The present assignment of families and subgroups is primarily based on transposase and DNA end sequence data. Analysis of transposases was performed by multiple sequence alignments and clustering methods followed by dendrogram construction to set up clusters of related proteins (see Methods). Left and right DNA extremities flanking transposase genes were aligned (i) to each other to facilitate observation of terminal inverted repeats (TIRs) and (ii) with TIRs from other ISs to detect DNA extremity conservations. Together, these approaches split the initial IS set into ten groups. In each of them, the percentage of residue identity among transposase pairs often varied between 20 % and 50 %, highlighting the magnitude of divergence occurring among elements of a same IS group. The length of TIRs ranged from 10 to 40 bp and many of them were imperfect. Comparison of TIRs from different elements of a given IS group uncovered unique and conserved signatures in each of them, illustrating the relationship between the transposase sequence and the IS terminal repeats (Figure 1).

To validate this clustering, established groups were analysed independently for conservation of specific transposase domains as well as length and specificity of target site duplications. Analysis of transposase DDE catalytic regions (N2, N3 and C1) revealed a perfect conservation of the DDE motif (Figure 2), while the size and sequence of the spacers that separate these catalytic residues differed among the ten groups. The YREK motif was partially lacking in three groups, where either its tyrosine, arginine or lysine residue was not conserved. According to the definition of IS families, IS groups sharing the same catalytic site structure were assigned together. Seven clusters displayed the complete YREK motif; and were named subgroups IS231, ISH8, IS4Sa, IS4, ISPepr1, IS10 and IS50. The three remaining groups all displayed a distinct variation of the YREK motif and were therefore assigned to different and new families, namely IS701, ISH3 and IS1634, referred as emerging families. Finally, while IS4 family members generated about 10 bp target duplications, i.e. the approximate length of a complete DNA helix turn, those from emerging families displayed direct repeats (DRs) of around five bps, which corresponds approximately to half a DNA helix turn (Figures 1 &3).

Family IS4 included 153 distinct intact elements. The main hallmarks were the presence of D(60~110)D(100~150)E and Y(2)R(3)E(6)K motifs, single orf encoded transposases and target site duplication lengths corresponding approximately to one DNA helix turn. Considerable diversity was observed at the level of transposase regions (sequence and length) outside catalytic residues, DNA end signatures and target site specificity. The following description of established subgroups will summarize this diversity.

Subgroup IS231 was previously established (reviewed in reference 2) and was the most coherent one since many transposase pairs displayed sequence identity percentages above 50 (see also Figure 1A). Likewise, their DNA ends are among the most conserved. A clear relationship was observed between transposase and extremity conservation (Additional file 1). This is particularly true for left extremities (conventionally upstream of the transposase gene) where transposase-related conservation extended beyond the TIRs. The IS231 subgroup further distinguished itself by the fact that its members occurred almost exclusively in genomes of phylogenetically close bacteria composing the B. cereus sensu lato group (i.e. B. cereus sensu stricto, B. thuringiensis and B. anthracis, see Additional file 1 and Figure 4). As previously shown, an important hallmark is the large size range of its elements (Table 1). A significant fraction of them displayed additional DNA between the left TIR and the transposase gene 23, which is still unusual among ISs. In this study, novel putative passenger genes were found in these extra sequences coding for resistance, virulence or metabolism determinants as well as unknown orfs (results not shown). This modular aspect has not yet been observed in other IS families or IS4 subgroups, which may be due to the fact that, so far, no systematic and thorough searches for supplementary IS DNA were carried out in other families.

Subgroup ISH8 was initially included into subgroup IS4Sa 1. Yet, our results showed a closer proximity to subgroup IS231; not only at the transposase level, but also due to DNA extremity conservations (both displayed conserved 5'-CAT ends, Figure 1B and Additional file 2). ISH8 elements were initially discovered in archaea but recent genome projects uncovered elements of this subgroup in proteobacteria. Examination of ISH8 insertions did not reveal apparent target site specificity, contrary to the one observed for IS231 elements (Figure 3).

Members of Family IS701 were already considered distantly related to IS4 in reference 1 and this was confirmed here. Only four of the 27 considered IS701 elements displayed the tyrosine of the YREK motif (results not shown, see Figure 2); and almost all elements displayed a highly conserved target site duplication of exactly four bps (results not shown, see Figure 3). The diversity emanating from IS701 transposases and TIRs allowed identification of three distinct clusters (Additional file 8), which announced a possible division into subgroups. As indicated by their prokaryotic distribution, family IS701 seemed to be rather widespread (Figure 4).

Family ISH3 was a small group so far restricted to archaea (Additional file 9). Half of their transposases lacked the lysine residue of the YREK motif while all (except ISFac10) displayed a Y-(2)-R-(3)-E-(3)-(R) motif. DRs flanking ISH3-like insertions are typically five bps long and generally flanked by A at one end and T at the other end.

Family IS1634 was initially named IS1549 1. This new designation should prevent confusion with an emerging group called IS1595 13. Transposases were among the largest due to relatively long N3-C1 spacers, which were sometimes twice as long as those from family IS4. Only five out of 32 displayed the arginine residue of the YREK motif (results not shown, see Figure 2). This large group could be divided in (three) distinct clusters according to transposases (Additional file 10). Target site duplications were five to six bp AT-rich tracts while DNA ends were only poorly conserved. They were remarkably diverse in archaea and relatively widespread in bacteria (Figure 4).

Members of the IS4 family were found in 92 out of 540 (~17 %) complete genomes representing 65 prokaryotic species. Families IS701, ISH3 and IS1634 displayed lower incidences, being present in ~3 %, ~0.7 % and ~2 % of available genomes, respectively. 172 of these genomes included plasmids of which 22 harbored elements of these families. Figure 5 displays the distribution of IS genomic copy numbers as it was found on chromosomes and plasmids. The distribution of family IS4 uncovered a clear preference for single genomic copies, followed by a preference for two, three and six to eight copies per genome. The genomes of Photobacterium profundum SS9, Mycoplasma mycoides subsp. mycoides SC strain PG1, Sulfolobus solfataricus P2 and the two sequenced strains of Xanthomonas oryzae pv. oryzae displayed the highest amounts of elements from families IS4, IS1634, ISH3 and IS701, respectively (Figure 5). The most IS-rich extrachromosomal replicons were megaplasmids pNRC100 (191 kb) and pNRC200 (365 kb) from Halobacterium sp. NRC-1, which harbored together 20 copies of IS4 family elements and 18 copies of family ISH3 elements.

The distribution of individual elements among the different bacterial and archaeal phyla is shown in Figure 4. As indicated by the histogram in Figure 4B, current sequencing projects focus unequally on different prokaryotic phyla 27. The distribution presented in Figure 4A partially reflected this sequencing bias, which is why most gathered elements were found in proteobacteria. Globally, families IS4, IS701, ISH3 and IS1634 were detected in eight, seven, two and seven phyla, respectively.

Figure 4A shows that the distribution of individual IS families or subgroups as defined in this study could either be broad (as for subgroups IS4, IS4Sa and family IS701) or concentrated in a particular phylum (as for subgroup IS231 and family ISH3). Interestingly, two copies of an element belonging to subgroup IS4 were found in the genome of the brown algae virus, Ectocarpus siliculosus virus 1 28. So far, no IS4-related element was found in eukaryotes.

The availability of a large number of completely sequenced genomes on one hand, and the extensive manual characterization of an IS family on the other hand, allowed a direct evaluation of global presence of its putatively functional members. Family IS4 displayed a sporadic distribution among 17 % of explored genomes. Moreover, the behavior of these ISs was assessed by determining their genomic occurrence. This showed that numerous elements exhibited single or low copy numbers, which was in line with the assumption that transposable elements need to compromise between transposition for self maintenance and limited insertions to keep host genome integrity 29. This distribution also displayed a certain, less pronounced, prevalence for copy numbers between 6 and 8, although the biological relevance of this observation remains unclear.

While the global distribution of IS elements was shown to be rather sporadic, important expansion events were observed in a few pathogens as well as in some free-living extremophiles. One of these pathogens is a member of the facultative intracellular Mycoplasma, of which eleven genomes are available. 65 intact copies of family IS1634 elements were found together with 17 intact copies of the IS3 family in M. mycoides 30. The presence of family IS1634 was also detected in M. synoviae 31, M. agalactiae [GenBank:CU179680] and the three strains of M. hyopneumoniae 3132. Only M. genitalium 33, M. pneumoniae 34 and M. mobile 35 were devoid of ISs. Interestingly, the Mycoplasma species lacking ISs were also those displaying the smallest genomes (between 0,58 and 0,82 Mb), while the IS-rich M. mycoides had a relatively large genome (1,2 Mbs). A positive correlation between the total number of genes and the amount of mobile DNA in a genome has been reported for other facultative intracellular bacteria 36, and a recent study has extended this observation to other prokaryotes 37. However, this relationship must be seen as a trend with notable exceptions (e.g. the large genome of M. penetrans – 1,36 Mbs – has only 21 transposase genes 38).

Family IS701 was represented by 70 and 72 elements in the genomes of two sequenced pathovars of the rice pathogen X. oryzae 3940. This family was completely absent in phylogenetically related species with available genome sequences, including X. campestris (black rot in crucifers) and X. axonopodis (citrus canker) 3940414243. Additionally, a moderate presence of family IS4 was observed in all but one of the six Xanthomonas genomes (X. campestris pv. vesicatoria), together with other, more expanded families. So far, all sequenced xanthomonads are IS-rich, which is supposed to provide the necessary flexibility for adaptation to different plant hosts 40. Interestingly, each strain is characterized by the expansion of one specific IS family 43. The IS4 family was also observed to be relatively expanded in sequenced Shigella spp. where other families, both with higher and lower expansion levels, were also detected 44454647. These mobile elements are supposed to contribute to the emergence of variable epidemiological and pathological features among these phylogenetically close pathogens 45.

These observations are reminiscent of IS expansions observed together with emergence of pathogenicity in other host associated pathogens, such as Bordetella pertussis (whooping cough 48), Yersinia pestis (plague 49) and Francisella tularensis (tularemia 50).

A different relationship was observed in the genus Vibrio. The expansion of family IS4 in the non-pathogenic Photobacterium 51 was striking (Figure 5) as compared to those observed in the pathogenic Vibrionaceae, such as V. cholerae 52, V. parahaemolyticus 53 and V. vulnificus 54, that displayed relatively modest IS contents. Therefore, ISs apparently have a quite different role in Vibrio evolution, where some IS-rich species are non-pathogenic piezophiles. Other important IS expansions among free-living extremophiles were observed in one out of three Sulfolobus spp. 555657, where family ISH3 represents a significant fraction of global IS content, as well as in Halobacterium, which harbors numerous copies of IS4 and ISH3 elements together with other families 58.

Analysis of the distribution of individual elements among prokaryotes and among established families and subgroups was also performed regardless of copy numbers (Figure 4). Prokaryotic IS hosts were from world-wide origins and covered a large range of lifestyles and habitats, from free-living environmental germs and extremophiles to host associated symbionts and pathogens. Evaluation of emergence on evolutionary timescales, as well as estimation of horizontal gene transfer (HGT) extent prompted by this distribution remain both problematic because of low identity levels among transposases. However, many IS4-related elements were found on putative mobile DNA, including various plasmids (Figure 5), bacteriophages (results not shown) and one eukaryotic virus (Figure 4). Additionally, most IS4-related subgroups and emerging families were scattered over several phyla (Additional files 1, 2, 3, 4, 5, 6, 7, 8, 9, 10), suggesting an involvement of HGT in the dissemination of these elements.

Reports on prokaryotic IS distribution have shown that IS families are not clade-specific and our data regarding family IS4 were in agreement with this assumption. However, the distribution of IS4 subgroups and emerging families showed that the number of elements of some IS groups was not systematically proportional to the extent of genome sequencing (Figure 4). Subgroup IS231 (family IS4) was almost exclusively restricted to Firmicutes and 26 of its 32 elements were found among bacteria of the B. cereus s.l. group. Likewise, family IS1634, and to a lesser extent family ISH3 and subgroup ISH8, were preferentially found among Euryarchaeota. These observations may partially be due to insufficient genome data, but they do not rule out the possibility of existing clade specificity at the level of IS subgroups. Preferential IS occurrences among certain prokaryotic phyla or genera can be the consequence of isolated niches reducing or limiting horizontal transfer of hosted ISs. In this case, a series of extremophilic archaea were almost exclusive hosts of subgroup ISH8 and family ISH3, their preference for harsh ecological niches may represent a frontier for HGT with other phyla or with bacteria. Another possible explanation can be based on IS-associated features. It is possible that the presence of some ISs cannot be tolerated by certain hosts. Uncontrollable transposition behaviors, lack of target site specificity, preferred insertions into vital genes or regulatory regions can limit the host compatibility of some ISs for which a viable equilibrium can be found only in rare 'IS – host' combinations.

The co-crystal structure of the IS50R transposase binding the ends of the Tn5 transposon generated function assignments for numerous residues of this protein 9. A comparative sequence analysis based on 19 transposases was performed and compared with a previous report on seven transposases from subgroup IS50 by Reznikoff et al. in 2004 59. This confirmed the previous function assignment of conserved residues since most identified domains were still conserved among this subgroup, as were a series of residues with unknown function (further results available in Additional file 11).

The primary sequence data concerning the transposase motifs (Figure 2) and target site duplications (Figure 3) suggested divergent biochemical behaviors between IS4 transposases and those from the emerging families. The two 3'-OH transposon ends are known to attack the target helix at opposite sites in the case of IS4. For the emerging families, the length of target site duplications (5 bp) suggested a different target strand cleavage path: owing to the structure of B-DNA, this attack needs to come from the same helix side. It is worth noting that in the case of family IS1634, two elements were reported to display atypical target site duplication lengths. IS1549 and IS1634, from M. smegmatis and M. mycoides, respectively, have both been shown to produce long, variable-length DRs upon insertion 6061. They had lengths between 8 and 514 bps, depending on the copy. So far, no mechanistic model has been proposed for the generation of this variable target duplication size.

This search covered more than 500 complete and partial bacterial genomes. Only apparently full-length transposases were retained to avoid inclusion of partially deleted and thus inactivated transposases which can accumulate mutations of functionally important residues.

When we began our search for novel IS4-like elements, the ISfinder database contained about 70 ISs designated as 'member of family IS4'. First, representative elements (about twenty) covering most of the sequence diversity of this family were selected. Primary transposase sequence of each of these representatives was then used in a BLASTP search, either among microbial genomes only, or against all organisms. Since IS4 elements exhibit short TIRs and generate target site duplications upon insertion (DRs), the flanking DNA of resulting hits was checked for the presence of these repeats. This, together with DNA extremity comparison of various elements, allowed the determination of both ends of the collected elements. New ISs were checked for the existence of formerly registered iso-forms which we defined as elements with less than 2% divergence in the amino acid sequence of their putative transposases and/or less than 5% difference in their DNA sequences. ISs were submitted to the ISfinder database, which provided new names according to the current IS nomenclature 15.

BLASTP searches were performed on the NCBI BLAST online interface 27 without low complexity filter and with otherwise default parameters. Each transposase sequence retrieved a series of protein hits which were possible transposase candidates. The DNA encoding these candidates was downloaded together with 1000 bps up- and downstream regions. These DNA sequences were then verified for the presence of TIRs flanking the transposase candidate genes using BLASTN and the PALINDROME algorithm of the wEMBOSS package at the Belgian EMBnet Node 63. If TIRs were present, flanking DRs were looked for by eye.

The following multiple alignment algorithms were then evaluated with the resulting ISs for their performance to accurately align the catalytic D, D and E residues of transposases : Clustal W 64, Dialign 65, Parallel PRRN 66, Muscle 67, T-Coffee 68 and M-Coffee 69. Global alignment of the 227 transposases was made by merging existing M-Coffee alignments with Clustal W. The following order was applied where pre-existing alignments are represented by subgroup or family designations flanked by brackets and merging procedures by + symbols: (((IS231)+(ISH8)+(IS4Sa)+(IS4))+(IS10, IS50, ISPepr1))+((IS701, ISH3)+(IS1634)). Examination of resulting alignments and subgroup specific dendrogram construction (UPGMA, BLOSUM62) were performed using the Jalview alignment editor 70. Dendrograms were drawn with TreeView 71.

In order to facilitate visualization of transposase clusters, the TRIBE-MCL clustering algorithm 72 was applied to the complete set of transposases with inflation option (-I) set to 1,2 and default values for other parameters. These settings corresponded to those used by the ISfinder 15 and ACLAME 73 databases.

587 chromosomes and 363 plasmids, representing 540 individual completely sequenced bacterial and archaeal genomes, were screened for the presence of IS4, IS701, ISH3 and IS1634 family DNA. Therefore, the 950 molecules were used as input in BLASTN 74 searches against a homemade database containing the 227 DNA sequences encompassing these IS families.