Intracellular excision and reintegration dynamics of the ICEclc genomic island of Pseudomonas knackmussii sp. strain B13


*E-mail; Tel. (+41) 21 6925630; Fax (+41) 21 6925605.


Genomic islands are DNA elements acquired by horizontal gene transfer that are common to a large number of bacterial genomes, which can contribute specific adaptive functions, e.g. virulence, metabolic capacities or antibiotic resistances. Some genomic islands are still self-transferable and display an intricate life-style, reminiscent of both bacteriophages and conjugative plasmids. Here we studied the dynamical process of genomic island excision and intracellular reintegration using the integrative and conjugative element ICEclc from Pseudomonas knackmussii B13 as model. By using self-transfer of ICEclc from strain B13 to Pseudomonas putida and Cupriavidus necator as recipients, we show that ICEclc can target a number of different tRNAGly genes in a bacterial genome, but only those which carry the GCC anticodon. Two conditional traps were designed for ICEclc based on the attR sequence, and we could show that ICEclc will insert with different frequencies in such traps producing brightly fluorescent cells. Starting from clonal primary transconjugants we demonstrate that ICEclc is excising and reintegrating at detectable frequencies, even in the absence of recipient. Recombination site analysis provided evidence to explain the characteristics of a larger number of genomic island insertions observed in a variety of strains, including Bordetella petri, Pseudomonas aeruginosa and Burkholderia.


Strain-to-strain genomic comparisons have increasingly underscored that bacterial genome evolution and rapid adaptation occur via acquisition or loss of large discrete DNA fragments (> 10 kb). Such ‘regions of genome plasticity’ (Mathee et al., 2008) or genomic islands (GEIs), as they are more frequently named, form a large group of evolutionary and mechanistically very diverse DNA elements (Juhas et al., 2009). Among others, GEIs comprise Integrative and conjugative elements (or ICEs), conjugative transposons and phage-like elements; mobile DNA elements capable of interspecies gene transfer. For a large number of GEIs, however, no clear functional mobility has been shown, which may point to a state of evolutionary regression or to a lack of knowledge on potentially other mobility mechanisms.

Genomic islands can contribute a wide variety of different functions to the host bacterium with potential adaptive benefits, such as virulence, nutrient or metal scavenging factors (Larbig et al., 2002), antibiotic resistance determinants (Beaber et al., 2002a; Mohd-Zain et al., 2004), plant symbiosis factors (Sullivan et al., 2002) or specific metabolic enzymes (Gaillard et al., 2006). Several GEI operate sophisticated control mechanisms for self-transfer and host-cell maintenance, and evidence has been obtained that GEI may have evolved ways to minimize fitness cost on the host (Gaillard et al., 2008).

Of obvious interest among GEIs are those elements, which are still self-transferable, because they can help us to understand the general mechanisms of GEI behaviour. A number of ICE models have been studied hereto in different bacteria, among which are SXT in Vibrio cholerae (Beaber et al., 2002b), ICEBs1 in Bacillus subtilis (Auchtung et al., 2005), PAPI-1 in Pseudomonas aeruginosa (Qiu et al., 2006), ICEHin1056 in Haemophilus influenza (Mohd-Zain et al., 2004). ICEs, as their name implies, normally reside in one or more locations in the bacterial host chromosome, but can excise to form a closed circular molecule. Both ends from the integrated element fuse together in the circular form (Fig. 1A), and it is thought that the circular form can conjugate in a single-strand DNA rolling circle mode via a type IV secretion system similar as some conjugative plasmids (Juhas et al., 2007). The presence of excised molecules has been demonstrated by specific PCR or Southern hybridization for a number of GEI, such as SXT (Burrus and Waldor, 2003), the high-pathogenicity island in Yersinia enterolitica (Lesic et al., 2004) and in Escherichia coli strain ECOR31 (Schubert et al., 2004), PAPI-1 in P. aeruginosa strain PA-14 (Qiu et al., 2006; Mathee et al., 2008), ICEMlSymR7A in Mesorhizobium loti R7A (Ramsay et al., 2006), or ICEBs1 in B. subtilis (Auchtung et al., 2005). GEI excision has been shown to require the activity of the integrase and an additional excisionase or recombination directionality factor to facilitate the process (Nunes-Düby et al., 1998; Burrus and Waldor, 2003; Lesic et al., 2004; Ramsay et al., 2006). Despite the fact that GEI excision has been detected numerous times and is supposed to be the prerequisite for successful self-transfer, actually very little is known about the fate of such molecules in the cell in the absence of a suitable recipient. Evidence was obtained from P. aeruginosa clone K populations that the element pKLK106 rapidly exchanged between two att sites formed by tRNALys genes (Kiewitz et al., 2000).

Figure 1.

Schematic representation of the organization of tRNAGly-3, -4 and -5 in P. putida UWC1 and the resulting structures upon single or double integration of ICEclcB13. Positions of tRNAGly-3, -4 and -5 bordered by two XmaI sites at nucleotides 2101654 and 2103904 are based on the genome sequence of P. putida KT2440 (Accession No. AE015451). Thick lines correspond to ICEclc-DNA, thin lines to P. putida chromosome, dotted lines represent the same but not drawn to scale. Open arrows comprise the location of tRNA genes, the intB13 integrase (only partially) or the two tRNAGlu genes PP_t31 and PP_t33. Black triangles correspond to the 18 bp recombination sequences between ICEclc and the 3′-end of tRNAGlys. Relevant distances and sizes of expected restriction or PCR fragments are indicated. Names near small arrows point to primers in Table S2.
A. Left-right end junction in the ICEclc circular form.
B. Region of the three tRNAGly genes.
C. Situation of ICEclcB13 integrated in tRNAGly-3.
D. Integration in tRNAGly-4.
E. Double integration in tRNAGly-3 and -4.

During transfer into a new host cell, the single-stranded ICE-DNA is assumed to undergo complementary strand synthesis to a double-stranded DNA and integrate into the chromosomal target. This process of site-specific integration is catalysed by an integrase, the gene for which is usually encoded on the ICE itself. The role of the integrase and target site in ICE site-specific recombination process have been particularized from a number of studies (Hochhut and Waldor, 1999; Lee et al., 2007). Interestingly, because most ICE do not seem to replicate autonomously in the host cell, their successful establishment is dependent on efficient integration into the host's chromosome. An exception to this are the pKLK106 and pKLC102 episomal elements of P. aeruginosa, which can form up to 30 copies in the cell (Kiewitz et al., 2000; Klockgether et al., 2007). The integration reaction could in principle also drive reintegration of the excised ICE molecule in the donor cell, but not much is currently known about the frequency and types of intracellular reverse integration reactions.

In our studies we have been using the clc element (ICEclcB13), a GEI first discovered in Pseudomonas knackmussii B13 (Ravatn et al., 1998a). ICEclcB13 can provide to its host the capacity to metabolize 3- and 4-chlorocatechol, and 2-aminophenol (although strain B13 itself can not grow on 2-aminophenol) (Gaillard et al., 2006). The element is 103 kb in size and two copies reside in the B13 chromosome (Ravatn et al., 1998a). Functional and DNA sequence comparisons have shown ICEclcB13 to be a good model for a large group of GEIs found in Beta- and Gammaproteobacteria, including P. aeruginosa, Xylella fastidiosa, Burkholderia xenovorans, Ralstonia metallidurans, H. influenza (Mohd-Zain et al., 2004; Gaillard et al., 2006; Juhas et al., 2007), Bordetella petri (Gross et al., 2008) and Herminiimonas arsenicoxydans (Muller et al., 2007). Homologies among this group are mainly restricted to a region of ∼40 kb, which encodes a type IV secretion system necessary for self-transfer in the case of ICEHin1056 (Mohd-Zain et al., 2004; Juhas et al., 2007). ICEclcB13 can transfer itself to other recipient bacteria with frequencies of up to 5% per donor (Sentchilo et al., 2003a). The integration site for ICEclcB13 (attB) is formed by an 18 bp sequence at the 3′-end of the gene for glycine tRNA (tRNAGly), and is identical to a region on the excised ICEclcB13 itself (attP, Table 1). Integration is mediated by the P4-family tyrosine recombinase-type IntB13 integrase, the gene for which is located 202 bp downstream of the integration site (Ravatn et al., 1998b). No gene for recombination directionality factor/excisionase has so far been found on ICEclc.

Table 1.  Proportion of supergreen cells among transconjugants of C. necator Pint-egfp as function of copy number and position of ICEclc type elements.
10 generations40 generations
  • a. 

    Proportion as number of supergreen cells per million of total in culture after 10 or 40 generations of growth, as counted per flow cytometry.

  • b. 

    Values indicate the average from independent biological triplicates; calculated standard deviation from the average is given within brackets.

C. necatorICEclcJS705tRNAGly-33.7 (±3.2)b3.5 (±6.0)6.7 (±5.8)120 (±30)
C. necatorICEclcJS705tRNAGly-52.0 (±3.5)7.1 (±6.2)1043 (±50)
C. necatorICEclcB132 + 3< 1< 13.3 (±5.8)3000 (±2420)
C. necatorICEclcB133 + 55.3 (±5.2)5.8 (±9.9)3.3 (±5.8)160 (±80)

Here we focused on the dynamics of excision and intracellular reintegration of ICEclcB13 from P. knackmussii sp. strain B13 and of an element very similar to it named ICEclcJS705, which has been found in Ralstonia sp. strain JS705 (Müller et al., 2003). We examined the different possible integration sites for the ICEclcs in two non-cognate host strains Pseudomonas putida UWC1 and Cupriavidus necator JMP289, and studied the excision and reintegration dynamics with the help of a conditional enhanced green fluorescent protein (EGFP)-producing trap. The results demonstrate that depending on the host strain and growth conditions, ICEclc is undergoing lively intracellular excision and reintegration, occupying various different integration sites or – in seldom cases – producing multiple copies of itself in primary ‘clonal’ cultures. Importantly, the observed ICEclc target site recombination products in P. putida and C. necator (i.e. attR and attL) are very similar to sequence motifs flanking integrated ICEs in multiple bacterial genomes, suggesting that they behave(d) mechanistically similar to ICEclc.


ICEclc integrates into multiple tRNAGly targets

To study target site preferences of ICEclcB13, the element was transferred in filter matings from P. knackmussii strain B13 to P. putida UWC1 and C. necator JMP289. At least 12 individual transconjugant colonies of each mating were purified and the position of ICEclcB13 was examined by PCR. Primers were designed to detect all possible junctions between ICEclcB13 and the potential tRNAGly targets, i.e. six tRNAGlys in P. putida UWC1 and five in C. necator JMP289 (Tables S1 and S2). Four distinct amplification patterns were observed among 16 P. putida UWC1-ICEclcB13 transconjugants (Figs 1B–E and 2A–E, Table S3). Concentrating on the most abundant PCR products (Fig. 2A and D), the patterns suggested integration either into tRNAGly-3 (three hits), -4 (seven hits), -5 (two hits) or -6 (four hits). No insertions were detected in tRNAGly-1 or -2, not even among 50 further transconjugant colonies tested (Table S4). Southern hybridization on total DNA of one of each representative clone against a probe targeting the right-end of ICEclcB13 indicated that a single main ICEclcB13 insertion had taken place in each transconjugant clone (Fig. 2F, Table S5). Maximum specific growth rates on 3-chlorobenzoate (3CBA) of P. putida UWC1 cultures with a single main integration type of ICEclcB13 in either tRNAGly-3, -4 or -5 were not significantly different in a two-tailed, homoscedastic t-test (> 0.05, μmax 0.24 ± 0.002 h−1, 0.23 ± 0.008 h−1, 0.23 ± 0.003 h−1 respectively), but were significantly different between strains with integrations in tRNAGly-4 or -5, and tRNAGly-6 (< 0.05, μmax P. putida UWC1 tRNAGly-6::ICEclcB13 0.25 ± 0.007 h−1).

Figure 2.

ICEclcB13 insertions in P. putida UWC1 transconjugants.
A. PCR amplification using primers for integration in tRNAGly-3.
B. As A, but for tRNAGly-4.
C. As A, but for tRNAGly-5.
D. As A, but for tRNAGly-6.
E. Amplification of the junction of the ICEclcB13 circular form.
F. Southern hybridization of four representative clones each carrying a single ICEclcB13 copy primarily in one integration site. Lanes 9–24, P. putida UWC1 (ICEclcB13); lane U, P. putida UWC1 (negative control); lane B, P. knackmussii B13; lane (–), water only; lane M, 100 bp DNA ladder (MassRuler, Fermentas). Primer pairs and expected amplicon sizes are shown on the right (Table S3). As explained in the text, each clonal population produced amplicons with more than one pair of target-specific primer. Note also that in some PCRs for the circular form (E) a second PCR product of ∼400 bp is produced (marked by white arrowhead) with a size expected for tandem insertion into tRNAGly-3 and -4.

In C. necator JMP289, all insertions were mapped to tRNAGly-2, -3 and -5 with equal occurrence. No insertions were found in tRNAGly-1 or -4 (among 23 clones tested, not shown). Taken together this analysis suggested that ICEclcB13 can target several alternative integration sites in the same genome, and, second, that not all tRNAGlys are used with the same integration efficiency.

Interestingly, despite the fact that Southern hybridization had shown one major insertion type, most P. putida UWC1-ICEclcB13 clones actually simultaneously produced several amplification products when targeting possible ICEclc junction sites (Figs 1 and 2, Table S3). For example, clone 12 produced an abundant PCR-product corresponding to the junction between tRNAGly-3 and ICEclcB13 (Fig. 2A), but also PCR products specific for junctions with tRNAGly-4, -5 and -6 (Fig. 2B–D). Purified DNA from all transconjugant cultures grown on 3CBA also amplified a 166 bp product representative for the excised ICEclcB13 molecule and in some cases even an amplicon slightly larger than 400 bp (Figs 1A and 2E). Such a larger amplicon may be the result of amplification of junctions from two neighbouring copies of ICEclcB13 inserted simultaneously into tRNAGly-3 and -4 (expected size 413 bp, Figs 1E and 2D, Table S3). These data therefore illustrated that ICEclcB13 is not stable in clonal populations, but can excise at low frequencies and reinsert into a different chromosomal target.

ICEclcB13 produces multiple insertions in C. necator when transconjugants are selected on 3CBA

To test whether the growth medium used for transconjugant selection can influence the final ICEclcB13 copy number we compared insertion site occupancy in transconjugants obtained under three different conditions. Because ICEclc transfer to recipients is usually selected for by growth on 3CBA (as a result of metabolic complementation by the clc genes on ICEclc) and a counterselection against the donor by antibiotic resistance of the recipient, we used slightly modified ICEclc elements for non-3CBA selection. One of these was a derivative of strain B13 in which ICEclcB13 was tagged with a gene for spectinomycin (Spc) resistance in between orf48922 and orf50240; the other was Ralstonia sp. strain JS705, which carries the related ICEclcJS705 element that enables growth on 3CBA and monochlorobenzene (MCB, Müller et al., 2003). P. putida transconjugants in all cases showed the presence of a single major ICEclc insertion, independent of the selection conditions (not shown). The situation was different for C. necator, however (Table S6). Almost all C. necator transconjugants selected on fructose plus Spc or on MCB carried a single insertion of ICEclcB13-Spc or ICEclcJS705, except for one (Fig. 3B, lane 21). On the contrary, all transconjugants selected in the presence of 3CBA had multiple copies of ICEclcB13, with a maximum of four simultaneous integrations (Fig. 3B, lane 16). As before, all ICEclcB13 insertions mapped into tRNAGly-2, -3 and -5, and none into tRNAGly-1 and -4. Transconjugants with ICEclcJS705 had never inserted in tRNAGly-2 either.

Figure 3.

Insertions and copy numbers of ICEclc in C. necator depending on transconjugant selection conditions.
A. Southern hybridizations of XmaI-digested total DNAs of transconjugants of C. necator (Pint-egfp) with ICEclcB13-Spc selected on fructose medium with streptomycin, probed against a digoxigenin-labelled DNA fragment at the outermost right side of ICEclc.
B. As A, for transconjugants with ICEclcB13 obtained by selection on 3-chlorobenzoate (3CBA), or with ICEclcJS705 obtained by selection on monochlorobenzene. Supergreen colonies are marked with asterisks. The inferred number of ICEclc copies in each strain is indicated as a number below the respective lane and panel. Bands are explained in between both panels. Lanes: 1–22, different transconjugant colonies; C, C. necator JMP289 (Pint-egfp) total DNA. Compare band sizes to theoretical digest in Table S7. Note that the C. necator JMP289 recipient has two copies of the Pint-egfp mini-Tn5.

Integrations into an artificial target

In order to more rigorously detail and quantify ICEclc excision and integration frequencies, and exclude possible PCR artefacts, we developed an artificial integration target with a conditional EGFP reporter. This target consisted of the attR sequence of ICEclcB13 in P. putida (containing a 60 bp 3′-part of the tRNAGly plus the Pint-promoter) transcriptionally fused to the egfp gene (Fig. 4A). Because Pint is a very weak and bistable promoter (Sentchilo et al., 2003b; Minoia et al., 2008), EGFP expression in the cells is very low. However, when ICEclcB13 integrates into attR, the outward facing constitutive and strong promoter Pcirc is placed in front of the egfp gene, overruling Pint (Fig. 4B). Matings with P. knackmussii B13 as donor for ICEclcB13 or ICEclcB13-Spc, and either P. putida UWC1 or C. necator JMP289 as recipient (both equipped with the Pint-egfp fusion) indeed produced two types of transconjugant colonies: those with weak and those with high egfp expression (Fig. 5A). PCR analysis of the weakly fluorescent transconjugants of both P. putida and C. necator origin, produced a 478 bp large PCR product with primers targeting the integrity of the Pint-egfp construct (Figs 4A and 5C). Southern hybridization patterns of weakly fluorescent C. necator Pint-egfp ICEclc transconjugants precisely matched in silico predictions for integrations having occurred into either tRNAGly-2, -3 or -5 (Fig. 3, Table S7). On the contrary, Southern hybridization of XmaI-digested total DNA isolated from highly fluorescent C. necator transconjugants against a right-end ICEclcB13 probe revealed bands matching the predicted length (5.6 kb) of a fusion between the left-end of the GEI and egfp (Fig. 3, lanes marked with asterisks, Table S7). A specific PCR product was amplified from such DNA encompassing the junction between the ICEclc left end and egfp (Fig. 4B). High egfp expression was therefore the result of ICEclcB13 integration into the artificial target, as explained above (Fig. 4), and the emergence of ‘supergreen cells’ demonstrated that the Pint-egfp fusion served as artificial target and easily visible conditional trap for ICEclc.

Figure 4.

Principle of the gene trap for ICEclc.
A. Artificial integration target Pint-egfp is comprised of the 3′ most 60 bp of the tRNAGly (including the 18 bp recombination site) joined to the right end of ICEclcB13 until the start of the intB13 gene, fused to a promoterless egfp (Sentchilo et al., 2003b). The ΔPint -egfp trap essentially lacks the upstream part of intB13 until the 18 bp site. Ends of the mini-Tn5 transposon used for construct delivery are depicted by black vertical bars. The ICEclcB13 integration site is marked with a black triangle. Primers used for analysis of transconjugants are indicated; names correspond to Table S2.
B. Schematic situation of ICEclcB13 being integrated in the Pint-egfp construct. Expression of egfp via the Pint-promoter is now overruled by the strong constitutive promoter Pcirc. Note that upon integration the 478-bp-long Ter_fw—gfp_rev PCR amplicon disappears and two instead of one XmaI fragment will hybridize to a probe against the ICEclcB13 right end.

Figure 5.

ICEclcB13 insertions in transconjugants of C. necator JMP289 or P. putida UWC1 containing the Pint-egfp construct.
A. Epifluorescent images of transconjugant colonies as indicated on the top of the panels, grown on selective plates. Note the lower frequency of fluorescent colony appearance for P. putida Pint-egfp as host.
B. Detail of fluorescent colonies showing sectors (dark within bright, or bright appearing in dark), likely due to internal ICEclcB13-Spc movement during colony growth.
C. PCR analysis of DNA from non-fluorescent and highly fluorescent clonal cultures of P. putida (Pint-egfp)-ICEclcB13 transconjugants grown on 3CBA. Upper panel: Amplification of intact Pint-egfp cassette. Middle: Amplification of junction between ICEclcB13 and the construct (see Fig. 4 for schematic outline). Lower: Amplification of the circular form of ICEclcB13. Controls: J, P. putida UWC1 (Pint-egfp); B, P. knackmussii B13; U, P. putida UWC1.

Three highly fluorescent (‘supergreens’) and two weakly fluorescent clones of P. putida in which ICEclcB13 had integrated were purified on minimal medium agar plates with 5 mM fructose. DNAs from pure clonal cultures were then isolated, purified and screened by PCR, cloning and sequencing for the presence and nature of, first, the intact attR-egfp integration site (indicative for target site repair), and second, the ICEclcB13 left-to-right end junction (indicative for excision). Cultures of all five clones probed positive in the PCR for the left-to-right end junction (Fig. 5C), which is indicative for the presence of excised ICEclcB13 among cells in the culture. Second, DNA from supergreen clones amplified detectable amounts of an intact Pint-egfp fragment (Fig. 5C). Subsequent cloning and sequencing of this amplified fragment gave an exact match to the original inserted Pint-egfp gene cassette. The sequence of the integrated junction was in perfect agreement to the in silico predicted integration of ICEclcB13 into attR of Pint-egfp (data not shown). The same results were obtained when using supergreen transconjugants of ICEclcJS705 in C. necator Pint-egfp. Images of primary transconjugant colonies in fluorescent light showed development of bright and dark sectors (Fig. 5A and B), suggesting that ICEclc is excising in clonal populations that started from a single ICE insertion. Bright colonies with dark sectors thus represent ICEclc excision from the Pint-egfp target site into a tRNAGly, and dark colonies with bright sectors the reverse (excision from tRNAGly and insertion into attR of Pint-egfp). Furthermore, these results demonstrated that ICEclc excision and reintegration are conservative (i.e. no insertions or deletions were introduced into the target site after excision).

Intracellular ICEclc excision and reintegration dynamics

Interestingly, we observed that the proportion of supergreen colonies among transconjugants was strongly dependent on the recipient (Fig. 5A). Integration of ICEclcB13 in C. necator JMP289 Pint-egfp produced 19.7% (± 3.5%) and in P. putida UWC1 Pint-egfp 0.5% (± 0.3%) of supergreen colonies among the rest. From these frequencies and given that in C. necator three tRNAGly genes (no. -2, -3 and -5) were used as integration site, plus two copies of the Pint-egfp insertion (leading to a total of five potential integration sites), we conclude that in C. necator the choice between the tRNAGly and Pint-egfp targets is weakly biased towards the genes for tRNAGly (expected frequency of insertion into Pint-egfp: two in five, 40%; observed: one in five, 19.7%). In P. putida, on the other hand, a strong preference exists for (four of six) tRNAGly integration sites compared with Pint-egfp (expected frequency of insertion into Pint-egfp in UWC1: one in five, 20%; observed: 0.5%).

To ensure that the position of the Pint-egfp in the recipient's chromosome was not the major reason for the low insertion frequency in P. putida, the experiment was repeated with a second clone of P. putida UWC1 Pint-egfp with a different insertion position. The frequency of fluorescent transconjugant formation remained virtually unchanged (0.4% ± 0.1%), therefore suggesting little or no position effect.

We then examined whether the sequence downstream of the 18 bp recombination site would be important for ICEclc insertion. For this purpose, one further conditional trap was designed, in which the region containing Pint was deleted leaving only 4 bp downstream of the 18 bp recombination site (named ΔPint-egfp). To our surprise, the ΔPint-egfp trap was a much better target for ICEclc in P. putida than Pint-egfp or either of the tRNAGlys. Up to 40% of all transconjugant colonies appeared brightly fluorescent (not shown).

To test whether ICEclc copy number and chromosomal location influence the frequency of intracellular excision reintegration, we measured the relative proportions of supergreen cells by flow cytometry in four (non- fluorescent) C. necator JMP289 Pint-egfp transconjugant cultures: tRNAGly-3::ICEclcJS705 (single copy of ICEclcJS705), tRNAGly-5::ICEclcJS705(single copy of ICEclcJS705), tRNAGly-2::ICEclcB13-tRNAGly-3::ICEclcB13 (double copy of ICEclcB13 in tRNAGlys -2 and -3) and tRNAGly-3::ICEclcB13-tRNAGly-5::ICEclcB13 (double copy of ICEclcB13 in tRNAGlys -3 and -5). Cultures were grown for a maximum of 40 generations on either 5 mM 3CBA or 5 mM fructose. After the first passage (96 h into stationary phase, or 10 generations) the proportion of bright fluorescent green cells as a result of ICEclc integration into the conditional trap in all cultures was low (between 1 × 10−7 and 6 × 10−7, Table 1). This proportion did not significantly increase in any of the transconjugants grown on fructose, even after 40 generations (four passages). On 3CBA, however, the proportion of brightly green cells increased by as much as three orders of magnitude (Table 1). C. necator cultures with two inserted copies of ICEclcB13 in tandem tRNAGly genes produced the highest number of supergreen cells over time (3 ± 2 × 10−4, Table 1). This showed that ICEclc copy number per se did not, but carbon source and the insertion position strongly influenced the frequency of excision and reintegration.

In vivo target site recombination

Finally, we investigated the nature of the in vivo target site recombination by ICEclc. Because of a two base-pair difference between the 18 bp attB sequences in C. necator compared with P. putida UWC1 and the attP sequence of ICEclcB13 (Table 2), site-specific recombination might produce different right and left ends. We thus amplified and sequenced the left and right ends, and the attP of the excised circular form of all C. necator transconjugants with a single copy ICEclcB13-Spc integration into tRNAGly-2, -3 or -5. All transconjugants produced the same sequences (Table 2). Interestingly, the right-end 18 bp sequences appeared to be mixtures of attB and attP (A at position 2, and T at position 8), whereas all attL sequences were indistinguishable from attP (i.e. T at position 2 and 8). All excised ICEclcB13-Spc molecules carried the original attP sequence (2T-8T), but the restored chromosomal target (attB′) maintained the hybrid right-end signature (i.e. 2A-8T).

Table 2.  Sequences of the 18 bp recombination sites on ICEclc and in C. necator transconjugants.
SiteOriginSequence 5′-3′a
  • a. 

    Sequences from three independent transconjugants were identical for the sites reported here.

  • b. 

    18 bp sequences of tRNAGly-2, -3 and -5 in C. necator are identical.



In this work we showed that the elements ICEclcB13 and ICEclcJS705 can target different tRNAGly genes in both P. putida and C. necator genomes. We observed that both ICEclc preferentially inserted into tRNAGly genes of 76 bp length and with the GCC anticodon sequence, but not into those of 74–75 bp with anticodons TCC or CCC. In most transconjugants and for both ICEclc elements, single primary insertions were detected, except when C. necator transconjugants were selected for growth on 3CBA. During subsequent culturing (and even during multiplication of the primary transconjugant cell to a colony, Fig. 5B), however, clonal populations in a few generations' time developed into a heterogenous mixture in which ICEclc had moved at low frequency from its original insertion site to another. Intracellular excision and reintegration had been previously observed for the episomal ICE pKLK106 from P. aeruginosa, which can occupy either of two tRNALys gene targets in the cell (Kiewitz et al., 2000). As pKLK106 can replicate in the cell up to 30 copies (Kiewitz et al., 2000), this may explain why the reintegration frequency is much higher than observed with ICEclc. The related ICEs from P. aeruginosa PAPI-1 and pKLC102 were also shown to excise in clonal populations, but no reintegration was observed (Qiu et al., 2006; Klockgether et al., 2007; Mathee et al., 2008). By contrast, no excision has yet been detected of the P. aeruginosa elements PAGI-2 and PAGI-3, which are strongly related to ICEclc (Klockgether et al., 2007). Results from Southern analysis on P. putida UWC1 indicated that the majority of cells in the culture carried a single ICEclc copy, suggesting that ICEclc does not replicate upon excision, but leaves its target site and reintegrates into the same or another. Only in one case of P. putida was clear evidence obtained from PCR for two neighbouring ICEclc copies (Fig. 2E, lane 10). Further sequencing data illustrated that excision does not give rise to deletions or sequence duplications in the liberated target site.

Detailed analysis of recombination sites in transconjugants of C. necator, a strain, which has two mismatches in the 3′ 18 bp target sequence of three of the tRNAGly sites compared with those of P. putida and with ICEclc attP, demonstrated that modification of the target site can occur, but is dependent on the target sequence. The corresponding 18 bp sequences of original targets (attP of ICEclc, attL and attR), and of the excised products (attP′ and attB′) provided evidence that the recombination reaction should proceed via a staggered-cut mechanism and strand-swapping reminiscent of the phage lambda system (Fig. S1) (Azaro and Landy, 2002). Due to the position of the nucleotide changes in the 18 bp recombination target we can conclude that the template must be cut in between position 2 and 9 of the 18 bp sequence (notation of attR), with a 7 bp swapping region in between like lambda integrase (Azaro and Landy, 2002) or IntDOT (Malanowska et al., 2007). We can draw a number of conclusions as to what defines a good target site for ICEclc and highly related ICEs, from the different observed target site occupancies between tRNAGly genes and the artificial target sites (gene traps) in P. putida and C. necator. The first determinant for efficient integration must be the nucleotide sequence conservation (order and bp identity) between the target 18 bp region and attP. For example, all the successfully integrated 18 bp regions in tRNAGly-3, -4, -5 and -6 of P. putida UWC1 matched ICEclc's attP perfectly, but sequences of tRNAGly-1 and -2 carried up to four mismatches and were not detectably used by ICEclc (Table S8). The tRNAGly-2, -3 and -5 in C. necator carried two substitutions (A2T and C8T) compared with attP of ICEclc, but were still used as integration site. However, the two non-targeted tRNAGly in C. necator carried four and five mismatches (Table S8). ICEclc integration site dependency would thus be in agreement with the model of phage lambda integration, where the sequence similarity between recombining attB and attP was shown to be crucial for productive recombination (Nunes-Duby et al., 1997), although a certain degree of dissimilarity decreased but not abolished integration in a site (Rutkai et al., 2006). The site specificity seen for ICEclc in vivo is also similar to that of other tyrosine recombinases, which all tolerate a certain number of mismatches within the recombination core region (Burrus and Waldor, 2003; Rajeev et al., 2007; Doublet et al., 2008), but not as non-specific as the integrases of CTn916 or CTnBST (Scott and Churchward, 1995; Song et al., 2007).

The second determinant for efficient integration is the attB sequence downstream of the 18 bp recombination target. This we conclude from the observations that ICEclc integration frequencies into an artificial target (Pint-egfp) in P. putida were at least an order of magnitude lower (0.5% of all insertions) than those into the most optimal tRNAGly gene targets, despite the fact that both types of insertion targets shared 47 bp among them (including the 18 bp recombination target). This has been observed for other minimal attB site determinations (Song et al., 2007), and might be due to the presence of binding sites for auxiliary proteins in the intasome to the DNA flanking the actual recombination site (Biswas et al., 2005). Interestingly, an artificial trap which did not carry any sequence downstream of the 18 bp region (ΔPint -egfp) served as even more favourable integration site than any of the tRNAGly targets in P. putida. This is in contrast to observations with tyrosine recombinases like IntBST from CTnBST in Bacteroides, where deletions of the attB region below 270 bp caused a strong decrease in integration frequency (Song et al., 2007). This suggests that the region downstream of the 18 bp recombination site for ICEclc reduces rather than increases successful integration rates and that thus, perhaps, ICEclc integration does not require further host auxiliary proteins. Results from comparisons of integrations in C. necator Pint-egfp and P. putida Pint-egfp showed that a less optimal downstream sequence can be compensated by a more optimal 18 bp recombination target.

A screening of GEI insertion sites in bacterial genomes with integrases displaying > 75% amino acid identity to IntB13 of ICEclc revealed that all of those also located downstream of the 76 bp long, GCC-type tRNAGlys. By contrast, TCC- or CCC-type tRNAGlys were either ‘free’ or occupied by more distantly related integrases (Table S8). We therefore conclude that this type of elements related to ICEclc have a preference for GCC-type tRNAGly targets. Comparison of putative attL and attR sequences of those GEIs defined on the basis of direct nucleotide repeats on either side, provided further evidence that target site hybrids may arise as a result of mispaired attP of the incoming element and the attB of the host (Table 3). For example, the B. petri DSM12804 genome has a GEI with an integrase gene 100% identical to that of ICEclc (Bpet_1545). The proposed attL and attR sequences differ in the same signature type as seen with ICEclc in C. necator. The history cannot be completely traced back, however, because the original 18 bp sequence of the target can only be inferred indirectly, as all three GCC-type tRNAGly in B. petri DSM12804 are now occupied by GEIs. Also other genome sequences bear traces for events observed here experimentally for ICEclc, albeit being difficult to reconstruct attachment sites and GEI extremities exactly. For example, in P. aeruginosa isolate PA2192 two GEIs related to ICEclcB13 (PAGI-2 and an element named ‘Dit’) were found, which, however, had integrated into the tRNAGly in a head-to-tail arrangement (Mathee et al., 2008). Because each ICE is flanked by a pair of 18 bp repeats, with the one in the middle being shared, the result is a hybrid sequence.

Table 3.  Comparison of potential ICEclc-type recombination sites in a variety of bacterial genomes.
Organism, GenBank accession No.Integrase5′-adjacent tRNAGlyIntegration site (attB)Right end (attR)Left end (attL)Left end co-ordinate
  • a. 

    Bpet_1545 and Bxe_1097 integrase genes are indistinguishable from intB13 on ICEclcB13.

  • b. 

    Originally B. xenovorans was likely to have two identical tRNAGlys [Bxe_AR0056 and AR0057 (C-C type, anticodon CCG, 76 bp)], adjacent to pgsA[Bxe_A1096]) and orthologous to the B. phytofirmans PsJN (CP001052) loci Bphyt_R0038 and Bphyt_R0039, which are devoid of insertions. Alternatively, like those in B. multivorans ATCC17616 (AP009385) it could have an A–C signature (orthologous tRNAGlys BMULJ_01082, _01083 and _01084).

  • c. 

    In B. petri all tRNAGlys carry insertions of ICEs related to ICEclcB13, thus determination of the original 18 bp sequence at attB is problematic. In the related strains B. bronchioseptica str. RB50 (BX470250) and B. parapertussis str. 12822 (BX470249) orthologous tRNAGly have an A–T signature, whereas in B. avium strain 197N (AM167904) it is A-C.

  • d. 

    In X. campestris pv vesicatoria str.85–10 (AM039952) tRNAGlys are likely to have T–T signature as judged from orthologous tRNAGlys in X. campestris pv. Citrii str. 306 (AE008923) and X. campestris pv. campestris ATCC33913 (AE008922).

P. knackmussii B13, attP ICEclcB13, AJ617740   GTCTCGTTTCCCGCTCCA 
C. necator = R. eutropha JMP134, CP000090 and CP000091IntB13Reut_AR0037, AR0038 and BR0068GACTCGTCTCCCGCTCCAGACTCGTTTCCCGCTCCAGTCTCGTTTCCCGCTCCA
Acidovorax sp. JS42, CP000539Ajs_1340Ajs_R0017 GACCGATTTCCCGCTCCAGTCTCGTTTCCCGCTCCA1516424
H. arsenicoxydans, CU207211HEAR2054HEARtRNA30 GACTCTTCTCCCGCTCCANot definedNot defined

The fact that in some cases two or more ICEclc copies were found in clonal populations suggests that ICEclc may replicate at low frequencies, although also other mechanisms may be responsible for the appearance of multiple inserted copies, as we will argue below. By PCR we found indications for neighbouring ICEclc repeats integrated in P. putida tRNAGly targets -3 and -4, although Southern showed overall a single ICEclc copy in the DNA. Even more striking were multiple ICEclc copies occurring in transconjugants of C. necator selected for growth on 3CBA. These results suggest that at low frequencies ICEclc may produce multiple insertions, which we can observe by PCR or which can become selected under conditions during which only transconjugants with an elevated ICEclc copy number can grow. The copy number effect on growth of C. necator with 3CBA has been observed before, but with the tfd genes, which also encode a pathway for chlorocatechol degradation (Trefault et al., 2002; Pérez-Pantoja et al., 2003). Apparently, an increased copy number of the clc (or tfd) genes is necessary to achieve the enzyme levels needed for productive growth with 3- and 4-chlorocatechol generated from 3CBA. Three mechanisms could potentially explain the arisal of multiple ICE copies: (i) transfer at low frequencies of multiple copies during a single conjugation event; (ii) arisal of multiple copies as a result of (limited) ICE replication in the host; and (iii) secondary cell-to-cell transfer within culture due to lack of immunity. Perhaps at low frequencies, concatenated copies are produced from the single-stranded incoming ICE molecule by rolling-circle replication, which can insert at independent locations or as a tandem array (whereby the attR may be serving as a new target site for insertion). In fact, our results with the artificial attachment site Pint-egfp demonstrated that attR is a functional insertion target itself. This may explain previous observations during which tandem arrays of up to 10 ICEclc copies were observed in P. putida transconjugants selected for growth on MCB (Ravatn et al., 1998a). Also others have demonstrated tandems of integrative elements employing the same attachment site, such as for the Salmonella mobilizable GEI SGI1 (Doublet et al., 2008), for Vibrio cholerae SXT (Burrus and Waldor, 2004), for SXT and R391 (Hochhut et al., 2001, p. 736), and for the double-ICE tandem PAGI-2/Dit Island in P. aeruginosa isolate PA2192 (Mathee et al., 2008). Multiple independent inserted copies have also shown to be formed at low frequencies for the conjugative transposon CTn916 and related elements (Poyart et al., 1995). Even though the formation of tandems occurs at low frequencies we showed here that it can become selected during specific growth conditions, and may be the starting point for further recombinatorial changes of the GEI in question as in the Dit island (Mathee et al., 2008).

Interestingly, one particular arrangement (i.e. tRNAGly-2::ICEclcB13-tRNAGly-3::ICEclcB13) in C. necator resulted in a more than 10-fold higher observed frequency of excision and reintegration into the Pint-egfp trap than with singular ICEclc integrated copies. In this arrangement the left end constitutive promoter (Pcirc) of the upstream ICEclcB13 element may read-through into the integrase gene of the downstream ICEclcB13 copy, causing a higher intB13 expression and excision of this copy (Fig. 1E). It is interesting to note that ICEclc insertion into tRNAGly-3 in P. putida would in principle interrupt expression of other downstream tRNA genes, such as PP_t33, tRNAGly-4 or -5, when assuming that expression of those is driven by a single canonical σ70 promoter upstream of PP_t31 (TTGACG −17 bp– TAGAAT). However, the maximum specific growth rate of P. putida UWC1 with integrations of ICEclcB13 in either tRNAGly-gene -3, -4 or -5 was not significantly different. This suggests that the presence of Pcirc at the outer end of ICEclc (which has a −10/−35 sequence very similar to the promoter upstream of PP_t31) may ensure that expression of downstream located tRNA genes is taken care of, which is another example of the curious and intricate coevolution of ICE with their host.

Experimental procedures

Strains and plasmids

Escherichia coli DH5α (Gibco BRL, Life Technologies), E. coli HB101 (pRK2013) (Ditta et al., 1980), E. coli CC118 λpir and E. coli S17-1 were used as described previously (Sentchilo et al., 2003a). P. knackmussii strain B13 (Dorn et al., 1974; Stolz et al., 2007) is the original host of the clc element (here specified as ICEclcB13). Ralstonia sp. strain JS705 (van der Meer et al., 1998; Müller et al., 2003) was used as donor for the ICEclcJS705 element. C. necator JMP289 is a plasmid-free and rifampicin-resistant derivative of C. necator JMP134. P. putida UWC1 is a rifampicin-resistant derivative of P. putida KT2440. Luria–Bertani broth was routinely used for growing E. coli, Pseudomonas and Ralstonia strains. As a defined mineral medium (MM) the type 21C mineral medium (Gerhardt et al., 1981) was used, supplemented either with 5 or 10 mM 3-chlorobenzoate or with 10 mM fructose. When required 50 mg ml−1 of ampicillin, kanamycin, rifampicin, streptomycin, spectinomycin and/or 5 mg ml−1 tetracycline were added. Strains of Pseudomonas, Cupriavidus and Ralstonia were grown at 30°C; E. coli at 37°C.

ICEclc self-transfer assays

Self-transfer assays with donors containing ICEclcs were performed as described previously (Sentchilo et al., 2003a). As donors we used here P. knackmussii B13 (ICEclcB13) or (ICEclcB13-Spc), and Ralstonia sp. strain JS705 (ICEclcJS705). Recipients were P. putida UWC1 or C. necator JMP289 or one of the Pint-egfp or ΔPint-egfp tagged derivatives of those. Selection of transconjugants carrying ICEclcB13 was regularly performed by growth on MM plates with 5 mM 3CBA plus rifampicin (to counterselect against strain B13), whereas for ICEclcJS705 selection was performed by growth on MM plus rifampicin with chlorobenzene vapour. Selection for transconjugants carrying ICEclcB13-Spc was carried out on MM plates with rifampicin, streptomycin and 5 mM fructose as carbon source. In all self-transfer assays, donor and recipient alone were incubated under the same conditions to correct for the number of spontaneous mutants appearing on the selective plates.

EGFP reporter trap construction

The 265 bp Pint integrase promoter region encompassing the 3′ end of the tRNAGly gene (i.e. the attR region: positions 31–295 of the ICEclcB13 complete sequence, AJ617740) fused to the promoterless egfp gene on a mini-Tn5 delivery plasmid pCK218 [designated Pint-egfp (Sentchilo et al., 2003b)] was randomly inserted into the genomes of C. necator JMP289 and P. putida UWC1 via mobilization from E. coli CC118λpir as donor and transposition in a triparental filter mating according to Herrero et al. (1990). A shorter trap (ΔPint -egfp) was constructed in a similar way on a mini-Tn5 delivery plasmid and contained a 43 bp region comprising the 3′-end of the P. putida UWC1 tRNAGly-4 gene plus 4 bp downstream of it (AE015451, co-ordinates 2102991–2103041).

Correct insertion of the trap-egfp fusions and the absence of the plasmid backbone in transconjugants was verified by antibiotic resistance profiling and by using the PCR with construct-specific primers, Ter_fw and GFP_rev (Table S2). Clones with correct insertions were assigned C. necator JMP289 (Pint-egfp) or P. putida UWC1 (Pint-egfp), and were used as recipients for matings with Pseudomonas sp. strain B13 (ICEclcB13) or (ICEclcB13-Spc) and Ralstonia sp. strain JS705 (ICEclcJS705). For unknown reasons, C. necator JMP289 (Pint-egfp) harboured two mini-Tn5 insertions.

Screening for EGFP producing colonies was done on a Safe ImagerTM blue light transilluminator (Invitrogen). High-resolution images were recorded on a Leica MZ16 FA high-performance stereomicroscope (Leica Microsystems, Switzerland) equipped with the colour digital camera Leica DC300F and Plant fluorescence filter set (excitation wavelength 470/40 nm, emission 525/50 nm). To obtain a liquid culture for microscopy a single colony was inoculated into 5 ml of MM supplemented with 5 mM 3CBA or fructose in a 15 ml glass tube and shaken at 180 r.p.m. Typically, exponentially growing cultures were obtained within 16 h, and after between 24 and 48 h cultures reached stationary phase.

DNA techniques

DNA cloning in E. coli, PCR and DNA sequencing were performed according to established procedures. Total DNA purification from Pseudomonas and Cupriavidus was performed by the xanthogenate method (Tillett and Neilan, 2000). DNA templates for PCR were prepared from 106 to 107 bacterial cells washed in pure sterile water, incubated at 95°C for 10 min, vortexed and centrifuged at 15 000 g for 5 min. An ICEclcB13 right-end probe for Southern analysis was produced by PCR amplification of a cloned 212 bp ICEclcB13 DNA fragment (AJ617740, positions 84–296) upstream of the intB13 gene (Sentchilo et al., 2003b) and Digoxigenin-labeling (Roche).

To tag ICEclcB13 with a spectinomycin-resistance gene, we amplified the aadA gene under control of a constitutive promoter from pHP45omega by PCR (Prentki and Krisch, 1984). The aadA gene was ligated in between PCR-amplified orf48922 and orf50240 of ICEclcB13 and into the suicide plasmid pME3087 (TcR) (Minoia et al., 2008). The resulting plasmid was mobilized into P. knackmussii B13 via triparental mating using pRK2073 as helper. Single recombinants were selected for tetracycline resistance, purified and verified by PCR. Double recombinants in which the plasmid vector was excised and aadA integrated in between orf48922 and 50240 on ICEclcB13 were enriched in a procedure to screen for tetracycline sensitivity as described earlier (Minoia et al., 2008). Putative double recombinants were purified and verified for correctness of the integration by using PCR.

Identification of the integration sites in P. putida UWC1 and C. necator JMP289

Pseudomonas putida UWC1 is a derivative of P. putida KT2440 and has six tRNAGly copies (accession number: AE015451, Table S1). Three of those (tRNAGly-3, -4 and -5) are part of the same operon (Fig. 1), whereas the others (tRNAGly-1, -2 and -6) are separated. ICEclc integration sites were mapped by PCR using either its right- or left-end as one primer anchor and the different tRNAGly targets as the other (Table S3, Fig. 1). Because of physical proximity and co-orientation of tRNAGly-3, -4 and -5, forward primers Pgly3_fw, Pgly4_fw and Pgly5_fw can detect three, two or one junction to an inserted ICEclc respectively (Fig. 1C and D, Table S3). C. necatorJMP289 (accession numbers CP000090 and 000091) has five tRNAGly genes in its genome, of which tRNAGly-2 and -3 are in tandem. Similar primers as for UWC1 were developed that amplified the possible junctions between the respective tRNAGly and the right- or left-end of ICEclc (Tables S2 and S9). Integration of ICEclc into Pint-egfp or ΔPint-egfp was confirmed by PCR amplification with primers located on ICEclc and in the construct (Table S2, Fig. 5).

Quantification of intracellular mobility of ICEclc

Intracellular mobility of ICEclcB13 and ICEclcJS705 was determined on four clones of C. necator JMP289 (Pint-egfp). Two of those carried a single copy of ICEclcJS705 integrated to tRNAGly-3 or -5, and two others contained two copies: in tRNAGly-2 and -3 or tRNAGly-3 and -5. Exponentially growing cultures of each clone with fructose as sole carbon and energy source were diluted 1:1000 in triplicate fresh MM with 5 mM fructose or 5 mM CBA in a 24-well plate, and regrown until late stationary phase (for 96 h, ∼10 generations). This procedure was repeated four times to have ∼40 generations in total. Culture dilutions (to have ∼106 cells per ml) at the end of each cultivation cycle were sampled by flow cytometry on a FACS Calibur (Becton-Dickinson). Primary triggering was done on the FL1 channel to count the number of total and of supergreen EGFP-containing cells, which emerged as a consequence of integration of ICEclc into the Pint-egfp trap.


We thank M. Stojanov for his help in analysing recombination products. The work of M.M. and J.vdM. was supported by grants from the Swiss National Science Foundation (3100–67229, 3100A0-108199). R.M. was supported by a postdoctoral fellowship from the Japanese Society for the Promotion of Science for Research Abroad.