Metal-resistant Cupriavidus strains

Strain CH34; from Alcaligenes eutrophus to Cupriavidus metallidurans

Soon after the isolation of strain CH34, from metallurgical sediments in Belgium (Mergeay et al. 1978, 1985), attention focused on its resistance to multiple heavy metals, especially cadmium, zinc, cobalt, and nickel, for which resistance determinants were unknown, and for the presence of two large plasmids named pMOL28 and pMOL30 (Mergeay et al. 2009; Monchy et al. 2007; Nies et al. 1987, 1989; Taghavi et al. 1997). Further research focused on unraveling the determinants of metal resistance, on its genetics, especially since CH34 was a good recipient of foreign genes, on biotechnological- and ecological-applications of these determinants (Diels and Mergeay 1990; Mergeay 1991, 2000; Mergeay et al. 2003), even on the frequently changing taxonomy (Brim et al. 1999; Goris et al. 2001; Sato et al. 2006; Vandamme and Coenye 2004; Yabuuchi et al. 1995), and finally on genome sequencing.

Genomic approaches in the Cupriavidus/Ralstonia genus

To date, six genomes are fully sequenced in the closely related Cupriavidus/Ralstonia genera: the plant pathogen R. solanacearum GMI1000 (Salanoubat et al. 2002), the β-rhizobium legume symbiont C. taiwanensis LMG19424 (Amadou et al. 2008), the facultative chemolithotroph C. eutrophus H16 (Pohlmann et al. 2006; Schwartz et al. 2003), the catabolic strain C. pinatubonensis JMP134 (GenBank Accession CP000090-CP000094), the opportunistic pathogen R. pickettii 12J (GenBank Accession CP0001068-CP0001070), and the multiple metal resistant strain C. metallidurans CH34 (Mergeay 2000; Genbank Accession CP000352-CP000355). At least three other strains of R. solanacearum are being sequenced as well as another strain of R. pickettii (strain 12D). There is an updated annotation of the different replicons of C. metallidurans CH34 available on GenoScope’s MaGe system (Vallenet et al. 2006; www.genoscope.cns.fr/agc/mage/wwwpkgdb/Login/log.php?pid=33).

All of these strains typically carry two chromosomes; the larger (3–4 Mb) shelters most of the housekeeping genes, while the second (2–3 Mb) seemingly has plasmid-like traits for its replication; hence sometimes it is called a megaplasmid. This designation might, however, be confusing since other native plasmids can be very large, and the second chromosome generally is considered to be indispensible since it carries essential genes. This two-chromosome arrangement is also characteristic of the Burkholderia genus, including many strains with (multiple) large plasmids and in some up to nine coexisting replicons (Fricke et al. 2009). The native plasmids have often specific traits that are linked to the very different ecological niches occupied by these closely related bacteria. C. eutrophus H16 has a large plasmid pHG1 carrying genes for chemolithotrophy and anaerobic growth on nitrates (Schwartz et al. 2003), plasmid pRALTA in C. taiwanensis LMG19424 is involved in legume symbiosis and nitrogen fixation, while plasmids pMOL28 and pMOL30 in C. metallidurans CH34 carry many genetic determinants for high resistance to heavy metals (Mergeay et al. 2009; Monchy et al. 2007).

Mobile genetic elements and niche specialization of Cupriavidus/Ralstonia strains

Mobile genetic elements such as composite, unit, and conjugative transposons, plasmids, and genomic islands often carry key determinants ensuring survival in environmental niches, such as hydrogenotrophy, degradation of chloroaromatic compounds, heavy metal resistance, or nitrogen fixation. In silico comparisons of the Cupriavidus/Ralstonia genomes can shed light on the role of mobile genetic elements in adaptation to specific environmental niches, such as symbiotic association with plants, chemolithotrophy, plant pathogenesis, or the colonization of industrial sites. Furthermore, we can assess the evolution and dissemination of different mobile genetic elements in the genera Cupriavidus/Ralstonia and beyond. Our goal was to make an inventory of the genomic islands and associated mobile genetic elements in C. metallidurans CH34 using the following tools: the MAGE GenoScope platform (Vallenet et al. 2006) to evaluate the synteny between the analyzed genomes and other genomes or replicons; the ACLAME database for mobile genetic elements (http://aclame.ulb.ac.be/; Leplae et al. 2004, 2006); the IS Finder database for IS elements (http://www-is.biotoul.fr/; Siguier et al. 2006); and IslandPath (http://www.pathogenomics.sfu.ca/islandpath/). The latter graphically presents the genome of an organism, highlighting features such as GC content, dinucleotide bias, proximity to tRNA or other structural RNA genes. Also, it denotes the presence of mobility genes (transposases and integrases) to identify regions likely to be acquired through horizontal gene transfer, since these features commonly are associated with genomic islands (Hsiao et al. 2003). This inventory can confer a better understanding of the role of genomic islands in bacteria, and how bacteria recruit advantageous genes. In addition, it can illuminate the origin and dispersion of these genes, and the evolutionary forces shaping this bacterium.

Defining genomic islands

Genomic islands were mainly looked for in regions with extensive lack of synteny and in prokaryotic genomes these islands are commonly defined as clusters of genes with one or a set of following properties:

  • a tyrosine-based site-specific recombinase gene and an adjacent tRNA gene at one extremity

  • flanking insertion sequence elements

  • a base composition and/or phylogeny differing from the bulk of the genome, indicating a foreign origin and, hence, acquisition through horizontal transfer

  • a higher content of hypothetical genes than neighboring regions

  • clustering of genes characteristic of mobile genetic elements: recombinase genes, IS elements

  • conservation of the genomic island between different (unrelated) hosts

  • a high concentration of genes specialized for resistance, catabolism, unusual metabolisms

Most often, however, as for most pathogenicity islands (reviewed by Hacker et al. 2004), intercellular mobility has not been demonstrated experimentally. Under this broad definition, the term genomic island covers a wide spectrum of integrated mobile genetic elements that are not well-defined. They therefore may be defective, or have lost their determinants of intercellular mobility, but are inferred to be or to have been mobile (or mobilized in trans by a proficient mobile genetic element; E.M. Top and A. Toussaint, personal communication). Thus, the term genomic island is broader than the integrative and conjugative elements that are transferred through conjugation, integrate into and replicate with the host genome. In addition, the term genomic island does not necessarily include structures defined as regions of genomic plasticity that do not reflect their evolutionary origin or genetic basis (Mathee et al. 2008).

Islands in chromosome 1

CMGI-1, a genomic island identical to PAGI-2C of P. aeruginosa clone C

The first observation attracting attention to the genomic islands in C. metallidurans CH34 was that this strain, as well as some related strains had a large genomic island comparable to the one of P. aeruginosa clone C strains isolated from cystic fibrosis patients (Klockgether et al. 2006; Larbig et al. 2002). This 109 kb PAGI-2C island that begins with an int gene (tyrosine-based site-specific recombinase) is fully conserved in C. metallidurans CH34 (with only a few base pairs different; Table 1). The CH34 version, adjacent to a tRNA-Gly gene, contains an additional endogenic IS element IS1088 (IS30 family), and its int gene is inactivated by transposon Tn6049 (see below) although no spontaneous excision from P. aeruginosa clone C has been demonstrated and most of the PAGI-2C genes are transcriptionally silent in P. aeruginosa clone C (Klockgether et al. 2008). However, in CH34 expression of some genes (cdfX, pbrR2, cadA, pbrC2 see below) has already been demonstrated in the presence of metals and putatively transcription of the int gene might differ in CH34 (and other genetic backgrounds). The island could therefore be prone to excision and insertion of Tn6049 fixed the island in the genome. Parts of the CMGI-1/PAGI-2C genomic island are highly conserved in various β-Proteobacteria (e.g., Burkholderia xenovorans, Delftia acidovorans, and Herminiimonas arsenicooxydans) and in an α-Proteobacterium (Parvibaculum lavamentivorans strain DS-1; Schleheck et al. 2004b, 2007) with protein identities higher than 80%. These regions are even conserved in γ-Proteobacteria with significant identities (>35% identity). Synteny reveals that blocs of 10–20 genes, mostly encoding hypothetical proteins, are highly conserved (throughout α-, β-, and even γ-Proteobacteria) and separated by identifiable functions, such as the metal-resistance or cytochrome biosynthesis genes, that are present in CMGI-1 but lacking in other bacteria. Nevertheless, the presence of an int gene (tyrosine-based site-specific recombinase) near a tRNA gene still provides the hallmark of a genomic island. Figure 1 shows an example for CH34 versus R. pickettii 12D, H. arsenicooxydans HEAR, B. xenovorans LB400, and P. lavamentivorans DS-1 indicating that the CMGI-1/PAGI-2C family of islands consists of conserved (with many hypothetical genes up to now refractory to annotation) and variable modules with various putative functions. One conserved module at the end of the genomic island (opposite the int gene) seemingly contained replicon maintenance genes such as soj (plasmid partitioning), ssb (single strand DNA binding), and topB (DNA topoisomerase).

Table 1 Characteristics of the genomic islands on chromosome 1 of C. metallidurans CH34
Full size table
Fig. 1
figure 1

The 109 kb genomic island CMGI-1 on chromosome 1 of strain CH34 and synteny with R. picketti 12D, H. arsenicooxydans HEAR, B. xenovorans LB400 and P. lavamentivorans DS-1 (HM: genes involved in heavy metal resistance). Only conserved gene clusters or synteny groups are shown and highlighted in different shades of gray. These groups are defined as clusters of orthologous genes (with a minimum threshold of 35% protein sequence identity on 80% of the length of the smallest protein)

Full size image

Importantly, nosocomial infections and environmental strains from industrial sites share in common this large piece of mobile genetic material, which suggests a recent horizontal gene transfer, and also implies that the barriers between (industrial) environmental reservoirs and clinical environments can be crossed. Acquiring genes by horizontal transfer may thus be enhanced significantly by potential co-inhabitation as well as by genetic relatedness. Mathee et al. (2008) showed that 42% of the open reading frames (ORFs) in the core genome of P. aeruginosa had best hits outside the genus Pseudomonas and these hits belonged to the genus Azotobacter, a closely related γ-Proteobacteria, and Burkholderia and Cupriavidus, two β-Proteobacteria. Strains of the β-Proteobacterial genera Cupriavidus, Ralstonia, and Burkholderia have been identified in cystic fibrosis patients. Similarly, P. aeruginosa and C. metallidurans may coexist in industrial sites with high levels of heavy metals.

Metal-response and resistance genes in CMGI-1 and PAGI-2C

CMGI-1 contains metal-resistance genes supporting the recognizable functions carried by this genomic island. However, the particular selective pressures involved in acquiring the island (in Cupriavidus or in Pseudomonas) are unknown, as are its selective advantage. Figure 2 shows the gene cdfX (Rmet_2299) encoding for a putative metal-efflux system that is distantly related to cation diffusion factors, such as CzcD (Anton et al. 1999) and CnrT, and the pbrR2 cadA pbrB2 genes (Rmet_2302–2304). The latter resemble three genes in the lead resistance operon, pbr, of pMOL30 (Borremans et al. 2001), which, respectively encode for a MerR family regulator (Chen et al. 2005, 2007), a metal-efflux ATPase (for lead, zinc, cadmium; Monchy et al. 2006) and a lipoprotein/peptidase. Microarray data (Gene Expression Omnibus accession number GSE7272) revealed the overexpression of cdfX, pbrR2, cadA, and pbrC2 after metallic challenge, implying a synergy between CdfX and CadA for metal detoxification from the cytoplasm to the periplasm, although little is known about CdfX. Other accessory CMGI-1 genes include mercury resistance genes (the mercury reductase is likely not operational), efflux ATPases, and the ccm and coo genes involved in the biosynthesis of major cytochromes.

Fig. 2
figure 2

Local synteny between CMGI-1 of CH34 and B. xenovorans LB400. For LB400 the structure consists of a four-gene based transposon with eight copies in the genome of strain B. xenovorans LB400. Orthologous genes are connected with a line (with a minimum threshold of 35% protein sequence identity on 80% of the length of the smallest protein). Genes that are shown encode for a cation diffusion factor (cdfX), a MerR family regulator (pbrR2), an ATPase (cadA), a lipoprotein/peptidase (pbrC2), and a transposase (tnpA)

Full size image

New transposons carrying metal response genes found in CMGI-1/PAGI-2C

Syntenic analysis of various β- and γ-Proteobacteria identified the presence of a locus similar to that found in CMGI-1/PAGI-2C but in which cdfX replaced cadA forming pbrR2 cdfX pbrC2. Next, to this conserved gene arrangement a transposase (ISL3 family) was located and this complete four-gene set was identified in Pseudomonas putida plasmids as ISPpu12 (Li et al. 2004; Weightman et al. 2002; Williams et al. 2002) and in B. xenovorans LB400 (Chain et al. 2006). The latter contained eight copies of this transposon, registered here as Tn6052, with one disrupted by Tn6053, a Tn6049-equivalent in B. xenovorans LB400 (see below). In B. xenovorans LB400 this transposon apparently neighbors catabolic functions. Furthermore, two slightly modified copies were identified in Idiomarina loihiensis L2TR (a marine bacterium found in a deep sea hydrothermal vent near Hawaii; Hou et al. 2004) and two almost identical copies were identified in Acinetobacter baumannii AYE. The phenotype associated with this mobile genetic element is undefined, and at least for Pseudomonas putida mt-2 the wild-type strain carrying ISPpu12 on plasmid pWW0 and plasmid-free derivatives (no ISPpu12) showed no differences in resistance to Hg, Zn, Cd, Cu, and Pb (P.A. Williams, personal communication). However, this could perhaps point to the origin of the element since many β-Proteobacterial determinants of heavy metal resistance are not or only slightly expressed in γ-Proteobacteria.

Finally, this listing of transposons clearly extends the set of metal-response genes associated to transposons, now supposedly mainly populated by mercury- and arsenic resistance genes. This example adequately illustrates the power of synteny for detecting new mobile genetic elements outside the bias of a selectable marker, similar to other examples described below (see CMGI-4).

CMGI-2 and CMGI-3: genomic islands of the Tn4371 family: a major contribution to strain CH34, chemolithotrophy and catabolism of aromatic compounds

Cupriavidus metallidurans CH34 contains islands belonging to the Tn4371 family. The first identified member of this family, Tn4371 from the strain C. oxalaticus A5 (formerly R. eutropha A5), was described as a catabolic transposon carrying genes that degraded biphenyl- and chlorobiphenyl-compounds (Merlin et al. 1997, 1999). Soon similar transposons were found in other bacteria including R. solanacearum (but not in Cupriavidus strains H16 or JMP134; Toussaint et al. 2003).

Analysis showed that Tn4371-like elements comprised a large family with a quadripartite structure corresponding to an integrated conjugative plasmid with basic genes (parAB, repA, traF, virD2, traR, and traG), the transfer genes (trb), accessory genes systematically present between virD2 and traR, and a generally small int module at the parB side. Previous descriptions of the Tn4371 family considered the int gene as part of the parB-virD2 module but synteny studies indicated there were variable, poorly conserved int modules within the Tn4371 islands. Undoubtedly, the whole family is defined by a quadripartite structure, consisting of an integrase module, a region with plasmid/phage/genomic island maintenance genes, the accessory genes, and the conjugative genes. These conjugative genes strongly resembled those of the broad-host-range plasmids Ti/IncP1, and while the transfer regions of these genomic islands were conserved, the parAB virD2 sections were much more variable. At the extremity of this sector (with the maintenance genes), lied an int gene that likely directed the integration of the conjugative plasmid and controls further replicon-based movements of the corresponding genomic island. Mobility of the Tn4371 family of genomic islands was illustrated by Tn4371 itself: in C. oxalaticus A5, the biphenyl genomic island could easily move into the broad-host-range IncP1 plasmid RP4, and after transfer into C. metallidurans CH34, could stably integrate into its genome. Annotation of the C. metallidurans CH34 genome revealed the presence of at least two complete genetic islands belonging to the Tn4371 family: CMGI-2 and CMGI-3. In both cases, int-traF-virD2 and trb modules flank the variable sector with accessory genes (Fig. 3).

Fig. 3
figure 3

Quadripartite structure of the Tn4371-family of conjugative transposons, consisting of an integrase module, a region with plasmid/phage/genomic island maintenance genes, the accessory genes, and the conjugative genes. CMGI-2, CMGI-3 and CMGI-4: C. metallidurans CH34 GIs; DAGI-1 and DAGI-2: D. acidovorans SPH-1; BPGI-7: Bordetella petrii DSM12804

Full size image

CMGI-2, primarily a “catabolic transposon”

CMGI-2 contained 25 genes involved in toluene degradation, underlying CH34’s ability to grow using toluene, benzene, or xylene as the sole carbon source (Table 1; Fig. 3). Besides the biphenyl transposon Tn4371 (Toussaint et al. 2003), CMGI-2 is the second catabolic transposon of this family. In addition, other accessory genes were inserted near the catabolic cluster including the hydrogenases involved in chemolithotrophy that were flanked by two copies of ISCme5 (four copies in the complete genome), and a bloc of mostly IS elements or genes encoding integrases/tyrosine-based site-specific recombinases with some integrases organized in “trios” (see section below). Both ISCme5 copies (IS481 family) in CMGI-2 occured in a region with multiple transposases and fragments thereof; probably, they were the driving force integrating the hyp/hox cluster modulating chemolithotrophy in CMGI-2. An extensive inventory and analysis of the IS elements in CH34 is beyond the scope of this study.

CMGI-3 and chemolithotrophy

CMGI-3, the second Tn4371-family genomic island in CH34, shelters accessory genes involved in chemolithotrophy between virD2 and traR (Table 1; Fig. 3). They constitute two adjacent blocs flanked by IS1071: one contained genes for fixing CO2 (cbb genes coding for ribulose phosphate kinase and ribulose biphosphate carboxylase), and the other harbored genes for chemolithotrophy (hyp and hox genes coding for hydrogenases). In fact, all the genes for aerobic chemolithotrophic growth on CO2 and H2 via two kinds of hydrogenases (one cytosolic and one membrane-bound, respectively encoded by the hox and hyp clusters) were located on CMGI-2 and CMGI-3. Together with the numerous genes located on the four replicons involved in the response to heavy metals, they participate in the survival of strain CH34 in mineral niches. The closely related bacteria C. eutrophus H16 also grows chemolithotrophically, but the essential genes are organized differently. Table 2 summarizes these organizational differences between H16 and CH34.

Table 2 Organization of the genes involved in chemolithotrophy
Full size table

All the genes in CMGI-2 and CMGI-3 directly linked to hydrogenases may be recognized in the 452 kb megaplasmid pHG1 in C. eutrophus H16 (Pohlmann et al. 2006; Schwartz et al. 2003), suggesting that strains H16 and CH34 have inherited them from the same source, but have captured them on different vectors: an autonomous plasmid (pHG1) in H16 and genomic islands (CMGI-2 and CMGI-3) in CH34 (Fig. 4). Yet, while the hydrogenotrophy genes are closely related and similarly organized in H16 and CH34, the cbb (CO2 fixation) genes are very different and originated from different sources. In CH34, the cbb genes are phylogenetically much closer to those of photosynthetic- or nitrifying-organisms than to the cbb of C. eutrophus H16 (Pohlmann et al. 2006). In addition, regions encompassing integrases in trios and even a region where remnants of four trios (twelve partially deleted xerD/int genes) were located, are equivalent in pHG1 and CMGI-2 and CMGI-3 (Fig. 4). It should be reminded that C. metallidurans CH34 plasmid pMOL28 has a gene core strongly identical to that of pHG1 and later recruited genes involved in the resistance to heavy metals.

Fig. 4
figure 4

Map of C. eutrophus H16 plasmid pHG1 showing the various regions displaying high identity with parts of C. metallidurans CH34 plasmid pMOL28, and genomic islands CMGI-2 and CMGI-3

Full size image

This picture of gene mobility again illustrates the bacterial versatility in using the “mobile toolbox” to recruit genes, allowing the colonization of specific niches. Probably the hydrogenotrophy genes in CMGI-3 were recruited via IS1071 elements. Vice versa, the capability of strain CH34 for autotrophic growth can be lost by IS1071-mediated deletion. IS1071 (Tn3 family) was also detected in several catabolic strains (Peel and Wyndham 1999; Providenti et al. 2006; Wyndham et al. 1994). In all cases, the sequence of IS1071 was extremely conserved and the dissemination of IS1071 apparently had infectious traits. This analysis suggests that the hydrogenotrophy determinants were recruited recently, as was the colonization by C. metallidurans strains of industrial sites that dates from the modern industrial development. IS1071 also occured in plasmid pMOL28 where it also participated to recruit a genomic island carrying determinants for heavy metal resistance (Mergeay et al. 2009; Monchy et al. 2007).

Chromosomally encoded mobilizing capacity in C. metallidurans CH34: a possible intervention of CMGI-2 or CMGI-3

The presence of transfer genes, exhibiting strong similarity to those of the broad-host-range plasmids Ti/IncP1, on the Tn4371-like elements CMGI-2 and CMGI-3 might explain previous observations of horizontal gene transfer between derivatives of C. metallidurans CH34. They included the transfer of RSF1010 based plasmids (IncQ, Mob+) and pLAFR3 (IncP, Mob+) from CH34 and plasmid-free derivative strains at a frequency of 10−6 or 10−5 transconjugants per recipient to C. metallidurans or E. coli without the presence of a Tra+ helper plasmid, which implies the presence of functionally active transfer functions on the C. metallidurans CH34 chromosome (Taghavi 1996; S. Taghavi and D. van der Lelie, personal communication).

CMGI-4: a large composite genomic island with metal response genes and a full Tn4371-like counterpart in Delftia acidovorans SPH-1:

Analysis of CMGI-4 revealed a puzzling observation. Like CMGI-2 and CMGI-3 it contained the int-parB-traF-virD2 region, but lacks a trb region with plasmid conjugative genes (Table 1; Fig. 3). Yet synteny studies demonstrated that the putative island would extend much further than the virD2 gene, suggesting that accessory genes were present. Indeed, three successive blocks were observed: seven genes including an hmz gene cluster, the transposon Tn6048 (Van der Auwera et al. 2009), and a more mosaic region. The first block includes a putative ferritin reductase, and a hmzBA hmzRS cluster similar to the czc metal-efflux genes of the RND family. Genes encoding for HmzB and HmzA are canonical paralogs, respectively of the membrane fusion protein-encoding czcB and the efflux pump-encoding czcA. The cation pump HmzA was previously allocated to the HME3b family (Nies et al. 2006). However, the czcC equivalent, encoding for the outer membrane protein, is absent and is replaced by two small open reading frames. Next to this small hmz gene cluster lied Tn6048, a twelve-gene transposon with four genes for transposon maintenance and mobility, and eight genes apparently responding to heavy metals since they were induced by zinc and lead according to microarray data. The last block of eleven genes adjacent to Tn6048 include genes for a truncated transposase, a cluster involved in breaking down phosphite, another IS element, and three genes involved in methionine biosynthesis. This cluster has a 100% identical counterpart in Delftia acidovorans strain SPH-1 (Denger et al. 2008; Rosch et al. 2008; Schleheck et al. 2004a). However, strain SPH-1 carries an untruncated transposase (Daci_2679), which is part of a Tn3-related transposon, registered here as Tn6051, together with a resolvase (Daci_2680), and putative transport genes. This element occured almost 100% conserved in Pelobacter propionicus DSM 2379 (3 copies) and Marinobacter aquaeolei VT8 (1 copy). The gene cluster involved in phosphite metabolism (Rmet_2992 to Rmet_2996) is syntenic with a region in Herminiimonas arsenicoxydans (HEAR3211-3215) (Muller et al. 2006, 2007) where this cluster is flanked by two IS elements (IS30 family), thus defining a potential composite transposon. Furthermore, this complete element is strictly conserved in Janthinobacterium sp. Marseille (mma_3039 to mma_3045; Audic et al. 2007). In both species, the neighboring genes apparently modulate heavy metal resistance. Considering precipitation of metal-phosphates, oxidizing phosphite to phosphate could be involved in the resistance to heavy metals. However, microarray data do not support the induction of these genes under metallic challenge. Instead, constitutive expression, or a slight down regulation are observed. At the moment, the participation of CMGI-4 to the metabolism of CH34 is not clear.

In D. acidovorans SPH-1, transposon Tn6051 is close to a complete Tn4371-like island that could be called DAGI-1 (Fig. 3). From hmzA (Daci_2709) to intA (Daci_2745), DAGI-1 is almost identical to CMGI-4 but it also contains a full traG-trb conjugative module like CMGI-2 and CMGI-3 as well as a few accessory genes adjacent to hmzA in particular Daci_2708 that is transcriptionally coupled with hmzA. The loss of this gene in CH34 may explain the weak expression of hmzA in CH34. Therefore, it seems that two events occurred in an undefined order: a transfer of the whole region from Daci_2745 to Daci_2669, and a deletion going from part of Daci_2679 to Daci_2709. This deletion (unspecified in which bacterium it occurred) would have removed some accessory genes, the conjugative module of DAGI-1 and most of Tn6051 in CH34.

Further examination of CMGI-4 and DAGI-1 showed the latter has a full counterpart, CTGI-1, in the β-Proteobacteria Comamonas testosteroni KF-1 (Rosch et al. 2008; Schleheck et al. 2004a). Both D. acidovorans and C. testosteroni were earlier considered to belong to the same genus (Denger et al. 2008; Wen et al. 1999). Furthermore, strains KF-1 and SPH-1 were isolated from the same environment, and their fully sequenced genomes were the first ones to be sequenced in their corresponding genera. The only difference with CTGI-1 is that DAGI-1 contains a four-gene insertion (with two putative integrases) in the parB module that might be a mobile element because it also appears elsewhere in the genome of D. acidovorans SPH-1.

Other Tn4371-family genomic islands carrying metal resistance genes as revealed by syntenic analysis

DAGI-2 in D. acidovorans SPH1

The D. acidovorans SPH1 genome surprisingly contained a new Tn4371-like genomic island with all the accessory genes for metal resistance. Indeed, DAGI-2 carried five clusters of resistance genes, to copper, silver/copper, cadmium/lead, mercury, and arsenic, with cop genes sited between the parAB and the traR-trb module similar to the biphenyl genes in the original Tn4371. The clusters, silABCDR, pbrRcadApbrC, merRTPA, and arsRIC2BC1H were flanked by the parAB cluster and the int gene (Fig. 3). All these genes have counterparts in strain CH34: sil on plasmid pMOL30, pbr and mer in genomic island CMGI-1, and ars in CMGI-7. The assembly of the ten cop genes is remarkable: copK, copDC, copGFO, and copBARS. Some genes are more closely related to the chromosomal cop genes, while others are similar to the plasmid pMOL30 version. DAGI-2 also contains a czcD/dmeF-like gene (Anton et al. 1999; Munkelt et al. 2004) inserted with a strongly conserved companion gene in the parAB module. Several bacteria contain the czcD-like gene (Daci_0464) and its companion gene (Daci_0465). In DAGI-2, this gene pair resides in a module (partition/maintenance module) of the Tn4371 family of genomic islands that normally does not shelter accessory genes (Toussaint et al. 2003). The presence of this czcD-like gene would add a supplementary tool to the diversified arsenal of metal resistance genes carried by DAGI-2 (Fig. 3).

BPGI-7 in the soil bacterium Bordetella petrii DSM12804

Bordetella petrii strain DSM12804 was initially isolated from river sediment (Gross et al. 2008). Unlike other Bordetella species, it is a facultative anaerobe, not known to be associated with humans or other warm-blooded animals. Analysis of its genome (a 5.3 Mb unique replicon) revealed seven acquired genomic islands (Gross et al. 2008). Although one of them, BPGI-4, was similar to Tn4371 (Gross et al. 2008), the genome clearly contains two Tn4371-like genomic islands displaying a complete quadripartite structure (int, par, accessory genes and trb modules). The second Tn4371-like genomic island is actually BPGI-7 (Bpet4544-Bpet4630) and carries ars genes in the int module as in D. acidovorans, as well as cop and czc-like genes in the accessory module inserted between virD2 and traR (Fig. 3).

Other small genomic islands on the chromosome: CMGI-5 to -11

The characteristics of these small islands correspond to the main criteria discussed above: lack of synteny, divergence from the current gene model, location near a tRNA gene, presence of an int gene (tyrosine-based site-specific recombinase), high content of hypothetical genes (often small ones), and presence of IS and/or transposons. Their contribution to the metabolism of C. metallidurans is not apparent, yet each has a distinctive feature of interest in understanding the diversity of genomic islands. Their characteristics are summarized in Table 1 and their location on chromosome 1 is shown in Fig. 5 (as for other islands).

Fig. 5
figure 5

Schematic presentation (not to scale) indicating the location of the genomic islands CMGI-1 to -11 and the transposons Tn6048, Tn6049, and Tn6050. For Tn6049, filled triangles represent inactivation of an int gene

Full size image

CMGI-5: a plasmid remnant?

This island comprising 24 ORFs (from Rmet_2847 to Rmet_2824) begins with an int gene and is flanked at the other extremity by a tRNA gene. It contains copies of IS1086 (Dong et al. 1992; Taghavi et al. 1997) and Tn6049 (Mergeay et al. 2009). Besides the hypothetical genes, some typical plasmid-related genes are present, such as repA, traY, mobA (likely knocked out by Tn6049), and mobB. The int gene is coupled transcriptionally with an exonuclease gene that putatively could exert a resolvase-like function. This coupling of int genes with a second gene was also observed in CMGI-4, pMOL28 (bimAB) and pMOL30 (bimAB), but the second gene is very variable and poorly defined. The repA gene displays some identity with orthologs of various broad-host-range plasmids like pSa (IncW), and mainly pSB102 (Schneiker et al. 2001), pIPO2T, and pMOL98 (all from the PromA family; Mela et al. 2008; Van der Auwera et al. 2009). Other plasmid-related genes of CMGI-5 have orthologs in different plasmids from various origins, pointing to mosaicism. The limited size of this putative integrated plasmid in CMGI-5 suggests that it probably is a remnant of a larger and more complex genomic island.

CMGI-6, a cryptic genomic island sheltering Tn6049

This island (Rmet_1997 to Rmet_2020) is located near a tRNA-Met gene. It contains two copies of Tn6049, genes coding for hypothetical proteins (mostly small sizes), and a truncated gene encoding for a putative major phage capsid protein. The latter is one of the rare C. metallidurans CH34 genes encoding for a bacteriophage structural protein. The terminal int gene is disrupted by one of the Tn6049 elements and putatively immobilized the island (Fig. 5). Finally, next to the island a copy of IS1087 (IS3/IS911 family) (Collard et al. 1993; Tibazarwa et al. 2000) was recognized.

CMGI-7: an arsenic gene island

Features characterizing this small genomic island are lack of synteny, loss of current gene model, presence of resistance genes, presence of an integrase trio, and one Tn6049 inactivating the central gene of this trio; its extremity, formed by Rmet_0333, is not well-defined. Seven genes (from Rmet_0327 to Rmet_0333) are involved in resistance to arsenic (Zhang et al. 2009) and are highly overexpressed by arsenate and bismuth exposure and by zinc and lead. Further, expression of the regulatory gene arsR is up-regulated by Cd, Hg, and Au. The extremity at the int side seems to correspond to a four-gene structure, including a putative trio of recombinases, and is similar to one extremity of CMGI-28b on pMOL28 (further described below). The second CMGI-7 gene is interrupted by a Tn6049, which likely inactivated the mobility as in many other C. metallidurans CH34 mobile genetic elements (Fig. 5).

CMGI-8 is also a cryptic genomic island sheltering IS1087 and Tn6049

This small island (13 genes from Rmet_2549 to Rmet_2561) is defined by the presence of a tRNA-Leu gene at one extremity. Besides the int gene, there is, in the following order, a gene encoding for a putative serine-based recombinase, a copy of IS1087 and a copy of Tn6049. Other genes present are mostly very small and hypothetical. The linkage between an int gene and a tnpR-like (resolvase) gene, also observed in other bacteria such as in P. putida W619, may correspond to a defined mobility pattern.

CMGI-9 shares features with the int module of CMGI-3

This genomic island with 17 ORFs (from Rmet_2156 to Rmet_2172) fulfills most criteria (i.e., location by a tRNA gene, lack of synteny, divergence from the current gene model, int gene at one extremity). The absence of IS elements or copies of Tn6049 suggests that it might still be mobile. Most CMGI-9 genes are hypothetical, except for a four-gene operon that is well conserved through the plasmids of Deinococcus radiodurans, α-, β- and γ-Proteobacteria, as well as in the int module of CMGI-3. This operon may be involved in repairing or degrading DNA, since it contains genes encoding, respectively for, a nucleoside triphosphate hydrolase, and an endodesoxyribonuclease.

CMGI-10

This island with 22 ORFs (Rmet_3347 to Rmet_3368) was identified mainly by divergence from the gene model and lack of synteny. Besides a gene encoding for an ATPase and hypothetical genes, it contains three IS elements (ISCme3, ISCme1 and ISCme4). It does not have an int gene.

CMGI-11: an island flanked by IS and carrying genes for fimbrial biosynthesis

This small element (9 ORFs) contains an operon of five genes encoding proteins involved in fimbriae biosynthesis. Only one similar structure was found in closely related genomes, namely on the 301 kb megaplasmid pBMC401 of Burkholderia ambifaria MC40-6, a soil-borne strain associated with maize roots, where the gene cluster also is flanked by IS elements. However, at least 16 counterparts were uncovered in genomes of γ-Proteobacteria (mostly enterics). Acquiring an extra set of genes involved in the biosynthesis of fimbriae or similar surface components might be relevant to adaptation to harsh soil conditions. The island is flanked by 2 copies of ISCme7 that differ by three base pairs and their transposase genes ended at the stop codon present in the flanking sequences. ISCme7 belongs to the IS6 family. In Firmicutes and γ-Proteobacteria, IS6-related IS appear to be often present as arrays (up to 5 in a 20 kb fragment) flanking captured genes involved in resistance to antibiotics, encoding for toxin/antitoxin, and for pili biosynthesis. These duplicated elements are regularly deleted, probably to fix the captured genes.

Genomic islands of chromosome 2

The structure of the second chromosome in Cupriavidus, Ralstonia, and Burkholderia strains differs markedly from that of the first chromosome. The structure of chromosome 2 is a patchwork, so discarding lack of synteny and divergence from the gene model as significant parameters to define a genomic island. In addition, chromosome 2 contains fewer tRNA genes, only eight compared to 54 in chromosome 1, and very few xerD/int genes encoding for tyrosine-based site-specific recombinases. The relative high number of genes involved in the response to heavy metals compared to chromosome 1 and 2 of related Cupriavidus and Ralstonia species might indicate that their acquisition event occurred long ago. The acclimatization/integration process of these acquired genes might even be completed to the point of eliminating all clear traces of the lateral acquisition. These regions certainly meet the broader terminology of regions of genomic plasticity but care should be taken when assigning them as genomic islands.

Genomic islands in plasmids pMOL28 and pMOL30 and metal resistance genes

Cupriavidus/Ralstonia strains often carry one or more megaplasmids with very specific traits seemingly linked to niche specialization. For strain CH34, both megaplasmids pMOL28 and pMOL30 carry genetic determinants conferring high resistance to a wide diversity of heavy metals (Mergeay et al. 2009; Monchy et al. 2007). The plasmid backbones and the genomic islands on pMOL28 and pMOL30 carrying these determinants were recently detailed (Mergeay et al. 2009; Monchy et al. 2007).

Plasmid pMOL28 carries three genomic islands. Island CMGI-28a containing all the heavy metal resistance genes of pMOL28, is flanked by IS elements (Tn3 family) inactivated either by partial deletion or by insertion of Tn6049. Island CMGI-28b encompasses at one extremity three rhs-like genes of unknown function that code for products rich in tyrosine-aspartate motifs. Finally, CMGI-28c sited adjacent to CMGI-28a, contains mostly hypothetical genes: their proximity suggests that these two might constitute one single genomic island.

In plasmid pMOL30, most genes involved in the response or resistance to heavy metals occur in two large genomic islands, CMGI-30a with the czc, pbr, and mer genes, and CMGI-30b with the sil and cop genes. Partially deleted IS elements or transposons exist at one extremity of both islands. Furthermore, both contain “nested islands” flanked by (partial) tyrosine-based site-specific recombinases, which delimit specific gene clusters.

New transposons in the genome of C. metallidurans CH34

The new transposons reported during this study, both in strain CH34 and other strains, were all registered according to the new revised nomenclature for transposable genetic elements (Roberts et al. 2008). The transposons on the chromosomes of CH34 are shown in Fig. 5.

Tn6048, containing 8 genes highly induced by zinc and lead

This transposon present at three copies (one in chromosome 1 and two in chromosome 2) contains twelve genes. It has a module of four genes involved in mobility, and eight multiple metal phenotype (mmf) genes; although the function of the latter is unknown, they apparently participate in the response to heavy metals as they are strongly induced when zinc and lead are present. The four mobility genes share some similarity with Tn7 genes. Interestingly, thallium- and cesium-resistant clones from a CH34 cosmid library contained one of the Tn6048 copies located on chromosome 2, but this phenotype looks to be linked to genes neighboring Tn6048.

Tn6049: a role in stabilizing genomic islands?

This four-gene-based transposon of 3,517 bp was recently described in plasmids pMOL28 and pMOL30 as TnCme2 (Mergeay et al. 2009). It contains the tnmB, gspA, tnmA, and tnmC genes but the functions of tnmB and tnmC are unknown. The tnmA gene codes for a Tn7-like transposase. While gspA initially was annotated as an equivalent of a major gene encoding for an ATPase acting in a type II-secretion system, additional analyses via the ACLAME database revealed its possible equivalents in mutator phage B3 and suggested that its gene product shares characteristics with the MuB protein (A. Toussaint, personal communication). The gspA gene of Tn6049 therefore was re-annotated as a putative activator of transposition. The genome of strain CH34 has twelve copies of Tn6049. Similar four-gene structure were observed in B. xenovorans LB400 (i.e., two identical copies of Tn6053 with around 60% protein identity with their corresponding Tn6049 orthologs, except for tnmC where the 72 amino acid gene product of the B. xenovorans transposon shares only 24% identity with TnmC of CH34), in B. phymatum STMB15 (one copy), in B. phytofirmans PsJN (6 copies) and in B. petrii (one copy with an insertion in the gspA-equivalent). The identities between these Burkholderia and Bordetella Tn6049-like transposons are much higher (up to 100% for tnmC) than with Tn6049.

Tn6049 appears to play the peculiar role of “stabilizing” genomic islands in strain CH34. Table 3 shows the insertion sites of this transposon. Tn6049 inactivates the terminal int gene of CMGI-1, CMGI-7, and CMGI-6. Tn6049 is present in CMGI-5, lies at the extremity of the putative island CMGI-28a and inserted inside IS1071. The latter is a very active IS element often observed in catabolic plasmids. Again, we hypothesize that the inactivation of IS1071 by Tn6049 helps to stabilize the genomic island rich in genes involved in responding to heavy metals.

Table 3 Insertion sites Tn6049
Full size table

Tn6050 in chromosome 2: peculiar passenger genes

There are two copies of this peculiar transposon in chromosome 2 lying in opposite orientations and separated by 143 kb. The core (tnpA tnpR) is very similar to that of the mercury transposons Tn4378 and Tn4380 located, respectively, on plasmid pMOL28 and pMOL30. As for Tn4378 and Tn4380, Tn6050 is bound by 38 bp inverted repeats (IR) and clearly belongs to the Tn501/Tn3 family. However, the accessory genes are not involved in mercury resistance and encode proteins not generally associated with transposons, namely a sulfate permease, a universal stress protein (UspA), and a DksA-like DnaK suppressor protein. Their role is unclear but we assume functional associations since these accessory genes (or a part of it) are found in other Tn3-related transposons. In these transposons they are either in combination with antibiotic resistance determinants like for Tn6001, Tn1403, and Tn1404* derived from clinical and environmental Pseudomonas strains (Stokes et al. 2007; Tseng et al. 2007), or occur alone like in Tn1013 from the IncP-1α plasmid pBS228 (Haines et al. 2007).

Closer scrutiny of the chromosome 2 regions flanking the Tn6050 elements revealed the formation of a chromosomal inversion by recombination between the pair of Tn6050 transposons (Fig. 5). The first copy is located upstream, a partial heavy metal cation tri-component efflux system (czc-like), comprising a two-component transcriptional regulator czcR2 (Rmet_4465) and a sensor czcS2 (Rmet_4466), the efflux pump czcA2 (Rmet_4468), and the membrane fusion encoding czcB2′ (Rmet_4469). The latter is truncated by Tn6050, and its apparent 5′ part czcB2′′ (Rmet_4597) is found downstream of the second copy of Tn6050 (oriented in the opposite direction). Further downstream of czcB2′′, the genes czcC2 and czcI2 are found, encoding, respectively, for an outer membrane porin, and for a regulator, thus complementing the heavy metal cation tri-complex efflux system (Fig. 5). Additional evidence for this inversion is found through analysis of the direct repeats generated by the transposition of the two Tn6050 copies.

Integrases and site-specific recombinases associated with genomic islands

The number of genes (complete or partial) annotated as transposases (with a catalytic site bearing non-adjacent DDE residues) or as integrases, more accurately as tyrosine-based site-specific recombinases, in the genome of C. metallidurans CH34 is very striking, especially compared to those in the closely related genomes of C. eutrophus H16, C. pinatubonensis JMP134 and C. taiwanensis LMG19424. The web-interfaced IslandPath software (http://www.pathogenomics.sfu.ca/islandpath/) clearly allowed to visualize this difference on the chromosomes. Furthermore, the number of int/xerD-like genes (many partial ones) attracts attention, as well as the previously unnoticed diversity of gene combinations around the int/xerD genes in association with the genomic islands of C. metallidurans CH34. The most notable associations were described here, but indications of their functional significance are lacking. The classification of recombinases and their role in genomic islands is not straightforward because of the confusion between transposases, phage integrases catalytic site, and integrases sensu stricto caused by automatic annotations. Hence, we adopted the denomination, tyrosine-based site-specific recombinase, with int most frequently used. A new consensus is needed for denominating int genes of genomic islands, especially when they are associated with a transcriptionally or closely linked accompanying gene, or when multiple putative tyrosine-based site-specific recombinases are in tandem. Lastly, we have few clues about the extent to which some of these int genes are linked to the well-known integrases of the sensu stricto integrons (Mazel 2006). In this descriptive section, we focus attention on the diversity of int genes in C. metallidurans CH34 and some related bacteria, without making mechanistic implications.

Integrases of genomic islands (chromosome 1 and plasmids pMOL28 and pMOL30)

When associated with a genomic island, most of the putative int genes are at a terminal position, like for chromosomal islands CMGI-1 to CMGI-9, and for the islands CMGI-28b and CMGI-28c on plasmid pMOL28. Other plasmid-nested islands are flanked at both sides by partially deleted genes encoding for tyrosine-based site-specific recombinases like CMGI-28a on pMOL28 carrying the bimA1 and chr (chromate resistance) genes, CMGI-30a on pMOL30 carrying the bimA2 and czc (cadmium, zinc, cobalt resistance), and CMGI-30b on pMOL30 carrying the cop (copper resistance) genes. Among the exceptions are the rit genes (recombinase in trio or triad) observed in CMGI-2 and described below. Contrarily, the putative int/xerD genes of chromosome 2 are syntenic (Rmet_5821, Rmet_3675 and less clear for Rmet_4514) or partial with no clear clues for a genomic island (Rmet_4534 and Rmet_4556).

rit genes: site-specific recombinases in triads, a poorly noticed, but common element, except in γ-Proteobacteria

Two groups of int-like genes in CMGI-2 that are annotated via ACLAME as tyrosine-based site-specific recombinases are defined as trios with 3 bp overlaps. These elements are denominated here as RIT elements, with RIT representing recombinase in trio or triad. One trio (ritA2B2C2; Rmet_1239 to Rmet_1241) belongs to the integrase module of CMGI-2 and inactivates a radC-like gene (putatively involved in DNA repair). A second trio (ritA1B1C1; Rmet_1271 to Rmet_1273) belongs to the complex accessory module of CMGI-2 where it also inactivated a gene, in this case a tnpA gene of the IS66 family. Thus, both trios seem to behave as mobile genetic elements. The protein identity between them is low but both share higher protein identities with other trios, some being sheltered in bacteria taxonomically very distant. These RIT elements apparently occur in various β-Proteobacteria, such as Aromatoleum aromaticum EbN1 and Polaromonas sp. JS66, in plasmid pHG1 of C. eutrophus H16, and in several α-Proteobacteria, for example, Mesorhizobium loti MAFF303099, Sinorhizobium meliloti 1021 and S. medicae WSM419, Dinoroseobacter shibae DFL12, and Caulobacter sp. K31. In these two latter species, there are two identical copies of the RIT element, quite distant on the genome, again recalling transposon-like behavior. Furthermore, the RIT elements with substantial protein identities are found in at least one Firmicute of the genus Heliobacterium, and in various Actinobacteria such as Bifidobacterium longum NCC2705, Rhodococcus jostii RHA1, Frankia sp. EAN1pec, and Mycobacterium vanbaaleni PYR-1. Notably, so far, no rit trio was detected via synteny in γ-Proteobacteria. We propose to name these elements in a way similar to the nomenclature of transposons: RITCme1 and RITCme2 for C. metallidurans CH34, RITMme1 for Mesorhizobium meliloti, and so on.

Inside the RIT elements, the recombinases display very low protein identity, but exhibit higher identities with int genes of other trios sharing the same position in the trio/triad. Seemingly, each member of the element evolved for a specific role in the mobility behavior of the RIT element. Their lengths are not the same; the central rit gene sometimes is shorter. To understand their role in horizontal gene transfer, and bacterial biology, a systematic study is needed, comparing all existing RIT elements and conducting in vivo experiments. A preliminary examination of RIT targets showed that RIT elements lie mainly near transposase genes or remnants, and near radC-like genes in at least two other (very different) bacteria.

For the C. metallidurans CH34 rit genes from CMGI-2 (the genomic island involved in chemolithotrophy that shares various genes with pHG1), the RITCme2 element (Rmet_1239 to Rmet_1241) displays the highest protein identity with an orthologous element in B. petrii while the RITCme1 element (Rmet_1271 to Rmet_1273) displays very high homology with its various pHG1 counterparts (Table 2; Schwartz et al. 2003).

rit genes in plasmid pHG1 (C. eutrophus H16): a collection of RIT elements

Plasmid pHG1 contains 21 rit-like genes or fragments encoding for putative tyrosine-based site-specific recombinases, mostly organized in trios and linked by 3 bp overlaps. These pHG1 counterparts of rit genes lie in two clusters: from PHG034 to PHG048 (with 14 rit-like genes or (very small) fragments); and, from PHG129 to PHG141 with seven rit genes. In the first region, three full RIT elements perhaps orthologous to RITCme1 are recognizable, but each seems to have a rearrangement (short deletions in PHG035 and PHG044, gene fusion in PHG042; Fig. 6). This region is located not far downstream from the hox genes that are involved in the biosynthesis of the soluble hydrogenase, and that are also flanked on both sides by 2–4 remnants of transposase genes. The second region PHG129 to PHG141 contains two highly identical homologs (more than 92%) of ritA1B1 (of RITCme1) associated with a more distant ritC, and contains a homolog of ritC1 (of RITCme1) associated to another element RIT element. This region also is flanked by fragments of an IS66-like tnp (transposase) as observed for RITCme1 in CMGI-2 in C. metallidurans CH34, again intimating possible affinity for specific genes as targets.

Fig. 6
figure 6

Synteny between RitCme1, located on chromosome 1 of CH34, and plasmid pHG1 of C.eutrophus H16. Orthologous genes are similarly highlighted (with a minimum threshold of 35% protein sequence identity on 80% of the length of the smallest protein)

Full size image

At least four varieties of RIT elements seem to be represented in both pHG1 rit clusters. It is not clear if one or more of them are functional or if these clusters act as storage of remnants, as it could be the case for pMOL30.

Between both rit clusters of pHG1, there also is an isolated RIT element (PHG056 to PHG058) that has orthologs in a variety of bacteria, and especially with the three identical copies of a B. petrii RIT-like element that are flanked by fragments of non-identical DNA-cytosine methyl-transferase genes (Gross et al. 2008). Again, this arrangement suggests that the RIT elements may recognize some genes as preferential targets.

Both CMGI-2 and pHG1 have in common RIT elements and genes involved in chemolithotrophy but the role of RIT in the mobility of the hyp and hox genes clusters remains unclear since other mobile elements are associated with these gene clusters.

CMGI-28b of pMOL28: a four-gene element with a regulatory gene

Genomic island CMGI-28b, located on pMOL28, mainly contains rhs genes (Mergeay et al. 2009). CMGI-28b contains a combination of four closely linked genes including three putative ones encoding for int-like site-specific recombinases (Rmet_6252 to Rmet_6254) and a regulatory gene (Rmet_6255) from to the TetR family of repressors. This four-gene structure recalls the beginning of the small ars element (CMGI-7) for which the second gene is knocked out by Tn6049. Therefore, these four-gene structures might play a role in the mobility of some genetic elements. They do not display any similarity with RIT elements. Both have counterparts in various bacteria, including some γ-Proteobacteria such as Shewanella frigidimarina NCIMB400 and Vibrio cholerae O1 biovar el tor str. N16961 for the four-gene structure contrary to RIT elements. It is not known if these counterparts are located in the extremities of mobile genetic elements in other bacteria, as it is the case for CMGI-28b and CMGI-7.

bimA1B1, chromate genes in pMOL28 and a chromate island in B. vietnamiensis

Throughout this study, we found that the position of int genes gave valuable pointers to detect genes from the horizontal gene pool. In plasmid pMOL28, the tyrosine-based site-specific recombinase bimA gene (Rmet_6193), and its companion bimB (a putative repressor gene) lie close to the mercury resistance transposon, Tn4378. This pair is separated from the chr genes by five genes coding for hypothetical proteins with unknown function or phenotypic clues. The chr genes, in turn, are followed by the cnr genes. Synteny analysis shows that the first two of these five genes, orf44 (p10049) and lppY (p10050), are systematically associated with chromate genes in bacteria as different as Burkholderia pseudomallei, Methylobacterium sp.4-46, and Arthrobacter sp. FB24. Microarray analysis revealed that the next three genes preceding the chromate cluster chrFECABI are among those most induced in the presence of chromate. Rmet_6196 encodes a permease of the Major Facilitator Superfamily and will be called chrP, while Rmet_6197 and Rmet_6198 encode unknown proteins are denominated chrN and chrO.

These observations suggest that more previously unnoticed genes must be taken into consideration in detailing the response to or the resistance to chromate in C. metallidurans CH34 and other bacteria.

Similarly, the bimAB module is present in B. vietnamiensis G4 (O’Sullivan et al. 2007) at four identical copies together with at least 12 genes including lppY and the chromate genes chrNOFECAB. These bimAB-chromate structures make two pairs, one pair wherein one putative Tn3-related element is inserted (located, respectively, on G4 chromosome 1 and plasmid pBVIE01). For the second pair, two putative Tn3-related elements are integrated (located, respectively, on G4 chromosome 3 and plasmid pBVIE03). Again, a more detailed, systematic analysis and in vivo experiments are needed to clarify the role of these bimAB modules in horizontal gene transfer. Nevertheless, there are two noteworthy observations. First is the putative relation between Tn3-related elements and bimAB modules, such as the insertions of Tn3-related elements in the four bimAB structures in B. vietnamiensis G4, the Tn3-family mercury resistance transposons (complete or partial) near bimA1B1 on pMOL28 and bimA2B2 on pMOL30 of strain CH34, and the Tn3-related element close to a bimAB module in plasmid 2 of Polaromonas sp. JS66. Second is the neighboring heavy metal resistance determinants, such as those for chromate in CH34 and B. vietnamiensis G4, for copper and mercury resistance in Ralstonia pickettii 12J (also CH34 if mercury resistance transposons are considered) and for arsenate/arsenite resistance in Burkholderia multivorans ATCC17616.

General discussion

Cupriavidus metallidurans strain CH34 belongs to a species mainly defined by multiple metal-resistant (and often hydrogenotrophic) strains isolated from industrial sites mainly linked to mining-, metallurgical-, and chemical-industries. The sites are characterized by high levels of bio-available heavy metals, often above the maximum tolerated concentrations. The sequenced genome of C. metallidurans CH34 shows much similarity to the main chromosome (chromosome 1) of the corresponding replicons of C. eutrophus H16, C. taiwanensis LMG19424, and C. pinatubonensis JMP134 (Sato et al. 2006). Chromosome 2 is more of a patchwork, as it also was observed in various Burkholderia species (Fricke et al. 2009) and contains a variety of heavy metal resistance genes that are not evident in other Cupriavidus strains (but many are present in R. solanacearum; Mergeay et al. 2003). Niche specialization seems to be linked to the presence of megaplasmids. At first glance, plasmids/megaplasmids seem to gather most of the horizontal gene pool required for niche specialization: this applies to C. taiwanensis LMG19424, C. eutrophus H16, and likely to C. pinatubonensis JMP134.

However, a glance at the genomic maps as displayed by IslandPath (http://www.pathogenomics.sfu.ca/islandpath/) shows a striking difference between C. metallidurans and the other related genomes due to the much higher abundance of ORFs (complete or fragments) related to transposases or site-specific recombinase genes. For example, in chromosome 1 of C. metallidurans CH34 there are 69 ORFs compared with 9 (5 fragments, 3 bona fide IS elements and 1 bona fide int gene) for C. taiwanensis LMG19424 (Amadou et al. 2008).

The differences between C. taiwanensis LMG19424 and C. metallidurans CH34 may be attributed to their respective environments, namely the symbiotic nodules of legumes, and industrial polluted soils. The nodules might be considered as protected from external microbial influences, or extensively changing physicochemical pressures. From an evolutionary perspective, the industrial sites where the metal resistant Cupriavidus strains occur are recent, their advent mainly arose after the industrial revolution a few centuries ago (a few millennia ago if we consider earlier metallurgy). As a species, C. metallidurans has undergone part of its constituent evolution quite recently, relying on horizontal gene transfer with recognizable and still active tools. Here, a valuable finding is the presence of mobile genetic elements that are conserved fully between taxonomically distant species. There are examples of full conservation in strain CH34. The full conservation of CMGI-1 between CH34 and P. aeruginosa clone C strains, of CMGI-4 between CH34 and D. acidovorans SPH-1 as reported here, of ISCme3 between CH34 and pHG1 (data not shown), and of IS1071 between CH34 and various environmental bacteria able to degrade aromatic xenobiotic compounds (Di Gioia et al. 1998; Nakatsu et al. 1991; Peel and Wyndham 1999; Providenti et al. 2006; Wyndham et al. 1994). Possibly, to obtain IS1071 the progenitors of C. metallidurans might have coexisted with xenobiotic degraders in polluted soils (many of these strains being β-Proteobacteria).

The rearrangements found in the IS1071 copy on pMOL28 might function by stabilizing the captured genes: this concept may be applicable in the many genetic islands of CH34 (CMGI-1, CMGI-5 to 8, CMGI-28a) and especially the transposon Tn6049 seems to play a major role in this inactivation of genes involved in mobility (mainly, but not exclusively, site-specific recombinase genes). Inactivation of the genes involved in mobility is not only observed by insertion of Tn6049, but also by insertion of IS elements or through accumulation of point mutations or deletions in int genes and IS elements. For instance, in pMOL30 deleted tyrosine-based site-specific recombinases flank the cop genes (Mergeay et al. 2009; Monchy et al. 2007). All these observed events suggest primary acquisition via a mobile genetic element and successive stabilization or domestication via the rearrangement of the key mobility genes.

A curious feature possibly linked to this process is the clustering of tnp and int gene fragments in some plasmid regions of pMOL30 and even in chromosomal genomic islands. It was also observed in pHG1 of C. eutrophus H16 (Fig. 4) and in pRALTA of C. taiwanensis that contains up to 120 ORFs linked to IS elements. All these deleted or rearranged mobility genes may represent scars of events occurring at various periods in the evolutionary history of these strains, and likely encompassing a long time up to rather recent events. An interesting question is if whether these clusters would dissolve completely in time or would they be a kind of storage of deleted or “eroded” mobility genes after usage?

The genome as a source of unknown or poorly noticed mobile genetic elements

The main islands found in the chromosome of C. metallidurans CH34 belong to the KLFC2/PAGI-2 family (CMGI-1) (Klockgether et al. 2006; Larbig et al. 2002) or to the Tn4371-family (CMGI-2, CMGI-3, and CMGI-4). Some small islands of interest are the arsenic island (CMGI-7) that expands our view on heavy metal responses, and CMGI-9 that shares characteristics with the int module of CMGI-3. We described new transposons of special interest because they carry genes that are never, or rarely represented in the current transposon collection from clinical- or environmental-studies such as the eight still undefined genes of Tn6048 for which no phenotype is recorded but that are induced by heavy metals, and the unusual but conserved association of Tn6050 genes encoding for an universal stress protein (UspA), a suppressor of DnaK and a permease of sulfate. The latter one is a good example of passenger genes from well disseminated mobile genetic elements for which we do not yet have good phenotypic clues.

Maybe the advantageous power of synteny in annotation studies is the ability to reveal the diversity of mobile genetic elements outside the bias of selection not only in the studied genome but also in a variety of other genomes. We can recall here numerous cases like transposon Tn6053 of B. xenovorans LB400 very similar in structure to Tn6049, the four-gene transposon Tn6052 (pbrR cdfX pbrC tnpA) similar to ISPpu12 (from P. putida plasmids) and found in very distant bacteria, a genetic element carrying chromate genes in B. vietnamiensis G4 and particularly the Tn4371 family genomic islands of D. acidovorans and B. petrii carrying various metal responding genes. In this respect, the D. acidovorans DAGI-1 genomic island is remarkable as a carrier of metal resistance genes that are gathered in a compact collection: cop genes in the accessory module, czcD-like in the replication module and sil, pbr, mer, and ars genes in the integrase module. Annotation using synteny allows also putting in evidence the great diversity in tyrosine-based site-specific recombinase genes (integrases) going beyond the diversity already known for the bacteriophage and the integron integrases. Among these genes, there are the int genes of the genomic islands of the PAGI-2 family and of the integrase modules of the Tn4371 family, the well conserved four-gene elements found in CMGI-7 (ars) and CMGI-28b, and the peculiar RIT elements (triads of site-specific recombinase genes with 3 bp overlaps) that are present in CMGI-2 and especially in pHG1. Moreover, the latter are present in many bacteria and especially in Actinobacterial strains belonging to the genera Frankia, Mycobacterium and Rhodococcus.

At this stage, it may be useful to mention that antibiotic resistance genes are practically absent of the mobile genetic elements reported in this study. The diversity of the mobile genetic elements so found is representative of this part of the horizontal genetic pool that was at work when β-Proteobacteria (genera Cupriavidus, Ralstonia, Burkholderia, Delftia, Bordetella, Herminiimonas, and Janthinobacterium) began to colonize anthropogenic sites (soils, sediments) characteristic of the modern industrial revolution. The annotation of C. metallidurans CH34 allows then a new and refreshing insight on the diversity of bacterial mobile genetic elements.