenetic relative (Figure 2) [21]. Those two NRPS genes share high protein identity (82%) and high similarity (87%) on the DNA level. Beside the NRPS gene and the efflux protein (lubB), there are however no further similarities in the gene neighborhood between the two cosmid clones.

The genomic organization of the NRPS-containing cosmid clone is shown in Figure 3. At least 26 putative ORFs were identified (Table 1). Of these, four ORFs are proposed to represent the gene cluster coding for NRPS related proteins, which were termed lubA, B, C, D. The proposed gene cluster consists of 15752 bp and has a G + C content of 68%. The gene lubA encodes for a putative transcriptional regulator of NRPS expression, and is located upstream of the putative efflux protein-encoding gene lubB. The NRPS structural gene lubC contains two complete NRPS modules and is therefore predicted to encode biosynthesis of the dipeptide [28]. Both A1 and A2 adenylation domains probably use aromatic amino acids, such as phenylalanine and tyrosine, as substrates (NRPSpredictor2; [26]). Furthermore, three TonB-dependent receptors and a phosphopantetheinyl transferase (lubD) are contained on the metagenomic cosmid fragment (Table 1). TonB-dependent transporters are frequently involved in iron uptake via siderophores, as well as other substrates including heme, vitamin B₁₂, proteins or polysaccharides [29]. The phosphopantetheinyl transferase enzyme (PPT) activates a carrier protein by the transfer of a phosphopantetheinyl moiety to a serine residue.

The postulated chemical product of this novel NRPS gene cluster remains elusive, owing to a lack of related NRPS genes in the databases and the lack of robust prediction tools. However, it has become clear that NRPS gene clusters are widespread in actinomycete strain collections with recoveries of NRPS genes from more than half of the isolates screened [18,19] as well as being abundant in microbial genomes/metagenomes. With ever cheaper sequencing technologies and improved bioinformatic prediction tools, genomic mining approaches will undoubtedly be instrumental for the identification of sources suitable for natural product discovery.

Marine sponges were collected by SCUBA diving at a depth of 5 to 15 m: Aplysina aerophoba offshore from Banyuls sur mer, France (GPS: 42°29′N, 03°08′E); Agelas citrina, Aplysina archeri, Aplysina cauliformis, Aplysina insularis, Callyspongia vaginalis, Niphates digitalis, Plakortis sp., Smenospongia aurea, Tedania ignis, Verongula gigantea, Xestospongia muta offshore from Patch Reef, Bahamas (GPS: 24°14′N, 74°32′W). Additionally, the ascidian Ecteinascidia turbinata, sediment and seawater samples were collected from the sampling site at Patch Reef, Bahamas. Individual specimens were placed separately in plastic bags and brought to the surface. The sponge and ascidian tissues were cut into pieces and stored at −80 °C until use.

Strain Streptomyces sp. Aer003 was cultivated from the sponge Aplysina aerophoba using M1 [30] culture medium and identified by 16S rRNA gene sequencing as described previously by Hentschel et al. [31].

Genomic DNA was isolated from freshly collected sponges, ascidian and seawater following the method as described previously by Fieseler et al. [22] using the FAST DNA Spin kit for Soil (Q-Biogene). Amplification of the A domains of NRPS gene fragments was performed as described previously [25] using degenerate primers A7R (5′-SASGTCVCCSGTSCGGTAS-3′) and A3 (5′-GCSTACSYSATSTACACSTCSGG-3′). PCR amplification products of ca. 750 bp in size were cloned into a pGEM-Teasy vector (Promega) and transformed into electrocompetent Escherichia coli XL1-Blue cells. Plasmid minipreps by alkaline lysis procedures and sequencing of the inserts were performed as described previously [22]. The same protocol was followed for the strain Streptomyces sp. Aer003.

Two metagenomic libraries constructed from microbial cells of the marine sponge Aplysina aerophoba [22] using an E. coli-Streptomyces shuttle cosmid vector, pAY1, [32] were used for NRPS screening. The libraries represented a total of ca. 2.4 Gb of sponge-associated microbial DNA. Library pools were screened by PCR following the method of Piel et al. [33] using the degenerate primers targeting the A domains of NRPS genes (A7R and A3; see sequences above). Two PCR-positive overlapping cosmid clones were sequenced (pANRPS19p18 and pANRPS32i21).

Sequencing analysis was performed by Agowa/LGC Genomics, Berlin, Germany. Sequence data were assembled and annotated using the Vector NTI software (Invitrogen) and analyzed using EMBOSS-Transeq and BLAST algorithms [34].