Conjugative transposons: the tip of the iceberg


  • Vincent Burrus,

    Search for more papers by this author
    • Present address: Division of Geographic Medicine/Infectious Diseases, New England Medical Center and Tuft University School of Medicine, Boston, MA 02111, USA.

  • Guillaume Pavlovic,

    1. Laboratoire de Génétique et Microbiologie, UMR INRA-UHP 1128, IFR110, Faculté des Sciences et Techniques, Université Henri Poincaré (Nancy 1), BP239, 54506 Vandœuvre-lès-Nancy, France.
    Search for more papers by this author
  • Bernard Decaris,

    1. Laboratoire de Génétique et Microbiologie, UMR INRA-UHP 1128, IFR110, Faculté des Sciences et Techniques, Université Henri Poincaré (Nancy 1), BP239, 54506 Vandœuvre-lès-Nancy, France.
    Search for more papers by this author
  • Gérard Guédon

    Corresponding authorSearch for more papers by this author


Elements that excise and integrate, such as prophages, and transfer by conjugation, such as plasmids, have been found in various bacteria. These elements appear to have a diversified set of characteristics including cell-to-cell contact using pili or cell aggregation, transfer of single-stranded or double-stranded DNA, low or high specificity of integration and serine or tyrosine recombinases. This has led to a highly heterogeneous nomenclature, including conjugative transposons, integrative ‘plasmids’, genomic islands and numerous unclassified elements. However, all these elements excise by site-specific recombination, transfer the resulting circular form by conjugation and integrate by recombination between a specific site of this circular form and a site in the genome of their host. Whereas replication of the circular form probably occurs during conjugation, this replication is not involved in the maintenance of the element. In this review, we show that these elements share very similar characteristics and, therefore, we propose to classify them as integrative and conjugative elements (ICEs). These elements evolve by acquisition or exchanges of modules with various transferable elements including at least ICEs and plasmids. The ICEs are probably widespread among the bacteria.


Conjugative plasmids encode the proteins involved in their own transfer from a donor cell to a recipient cell. First of all, the pili encoded by the conjugative plasmids from proteobacteria (Silverman, 1997) or the aggregation factors encoded by the pheromone conjugative plasmids of the low G+C Gram-positive genus Enterococci (Clewell, 1993) promote cell-to-cell contacts. Then, a rolling circle replication-like system (relaxosome) nicks the transferred strand at the origin of transfer, unwinds the nicked strand, replicates the remaining strand from the free 3′-OH and rejoins both ends (Lanka and Wilkins, 1995; Llosa et al., 2002). The nicked strand is transferred through a mating pore related to type IV secretion systems and is replicated in the recipient cell. In the past, all described conjugative elements were plasmids. More recently, various chromosomal conjugative elements have been identified that, unlike conjugative plasmids, cannot be isolated as circular replicative molecules. Site-specific recombinases en-coded by these elements promote their excision and integration (Churchward, 2002). Site-specific recombinases can catalyse the integration and excision of circular DNA molecules (integrases), resolution of co-integrates (resolvases) or inversion of specific DNA fragments (invertases). Two unrelated families of site-specific recombinases are known, the serine recombinase family and the tyrosine recombinase family. Related serine recombinases were found to be involved in insertion or excision of some prophages and some other integrative elements (Smith and Thorpe, 2002), whereas the tyrosine recombinases are involved in integration and excision of a very large array of integrative elements including most of the prophages such as λ (Argos et al., 1986). Most of the integrases belonging to the tyrosine recombinase family catalyse site-specific integration into the 3′ end of a gene encoding a tRNA.

The best studied of the chromosomal conjugative elements are the conjugative transposon Tn916 from Enterococcus faecalis (Clewell et al., 1995; Scott and Churchward, 1995) and pSAM2 from Streptomyces ambofaciens, an element that was first described as an integrative plasmid (Pernodet et al., 1984). The present review focuses on the comparison between the chromosomal conjugative elements and their evolution. As these elements share similar properties and evolve by acquisition/exchange of modules, we propose to group them in a large class.

What is a conjugative transposon?

The term ‘conjugative transposon’ was first used to describe Tn916 from the low G+C Gram-positive bacterium E. faecalis (for reviews on Tn916, see Clewell and Flannagan, 1993; Clewell et al., 1995; Salyers et al., 1995a ; Scott and Churchward, 1995; Churchward, 2002). Tn916, like all other conjugative elements from Gram-positive bacteria, does not encode pili. The mechanism of cell-to-cell contact remains unknown. This element encodes an integrase belonging to the tyrosine site-specific recombinase family. The first step in Tn916 conjugative transfer is an excision event catalysed by the integrase and the excisionase. It leads to the creation of a covalently closed circular molecule containing an attachment site attTn resulting from recombination between the two attachment sites attL and attR flanking Tn916. The second step is the conjugative transfer itself, which is likely to be similar to that of conjugative plasmids. The genes carried by Tn916 encode all the proteins needed to form the mating pore and to mediate its transfer. A nick at the origin of transfer (oriT) would initiate the transfer of a single strand of the circular intermediate (Jaworski and Clewell, 1995). After completion of the transfer and replication, donor and recipient cells would each contain a copy of the double-stranded circular molecule. The tyrosine recombinase catalyses the integration between the attachment site attTn of the circular form of Tn916 and a very large array of sites in E. faecalis and in numerous other bacteria, preferentially into A+T-rich sequences. Site-specific excision and low specificity of integration in the same cell lead to the intracellular transposition of the element. However, this transposition mechanism is very different from that of the mobile elements coding for a genuine DDE transposase, i.e. insertion sequences, and type I and type II transposons. As Tn916 is able to transpose and conjugate, it was described as a conjugative transposon.

Conjugative transposons unrelated to Tn916 have higher specificity of integration than this element (Table 1). The conjugative transposon Tn5276 from the low G+C Gram-positive bacterium Lactococcus lactis integrates into at least five chromosomal sites in L. lactis MG1614 (Rauch and de Vos, 1992). A large array of conjugative transposons was found in Bacteroides, a genus belonging to the Cytophaga–Flexibacter–Bacteroides (CFB) division (Salyers et al., 1995a,b; Smith et al., 1998). These elements (e.g. CTnDOT or TCrERL) have a narrow specificity of integration, integrating into only three to seven chromosomal sites depending on the element (Table 1). Two other types of conjugative transposons, CTnscr94 from the proteobacterium Salmonella enterica and Tn5252 from Streptococcus pneumoniae, are site-specific integrative. Other chromosomal conjugative elements, such as bph–sal from Pseudomonas putida, clc from Pseudomonas sp., the symbiosis islands from Mezorhizobium loti and SXT from Vibrio cholerae, were not initially classified as conjugative transposons and were named according to the functions they carry (Table 1) (Churchward, 2002). Nevertheless, these elements, like Tn5252 and CTnscr94, are site-specific integrative and conjugative.

Table 1. . Putative and proved integrative and conjugative elements. a,b
ElementsGenus or speciesSize(kb)Other carried functionsIntegrasetypeIntegration sitesReferences
  • a

    . Other related integrative and conjugative elements have not been indicated in this table.

  • b

    . The names of the proven ICEs are not underlined. The names of the putative ICEs are underlined.

  • c

    . pSE101 and Tn5397 were not found in S. lividans and B. subtilis respectively.

  • Cam, chloramphenicol; Erm, erythromycin; Hg, mercury; Kan, kanamycin; ND, not determined; Ser, serine recombinase; Str, streptomycin; Suf, sulphamethoxazole; Tet, tetracycline; Trm, trimethoprim; Tyr, tyrosine recombinase; Van, vancomycin;

TCrERLa Bacteroides fragilis  80TetrNDFew sites Salyers et al. (1995b)
CTnDOTa Bacteroides  65Tetr ErmrTyrFew sites (GTTNNTTTGG) Cheng et al. (2001)
Symbiosis islanda Mezorhizobium loti R7A502SymbiosisTyr3′ of a gene encoding tRNAPhe gene Sullivan et al. (2002)
CTnscr94 Salmonella enterica 100Sucrose utilizationND3′ of two genes encoding a tRNAPhe Hochhut et al. (1997)
SXTa Vibrio cholerae  99.5Sufr Trmr Camr StrrTyr5′ of a gene encoding RF3 Hochhut and Waldor (1999)
R391a Providencia rettgeri  94Kanr HgrTyr5′ of a gene encoding RF3 Murphy and Pembroke (1999)
clc Pseudomonas sp. B13105Chlorocatechol degradationTyr3′ of two genes encoding a tRNAGly Ravatn et al. (1998b)
bph–sal Pseudomonas putida  90Biphenyl and salicylate degradationNDND Nishi et al. (2000)
pSAM2a Streptomyces ambofaciens  10.9None identifiedTyr3′ of a gene encoding a tRNAPro Pernodet et al. (1984)
pSE101a Saccharopolyspora erythraea  11.3None identifiedTyr3′ of a gene encoding a tRNAThr Brown et al. (1994)
Streptomyces lividans c    Numerous sites 
Tn1549a Enterococcus sp. 34VanrTyrA+T rich Garnier et al. (2000)
Tn916a Enterococcus faecalis  18TetrTyrA+T rich Clewell et al. (1995)
Clostridium difficile    One preferential site Wang et al. (2000a)
Tn5397a C. difficile  21TetrSerSingle site Wang et al. (2000b)
Bacillus subtilis c    Numerous sites 
Tn5252a Streptococcus pneumoniae  47CamrTyrSingle intergenic site Vijayakumar and Ayalew (1993)
Tn5276a Lactococcus lactis  70Sucrose utilization and nisin synthesisTyrFew sites (TTTTTG) Rauch and de Vos (1992)
pRS01/sex factor L. lactis  48.4Tellurium resistanceNDSingle site Gasson et al. (1995)
EfaC1 a E. faecalis  25.3NDTyr3′ end of a gene encoding a ThrtRNA Burrus et al. (2002)
EfaC2 a E. faecalis  32.7NDTyr3′ of a gene encoding a GMP-synthase Burrus et al. (2002)
EfaD2 a E. faecalis NDNDTyrND Burrus et al. (2002)
Tn 5801 a Staphylococcus aureus  25.8TetrTyr3′ of a gene encoding a GMP-synthase Kuroda et al. (2001)
CW459 tet(M) a Clostridium perfringens NDTetrTyr3′ of a gene encoding a GMP-synthase Roberts et al. (2001)
ICE Lm1 a Listeria monocytogenes  21.3Putative Cd2+ resistanceTyr3′ of a gene encoding a GMP-synthase Burrus et al. (2002)
ICE St1 a Streptococcus thermophilus  34.7Restriction-modification Sth368ITyr3′ end of a gene encoding afructose-1,6-diphosphate aldolase Burrus et al. (2002)

The mechanisms of conjugation of these site-specific integrative and conjugative elements/conjugative transposons are poorly known. Only one of these elements, SXT, was found to encode pili related to those encoded by the conjugative plasmids from proteobacteria (Beaber et al., 2002). Likewise, the origin of conjugation of only three elements, the conjugative transposon TcrEmr DOT from Bacteroides (Li et al., 1995), the conjugative transposon Tn5252 (Srinivas et al., 1997) and the element SXT (Beaber et al., 2002), were identified, suggesting that single-strand DNA is transferred during conjugation of these elements.

Plasmids or not plasmids? That is the question!

Some elements were described as plasmids because they can spread by conjugation or were isolated as circular molecules in some conditions. Twelve conjugative elements from proteobacteria, including R391 from Providencia rettgeri and R997 from Proteus mirabilis, were previously classified as ‘IncJ plasmids’, but strains harbouring these elements appear to lack extrachromosomal DNA. The elements R391 and R997 only remain as replicative circular molecules in recA strains when their integration site already contains an IncJ element (Pembroke and Murphy, 2000). These two elements integrate into the same locus, the 5′ end of prfC encoding the peptide release factor RF3 (Hochhut et al., 2001a). In fact, these elements are site-specific integrative and conjugative elements related to SXT.

pRS01 from L. lactis ML3, a strain deriving from NCDO712, was described as a plasmid because it can mobilize co-resident plasmids by co-integration, and a circular form was isolated (Anderson and McKay, 1984). However, the low copy number of the free form (0.25 copies by genome) suggests that replication is not involved in the maintenance of this element. Furthermore, a site-specifically integrated element, the ‘sex factor’, which was characterized from another strain derived from NCDO712, was found to mobilize a co-resident plasmid by co-integration (Gasson et al., 1995). Also, partial sequences of these elements are identical. Therefore, pRS01 and the ‘sex factor’ very probably correspond to the same site-specific integrative and conjugative element. The oriT of pRS01 was identified, suggesting that single-strand DNA is transferred during conjugation (Mills et al., 1998). The gene cluA of the sex factor encodes a surface-presented protein that is responsible for cell aggregation and, therefore, for cell-to-cell contact during conjugation. This protein is related to the cell aggregation factors of pAD1 and other pheromone conjugative plasmids from Enterococci (Godon et al., 1994).

The high G+C Gram-positive bacterium genus Streptomyces and its relatives Saccharopolyspora and Amycolatopsis are soil bacteria that display complex morphological differentiation. The vegetative mycelium consists of septate, multinucleoid, branched hyphae that grow into the solid nutrient substrate. Nine elements from Streptomyces and its relatives encode integrative and conjugative functions and were first described as integrated conjugative plasmids (Raynal et al., 1998 and references therein). At least six of them are site-specific integrative elements. The best studied of these elements is pSAM2 from Streptomyces ambofaciens. The wild-type pSAM2 is integrated into a unique site, the 3′ end of a gene encoding a tRNAPro (Raynal et al., 1998), and its circular form was never detected in the absence of a recipient mycelium. The genes repSA (encoding the rolling-circle replicase), xis (encoding an excisionase) and int (encoding a tyrosine recombinase) are organized into a single operon activated by the pra gene product. pra transcription is repressed in the absence of recipient cells, thus ensuring that the wild-type pSAM2 remains integrated (Sezonov et al., 1995 and references therein). A pSAM2 mutant, resulting from a point mutation in the promoter of the gene pra, is found simultaneously as one integrated copy and as 5–10 circular replicating copies per genome (Sezonov et al., 1995). Furthermore, pSAM2 derivatives, defective for integrative functions (deletion of attP site or disruption of the integrase gene), are maintained as free replicons (Smokvina et al., 1991).

After physical contact between donor and recipient mycelia, the operon repSA–xis–int is activated, leading to pSAM2 excision and its rolling circle replication (Possoz et al., 2001). The intermycelial transfer is then mediated by only one gene, traSA. pSAM2 was recently found to transfer as double-stranded DNA (Possoz et al., 2001). traSA encodes a protein related to SpoIIIE from Bacillus subtilis that acts as a double-stranded DNA pump during sporulation, actively transporting one of the replicated pair of chromosomes from the mother cell to the prespore. The transferred copies of pSAM2 then replicate in the recipient mycelia. The intramycelial transfer within the recipient strain is mediated by the four genes spdABCD. The transfer of pSAM2 and other integrative conjugative elements of Streptomyces and related genera is similar to that of genuine conjugative plasmids from Streptomyces both in the transfer proteins involved and in physical appearance: these conjugative elements transiently inhibit the growth of the recipient mycelia, yielding to zones of slow growth called pocks (Hopwood and Kieser, 1993). After completion of the transfer, unlike plasmids, pSAM2 is found site-specifically integrated in the donor and recipient mycelia. Although pSAM2 is able to replicate, the replication of the circular form is not involved in the maintenance of the element, but is necessary for its transfer (Hagège et al., 1994). Therefore, pSAM2 is not a genuine plasmid, but should be considered as a site-specific integrative and conjugative element.

It must be emphasized that rolling circle replication also occurs during conjugation events involving single strands (Llosa et al., 2002). However, whereas the replication of single-strand DNA probably occurred during the conjugative transfer of elements possessing an oriT, the maintenance of the various conjugative transposons and site-specific integrative and conjugative elements in the bacteria progeny do not require the replication of the double-strand circular molecule but its integration. Therefore, none of these elements should be considered as a plasmid.

Low or high integration specificity? That is another question!

A large array of integrative and conjugative elements, including some elements usually described as conjugative transposons, are site specific and would not be able to transpose within the bacterium where they were isolated. However, most of the conjugative transposons integrate in various sites and are therefore able to transpose within the genome of their host.

The study of various prophages, such as λ, and other site-specific integrative elements that encode a tyrosine recombinase revealed the existence of secondary integration sites. Furthermore, some ‘conjugative transposons’ and some ‘site-specific integrative and conjugative elements’ have a high specificity of integration in one host and a lower specificity of integration into another host. The element pSE101 integrates into the 3′ end of a tRNAThr-encoding gene in Saccharopolyspora erythraea, where it was isolated, but into numerous other sites in Streptomyces lividans (Brown et al., 1994). In the same way, the conjugative transposon Tn916, which integrates into a large array of sites in E. faecalis, integrates very preferentially into a unique site, att916, in Clostridium difficile (Wang et al., 2000a) . Additionally, an element indistinguishable from Tn916, Tn916CD, was found to be integrated in att916 in an environmental isolate of C. difficile (Wang et al., 2000a). In the same way, the conjugative transposon Tn5397 integrates in one preferential site in C. difficile, where it was isolated, but integrates into a large array of sites in Bacillus subtilis (Wang et al., 2000b). As some of these elements, such as Tn916, were found to transfer between different bacterial species, genera or divisions (Clewell et al., 1995; Scott and Churchward, 1995), the distinction between site-specific integrative and conjugative elements and conjugative transposons is not relevant.

Most of the integrated prophages encode a tyrosine recombinase and only some a serine recombinase. In the same way, almost all site-specific integrative and conjugative elements/conjugative transposons encode a tyrosine recombinase (Table 1). Only one, Tn5397, was found to encode a serine recombinase (Wang et al., 2000b). However, sequence analysis revealed five putative site-specific integrative and conjugative elements/conjugative transposons encoding a serine recombinase in the incompletely sequenced genomes of C. difficile, E. faecalis and Streptococcus equi (Burrus et al., 2002).

Modular evolution

The genes of various mobile elements (e.g. prophages) involved in the same function, such as lysogeny or capsid formation, are grouped in modules. The prophages (and probably the other types of mobile elements) evolve by exchange of these functional modules (Toussaint and Merlin, 2002 and references therein). Comparison of the structure and organization of genes of Tn916 and its relatives revealed that they also evolve by acquisition and/or exchange of modules. The conjugation, tetracycline resistance and regulation modules of Tn916 and Tn5397 are closely related (Roberts et al., 2001) (Fig. 1A). Nevertheless, their recombination modules are unrelated: Tn5397 encodes a serine recombinase, whereas Tn916 encodes a tyrosine recombinase. Likewise, the Tn1549 recombination module is closely related to that of Tn916 (Garnier et al., 2000), but their conjugation modules are unrelated (Fig. 1A). Comparison of sequenced elements from low G+C Gram-positive bacteria, including Tn916, Tn1549, Tn5397, ICESt1 and 17 putative elements found in completely or incompletely sequenced genomes of low G+C Gram-positive bacteria, also reveals multiple exchanges of integration and conjugation modules (Burrus et al., 2002).

Figure 1.

Comparison of the gene organization of some putative or proven conjugative elements.

A. Conjugative transposons Tn5397, Tn916 and Tn1549.

B. Putative site-specific integrative and conjugative elements EfaC1 and EfaC2.

C. Putative integrative conjugative element/conjugative transposon EfaD2 and conjugative plasmid pAM373. Boxes represent the ORFs, with the proposed direction of transcription shown by the arrow. ORF names beginning with ‘orf’ (Tn5397, Tn916, Tn1549), ‘EF’ (EfaC1, EfaC2, EfaD2) or ‘EP’ (pAM373) in the GenBank annotation are abbreviated with the corresponding number. The colour of the boxes depicts the putative function of each ORF deduced from functional analysis or from blastp comparison: Blue, conjugative transfer; grey, antibiotic resistance; green, putative transcriptional regulator related to those of Tn916; orange, serine recombinase; red, excisionase and tyrosine recombinase; black, plasmid replication functions; white, other putative functions or no putative function.

In (A), the ORFs are connected by grey shading when significant nucleotide identity (given between brackets) is detected; the ORFs are connected by light yellow shading when the predicted proteins share significant sequence identity (<40% in blastp comparison).

In (B) and (C), the ORFs are connected by a yellow/orange shading when the predicted proteins share significant sequence identity; the colour of this shading is linked to the percentage amino acid identity. The whole sequence of each element is presented, except for EfaD2, as its ends are not known. In Tn5397, the group II intron is depicted by a square pattern. In Tn5397, orf5 is not described by Roberts et al. (2001), but is found by sequence comparison with Tn916. In EfaC1, the black line indicates ORFs closely related to those of the conjugative plasmid pAD1 (>90% identity in nucleotide sequence). Whereas the ORFs EF2546 from EfaC1 and EF0166 from EfaC2 encode distantly related putative integrases (29% amino acid identity), they are not connected in (B).

Exchanges or acquisitions of other types of modules have also occurred. The putative elements Tn5801 (SAV0392–SAV414 region of the sequenced genome of Staphylococcus aureus Mu50) and ICELm1 (lmo1097–lmo1115 region of the genome of Listeria monocytogenes EGD-e) have very closely related integration, regulation and conjugation modules (Burrus et al., 2002). However, the tetracycline resistance module of Tn5801 is replaced by a putative cadmium resistance module in ICELm1.

These module exchanges can also occur between different types of mobile elements, e.g. plasmids and conjugative transposons/site-specific integrative and conjugative elements. The site-specific integrative element ICESt1 from S. thermophilus encodes a recombination module distantly related to those of Tn5252 and Tn5276 (Burrus et al., 2000), a putative conjugation module distantly related to that of Tn916 (Burrus et al., 2002) and a restriction–modification system, Sth368I, related to LlaKR2I encoded by the plasmid pKR223 from L. lactis (Burrus et al., 2001). The conjugation modules of the putative site-specific integrative and conjugative elements EfaC1 (EF2512–EF2546 region of the sequenced genome of E. faecalis V583) and EfaC2 (EF0127–EF0166 region of the genome of E. faecalis V583) are closely related, but the conjugation module of EfaC2 possesses five open reading frames (ORFs) (Ef0145–Ef0149) closely related to conjugation genes of the enterococcal conjugative plasmids pAD1, including sea1 involved in surface exclusion (immunity) and asa1 encoding the aggregation factor (Fig. 1B). Furthermore, the conjugative module of the putative element EfaD2 (EF0479–EF0514 region of the genome of E. faecalis V583) is closely related to that of the enterococcal conjugative plasmids pAM373 and pAD1 (Fig. 1C).

The insertion of one element into another, followed by deletions, could explain at least some of the module exchanges/acquisitions. Indeed, composite conjugative transposons were found in Streptococci and Enterococci (Clewell and Flannagan, 1993; Scott and Churchward, 1995 and references therein). Tn5253 and Tn3701 are Tn5252-like elements containing the Tn916-like elements Tn5251 and Tn3703. The 70 kb conjugative transposon Tn3705 contains the Tn916-related element Tn3704. In the same way, the conjugative transposon CTnDOT from Bacteroides is virtually identical over most of its length to another conjugative transposon, CTnERL, except that CTnDOT carries a 13 kb region harbouring an element related to various mobilizable transposons from Bacteroides (Whittle et al., 2001). The conjugative transposons/site-specific integrative and conjugative elements can also acquire novel modules by insertion of a non-conjugative transposon. Indeed, the 23 kb region of SXT encoding sulphamethoxazole, trimethoprim, streptomycin and chloramphenicol resistance has a structure similar to that of composite type I transposons flanked by insertion sequences (Hochhut et al., 2001b).

The conjugative transposons of Bacteroides and the pRS01/sex factor from L. lactis can mobilize non-conjugative plasmids in cis. The insertion of conjugative transposons of Bacteroides within a non-conjugative plasmid (Salyers et al., 1995a,b) or the insertion of a non-conjugative plasmid within the sex factor by replicative transposition of ISS1 (Gasson et al., 1992) was found to create novel conjugative plasmids. However, Tn916 does not mobilize markers in cis, probably because the circularization of Tn916 is required for the transcription of the genes involved in its conjugation (Celli and Trieu-Cuot, 1998). The conjugative transposon Tn1549 from Enterococcus sp. is inserted into the traE1 gene of a conjugative plasmid closely related to the pheromone conjugative plasmid pAD1 (Garnier et al., 2000). Therefore, co-integrations of plasmids and conjugative transposons followed by rearrangements (e.g. deletions) are probably involved in the modular evolution of plasmids and conjugative transposons.

Tandem integration of site-specific integrative and conjugative elements could also be involved in module acquisition. The presence of a resident element Tn916, Tn1545, SXT or R391 in the recipient cell does not significantly reduce its ability to acquire a second related element (Hochhut et al., 2001a; Churchward, 2002). Tandem integration of the closely related elements R391 and SXT but not of Tn916 were reported recently, suggesting that R391 and SXT can integrate into the attachment sites attL or attR of a resident element instead of the attachment site attB (Hochhut et al., 2001a). However, the tandem arrangement is unstable, and one of the elements is frequently lost. Likewise, tandem integrated copies of clc were found in transconjugants cultivated on media containing chlorobenzene, but this organization is lost when the bacteria are grown in its absence (Ravatn et al., 1998a). Unlike Tn916 or SXT, a resident copy of pSAM2 greatly reduces the frequency of acquisition of a second copy by conjugation (Possoz et al., 2001). However, unstable tandem arrangements of pSAM2 were found in the rare transconjugants (Possoz et al., 2001). The Tn5276-related conjugative transposon Tn5481 contains an internal 187 bp sequence sharing 88% identity with the Tn5481 left end, i.e. attL (Immonen et al., 1998). The two attL-type sequences border a 13.3 kb module involved in nisin biosynthesis and resistance. This structure could have arisen by integration of a Tn5276-related element into the attR or attL attachment sites flanking another Tn5276-related element, followed by the deletion of one of the conjugation and recombination modules. This event would have led to the acquisition of the nisin synthesis module.

Consequences of the module exchanges

Integration and conjugation module exchanges probably play a significant part in the host specificity of the elements. Tn916 has a very broad host range (Clewell et al., 1995): it transfers from E. faecalis to a large number of species of low G+C Gram-positive bacteria and to various proteobacteria. The Bacteroides conjugative transposons transfer from Bacteroides to the closely related genus Prevotella (Salyers et al., 1995b; Smith et al., 1998). However, they can mobilize some elements in trans from Bacteroides to Escherichia coli, suggesting that the host range is only limited by the maintenance of the conjugative transposon in the recipient, not by the transfer functions (Smith et al., 1998). Other elements have a very narrow host range: the prS01/sex factor transfers only between strains of L. lactis (Gasson et al., 1995). Therefore, the exchange or acquisition of conjugation modules may change the transfer specificity.

The putative elements Tn5801 from S. aureus, ICELm1 from L. monocytogenes, CW459tet(M) from Clostridium perfringens and EfaC2 from E. faecalis have closely related conjugation, regulation and recombination modules and are site-specifically integrated into an identical 11 bp sequence corresponding to the 3′ end of genes encoding a GMP synthase (Burrus et al., 2002). Tn916, which has conjugation and regulation modules closely related to those of these four elements, but encodes a very different recombinase, integrates into a large array of sites in E. faecalis from which it was isolated. Furthermore, EfaC1 and EfaC2 have closely related conjugation modules (Fig. 1) but encode very different tyrosine recombinases. Sequence analysis showed that EfaC1 is site-specifically integrated in the 3′ end of a gene encoding a tRNAThr, whereas EfaC2 is site-specifically integrated in the 3′ end of a gene encoding a putative GMP synthase (Burrus et al., 2002). This suggests that module ex-changes or acquisition has led to variations in the integration specificity.


Most of the integrative and conjugative elements/conjugative transposons integrate into a unique site, usually into the 3′ end of genes encoding tRNAs (Table 1). These elements would not be able to transpose within the bacterium where they were isolated. These features are similar to the site-specific integration systems of prophages and not to those of type I or type II transposons. Only some of the conjugative transposons have a low specificity of integration and were actually found to transpose within the cell. However, the variations in integration specificity of Tn916, Tn5397 and pSE101 in different hosts and the exchanges of recombination modules strongly suggest that distinctions between site-specific integrative elements and conjugative transposons are not relevant. The highly heterogeneous nomenclature and classification hide very similar characteristics and even close relationships such as the relationship between SXT and some of the ‘plasmids’ of the IncJ group (Hochhut et al., 2001a).

By comparing the properties of SXT and numerous other elements, Hochhut and Waldor (1999) grouped the conjugative transposon Tn5252, Tn5276, CTnscr94, the Bacteroides conjugative transposons, the symbiosis islands, clc, R391, SXT and related elements within a class that they called CONSTIN for conjugal, self-transmissible, integrating element. However, Tn916, pRS01, pSAM2 and related elements were not included in this class, possibly because Tn916 was first found to have a very low integration specificity, and pRS01 and pSAM2 were long considered as plasmids.

It was proposed more recently that the site-specific integrative elements clc, bph–sal, SXT and the symbiosis island should be considered as conjugative transposons (Merlin et al., 2000). However, the name conjugative transposon could suggest not only that these elements have a low integration specificity like most bacterial transposons, but also that they could have some relationships with type I or type II transposons. These site-specific integrative and conjugative elements were neither initially nor usually reported as conjugative transposons, probably because these elements are site-specifically integrated (Churchward, 2002). Furthermore, the ‘integrative plasmids’ such as pSAM2 were never included in the conjugative transposons.

Recently, the term ‘genomic island’ was proposed as an extension of pathogenicity islands (Hentschel and Hacker, 2001). Depending on the functions that they encode, they would be named pathogenicity islands, symbiosis islands, metabolic islands or resistance islands. However, this class groups numerous elements showing very different structures (satellite phage-like, type I transposon-like, conjugative transposon-like, mobilizable transposon-like) and very different modes of acquisition (transport by a capsid, conjugation) or mobility mechanisms that generally remain to be elucidated.

The large grouping initiated by Hochhut and Waldor (1999) should be continued. We propose to establish a class of elements that contains all elements that excise by site-specific recombination into a circular form, self-transfer by conjugation and integrate into the host genome, whatever the specificity and the mechanism of integration and conjugation. These elements would also be able to replicate during the conjugation event, but this replication should not be involved in their maintenance. Hagège et al. (1994) renamed pSAM2 as an integrating conjugative element. We propose to name these elements in a similar manner, as ICEs for integrative and conjugative elements. The conjugative transposons would be a subset of ICEs that can transpose within the cell. ICE could also be used to name novel elements found in the future. In these cases, ICE would be followed by the initials of the name of the bacterium from which it was isolated and by a number, which may identify the strain or correspond to the rank of the discovery of the element, for example ICESt1.

Identified ICEs: the tip of the ICEberg

At present, ICEs are known to exist in a limited range of hosts belonging to the four major divisions of bacteria, i.e. the proteobacteria, the low G+C Gram-positive bacteria, the high G+C Gram-positive bacteria and the CFB division. However, the only ways to detect ICEs are the conjugative transfer of genes conferring a specific phenotype or the effect of their transposition. Furthermore, some ICEs were initially described as plasmids or transposons. Therefore, the relatively low number of described elements may greatly underestimate the number of elements that actually exist. However, recent genomic searches by comparison with sequenced ICEs revealed 17 putative ICEs in completely or incompletely sequenced genomes from various low G+C bacteria, including B. subtilis, C. difficile, E. faecalis, L. monocytogenes, S. mutans and S. aureus (Burrus et al., 2002). Furthermore, sequence analysis of the Streptomyces coelicolor A3(2) genome revealed six putative elements related to pSAM2 and the previously identified element SLP1 (Bentley et al., 2002). This suggests that the ICEs are widespread in bacteria. Therefore, like the conjugative plasmids and the prophages, the ICEs could be one of the main types of elements responsible for horizontal gene transfer. The transfer of the known ICEs allows a one-step acquisition of novel functions that can be advantageous for the bacteria, including antibiotic, heavy metals or phage resistance, sucrose, biphenyl and chlorocatechol degradation or bacteriocin synthesis (Table 1). Furthermore, the acquisition of an ICE, the symbiosis island, converts a soil saprophyte to a symbiont of the legume roots, changing the ecological niche of its bacterial host (Sullivan et al., 2002).


We are grateful to John Beaber and Dr Matthew Waldor from the New England Medical Center (Boston) for their insightful comments on this manuscript. This work was supported by grants from Institut National de la Recherche Agronomique, Université de Nancy 1 and Ministère de l’Education Nationale, de la Recherche et de la Technologie, Paris, France.