Comparative Genomics Highlights Symbiotic Capacities and High Metabolic Flexibility of the Marine Genus Pseudovibrio.

Pseudovibrio is a marine bacterial genus members of which are predominantly isolated from sessile marine animals, and particularly sponges. It has been hypothesized that Pseudovibrio spp. form mutualistic relationships with their hosts. Here, we studied Pseudovibrio phylogeny and genetic adaptations that may play a role in host colonization by comparative genomics of 31 Pseudovibrio strains, including 25 sponge isolates. All genomes were highly similar in terms of encoded core metabolic pathways, albeit with substantial differences in overall gene content. Based on gene composition, Pseudovibrio spp. clustered by geographic region, indicating geographic speciation. Furthermore, the fact that isolates from the Mediterranean Sea clustered by sponge species suggested host-specific adaptation or colonization. Genome analyses suggest that Pseudovibrio hongkongensis UST20140214-015BT is only distantly related to other Pseudovibrio spp., thereby challenging its status as typical Pseudovibrio member. All Pseudovibrio genomes were found to encode numerous proteins with SEL1 and tetratricopeptide repeats, which have been suggested to play a role in host colonization. For evasion of the host immune system, Pseudovibrio spp. may depend on type III, IV, and VI secretion systems that can inject effector molecules into eukaryotic cells. Furthermore, Pseudovibrio genomes carry on average seven secondary metabolite biosynthesis clusters, reinforcing the role of Pseudovibrio spp. as potential producers of novel bioactive compounds. Tropodithietic acid, bacteriocin, and terpene biosynthesis clusters were highly conserved within the genus, suggesting an essential role in survival, for example through growth inhibition of bacterial competitors. Taken together, these results support the hypothesis that Pseudovibrio spp. have mutualistic relations with sponges.

Introduction

Pseudovibrio is a genus of Gram-negative, heterotrophic, facultative anaerobic, marine α-proteobacteria. The genus was proposed by Shieh et al (2004), and presently six species have been described, including Pseudovibrio denitrificans (Shieh et al. 2004), Pseudovibrio ascidiaceicola (Fukunaga et al. 2006), Pseudovibrio japonicus (Hosoya and Yokota 2007), Pseudovibrio axinellae (O'Halloran et al. 2013), Pseudovibrio hongkongensis (Xu et al. 2015), and Pseudovibrio stylochi (Zhang et al. 2016). To date, members have been isolated from seawater (Hosoya and Yokota 2007), ascidians (Fukunaga et al. 2006), a flatworm (Zhang et al. 2016), a sea cucumber (Zhang et al. 2013), tunicates (Sertan-de Guzman et al. 2007; Riesenfeld et al. 2008), corals (Chen et al. 2012), and sponges (Santos-Gandelman et al. 2013; Versluis et al. 2017). Members of the genus Pseudovibrio are characterized as motile, rod-shaped marine bacteria. For energy production, they are capable of oxygen-dependent respiration and/or nitrate- and nitrite-dependent (anaerobic) respiration as well as fermentation. An analysis of the genomes of Pseudovibrio sp. FO-BEG1 and Pseudovibrio sp. JE062 showed that Pseudovibrio spp. are metabolically versatile, that is, they are capable of utilizing a wide range of organic and inorganic compounds to meet their carbon, nitrogen, phosphorous, and energy requirements (Bondarev et al. 2013).

The relationship between Pseudovibrio and sponges is particularly interesting because these bacteria have consistently been isolated from different sponge species (Lafi et al. 2005; Muscholl-Silberhorn et al. 2008; Menezes et al. 2010), while never simultaneously being isolated from nor detected in surrounding seawater (Webster and Hill 2001; Enticknap et al. 2006). Although it should be noted that the relative abundance of Pseudovibrio spp. in the sponge microbiota is generally low considering that their presence is rarely detected by cultivation-independent assays (Enticknap et al. 2006), these findings indicate that Pseudovibrio spp. are enriched in sponges, reinforcing the current hypothesis that they are symbionts of sponges. Further support for this hypothesis comes from the fact that Pseudovibrio spp. were reported to be vertically transmitted through sponge larvae of Mycale laxissima (Enticknap et al. 2006). Furthermore, Pseudovibrio spp. were found in reduced relative abundance in the cultured bacterial community of diseased specimen of Rhopaloeides odorabile (Webster and Hill 2001; Webster et al. 2002) and Callyspongia aff. biru (Sweet et al. 2015). This suggests that Pseudovibrio spp. may benefit sponge health.

The symbiotic role of Pseudovibrio in the sponge host is still unclear but it could include functions in nutrient uptake (Webster and Hill 2001), denitrification (Shieh et al. 2004; Fiore et al. 2010; Han et al. 2013), or host defence (Penesyan et al. 2011; Graca et al. 2013). In addition, genomic data indicated several mechanisms for establishing and maintaining symbiosis, for example by interactions of Pseudovibrio with the host immune system or cytoskeleton (Bondarev et al. 2013; Alex and Antunes 2015; Romano et al. 2016). Pseudovibrio spp. have received particular interest as sources of clinically relevant antimicrobials (Santos et al. 2010; O'Halloran et al. 2013), with multiple studies reporting that members of this genus are highly bioactive (Flemer et al. 2012; Graca et al. 2013). So far, Pseudovibrio-derived secondary metabolites that have been identified are phenazine (Schneemann et al. 2011), tropodithietic acid (TDA) (Penesyan et al. 2011; Harrington et al. 2014) and pseudovibrocin (Vizcaino 2011). Pathways of polyketide synthases (PKS) and nonribosomal peptide synthetases (NRPS), and/or their hybrids, are particularly important in the production of secondary metabolites by Pseudovibrio spp., albeit not for TDA biosynthesis (Kennedy et al. 2009; O'Halloran et al. 2011; Crowley et al. 2014).

In this study, we compared the genomes of 31 Pseudovibrio isolates, including 25 isolates obtained from sponges, 1 isolate retrieved from a bryozoan, and 5 type strains, by employing SAPP, the Semantic Annotation Pipeline with Provenance (Koehorst et al. 2016a, 2016b). Twenty-eight of these Pseudovibrio isolates were genome sequenced in this study, namely 22 isolates from sponges, 1 isolate from a bryozoan, and the 5 type strains. We resolved the phylogeny, and studied metabolic and secondary metabolite biosynthesis (SMB) pathways. In addition, we predicted features of antibiotic resistance and host–symbiont interactions, providing additional insights into the nature of the association between Pseudovibrio spp. and marine sponges.

Materials and Methods

Sample Collection and Data Deposition of All Analyzed Genomes

The type strains P. ascidiaceicola DSM 16392T, P. axinellae DSM 24994T, P. denitrificans JCM 12308T, and P. japonicus NCIMB 14279T were obtained from the respective culture collections. Sponge-associated Pseudovibrio strains were isolated from the sponges Corticium candelabrum (n = 2), Petrosia ficiformis (n = 10), and Aplysina aerophoba (n = 7) as described previously (Versluis et al. 2017). Pseudovibrio sp. AB108, Pseudovibrio sp. AB111, Pseudovibrio sp. AB113, and Pseudovibrio sp. BC118 were isolated from the sponges Ircinia sp., Chondrilla nucula, Acanthella acuta, and the bryozoan Cellepora pumicosa, respectively, collected in the Limski Channel in Croatia (45°7′54.56″N, 13°39′13.02″E) as described previously (Thiel and Imhoff 2003; Heindl et al. 2010). The type strain P. hongkongensis UST20140214-015BT was genome sequenced but not included in growth experiments. Publicly available genome sequences of Pseudovibrio sp. FO-BEG1 (GenBank accession numbers: CP003147 and CP003148), Pseudovibrio sp. JE062 (ABXL00000000), Pseudovibrio sp. POLY-S9 (LCWX00000000), Stappia stellulata DSM 5886T (AUIM00000000), Nesiotobacter exalbescens LA33BT (AUGS00000000), and Labrenzia alexandrii DFL-11T (ACCU00000000) were downloaded and reannotated applying the methods particular to this study. The assembled genomes and the corresponding GenBank files of the Pseudovibrio strains sequenced as part of this study were deposited under European Nucleotide Archive study accession PRJEB20602.

DNA Isolation and Sequencing

All Pseudovibrio strains were inoculated from glycerol stocks onto marine agar 2216 (Difco, Detroit, USA). Subsequently, a single colony was grown at 20 °C in marine broth 2216 (Difco), and DNA was isolated using the MasterPure™ DNA Purification Kit (Epicentre, Madison, USA). Genome sequencing was done using the Illumina MiSEQ (paired end, 2 × 300 bp reads, 500 bp average insert size) at GATC Biotech (Konstanz, Germany). As an exception, DNA from the liquid culture of P. hongkongensis UST20140214-015BT was isolated as described previously (Ausubel et al. 2002). In addition, the genome of this strain was sequenced on the Illumina Hiseq 2000 (paired end, 2 × 101 bp reads, 500 bp average insert size) (Shanghai Majorbio Bio-pharm Technology Co., Ltd, Shanghai).

Antibiotic Resistance

Resistance profiles were determined by inoculation of pregrown liquid cultures onto marine agar 2216 containing 1 of the following 17 (combinations of) antibiotics: (1) 20 μg/ml polymyxin B, (2) 20 μg/ml daptomycin, (3) 50 μg/ml vancomycin, (4) 50 μg/ml penicillin, (5) 20 μg/ml erythromycin, (6) 20 μg/ml ciprofloxacin, (7) 100 μg/ml cefotaxime, (8) 50 μg/ml tetracycline, (9) 50 μg/ml chloramphenicol, (10) 95 μg/ml sulfamethoxazole and 5 μg/ml trimethoprim, (11) 20 μg/ml lincomycin, (12) 100 μg/ml kanamycin, (13) 20 μg/ml rifampicin, (14) 20 μg/ml linezolid, (15) 50 μg/ml ampicillin, (16) 20 μg/ml imipenem, and (17) 50 μg/ml d-cycloserine. Antibiotic resistance was evaluated 2 days postinoculation and compared with growth on media without antibiotics as a reference. We defined three levels of antibiotic resistance: (1) “resistant”; growth of the bacteria was identical to their growth on media without antibiotics, (2) “intermediately resistant”; growth of the bacteria was slower than growth on media without antibiotics, and (3) “susceptible”; no growth.

Genome Assembly and Quality Control

The read quality and presence of Illumina Truseq adapter sequences was investigated with FASTQC (Andrews 2010), and Illumina Truseq adapter sequences were removed with Cutadapt 1.8.1 using default settings (Martin 2011). All genomes were assembled with the A5-miseq assembler (Coil et al. 2015) version 20150522 using default settings with as exception the P. hongkongensis UST20140214-015BT genome that was assembled with Velvet 1.2.10 (Zerbino and Birney 2008). Subsequently, Pilon 1.13 (Walker et al. 2014) was used to improve the assemblies based on the assembled scaffolds and error-corrected reads obtained as output from the A5 assembler. Assemblies were investigated for contamination by BLASTn v2.3.0 (Altschul et al. 1990) using the NCBI nr/nt database (June 2015) as a reference, and using default settings. Overall, seven <3 kb contigs were removed that were assigned to Clostridia and Bacilli spp. Sequences from Enterobacteria phage phiX174 were removed as well as DNA of this phage is routinely included as a positive control during sequencing. In order to determine the coverage, A5 error-corrected reads were aligned with the assemblies by Bowtie2 2.2.5 (Langmead and Salzberg 2012) using default settings. The resulting sequence alignment map (SAM) file was converted by Samtools 0.1.19 (Li et al. 2009) into a binary alignment map (BAM) file, after which the coverage per base was calculated using Bedtools 2.25.0 (Quinlan and Hall 2010).

Genome Annotation

To obtain a manageable and homogeneous annotation with provenance, the genomes of all Pseudovibrio strains were de novo annotated using a defined set of algorithms and an in-house pipeline for annotation and data storage (Koehorst et al. 2016a). In addition, the genomes of L. alexandrii DFL-11T, S. stellulata DSM 5886T, and N. exalbescens DSM 16456T were de novo annotated and added in order to serve as a reference. The complete genome sequences were used to detect rRNA genes by RNAmmer 1.2 (Lagesen et al. 2007), and to detect tRNA and tmRNA genes by ARAGORN 1.2.36 (Lagesen et al. 2007). Prodigal 2.6.2 (Hyatt et al. 2010) was applied to predict genes, and subsequently the InterPro Database 5.4-47.0 (Mitchell et al. 2015) was used to classify the genes and to predict domains and sites. At runtime, the following databases were selected: TIGRFAM, ProDom, SMART, PROSITE, PfamA, PRINTS, SUPERFAMILY, Coils, and Gene3d. PRIAM (Claudel-Renard et al. 2003), SignalP 4.0 (Petersen et al. 2011), and TMHMM (Krogh et al. 2001) were used to assign Enzyme Commission numbers (ECs), to detect signal peptides, and to detect transmembrane protein topology, respectively. Predictions and provenance were converted into the RDF data model with a self-defined ontology (van Dam et al. 2015). Results are stored as Turtle files and analyzed as described (Koehorst et al. 2016b).

Phylogeny, Functional Diversity, and Environmental Distribution

A phylogenetic tree was made in ARB (Ludwig et al. 2004) with the 16S rRNA gene sequences from the genomes. When multiple 16S rRNA genes were found in a genome that were >99% identical, a single gene was randomly picked to be included in the tree. The three closest non-Pseudovibrio relatives in the Silva guide tree (release 115) (Yilmaz et al. 2014) from which the genomes were publicly available were used as outgroup. Gene alignments were manually curated, and a Maximum Likelihood tree was constructed with 1,000 iterations of rapid bootstrapping. Average nucleotide identity (ANI) between the different genomes based on BLAST results (ANIb) was calculated using JSpecies v1.2.1 with default parameters (Richter and Rossello-Mora 2009). The dissimilarity in terms of protein diversity was analyzed and visualized in a distance tree. Therefore, to identify homologous proteins shared between the genomes, an all-against-all BLASTp was performed, after which OrthAgogue v1.0.3 (Ekseth et al. 2014) and MCL 14-137 (van Dongen 2000) were used to identify orthologous groups of proteins (OGPs), where the OGPs included proteins with both orthologous and inparalogous relations. A presence/absence matrix was created using all types of OGPs. With R version 3.2.2. (R development Core Team 2010), a NxN Jaccard distance matrix was created based on the presence/absence of the OGPs that are present in greater than two genomes. This distance matrix was used for complete linkage hierarchical clustering. To assess the uncertainty in the hierarchical cluster analysis, we used the R package pvclust 2.0-0 (Suzuki and Shimodaira 2006) with 10,000 bootstraps. Canonical (constrained) correspondence analysis as implemented in Canoco 5 (Šmilauer and Lepš 2014) was used to investigate which variables (i.e., membership of the Pseudovibrio genus, geographic origin, and sponge of origin) could explain the variation in OGP composition among the genomes. The OGP compositional table contained information about the presence/absence of the OGPs in the genomes. The presence of metabolic pathways was predicted using OGPs that had EC numbers assigned, and using the MetaCyc database (Caspi et al. 2016) as a reference. The presence/absence data of OGPs that are present in at least three genomes were used as input for SIMPER from the package PRIMER 6 v6.1.9 (PRIMER-E Ltd, Plymouth, UK) (Clarke and Gorley 2006) in order to calculate the contribution of each OGP to the observed dissimilarity between genomes grouped by (environmental) variables. No cut-off was used for low contributions. DIAMOND v.0.7.12 (Buchfink et al. 2015) was used to assign all protein sequences to Clusters of Orthologous Groups (COGs) with the 2014 edition of the COGs database as a reference (Galperin et al. 2015). The environmental distribution of Pseudovibrio spp. was investigated using the Integrated Microbial Next Generation Sequencing (IMNGS) platform (Lagkouvardos et al. 2016). Full-length 16S rRNA gene sequences of the type strains were used as query, and the similarity threshold was set at 99% nucleotide identity.

Annotation of Symbiosis-Related Genes

The standalone version of Interproscan 5.17-6.0 (Jones et al. 2014; Mitchell et al. 2015) was used to detect the following potentially symbiosis-related gene products: ankyrin-repeat proteins (ARPs; IPR020683 and IPR002110), proteins with SEL1 repeats (IPR006597), tetratricopeptide repeat proteins (TRPs; IPR011990, IPR019734, IPR013105, IPR001440, and IPR011717), invasion protein B (IPR010642), proteins with the TadE-like domain (IPR012495), and LuxR (IPR000792 and IPR005143). Proteins with SEL1 repeats and tetratricopeptide repeats are occasionally wrongly classified due to their high similarity. Therefore, we used TPRpred (Karpenahalli et al. 2007) to correctly categorize proteins with these repeats using an e-value of 1.0E−7 as cut-off. The presence of Type III secretion systems (T3SSs), Type IV secretion systems (T4SSs), and Type VI secretion systems (T6SSs) was investigated in the genomes with MultiGeneBlast 1.1.14 (Medema et al. 2013) using default settings with as query previously identified gene clusters that were described by Romano et al. (2016), that is, FO-BEG1_3657-03696 (T3SS), PsW64_00918-00938 (T4SS), FO-BEG1_01844-01855 (T6SS-I), and FO-BEG1_02827-02846 (T6SS-II).

Detection of Antibiotic Resistance Genes and SMB Clusters

The presence of the tdaABCDEFH genes and the genes paaI, paaJ, paaK, cysI, and malY in the genomes was assessed to predict the capacity for TDA biosynthesis (Geng and Belas 2010; Harrington et al. 2014), whereas the presence of the tdaR1, tdaR2, and tdaR3 genes (Wilson et al. 2016) was assessed to predict TDA resistance. Therefore, BLASTp using the sequences of UniProtKB accessions G8PKJ3_PSEUV (TdaA), G8PKJ1_PSEUV (TdaB), G8PKJ0_PSEUV (TdaC), G8PKI9_PSEUV (TdaD), G8PKI8_PSEUV (TdaE), G8PKH6_PSEUV (TdaF), G8PGC0_PSEUV (TdaH), I7DVM6_PHAI2 (tdaR1), I7EHC6_PHAI2 (tdaR2), and I7ES62_PHAI2 (tdaR3) was performed against the genome-encoded proteins. Furthermore, the presence of the genes paaI, paaJ, paaK, cysI, and malY was analyzed by investigating if genome-encoded proteins were assigned the EC numbers 3.1.2. (PaaI), 2.3.1.174 (PaaJ), 6.2.1.30 (PaaK), 1.8.1.2 (CysI), and 4.4.1.8. (MalY). The presence of gene clusters involved in curli fiber formation was investigated by BLASTp against the genome-encoded proteins using sequences of UniProtKB accessions G8PUY1_PSEUV (curlin associated repeat protein), G8PUY2_PSEUV (curlin associated repeat protein), G8PUY3_PSEUV (curlin associated repeat protein), G8PUY4_PSEUV (CsgG), and G8PUY5_PSEUV (CsgF). AntiSMASH version 3.0.4 was used to identify SMB clusters (Weber et al. 2015). Similarity among SMB clusters from the same type was investigated with MultiGeneBlast by calculating the number cluster-genes that shared ≥30% nucleotide identity (Medema et al. 2013). Antibiotic resistance functions were predicted with HMMER 3.1b2 (http://hmmer.org/) using the pHMMs of the Resfams database (core) v1.2 (Gibson et al. 2015). In addition, BLASTp of amino acid sequences against the CARD database (McArthur et al. 2013) was performed.

Results and Discussion

Metadata and Genome Characteristics

The genomes of 31 Pseudovibrio strains were analyzed (table 1) including 5 type strains that are currently characterized within the genus. The genomes of 28 Pseudovibrio strains are new and were sequenced as part of this study, whereas the genomes of strains Pseudovibrio sp. FO-BEG1, Pseudovibrio sp. JE062, and Pseudovibrio sp. POLY-S9 were obtained from previous studies (Bondarev et al. 2013; Alex and Antunes 2015). Twenty-five strains were isolated from sponges, four strains were isolated from other sessile marine animals, and two strains were isolated from seawater (table 1). The genome sizes of all Pseudovibrio strains, except for P. hongkongensis UST20140214-015BT, were large (6 Mb ± 0.6 [s.d.]). The genome of P. hongkongensis was only 3.75 Mb and correspondingly, the genome also contained fewer unique OGPs (2,878 in total). On average, Pseudovibrio genomes, excluding P. hongkongensis UST20140214-015BT, contained 4,857 ± 508 (s.d.) unique OGPs (table 1, supplementary table S1, Supplementary Material online). The GC-content of the Pseudovibrio strains ranged from 44.6% for Pseudovibrio sp. AB108 to 52.4% for Pseudovibrio sp. FO-BEG1, whereas the GC-content of the three close relatives was 63.9% for S. stellulata DSM 5886T isolated from marine sediment (Ruger and Hofle 1992), 54.6% for N. exalbescens LA33BT isolated from water in a hypersaline lake (Donachie et al. 2006), and 57.4% for L. alexandrii DFL-11T isolated from a culture of the dinoflagellate Alexandrium lusitanicum (Fiebig et al. 2013).

Table 1

Metadata, Genome Assembly, and Annotation Statistics of the 31 Pseudovibrio Strains and the Three Close Relatives (Outgroup Members) Analyzed in This Study

Genome Metadata			Assembly Statistics					Genome Annotation Statistics
Strain	Isolation Source	Geographic Location	Number of Scaffolds	Genome Size (Mb)	N50 (Mb)	GC content (%)	Coverage (X)	Number of ORFs	Number of OGP-assigned proteins	Number of unique OGPs
*P. hongkongensis UST20140214-015BT	Flatworm Stylochus sp.	Yung Shue O, Hong Kong	39	3.75	0.34	51.7	131	3,514	2,939	2,878
*P. ascidiaceicola DSM 16392T	Ascidian P. proliferus	Beach of Boso peninsula, Japan	40	5.88	0.42	50.3	178	5,365	5,112	4,992
*P. axinellae DSM 24994T	Sponge A. dissimilis	A marine lake, Ireland	78	5.20	0.39	50.7	230	4,786	4,122	3,982
*P. denitrificans JCM 12308T	Seawater	Nanwan Bay, Taiwan	27	6.12	0.64	50.7	199	5,428	5,123	5,033
*P. japonicus NCIMB 14279T	Seawater	Coast of Boso peninsula, Japan	17	4.97	0.65	50.9	245	4,554	3,945	3,870
*Pseudovibrio sp. 2G02	Sponge P. ficiformis	Coast of Blanes, Spain	25	5.87	0.41	50.2	246	5,369	5,204	5,108
*Pseudovibrio sp. 2H03	Sponge P. ficiformis	Coast of Blanes, Spain	39	6.00	0.47	50.0	233	5,524	5,349	5,210
*Pseudovibrio sp. 2H05	Sponge P. ficiformis	Coast of Blanes, Spain	28	5.94	0.42	50.1	137	5,389	5,234	5,138
*Pseudovibrio sp. 2H06	Sponge P. ficiformis	Coast of Blanes, Spain	21	6.09	0.57	50.7	221	5,474	5,328	5,200
*Pseudovibrio sp. 4B07	Sponge P. ficiformis	Coast of Blanes, Spain	52	5.13	0.38	50.4	226	4,692	4,281	4,164
*Pseudovibrio sp. 4H03	Sponge P. ficiformis	Coast of Blanes, Spain	26	5.79	0.88	50.3	212	5,295	5,196	5,078
*Pseudovibrio sp. 4H10	Sponge P. ficiformis	Coast of Blanes, Spain	14	5.78	0.86	50.8	250	5,218	5,116	5,029
*Pseudovibrio sp. 8G08	Sponge P. ficiformis	Coast of Blanes, Spain	36	5.94	0.48	50.3	165	5,361	5,240	5,144
*Pseudovibrio sp. 8H04	Sponge P. ficiformis	Coast of Blanes, Spain	18	5.68	0.50	48.2	295	5,174	4,653	4,512
*Pseudovibrio sp. 8H06	Sponge P. ficiformis	Coast of Blanes, Spain	40	6.06	0.40	49.7	264	5,554	5,342	5,207
*Pseudovibrio sp. 1F12	Sponge A. aerophoba	Gulf of Roses, Spain	28	5.84	0.49	50.9	253	5,289	5,272	5,291
*Pseudovibrio sp. 1D03	Sponge A. aerophoba	Gulf of Roses, Spain	22	5.77	0.50	50.3	711	5,239	5,126	5,051
*Pseudovibrio sp. 3C10	Sponge A. aerophoba	Gulf of Roses, Spain	36	6.05	0.85	49.9	316	5,539	5,188	5,082
*Pseudovibrio sp. 7F08	Sponge A. aerophoba	Gulf of Roses, Spain	27	5.78	0.48	51.3	124	5,243	5,127	5,021
*Pseudovibrio sp. 7G12	Sponge A. aerophoba	Gulf of Roses, Spain	29	5.85	0.45	51.0	138	5,288	5,271	5,192
*Pseudovibrio sp. 7H02	Sponge A. aerophoba	Gulf of Roses, Spain	28	5.84	0.52	50.9	119	5,288	5,272	5,192
*Pseudovibrio sp. 7H04	Sponge A. aerophoba	Gulf of Roses, Spain	23	5.59	0.58	50.4	119	5,041	4,884	4,815
*Pseudovibrio sp. 7A08	Sponge C. candelabrum	Coast of Blanes, Spain	22	5.76	0.59	50.8	126	5,176	5,116	5,014
*Pseudovibrio sp. 2GO5	Sponge C. candelabrum	Coast of Blanes, Spain	20	5.76	0.59	50.8	189	5,182	5,117	5,020
*Pseudovibrio sp. AB108	Sponge Ircinia sp.	Limski Channel, Croatia	49	5.91	0.44	44.6	200	5,369	5,202	5,098
*Pseudovibrio sp. AB111	Sponge C. nucula	Limski Channel, Croatia	36	5.92	0.42	49.8	228	5,423	5,211	5,074
*Pseudovibrio sp. AB113	Sponge A. acuta	Limski Channel, Croatia	13	5.37	0.88	51.0	191	4,886	4,796	4,727
*Pseudovibrio sp. BC118	Bryozoan C. pumicosa	Limski Channel, Croatia	23	5.81	0.52	51.3	237	5,304	5,169	5,070
Pseudovibrio sp. FO-BEG1	A scleractinian coral	Coast of Florida, USA	1	5.48	5.48	52.4	n/a	4,926	4,842	4,673
Pseudovibrio sp. JE062	Sponge M. laxissima	Key Largo, Florida, USA	19	5.73	0.51	52.1	n/a	5,138	4,928	4,842
Pseudovibrio sp. POLY-S9	Sponge P. penicillus	Coast of Portugal	271	6.60	0.12	51.0	n/a	6,331	5,530	5,308
S. stellulata DSM 5886T	Marine sediment		15	4.62	0.69	63.9	n/a	4,212	2,879	2,810
N. exalbescens LA33BT	Water in a hypersaline lake	Laysan Atoll, Hawaiian Islands	40	4.15	0.29	54.6	n/a	3,771	3,094	3,022
L. alexandrii DFL-11T	Culture of the dinoflagellate A. lusitanicum	Helgoland, Germany	6	5.50	5.30	57.4	n/a	5,128	3,276	3,154

Note.—An asterisk indicates that the genome was sequenced as part of this study.

Phylogeny and Taxonomy

Phylogeny of the strains was assessed based on 16S rRNA gene sequences retrieved from the respective genomes as well as the presence of OGPs. Phylogenetic analysis based on 16S rRNA gene sequences placed all Pseudovibrio type strains (DSM 16392T, DSM 24994T, JCM 12308T, NCIMB 14279T, and UST20140214-015BT) in separate branches of the corresponding tree (fig. 1). Most Pseudovibrio strains (23 of 31) could be classified at the species level to P. ascidiaceicola based on the phylogenetic tree and 16S rRNA gene identity values (supplementary table S2, Supplementary Material online). Further phylogenomic analysis of these strains based on the presence of OGPs placed P. hongkongensis UST20140214-015BT closer to the non-Pseudovibrio outgroup members when compared with the 16S rRNA gene-based phylogenetic analysis (fig. 2). In addition, the ANI between the P. hongkongensis type strain and all other Pseudovibrio spp. strains was <75%, which was comparable to ANI values observed for the three outgroup species that all belong to different genera (supplementary table S3, Supplementary Material online). As such, the data presented here suggest that P. hongkongensis UST20140214-015BT should not be considered a member of the Pseudovibrio genus and therefore might need to be placed in a separate genus. For this reason, it was treated as an outgroup member in all subsequent analyses. Furthermore, Pseudovibrio sp. 8H04 and Pseudovibrio sp. 4B07 were found in isolated branches in both trees (fig. 2), and 16S rRNA gene identity values did not allow species classification as both these strains shared between 97.6% and 98.6% nucleotide identity with the four most closely related type strains: P. ascidiaceicola DSM 16392T, P. denitrificans JCM 12308T, P. japonicus NCIMB 14279T, and P. axinellae DSM 24994T. These results, and the fact that the ANI values between these two strains and the Pseudovibrio spp. type strains are rather divergent (Richter and Rossello-Mora 2009) (Pseudovibrio strain 4B07 and strain 8H04 share, respectively, 94.3% and 97.1% ANI with the most similar type strain), indicate that they likely represent two distinct novel species within the Pseudovibrio genus. These strains would not have been recognized as novel species based on 16S rRNA gene similarity alone, and hence, 16S rRNA gene identity values underrepresent the genomic diversity within Pseudovibrio. This is also in line with the observation that P. japonicus NCIMB 14279T is closely related to P. denitrificans JCM 12308T based on the 16S rRNA gene tree (99% sequence identity) even though the two strains were highly dissimilar based on the OGP-based tree and ANI (81.4%). ANI scores showed strong r2 correlation values of >0.93 with DNA–DNA hybridization (DDH) values and therefore can confidently assess species delineation (Konstantinidis and Tiedje 2005; Goris et al. 2007). The 70% DDH cut-off value for species delineation corresponds to 94–95% ANI (Richter and Rossello-Mora 2009). It should be noted, however, that genome-to-genome sequence comparison methods such as ANI are not (yet) allowed to be used as substitute for DDH in the characterization of a novel taxon (Auch et al. 2010; Tindall et al. 2010; Kim et al. 2014).

. 1.

—Phylogenetic tree based on 16S rRNA gene sequences from the 31 Pseudovibrio spp. and 3 closely related species that were included in the comparative genomics analysis. Pseudovibrio type strains are in blue and the closely related species are in green. The tree was constructed in ARB by Maximum likelihood analysis using 1,000 iterations of bootstrapping. Bootstrap values <50 are not shown. The horizontal bar indicates the number of substitutions per site.

. 2.

—Functional similarity and phylogenetic distance of Pseudovibrio spp.. Comparison of two phylogenetic trees made by hierarchical linkage clustering based (1) on presence/absence of OGPs that are present in greater than two genomes and (2) the 16S rRNA gene alignment. Pseudovibrio type strains are depicted in blue and the outgroup members are depicted in green. Hierarchical clustering was performed using 10,000 iterations of bootstrapping. Dashed branches can be swapped.

Clustering by Metadata and Group-Differences

To identify variables that best explain the variation in the distribution of OGPs in the genomes, we performed a canonical correspondence analysis (CCA). Strains that are members of the outgroup (including P. hongkongensis UST20140214-015BT) clustered separately from members of the Pseudovibrio genus (P = 0.002) (fig. 3). Furthermore, we found clustering by geographic origin (Mediterranean, Caribbean, Pacific, or Atlantic, P = 0.002) (fig. 3). The fact that geographic origin can partially explain the variation in OGPs suggests that in different geographic regions adaptation has occurred to local environmental conditions (e.g., temperature, salinity, local flora and fauna, or nutrient availability). This is underlined by the fact that clusters in two of four geographic regions included different Pseudovibrio species. Namely, the Atlantic Ocean cluster includes isolates belonging to P. ascidiaceicola (POLY-S9) and P. axinellae (DSM 24994T), and the Pacific Ocean cluster includes P. japonicus NCIMB 14279T, P. denitrificans JCM 12308T, and P. ascidiaceicola DSM 16392T. Most studies find that microbial composition and biogeography correlate in the aquatic environment, suggesting selective pressure of at least one environmental variable (Hanson et al. 2012). However, to what extent (trait) selection plays a role in the genetic variation among Pseudovibrio spp. by geography, as opposed to other suggested major processes such as drift, dispersal, and mutation, remains to be elucidated. This question can be further investigated by attempting to link variation in OGPs to environmental variables, and reinforces the need for proper recording of metadata for environmental samples and derived microbial isolates (DeLong 2009; Sun et al. 2011). We also found that Pseudovibrio strains isolated near the coast of Spain clustered by sponge of origin, that is, A. aerophoba, C. candelabrum, or P. ficiformis (P = 0.006) (fig. 3). This might indicate that Pseudovibrio strains harbor functional genes that are associated with their ability to preferentially colonize certain sponge species. In turn, it might indicate that the Pseudovibrio strains selectively evolved in the sponge host after colonization. At least in some cases, sponge-associated bacteria are transmitted vertically (Sharp et al. 2007; Schmitt et al. 2008; Lee et al. 2009; Sipkema et al. 2015), and there is evidence that Pseudovibrio can be vertically transmitted via sponge larvae (Enticknap et al. 2006).

. 3.

—Clustering of strains by genus membership, geographic origin and sponge of origin. CCA was used to investigate which environmental variables could explain the variation in orthologous protein group composition among the genomes. All clustering shown in this figure is statistically significant (P < 0.01). Panel (A) shows results based on grouping of the genomes according to membership of the Pseudovibrio genus. In panel (B), the Pseudovibrio genomes are grouped according to their geographic origin. In panel (C), the Pseudovibrio strains that were isolated at the coast of Spain are grouped according to their sponge of origin.

To identify differences in substrate utilization and product formation between the strains, the presence of metabolic pathways was investigated using the MetaCyc database as a reference. This database contains pathways and enzymes that are predominantly found in microorganisms and plants. We analyzed pathways that contained at least three enzymes with an EC number (supplementary fig. S1, Supplementary Material online) and pathways that contained at least four enzymes with an EC number (supplementary fig. S2, Supplementary Material online). In terms of metabolism, all members of the Pseudovibrio genus were highly similar, that is, across all genomes the same pathways were detected with differences between strains nearly always amounting to the presence and/or absence of single genes. Pseudovibrio spp. are metabolically highly versatile, which may enable them to survive in distinct habitats containing different substrates. Furthermore, Pseudovibrio sp. FO-BEG1 has been shown to sustain growth in ultra-oligotrophic seawater by simultaneously degrading different compounds (Schwedt et al. 2015). Although high metabolic versatility and large genome sizes are not considered typical for symbiotic bacteria (McCutcheon and Moran 2011; Dutta and Paul 2012), it could be that these characteristics benefit Pseudovibrio spp. in a variable and nutrient-rich environment such as in sponges. In contrast to members of the outgroups, all Pseudovibrio spp. were predicted to be able to synthesize Coenzyme F420. Coenzyme F420 is best known as an essential coenzyme of methanogenesis (Hendrickson and Leigh 2008). However, because Pseudovibrio spp. are incapable of methanogenesis, F420 might be involved in SMB (Mccormick and Morton 1982; Peschke et al. 1995) or other metabolic activities such as aerobic catabolism of 2,4,6-trinitrophenol or as an electron carrier to an F420-dependent glucose-6-phosphate dehydrogenase (Purwantini et al. 1997; Ebert et al. 1999; Stover et al. 2000). In contrast to the three non-Pseudovibrio close relatives, all Pseudovibrio strains (including P. hongkongensis UST20140214-015BT) were predicted to have the capacity to synthesize biotin (vitamin B7). It has been suggested that sponges can benefit from the presence of bacteria that produce biotin, as they cannot synthesize this vitamin themselves (Webster and Thomas 2016).

High similarity between the Pseudovibrio spp. was also found with respect to the number of proteins assigned to COG functional categories (supplementary table S4, Supplementary Material online). The most pronounced difference between Pseudovibrio strains and the outliers was found for COG category “Mobilome: prophages, transposons,” to which Pseudovibrio strains had on average twice the number of proteins assigned compared with the members of the outgroup.

Next, the differences regarding the presence of individual OGPs were investigated with SIMPER to investigate which functional capacities define our analyzed genomes when grouped according to geographic origin, sponge of origin, and membership of the Pseudovibrio genus (supplementary table S1, Supplementary Material online). Genes encoding enzymes that have urease activity (EC 3.5.1.5) were exclusively predicted in the outgroup members, the three P. denitrificans strains and P. axinellae DSM 24994T. The low abundance of nitrogen may be growth limiting in the oceans (Antia et al. 1991), and regeneration of urea (i.e., by urease), which is a waste product of many animals, may alleviate this restraint. Following this hypothesis, bacteria in sessile marine animals could enhance their growth by using the urea excreted by their host as nitrogen source (Su et al. 2013). Alternatively, the bacteria could contribute to the nitrogen budget of the host by fixation of environmental urea (Wilkinson and Fay 1979). Given that the urease gene was detected in only a few (3 of 29) Pseudovibrio strains that reside in marine animals, we do not expect it to play an important role in the hosts–symbiont relationship, or this gene only plays a role in specific hosts. SIMPER analysis also showed that OGPs corresponding to enzymes involved in vitamin B12 (cobalamin) and vitamin B1 (thiamin) biosynthesis were exclusively found in Pseudovibrio spp., but were absent in the outgroup members. Bacteria may be an essential contributor to the nutritional requirements of marine animals through production of these essential vitamins that the animals cannot synthesize themselves (Bondarev et al. 2013). However, we found that only a few enzymes of the multienzyme pathways to produce vitamins B1 and B12 were present in Pseudovibrio indicating that the vitamins cannot be de novo synthesized. Regarding vitamin B1 biosynthesis, all Pseudovibrio strains lack a phosphotransferase (EC number: 2.7.4.7) and a thiazole tautomerase (5.3.99.10), both of which catalyse the final step to respectively produce pyrimidine and thiazole moieties that are required to form thiamine phosphate. Cobalamin cannot be synthesized de novo by Pseudovibrio spp. due to the absence of essential enzymes such as the adenosylcobinamide phosphate synthase (EC number: 6.3.1.10) and the adenosyl-cobric acid synthase (6.3.5.10). No further characteristic functional traits could be ascribed to Pseudovibrio genomes grouped by environmental variables because a large number of the idiosyncratic OGPs were hypothetical proteins or were proteins involved in routine processes.

Symbiosis

Eukaryotic-like Proteins, Virulence Factors, and Quorum Sensing

The interactions that are hypothesized to occur between Pseudovibrio spp. and the sponge host are still not well understood. Eukaryotic-like proteins (ELPs) were first predicted to play a role in symbiosis when they were discovered to be highly prevalent in sponge-associated bacteria (Liu et al. 2011; Siegl et al. 2011; Thomas et al. 2010; Fan et al. 2013). ARPs, which are ELPs, were predicted to interfere with phagocytosis of Pseudovibrio spp. by sponge host cells by retarding phagosome biogenesis, or by blocking fusion of the phagosome with the lysosome and its digestive enzymes (Nguyen et al. 2014). It has also been estimated that the average number of ARPs per sponge symbiont genome is between 10 and 40 when compared with an average of 2.5 ARPs in genomes of other symbiotic bacteria (Fan et al. 2012). Our analysis predicted ARPs in 23 of 25 sponge-associated Pseudovibrio strains, with an overall average of 1.57 ±1.03 ARPs per genome (fig. 4 and supplementary table S5, Supplementary Material online), which does rank them among known symbiotic bacteria based on the number of ARPs. Proteins containing SEL1 repeats are also predominantly found in eukaryotes, and these proteins have been predicted to be symbiotic factors by which resident bacteria can interact with the host cells (Mittl and Schneider-Brachert 2007; Alex and Antunes 2015). Our analysis predicted on average 6.93 ±1.51 proteins with SEL1 repeats in the Pseudovibrio genomes. In the genomes of the Pseudovibrio strains that were not associated with marine animals (i.e., those isolated from seawater), that is, P. japonicus NCIMB 14279T and P. denitrificans JCM 12308T, nine and eight proteins with SEL1 repeats were detected, respectively. ELPs with TRPs have been reported to participate in the delivery of virulence factors to the host cell (Cerveny et al. 2013). On average, we detected 34.27 ±2.43 proteins with TRPs in the Pseudovibrio genomes. We also detected on average 6.27 ±1.53 proteins with invasion associated locus B, which has been reported to be involved in host cell invasion (Coleman and Minnick 2003; Eicher and Dehio 2012). Furthermore, in each Pseudovibrio genome, except for Pseudovibrio sp. POLY-S9, one or two proteins were predicted that contain a TadE-like domain, suggesting that these might be involved in adherence of Pseudovibrio to host cells (Tomich et al. 2007). Based on membership of the above-mentioned symbiotic factors to OGPs, we can draw conclusions about their ancestry (supplementary table S5, Supplementary Material online). ARPs and proteins containing a TadE-like domain belonged to a few distinct OGPs that were each present in a subset of the Pseudovibrio strains, where no proteins containing a TadE-like domain and only one ARP were detected in the outgroup members. Therefore, we speculate that these proteins were either selectively maintained or acquired by Pseudovibrio spp.. A subset of the OGPs that encoded proteins with the SEL1 repeat (2/9), proteins with invasion associated locus B (2/10) and TRPs (15/59) were present in all genomes, including those of the outgroup members, indicating that these proteins were highly conserved. Subsequently, we investigated whether the LuxR/LuxI quorum sensing system was present in the Pseudovibrio genomes. The enzyme encoded by luxI catalyses the final step in the production of N-acyl homoserine lactone (AHL). AHL can bind to the regulatory protein LuxR, which together activate transcription of genes involved in AHL biosynthesis (auto-induction) as well as a wide array of other genes (Miller and Bassler 2001; Nasser and Reverchon 2007). AHL is secreted by the bacterium; therefore, as the population density increases, the concentration of AHL will increase, resulting in activation of gene expression across the bacterial community. In each Pseudovibrio genome, we detected more than nine copies of the luxR gene whereas in none of the genomes luxI was detected. An absence of luxI suggests that Pseudovibrio strains are dependent on quorum sensing molecules produced by neighbouring bacteria for expression of luxR/AHL-controlled genes, which has been previously hypothesized based on the genomes of Pseudovibrio sp. FO-BEG1 and Pseudovibrio sp. JE062 (Case et al. 2008; Bondarev et al. 2013). The genes regulated by quorum sensing may be involved in production of secondary metabolites (e.g., against microbial competitors or host pathogens) or biofilm formation (Dobretsov et al. 2009; Subramoni and Venturi 2009; Zan et al. 2012). In 29 of the 31 Pseudovibrio genomes analyzed here, we detected a cluster of genes putatively involved in biogenesis of proteinaceous extracellular fibres called curli. These extracellular curli are involved in biofilm formation, and as such may contribute to sponge colonization (Barnhart and Chapman 2006). Compared with the Pseudovibrio genomes, the genomes of the outgroup members encoded similar numbers of ELPs, LuxR regulators, proteins with invasion associated locus B and proteins containing the TadE-like domain. However, no outgroup member encoded proteins involved in curli fiber biosynthesis.

. 4.

—Symbioses-related proteins, secretions systems, and SMB clusters. Green squares indicate presence. Numerical values were included if functions were detected more than once. The tree on top was made by hierarchical linkage clustering based on the presence/absence of OGPs that are present in greater than two genomes. Type strains are in blue and the close relatives are in green. Hierarchical clustering was performed using 10,000 iterations of bootstrapping.

Secretion Systems

The distribution of T3SSs, T4SSs, and T6SSs was investigated (fig. 4) because these systems have been ascribed important roles in a range of symbiotic interactions. All three types of secretion systems allow bacteria to deliver effector molecules into a target cell via needle-like structures in the cellular membrane (Cornelis 2006; Coulthurst 2013; Low et al. 2014). Besides effector molecules, T4SSs can also translocate genetic material to the target cell. In all Pseudovibrio genomes, except the genome of P. japonicus NCIMB 14279T, at least two of these secretions systems were detected. Interestingly, in the genome of P. japonicus NCIMB 14279T, not a single of the aforementioned secretion systems was found indicating that the strain might lack equipment to form symbiotic relations. This notion is in line with the fact that P. japonicus NCIMB 14279T was isolated from the open sea, and that members of this species were so far never isolated from marine animals. On the other hand, the notion is contradicted by P. japonicus spp. being detected more often in sponge 16S rRNA gene data sets when compared with seawater data sets (fig. 5). T6SS-I and/or T6SS-II clusters were detected in all Pseudovibrio genomes (except P. japonicus NCIMB 14279T), and the genes in these clusters were all homologous and showed perfectly conserved synteny (supplementary table S6, Supplementary Material online). The fact that the clusters are conserved in Pseudovibrio strains from diverse geographical locations and species suggests that they are important for survival and propagation. T3SS clusters were limited to Pseudovibrio strains (27/28) belonging to the species P. ascidiaceicola and P. denitrificans, whereas T4SSs were detected in only 10 of all Pseudovibrio genomes. These results suggest Pseudovibrio spp. developed different strategies to carve a niche in a competitive environment. Romano et al. carried out a more detailed analysis of these clusters for Pseudovibrio sp. FO-BEG1 and Pseudovibrio sp. JE062, as well as for 10 other Pseudovibrio strains, by analyzing which effector molecules are putatively delivered into host cells (Romano et al. 2016). It was predicted that T3SSs were mainly used to reduce phagocytosis and to block the inflammatory response of the host, that T4SSs were mainly involved in DNA mobilization, and that T6SSs might be involved in evasion of the host immune response, or they could have an antagonistic role toward other bacteria. Only a single secretion system, namely the T4SS in P. hongkongensis UST20140214-015BT, was detected in the outgroup members. The members of the outgroup (except P. hongkongensis UST20140214-015BT) do not belong to genera that are implicated in forming symbiotic relations. In addition, none of the outgroup members, except L. alexandrii DFL-11T that was washed from the dinoflagellate A. lusitanicum (Biebl et al. 2007), were isolated from a host. Hence, the absence of symbiosis-related gene clusters in their genomes (e.g., those encoding curli fibre formation and secretion systems) likely indicates that these bacteria are not committed to forming symbiotic relations.

. 5.

—The presence of Pseudovibrio in 16S rRNA gene data sets from different niches. The percentage of 16S rRNA gene amplicon data sets in the SRA database in which the Pseudovibrio type strains were detected. The analysis was performed by the IMNGS platform, which divides the data sets in the SRA in 105 categories according to which biological niche was sampled. Full-length 16 S rRNA gene sequences of the type strains were used as query. The similarity threshold was set at 99% nucleotide identity. All environmental categories are shown in which at least one type strain was detected at > 0.1% relative abundance.

Secondary Metabolite Biosynthesis Clusters

Pseudovibrio spp. have received great interest as sources of novel bioactive compounds, and especially antimicrobials (Crowley et al. 2014). Not only are novel bioactive compounds interesting from a clinical perspective but they may also provide insights into possible host–symbiont relations. Sponges in part depend on the production of bioactive compounds by associated microorganisms for defence against predation and disease (Taylor et al. 2007; Webster and Taylor 2012). Recently, sponge-derived Pseudovibrio spp. were found to inhibit a Bacillus sp. with sponge-degrading activity that was isolated from the sponge genus Tedania (Esteves et al. 2017).

TDA is a compound that has antibacterial activity, and was found to be produced by Pseudovibrio sp. D323 (Penesyan et al. 2011). TDA has a strong inhibitory activity against a range of marine bacteria, and hence has been proposed to protect the host from unwanted microbial colonization (Harrington et al. 2014). It has been previously shown that Pseudovibrio sp. FO-BEG1 and Pseudovibrio sp. JE062 can produce TDA, which was linked to the presence of the tdaA-tdaF biosynthetic gene cluster in the respective genomes (Bondarev et al. 2013). Here we found the tdaA-tdaF biosynthetic cluster in 26 of 31 Pseudovibrio genomes suggesting that TDA production is an important feature of the genus Pseudovibrio (fig. 4 and supplementary table S5, Supplementary Material online). In all these cases, the TDA resistance-conferring genes tdaR1, tdaR2, and tdaR3 (Wilson et al. 2016) were detected immediately downstream of the tdaA-tdaF operon. However, the Pseudovibrio strains that lacked TDA biosynthesis also lack TDA resistance, and as such these strains likely do not benefit, but rather are harmed, in the company of coresident TDA-producing Pseudovibrio spp.. Further potential for SMB was analyzed by application of the antiSMASH pipeline (Weber et al. 2015). Eighteen of 43 types of SMB clusters defined by antiSMASH were detected in the Pseudovibrio genomes, with bacteriocin, NRPS, siderophore, t1pks-t3pks, t3pks-t1pks, and terpene SMB clusters being detected in >5 genomes (fig. 4). At least one bacteriocin biosynthesis gene cluster was identified in all Pseudovibrio genomes, except for P. axinellae DSM 24994T. The main role of bacteriocins, which are ribosomally synthesized antibiotic peptides, is to provide their producers with a competitive advantage by inhibiting bacterial growth (Riley and Wertz 2002; Balko 2012). Therefore, similar to TDA, bacteriocins could serve Pseudovibrio to establish a niche in the sponge host, and at the same time protect the host against pathogen invasion. So far, few bacteriocins have been identified in the marine environment (Desriac et al. 2010), and only one study has identified novel bacteriocins in a sponge (Phelan et al. 2013). Recently, based on genome data, bacteriocins were also predicted to be produced by members of the obligate marine genus Pseudoalteromonas (Bosi et al. 2017). Mutual comparison of bacteriocin biosynthesis gene clusters in Pseudovibrio and Pseudoalteromonas genomes predicted that they code for different bacteriocins (supplementary table S7, Supplementary Material online). AntiSMASH detected a terpene biosynthesis gene cluster in all Pseudovibrio genomes. Via mining of bacterial genomes, it has become clear that terpenes are not only regularly produced by plants and fungi but also by a large number of bacteria (Cane and Ikeda 2012). Terpenes may give Pseudovibrio a competitive advantage due to antimicrobial activity (Martin et al. 1970; Song et al. 2015). Alternatively, the fragrance of the terpenes might be an addition to sponge-derived secondary metabolites that are designed to repel predators (Proksch 1994; Epifanio et al. 1999; Ding et al. 2015). However, terpenes typically have limited water solubility. For the SMB clusters detected by antiSMASH in Pseudovibrio spp., no more than 26% of the genes in any cluster showed similarity to genes in known clusters. The diversity among SMB clusters from the same type was analyzed by investigating the total number of cluster genes that mutually shared  ≥30% nucleotide identity (supplementary table S8, Supplementary Material online). Based on the number of clusters from the same type that share <50% of the genes at  ≥30% nucleotide identity, we predict that NRPS, bacteriocin, terpene, and t1pks-t3pks clusters across all Pseudovibrio genomes each encode for at least three different products. Taken together, these findings highlight Pseudovibrio spp. as a largely untapped resource for the discovery of novel compounds with potentially clinical significance.

Antibiotic resistance profiles were determined for 27 of 31 Pseudovibrio strains that were analyzed in this study (supplementary table S9, Supplementary Material online). All 27 strains were resistant to vancomycin, lincomycin, polymyxin B, and d-cycloserine; however, these resistance phenotypes could not be linked to the presence of known resistance genes. Therefore, resistance to these antibiotics is probably intrinsic. Twenty-one of 27 strains were resistant to tetracycline, which could not be linked to the presence of resistance genes in the genomes. All strains were either resistant or intermediately resistant to the β-lactam antibiotics ampicillin and penicillin, which could be explained by the class A and class B β-lactamases that were predicted to be encoded in their genomes (supplementary table S10, Supplementary Material online). The predicted class A β-lactamase encoded in 22 P. ascidiaceicola genomes was highly similar (≥95% amino acid identity) to blaPSV-1 from P. ascidiaceicola strain 1D03 of which the resistance function has been experimentally verified by functional metagenomics (Versluis et al. 2016). Within the Pseudovibrio genus, almost all class A β-lactamases clustered by species (supplementary fig. S3, Supplementary Material online). The exception was a distinct class A β-lactamase (8H04_2067) predicted in Pseudovibrio sp. 8H04 of which the closest experimentally verified β-lactamase was blaPSV-1. However, as previously deduced, Pseudovibrio sp. 8H04 is expected to represent a distinct novel Pseudovibrio species as well. These results, together with the absence of associated mobilizing elements, suggest that the class A β-lactamases are not prone to horizontal gene transfer, and that diversity within this group of enzymes is predominately achieved by speciation. The β-lactamases in these bacteria could function as a defence against β-lactam antibiotics that are produced by other micro-organisms coinhabiting sessile marine animals. Still, these enzymes could also have other roles such as in disruption of cell signalling (Allen et al. 2009).

Environmental Distribution of Pseudovibrio

The 16S rRNA gene data sets in the Sequence Read Archive (SRA) database (https://www.ncbi.nlm.nih.gov/sra; last accessed April 14, 2016) were interrogated for the presence of the Pseudovibrio type strains using the IMNGS platform (Lagkouvardos et al. 2016) (fig. 5). IMNGS divides the 16S rRNA gene data sets in 105 categories according to sample origin. The highest percentage of data sets containing Pseudovibrio spp. was found for the category sea squirts (20.8%), followed by data sets acquired from sponges (13.7%) and symbionts (7.6%). Pseudovibrio spp. were only detected in 16S rRNA gene data sets from marine habitats or from generalized categories that can also contain data sets from marine habitats. This is consistent with the notion that Pseudovibrio spp. have been isolated exclusively from the marine environment and depend on seawater to thrive. The fact that Pseudovibrio spp. were detected more often in 16S rRNA gene data sets from sessile marine animals as opposed to aquatic environments suggests that indeed Pseudovibrio spp. form symbiotic relations. Pseudovibrio ascidiaceicola, P. denitrificans, and P. japonicus were each detected at >6% relative abundance in 16S rRNA gene data sets from at least two categories that represent distinct biological niches, whereas P. axinellae and P. hongkongensis were never detected at >1% relative abundance in any biological niche.

Conclusions

The analysis of 16S rRNA gene data sets revealed that Pseudovibrio spp. are predominantly found and may be specifically associated with sessile marine animals. Our analysis of a total of 31 Pseudovibrio genomes revealed that, except P. hongkongensis UST20140214-015BT, members of the genus Pseudovibrio are highly similar in terms of metabolic capacity, with striking differences regarding the presence of specific SMB clusters. We identified various genomic elements that are highly conserved within the genus and are expected to give Pseudovibrio a competitive advantage as a symbiont of sponges or other marine animals. These conserved elements encompassed systems involved in immune evasion (T4SSs and T6SSs), SMB clusters with products that might inhibit growth of competitors (TDA, bacteriocin, and terpene biosynthesis clusters), and a range of other factors relevant for symbiotic relations such as ELPs and vitamin biosynthesis. The importance of these elements for Pseudovibrio survival is underscored by their conservation across Pseudovibrio spp. and across strains isolated from diverse geographic locations. Interestingly, based on gene composition, Pseudovibrio strains from the Mediterranean Sea clustered by sponge species of origin, suggesting host-specific colonization or adaptation. Taken together, the findings presented here support the hypothesis that Pseudovibrio spp. have evolved as symbionts of sponges and other marine invertebrates.

Supplementary Material

Supplementary data are available at Genome Biology and Evolution online.

Click here for additional data file.