From the Flavobacterium genus to the phylum Bacteroidetes: genomic analysis of dnd gene clusters


Correspondence: Eric Duchaud, INRA, Virologie et Immunologie Moléculaires UR892, Jouy-en-Josas, France. Tel.: +33 1 34 65 25 88; fax: +33 1 34 65 25 91; e-mail:


Phosphorothioate modification of DNA and the corresponding DNA degradation (Dnd) phenotype that occurs during gel electrophoresis are caused by dnd genes. Although widely distributed among Bacteria and Archaea, dnd genes have been found in only very few, taxonomically unrelated, bacterial species so far. Here, we report the presence of dnd genes and their associated Dnd phenotype in two Flavobacterium species. Comparison with dnd gene clusters previously described led us to report a noncanonical genetic organization and to identify a gene likely encoding a hybrid DndE protein. Hence, we showed that dnd genes are also present in members of the family Flavobacteriaceae, a bacterial group occurring in a variety of habitats with an interesting diversity of lifestyle. Two main types of genomic organization of dnd loci were uncovered probably denoting their spreading in the phylum Bacteroidetes via distinct genetic transfer events.


Most members of the family Flavobacteriaceae (hereafter designated flavobacteria), a prominent member of the phylum Bacteroidetes, are free-living organisms occurring in a variety of temperate and polar habitats in terrestrial, freshwater, and marine environments (Bernardet, 2011). Because of their ecological significance (Brown & Bowman, 2001; Kirchman, 2002; Horner-Devine et al., 2003), a large effort was recently undertaken to unravel their life styles including the use of genomic and metagenomic approaches (Venter et al., 2004; Bauer et al., 2006; Gómez-Consarnau et al., 2007; González et al., 2008; Oh et al., 2010).

The original identification of a five-gene (dndA-E) cluster in Streptomyces lividans (Zhou et al., 2005), which caused DNA degradation during electrophoresis (Zhou et al., 1988), has led to the discovery of phosphorothioate (PT) DNA modification, in which sulfur replaces a nonbridging phosphate oxygen (Wang et al., 2007). Yet, dnd genes have been identified in very few, taxonomically unrelated, bacterial and archaeal species so far (Ou et al., 2009; Wang et al., 2011). Phylogenies of Dnd proteins correlated with the PT sequence context strongly support the spreading of dnd genes by horizontal transfer (Wang et al., 2011). However, the functional role of PT modifications still remains unclear. Some studies have proposed their involvement in (1) the regulation of gene expression (Eckstein, 2007; Xu et al., 2010), (2) a new restriction–modification system protecting endogenous bacterial genome against foreign DNA introduction (Xu et al., 2010; Wang et al., 2011), and (3) the protection of bacterial DNA against oxidation (Xie et al., 2012).

Our group has a long-standing history in the study of flavobacteria and is currently involved in genomic projects on fish-pathogenic and environmental Flavobacterium species. During distinct projects, we identified a Dnd phenotype during gel migration in two bacterial strains belonging to two different Flavobacterium species. As they had not been reported yet in flavobacteria, we investigated the presence of dnd gene clusters and performed genomic comparisons that allow to identify noncanonical dnd gene clusters in members of the phylum Bacteroidetes.

Materials and methods

Bacterial strains and growth conditions

Flavobacterium indicum CIP 109464T was isolated from a warm spring in India (Saha & Chakrabarti, 2006), and F. psychrophilum KU060626-59 was isolated from a diseased ayu (Plecoglossus altivelis) in Japan (Fujiwara-Nagata et al., 2012). F. indicum CIP 109464T was grown in AOBE (Bernardet & Kerouault, 1989) for 24 h at 37 °C with shaking. F. psychrophilum KU060626-59, JIP 02/86, and ten other strains used as references were grown in AOBE for 48 h at 18 °C with shaking.

Pulse-field gel electrophoresis (PFGE)

Preparation of DNA in agarose blocks and restriction enzyme digestions were performed according to the protocol described by Liu & Sanderson (1995). Briefly, approximately 108 bacteria cells were embedded in Seaplaque agarose plugs (Lonza, Rockland, MI) and digested with proteinase K (Euromedex) for 10 h at 50 °C. Plugs were then washed in Tris–EDTA buffer with 1X protease inhibitor cocktail, and agarose-plug-embedded DNA were digested with 30 U of I-CeuI or XhoI (New England Biolabs) for 8 h at 37 °C.

PFGE was performed using a CHEF-DR III System (Bio-Rad, Hercules, CA). Digested plugs were run on 1% Seakem agarose (Lonza, Rockland, MI) gels in 1X TBE at 14 °C with an electric field of−1 alternating in two directions at an incident angle of 120° with 1-pulse times ramped from 50 to 70 s over 17 h followed by pulse times ramped from 3 to 12 s over 6 h for the I-CeuI-digested genomic DNA analysis (Fig. 1a) and 2-pulse times ramped from 1 to 3 s over 18 h for the XhoI-digested genomic DNA analysis (Fig. 1b). Gels were stained with 0.5 mg mL−1 ethidium bromide for 30 min before digital analysis (Storm; GE Healthcare).

Figure 1.

(a) PFGE banding patterns of twelve Flavobacterium psychrophilum isolates after chromosomal DNA digestion with the I-CeuI restriction enzyme. Strain KU060626-59 (lane 11) displays the typical DNA degradation phenotype (a smear pattern). Lanes 1, 8, and 15 : DNA of Salmonella Braenderup cleaved by XbaI was used as molecular size marker. Numbers on the right indicate size marker in kbp. (b) PFGE patterns of XhoI-digested genomic DNA displaying the typical degradation phenotype (left part) and same DNA samples run with thiourea (right part). Lanes 2 and 6, F. psychrophilum JIP 02/86; lanes 3 and 7, F. indicum GPSTA100-9T; lanes 4 and 8, F. psychrophilum KU060626-59. Lanes 1 and 5, mid-range PFG molecular marker II (New England Biolabs).

Test of Dnd phenotype

To assess the Dnd phenotype, methods described by Ray et al. (1992, 1995) were used: that is, gel electrophoresis performed with Tris buffer and the use of 100 μM of thiourea to reverse the DNA degradation phenotype. The DNA of F. psychrophilum JIP 02/86 (Dnd and devoid of dnd genes) was used as a negative control.

RT-PCR analysis of the dnd genes transcripts

The total RNA of F. indicum CIP 109464T was obtained from 6 mL of early-stationary-phase cultures. RNA was isolated using a guanidinium thiocyanate–phenol–chloroform extraction (TRIzol; Invitrogen) and treated with DNase I (RNase-free; Ambion). PCRs using 16S rRNA gene-specific primers were performed to determine whether RNA was free of contaminant DNA. Reverse transcription was performed using Superscript II Reverse Transcriptase (Invitrogen) and PCR with GoTaq DNA polymerase (Promega). Ten nanograms of RNA was used in each reaction. Primers used are listed in Supporting Information, Table S1.

DNA sequencing and sequence analysis

The genome sequencing of F. psychrophilum KU060626-59 was performed using a Solexa (Illumina, Inc., CA) single-end strategy. Reads were cleaned by adaptive trimming (home-made tool available at: and assembled using velvet (Zerbino & Birney, 2008). ORFs predictions and genome annotation were performed using the agmial annotation platform (Bryson et al., 2006) as previously described (Touchon et al., 2011). The de novo assembly resulted in 55-fold coverage of a 2 588 113-bp draft genome encompassed in 254 contigs. The annotated sequence of F. psychrophilum KU060626-59 dnd gene cluster is available in GenBank under the accession number KF241852.

Identification of dnd gene clusters in Bacteroidetes and phylogenetic analysis

Flavobacterium indicum Dnd proteins were initially identified by blastp against the dedicated dnd database (Ou et al., 2009). Bacteroidetes Dnd proteins were then identified by blastp (with e-value < 1 × 10−10) using Dnd proteins of F. indicum as a query against the ‘nonredundant’ database (April 2013) restricted to members of the phylum Bacteroidetes. Loci containing at least two of the five Dnd-protein-encoding genes confined in less than ten genes from each other were selected. Proteins homologies were confirmed by global alignments with at least 50% of similarity in amino acid sequence and < 10% of difference in protein length.

The molecular phylogeny of Dnd proteins has been explored by the construction of multiple sequence alignments with muscle (version 3.6) (Edgar, 2004) and filtered with Gblocks (version 0.91b) (Castresana, 2000). The phylogenetic tree was reconstructed from the concatenated alignments of DndC and DndD proteins using the maximum-likelihood method implemented in the phyml program (version 3.0) (Guindon & Gascuel, 2003) with the WAG matrix and a gamma correction for variable evolutionary rates. The reference tree was reconstructed from the 16S rRNA gene sequences of species of the phylum Bacteroidetes carrying a dnd gene cluster. The phylogenetic tree was built using PhyML (version 3.0) under the GTR model with gamma correction. The robustness of the tree topologies was assessed with 100 bootstraps.


Identification of functional dnd gene clusters in Flavobacterium species

During genomic DNA preparations, we noticed that two Flavobacterium strains displayed a Dnd phenotype, which results in a smear pattern, following gel electrophoresis. It was the case of F. psychrophilum KU060626-59 after DNA restriction by I-CeuI, the only one strain displaying a DNA smear pattern of the 12 F. psychrophilum stains tested (Fig. 1a). This phenotype was also observed for F. indicum CIP 109464T. Indeed, F. indicum CIP 109464T and F. psychrophilum KU060626-59 after DNA restriction by XhoI and following gel electrophoresis display a degradation phenotype (Fig. 1b – left part). This phenotype was maintained even with formaldehyde fixation (not shown), indicating that DNA degradation was not caused by extracellular DNase (Soto et al., 2008). However, the observed degradation phenotype was completely abolished when the same DNA samples were run in the presence of thiourea (Fig. 1b – right part), as reported with PT DNA modification (Ray et al., 1992, 1995). Taken together, these results indicate that F.indicum CIP 109464T and F. psychrophilum KU060626-59 harbor a typical DNA degradation phenotype triggered by PT DNA modification.

During annotation of the F. indicum CIP 109464T complete genome (Barbier et al., 2012), we identified a gene cluster encoding proteins similar to Dnd proteins previously shown to be responsible of the Dnd phenotype (Zhou et al., 2005). This gene cluster, encompassed in a 9450-bp region, contains five genes encoding proteins homologous to DndABCDE (Table S2) and two additional genes of unknown function (Fig. S1). Strikingly, when comparing this locus with previously reported dnd clusters (Ou et al., 2009), despite protein sequence similarity, gene order is not conserved. According to Alien Hunter prediction (Vernikos & Parkhill, 2006), this locus lies within a typical genomic island of 26 kbp inserted at a tRNA-Tyr gene suggesting a horizontally acquired origin.

To confirm our hypothesis that the degradation phenotype observed in F. psychrophilum KU060626-59 was triggered by a dnd locus, we performed the whole-genome shotgun of this strain. As expected, we identified on the draft genome a dnd gene cluster. This gene cluster contains dndCDE homologous genes and three additional genes of unknown function, different from those found in the F. indicum genome (Fig. S1). This gene cluster is encompassed in a 7732-bp contig with no sequence homology to the genome of F. psychrophilum JIP 02/86 (Duchaud et al., 2007), suggesting its presence within a genomic island.

Transcriptional analysis of F. indicum dnd genes

Sequence analysis of the 9450-bp region of F. indicum CIP 109464T containing the five dndA-E homologous genes and two additional genes of unknown function (KQS_08915 and KQS_008925) revealed that dndA and the other genes are divergently transcribed, the latter in a likely operonic structure (Figs 2a and S1).

Figure 2.

Organization of the Flavobacterium indicum CIP 109464T dnd gene cluster and RT-PCR analysis. dnd gene transcripts were reverse-transcribed and amplified. The two genes of unknown function KQS_08915 and KQS_08925 are labeled X and Y, respectively. (a) Relative positions and directions of the corresponding primers are marked with blacks arrows. A putative Flavobacterium promoter (Chen et al., 2010) was identified 124-bp upstream the start codon of dndC. The DNA sequences between adjacent genes are indicated at the top, with start codon and stop codon underlined. (b) Amplification products with sense primer (FW), antisense primer (RV), and their corresponding lengths. Intra-dnd gene amplification products are indicated as dnd gene names, while products of regions between dnd genes are named linking two corresponding genes such as CD. (c) Electrophoresis analysis of RT-PCR products. The amplification products are labeled as described above. Reactions omitting the reverse transcription step (-RT) and reactions without DNase treatment (*) were included in each run as negative and positive controls. DNA markers are labeled ‘M’ (10-kb DNA Ladder Mix, Fermentas).

To evidence divergent transcription of dndA and the hypothetical dndCDEi, KQS_08915, B, KQS_008925 operonic structure in F. indicum CIP 109464T, we performed a transcriptional analysis by RT-PCR using different sets of primers (Table S1). Our RT-PCR results (Fig. 2c) suggest that dndCDEi, KQS_08915, B, KQS_008925 are cotranscribed as a single operon in F. indicum CIP 109464T. The absence of DNA amplicon using primers Y1 and A2 (Fig. 2c – lane YA) confirms an independent transcription of dndA compared with the rest of the cluster, as previously reported in the dnd gene cluster of Streptomyces lividans (Zhou et al., 2005; Xu et al., 2009).

Identification of dnd clusters in the family Flavobacteriaceae

The presence of dnd genes has already been reported in the genome of Microscilla marina ATCC 23134 (Xu et al., 2010), another member of the phylum Bacteroidetes. However, such dnd gene clusters have never been reported within flavobacteria so far. Using a dedicated homologs identification strategy, we identified the presence of dnd gene clusters in three other members of flavobacteria. We also observed dnd clusters within the genomes of three members of the family Prevotellaceae and nine other members of the phylum Bacteroidetes (Table 1 and Fig. S1). Analysis of the GC% of the dnd loci and their dinucleotide distribution bias differing from the chromosomal backbone (Table 1) suggests their laterally acquired origin.

Table 1. Identification of dnd genes homologs in members of the phylum Bacteroidetes
Number of dnd genes detectedOrganism source dndA dndB dndC dndD dndE GC% dnd locus [GC% whole genome]δ* × 103aGenome fragments with lower δ* (%)b
  1. GenBank accession numbers of Dnd homologs identified within species of the phylum Bacteroidetes by blastp and confirmed by global alignment.

  2. a

    Dinucleotide bias analysis was adapted from the method proposed by Karlin (2001). The value δ* denotes the dinucleotide relative abundance difference between the dnd locus and the complete genome. The δ* value was calculated with the δρ-Web program ( (van Passel et al., 2005). The high δ* values indicate a likely heterologous origin.

  3. b

    The percentage distribution of δ* is plotted using the δρ-Web tool with random host genomic fragments of equal length as input sequences (van Passel et al., 2005).

  4. c

    Extended version of the previously described DndE.

  5. d

    Detected above the initial cut-off, these proteins contain the DndB-conserved DGQHR domain. Schematic representations of these dnd gene clusters are shown in Fig. S1.

  6. e

    Identified by tblastn between genomic positions 1316314 and 1316682 (ACPR00000000.1) with amino acids similarity > 70% with the amino-terminal part (1–130) of DndEi (YP_005357774.1).

3Flavobacterium psychrophilum KU060626-59   AGR55439.1 AGR55440.1 AGR55441.1 28.58 [32.5]83.23493.956
5Flavobacterium indicum CIP 109464T YP_005357778.1 YP_005357776.1 YP_005357772.1 YP_005357773.1 YP_005357774.1 c 29.30 [31.4]94.92795.413
4Flavobacteriaceae bacterium 3519-10  YP_003095738.1 d YP_003095736.1 YP_003095737.1 Non detected originally e31.76 [42.7]114.54799.76
4Kordia algicida OT1  WP_007096424.1 d WP_007096428.1 WP_007096427.1 WP_007096426.1 c 28.97 [34.2]161.48499.192
3Riemerella anatipestifer DSM 15868   YP_004045227.1 YP_004045228.1 YP_004045229.1 c 30.27 [35]154.94899.767
4Paraprevotella xylaniphila YIT 11841  WP_008630087.1 WP_008630105.1 WP_008630103.1 WP_008630100.1 37.48 [48.5]142.562100
4Prevotella amnii CRIS 21A-A  WP_008450453.1 WP_008450502.1 WP_008450432.1 WP_008450280.1 c 34.61 [36.4]89.94293.416
4Prevotella bivia JCIHMP010  WP_004335941.1 WP_004335930.1 WP_004335934.1 WP_004335936.1 c 36.23 [39.7]78.25497.706
3Haliscomenobacter hydrossis DSM1100   YP_004448668.1 YP_004448667.1 YP_004448662.1 c 37.50 [47.1]96.66797.814
4Belliella baltica DSM 15883 YP_006406347.1   YP_006406357.1 YP_006406356.1 YP_006406355.1 34.20 [36.8]38.05770.358
5Cyclobacteriaceae bacterium AK24 WP_010854647.1 WP_010854648.1 WP_010854651.1 WP_010854650.1 WP_010854649.1 c 38.11 [46.8]62.26095.982
4Microscilla marina ATCC 23134 WP_002704309.1   WP_002704312.1 WP_002704313.1 WP_002704315.1 37.01 [40.6]78.98793.707
4Microscilla marina ATCC 23134 WP_004155710.1   WP_004155718.1 WP_004155719.1 WP_004155721.1 38.58 [40.6]76.52498.199
3Parabacteroides goldsteinii CL02T12C30   WP_007659347.1 WP_007659346.1 WP_007659344.1 35.76 [43.3]85.26695.249
4Bacteroides sp. 2_1_33B  WP_008772058.1 d WP_008772054.1 WP_008772055.1 WP_008772056.1 c 34.18 [44.5]97.34797.304
3Bacteroides xylanisolvens XB1A   YP_007793114.1 YP_007793115.1 YP_007793116.1 36.33 [40.7]90.25396.561
3Bacteroides finegoldii CL09T03C10   WP_007766197.1 WP_007766203.1 WP_007766204.1 34.61 [42.3]90.13395.83
3Bacteroides faecis MAJ27   WP_010537098.1 WP_010537099.1 WP_010537100.1 c 34.63 [42.4]77.88693.262

Incomplete (i.e. fewer than three Dnd-encoding genes) and probably not functional dnd clusters have been also found in Cecembia lonarensis LW9 (DndA : WP_009184828.1 and DndC : WP_009184832.1) and Capnocytophaga canimorsus Cc5 (DndC : YP_004740273.1 and DndE : YP_004740276.1) genomes. These dnd gene clusters were not further included in our analysis.

An unforeseen predicted hybrid DndE protein in members of the phylum Bacteroidetes

DndE was originally described as a small putative phosphoribosylaminoimidazole carboxylase of 126 residues homologous to NCAIR synthetases (Zhou et al., 2005). Despite its essential role in the PT DNA modification, the biochemical function of DndE has not been demonstrated so far. Recent structural studies of DndE from Escherichia coli and Salmonella enterica indicate that it might be a nicked dsDNA-binding protein (Chen et al., 2011; Hu et al., 2012).

The predicted length of the DndE homologs identified to date never exceeds 141 residues (Ou et al., 2009), but strikingly, the DndE protein from F. indicum CIP 109464T (hereafter named DndEi) is predicted to be a 520-residue protein. Therefore, DndEi corresponds to an extended version of the previously described DndE (i.e. about four times longer). Indeed, the amino-terminal part of DndEi (1–128) shows 50% similarity with DndE from Clostridium perfringens NCTC 8239 (Table S2) corresponding to the domain PF08870 in the Pfam database (Punta et al., 2012), while its carboxy-terminal (187–485) part contains an AAA+ ATPase superfamily domain found in many proteins with ATPase activity involved in a wide range of cellular processes (Iyer et al., 2004). This domain, PF12846 in the Pfam database, contains a P-loop NTPase region, which exhibits the conserved nucleotide phosphate-binding motif (194−GxxxxGKT−201, where x is any residue) and a (412−DEAH−416) helicase motif (Fig. S2).

We found such dndEi, included in a dnd gene cluster, on two other genomes of the flavobacteria: Kordia algicida OT1 and Riemerella anatipestifer DSM 15868 (Table 1 and Fig. S1). DndEi homologous proteins (more than 60% of similarity in amino acid sequence and < 10% of difference in protein length with DndEi) were also detected from the genomes of seven other species of the phylum Bacteroidetes (Table 1). All these DndEi homologs exhibit conserved residues predicting AAA+ ATPase domain in their carboxy-terminal part (Fig. S2). Moreover, remote dndEi-encoding homologs (above our homology threshold criteria) or dndEi homologs not included in a dnd gene cluster could also be found in other Bacteroidetes (e.g. Psychroflexus torquis, Capnocytophaga canimorsus Cc5, Leeuwenhoekiella blandensis MED217, Saprospira grandis DSM 2844, and the unidentified eubacterium SCB49) as well as other bacteria outside the phylum Bacteroidetes (e.g. Desulfovibrio africanus PCS, Methylobacterium mesophilicum SR1.6/6, and Xanthomonas axonopodis pv. citrumelo F1) suggesting a widespread distribution.


Identification of functional dnd genes clusters in two Flavobacterium species

We identified functional dnd gene clusters in the genome of two bacterial strains belonging to two different Flavobacterium species. In contrast with all previously described dnd loci, the typical gene order was not conserved in the F. indicum genome, and a hybrid DndEi-encoding gene was identified. DndEi may result from the fusion between a typical DndE and an AAA+ ATPase domain and possess a ‘nicked dsDNA-binding activity’ and a NTPase activity of unknown biological role. F. psychrophilum KU060626-59 contains only the dndCDE gene homologs, suggesting that the minimal functional dnd gene cluster is limited to dndCDE. It has been recently shown that IscS, another cysteine sulfurtransferase, could complement for DndA protein in Escherichia coli (An et al., 2012) to supply this PT modification of DNA, while disruption of dndB does not abolish the Dnd phenotype (Liang et al., 2007; Xu et al., 2009). In genome of F. psychrophilum KU060626-59, we identified a gene encoding an IscS homologous protein. One can speculate that this gene could substitute dndA.

Distribution of dnd clusters in members of the phylum Bacteroidetes

All previously reported dnd gene clusters show a conserved genetic organization with the dndBCDE genes invariably oriented in the same order and direction (Ou et al., 2009). The variety of genomic organization and gene composition of dnd loci within members of the phylum Bacteroidetes is obvious, and these loci seem therefore particularly prone to gene rearrangements in this phylum. Based on their gene composition and gene order, one might conclude that at least two main types of dnd gene clusters coexist within members of the phylum Bacteroidetes. One is found in the genomes of Bacteroides xylanisolvans XB1A, Bacteroides finegoldii CL09T03C10, Paraprevotella xylaniphila YIT 11841, Flavobacteriaceae bacterium 3519-10, and F. psychrophilum KU060626-59, while the other one occurs in the genomes of K. algicida OT-1, R. anatipestifer DSM 15868, Bacteroides sp. 2_1_33B, P. amnii CRIS 21A-A, P. bivia JCVIHMP010, H. hydrossis DSM 1100, P. goldsteinii CLT02T12C30, and B. faecis MAJ27.

The presence of dnd gene clusters seems to occur at low frequency in bacteria. They have never been reported so far among Flavobacterium species or other members of flavobacteria. In this study, dnd gene clusters were only identified into one Flavobacterium genome (i.e. F. indicum) of the 12 publicly available to date. Among the 28 draft genomes of F. psychrophilum strains from many worldwide geographic origins and different host fish (Duchaud et al., 2007 and E. Duchaud, unpublished data), only strain KU060626-59 contained a dnd gene cluster. As such a cluster has been found in only five among the 204 sequenced genomes in the family, the presence of a complete dnd gene cluster in flavobacteria is a rare event. Moreover, the phylogenetic tree of concatenated DndC and DndD protein sequences reported here and the phylogenetic tree constructed on 16S rRNA gene sequences are obviously noncongruent (Fig. S3). Together with the GC% and the dinucleotide distribution bias (Table 1), this confirms that dnd gene clusters are the result of horizontal genetic transfer events.

Although the process of evolution and dissemination of dnd gene clusters across different bacterial species is unknown so far, plasmids have been proposed to play a major role in the dissemination of these clusters. In particular, large plasmids have been suggested to serve as ‘natural depository’ for dnd loci that could be probably sourced from diverse bacterial donors (He et al., 2007). Indeed, dnd gene clusters have been never found on complete phage genomes so far. The only defined dnd island shown to be functionally mobile is the S. lividans SLG island and is known to function as a typical, self-circularizing, site-specific integrative element (He et al., 2007). Using probabilistic model (HMM profiles) (Eddy, 1996) searches across the NCBI plasmid database (that contains 3867 complete plasmid genomes), we detected remote DndC-, DndD- and DndEi-encoding homologs (YP_002967204.1, YP_002967205.1, and YP_002967206.1, respectively) on the Methylobacterium extorquens AM1 megaplasmid (Vuilleumier et al., 2009). The presence of this remote homologous gene cluster on a megaplasmid suggests that large plasmids could indeed serve as ‘natural depository’ and/or vectors for the spreading across bacterial phyla of dnd gene clusters, including the unusual cluster identified in this study.

In addition, the high degree of conservation of gene organization in all previously reported dnd gene clusters may suggest that these elements evolved from a common ancient ancestor (He et al., 2007; Ou et al., 2009). Our study reveals a contrasted situation: the variety of genomic organization and gene composition of dnd loci within members of the phylum Bacteroidetes is obvious and differs from those already described. However, the frequent absence of dnd islands in members of the same species and the presence of two distinct dnd loci within a genome, for instance in M. marina (Table 1 and Fig S1), confirmed that the diverse dnd clusters islands had been acquired independently on many occasions (Ou et al., 2009). As two main types of dnd gene clusters coexist within members of the phylum Bacteroidetes, one might suggest at least two independent ways of acquisition possibly through distinct horizontal genetic transfer events where large plasmids likely play an important role.


This work was supported in part by Grant ERA-NET EMIDA PathoFish. P.B. is a Université Evry Val d'Essonne Ph.D. fellowship. We are grateful to Christophe Habib and Tatiana Rochat for their useful advices. We thank the inra migale bioinformatics platform ( for providing computational resources. R.A.-H. acknowledges Grants FONDECYT 1110219 and the CONICYT/FONDAP/15110027 from the Comisión Nacional de Investigación Científica y Tecnológica (CONICYT, Chile).