The characterization of the IS elements were realized within the dataset of Lxc DSMZ46306 genomic DNA. The dataset is comprised of 9,766 reads, of which 5,854 were derived from the shotgun library and 3,912 from sequencing BAC ends, sub-cloning of inserts and primer-walking. All sequences were assembled into 1,064 contigs accounting for 1,368,731 non-redundant bases, representing approximately 50% of the Lxc chromosome. Comparing all the sequences of Lxc with the complete genome of Lxx, we found that nearly 70% share more than 80% nucleotide sequence identity considering a continuous segment of at least 200 bases. Two other genomes of related subspecies Clavibacter michiganensis subsp. michiganensis and C. michiganensis subsp. sepedonicus that were recently sequenced 2627 have nearly 80% of nucleotide identity using the same criteria. These figures were derived from the Artemis Comparative Tool 28.

Among all the contigs, 70 shared sequence similarity to transposase-encoding genes. Fifty-six of these represented copies of five distinct IS elements that were characterized based on the criteria proposed (Table 1). The remaining 14 contigs correspond to uncharacterized IS elements, since no IR, site of insertion and structural limits were identified within the available sequences. They were classified only tentatively based on homology searches with Lxx genome, and were placed within IS families (Table 2): IS3 family (eleven elements); IS256 family (two elements); and IS481 family (one element). Elements of the same families were not characterized in Lxx as well 223. These uncharacterized transposases may represent degenerated forms of old insertions, however they may also be used as site for rearrangements within Lxc genome.

The transposases of characterized IS elements shared less than 78% amino acid sequence identity with transposases of Lxx elements (Additional file 1). To be considered as iso-forms, transposases of a given element must share more than 95% identical amino acids 11; therefore, Lxc and Lxx have different sets of IS elements. Despite of that, most of the elements found belong to the same families (Table 2), with the exception of IS110 (ISLxx2), which was not identified in our Lxc dataset.

In addition to the three IS elements previously sequenced for Lxc: IS1237 (GenBank ID: X75973); ISLxc1 (GenBank ID: EF437436); and ISLxc2 (GenBank ID: EF176596), we identified two new elements, named ISLxc3 (GenBank ID: EF421582) and ISLxc4 (GenBank ID: EF433175) (Table 1, Fig. 1 and Additional file 2). A detailed characterization of all these Lxc-IS elements, and a comparison with related elements in Lxx, is presented below.

The average nucleotide sequence identity between homologous fragments of Lxc and Lxx is 93%, as calculated on 80 kbp of 25 continuous Lxc-contigs, which were physically linked based on scaffold orientation. Comparing the adjacent regions of all 56 Lxc-IS element with the Lxx sequence, 44 elements could be mapped onto the Lxx genome (Additional file 3). All the flanking sequences of IS elements were submitted to GenBank and the accession number is available in the additional file 3. Four of these (occurrences 8, 14, 15 and 38, additional file 3), which are three copies of ISLxc2 and one copy of ISLxc3, have only one of the adjacent sequences homologous to the Lxx genome. The other end is specific to Lxc genome, indicating an association with genomic rearrangements. Although some IS elements of both genomes belong to the same families and may recognize the same target site for insertion, none of the Lxc-IS insertion loci were the same as for elements found in the Lxx genome (Fig. 2 and Additional file 3). The 12 remaining Lxc-elements were inserted in specific regions of its chromosome (Additional file 4).

IS elements were randomly distributed throughout Lxx genome (Fig. 2), 25% of all insertions were located in what has been described as genomic islands 223, in particular within genomic island LxxGI3. Comparatively the distribution of the elements in Lxc also seems to be random; however our approach so far does not allow us to make the same inferences about islands in Lxc genome.

To assess the impact of IS elements in gene disruption, they were classified into three categories: IS insertions within predicted genes; insertions in non-coding sequences; and insertions in non-coding sequences, but with one or two truncated genes nearby (less than 100 nucleotides distant) (Fig. 3). Truncated genes were defined as those ORFs that have a disrupted coding sequence based on BlastX results. Most of the elements were found inserted within intergenic regions for both genomes. In the Lxx genome, disrupted genes were associated mainly with degradation of polysaccharides, transport and regulatory functions, while in the Lxc genome they were linked to cell structure, regulatory functions and hypothetical genes (Additional file 3 and 4). Five of the putative genes truncated in Lxc have truncated orthologs in the Lxx genome however, in Lxx no IS insertions were detected: three of them probably encoded hypothetical proteins, Lxx05470 (Lxx genome position: 552,768–555,523), Lxx17040 (1,767,422–1,768,564) and Lxx21675 (2,229,230–2,230,675); another one, Lxx01850 (180,918–182,055), presented a conserved acyltransferase domain, associated with lipopolysaccharide modification; and the last one, Lxx14740 (1,531,157–1,531,354), presented a DNA-binding domain. Within Lxc, those genes were invaded by IS1237, ISLxc2, ISLxc3, IS1237 and ISLxc4, respectively. No DRs were detected within Lxx orthologous that could be an indicative of a prior invasion and excision event. Therefore is more likely that these genes were truncated prior to the IS elements insertion in Lxc.

According to the ISfinder database, elements of the IS5 family are common in both Archaea and Eubacteria 12. Considering all IS families found in Leifsonia xyli, examining ISfinder submission reports and using sequence similarity searches, we found that IS5 elements were the least represented in the sequenced genomes of Actinobacteria, despite being the most highly represented IS element in the Lxc genome. As observed in L. xyli genomes, transposases from the IS30 and IS481 families are abundant in most Actinobacteria (Fig. 4). Each putative transposase of characterized IS elements of Leifsonia xyli genomes was used as query and compared to transposases of Actinobacteria genomes, and the best BlastX hit result was considered. Most of the transposases shared around 70% of amino acid identity. The exception was IS1237, which related elements of Actinobacteria genomes did not share more than 50% identical amino acids. Actinobacteria is one of the largest taxonomic units in terms of number and variety of species 34 and members of IS elements of most characterized families are represented within this group 12. As probably expected no correlation was found between phylogenetic relatedness and IS family distribution in Actinobacteria genomes (Fig. 4). Not even among subspecies of Leifsonia xyli and Clavibacter michiganensis, which are much related to each other. This observation is in concert with the cycles of expansion and extinction of IS elements proposed before 14.

To further examine the distribution of IS elements between genomic DNA of two isolates of Lxc to check the involvement of these elements in diversification of strains, hybridization experiments were performed using fragments of three IS elements of different families as probes. There are two BamHI sites within IS1237, one within ISLxc4, and none within ISLxc2. There are no sites for PstI, EcoRV or EcoRI within these elements. IS1237 is the only element showing the same hybridization pattern between the two isolates for all tested enzymes (Fig. 5). However, variable DNA band patterns were detected for ISLxc2 and ISLxc4 (Fig. 5). We believe that partial digestions can not justify such differences because the same stripped membrane was used in each experiment, and we do not see partials in the assay with IS1237. Also, polymorphism at the enzymes restriction site could not explain the differences since they were found for all enzymes used when probed with ISLxc2 and ISLxc4, and they were not detected for IS1237, mainly considering the abundance of the later in Lxc genome. Finally, we assume that these differences were probably not the result of large-scale rearrangements; otherwise they would also have been detected in the distribution of IS1237. It is plausible to assume that the observed variation may result from homologous recombination between copies of these elements not interspersed by IS1237 copies, or that it may be associated with the presence of a variable number of elements, as a result of transposition of ISLxc2 and ISLxc4 to new locations. To assure any of these two hypotheses the bands should be isolated and the DNA sequenced. In any case, the difference exists and is probably originating diversification in the isolates of Lxc genomes. Consistent with this are the data generated by Young and collaborators where isolates of Lxc showed different band pattern in DNA fingerprinting 35.

The bacterial strains used were: Lxx CTCB07 (NC_006087) 2 (Brazil); Lxc DSMZ46306 (Taiwan); and Lxc SB (Australia). The DSMZ46306 strain was grown in a MSC New modified liquid medium 32 at 28°C for 5–10 days, under agitation at 300 rpm. The SB genomic DNA was kindly provided by Dr. Steven Brumbley (Bureau of Sugar Experiment Station/Queensland, Australia).

Two DSMZ46306 genomic libraries were prepared. Genomic DNA extracted as previously described 2 was mechanically sheared and the resulting fragments cloned into pUC19 (Q-Biogene, Carlsbad, CA). A BAC clone library was prepared with 25 kbp-inserts obtained by partial digestion with BamHI 36 cloned into pIndigoBAC-5 (Epicentre^®). Both ends of each insert were sequenced using an automated sequencer (model 3700, ABI Prism, Applied Biosystems, Foster City, CA). Results were analyzed by ABI sequencing analysis software, and assembled using the phred/Phrap/consed package 3738. All consensus sequences were generated with phred quality ≥ 20.

Inserts containing end-sequences similar to transposases were subcloned or primer-walked. Homology searches were performed using BlastN and BlastX 39 at GenBank 40 and IS finder 41. Identities were considered significant only when the E-value was less than 10^-05. Positive matches for transposase/integrase were manually verified to determine the presence of the following determinants of a given family: element size; presence of terminal inverted repeats (IR) and conserved terminal based pairs; target site of insertion and number of bases associated to direct repeats (DR); number of ORFs; distance among amino acids of DDE; as well as comparisons with related elements 11. We also analyzed the presence of variants, domains, frameshifts and premature stop codons.

Transposases sharing more than 95% of amino acid identity were grouped. Their nucleotide sequences in addition to 300 bases up and downstream were aligned using ClustalW 42. For each cluster of sequences, the first sixty bases of the 5'-end was aligned with the reverse complement of the first sixty bases of the 3'-end. Best matches were considered as IR candidates and compared to IRs of other characterized IS elements. IRs were selected when the minimum number of non complemented nucleotides followed the pattern of a given family. In an attempt to identify the target sites of insertion and the DR, an alignment of up to 20 bases flanking each IS element was done and the sequence was compared to the already described target sites of other IS elements. Features such as stop codon in frame and purine rich sites related to ribosome frameshift were sought in ORFs whose alignment partially matches transposases, based on BlastX results. Domains were identified within the Blast results at the CDD database 40. DDE motifs were identified following IS finder pattern 41. Minimal numbers of occurrences were determined based on flanking sequences.

Lxc-IS element loci were compared to those of the Lxx genome using adjacent sequences as anchors and a cross_match program, which is part of the swat/cross_match/phrap package 38, with default parameters. Lxc-specific loci were annotated using SABIA 43.

Amino acid sequences of putative transposases were used as queries within completely sequenced genomes of Actinobacteria. Genomes were downloaded from GenBank (IDs in Fig. 4). Searches were performed using BlastX and tBlastN 39.

Lxc genomic DNA (1 μg; DSMZ46306 and SB) was digested completely with PstI, EcoRI, BamHI and EcoRV. Fragments were separated in a 0.8% (w/v) agarose gel and transferred to a Hybond N+ membrane (Amersham, Piscataway, NJ). Probe labeling, hybridization and detection were performed with an ECL kit (Amersham) according to the manufacturer's guidelines and under high stringency conditions. Inserts of shotgun clones containing the entire fragments of three Lxc-IS elements (IS1237, ISLxc2 and ISLxc4) were amplified and used as probes. The same membrane was stripped and re-used according to the manufacturer's recommendations.