Evidence that the cytolethal distending toxin locus was once part of a genomic island in the periodontal pathogen Aggregatibacter (Actinobacillus) actinomycetemcomitans strain Y4

Actinobacillus actinomycetemcomitans is a facultative Gram-negative bacterium that has long been associated with localized aggressive periodontitis (LAP) and may also contribute to chronic forms of the disease. This bacterium has recently been reclassified as Aggregatibacter actinomycetemcomitans (Norskov-Lauritsen & Kilian, 2006). The bacterium expresses a number of genes whose products can be defined as virulence factors based on their in vitro biological activities or similarities to proteins produced by other pathogens (Henderson et al., 2003). Most notable is the expression of two multi-gene cytotoxin loci, leukotoxin (lkt) and cytolethal distending toxin (cdt). While Lkt has been extensively studied (see Lally et al., 1999 for a review), Cdt is a relatively newly described A. actinomycetemcomitans virulence factor. Cdt is a heterotrimer composed of three gene products (CdtA, CdtB and CdtC) that have homologues in a handful of facultative or microaerophilic Gram-negative pathogenic bacterial genera (Mao & DiRienzo, 2002; Mayer et al., 1999; Sugai et al., 1998).

Although Cdt is not unique to A. actinomycetemcomitans, this bacterium is the only member of the oral microbial flora identified to date that carries and expresses the toxin locus (Yamano et al., 2003). Cdt is prevalent in A. actinomycetemcomitans strains. Forty of 45 isolates from periodontitis patients examined by Yamano et al. (2003) exhibited various levels of Cdt activity. Ahmed et al. (2001) found that 43 of 50 strains from periodontitis patients contained all three cdt genes and displayed Cdt activity. In another study, PCR of subgingival plaque samples revealed that 13 of 106 diseased sites in 146 aggressive and chronic periodontitis patients contained A. actinomycetemcomitans expressing all three cdt genes (Tan et al., 2002). Fabris et al. (2002) reported that 39 of 40 A. actinomycetemcomitans isolates from a mix of healthy and periodontal-diseased subjects displayed activity characteristic of Cdt.

Much of our work has been focused on characterization of the periodontopathic strain A. actinomycetemcomitans Y4. This strain was isolated from a LAP patient in 1979 at the Forsyth Dental Institute (Boston, MA, USA) and has served for many years as a prototype strain for laboratory studies (Tanner et al., 1979). Our initial description of Cdt in this strain revealed that the genetic locus for this toxin is flanked by incomplete sequences homologous to virulence-associated and integrating plasmids (Mayer et al., 1999). In the same study, we also identified a bacteriophage integration (att) sequence immediately downstream from the cdt genes. These observations are consistent with the possibility that strain Y4 acquired the cdt gene as part of a recombination event (most probably horizontal gene transfer). To obtain more conclusive evidence in support of this hypothesis, we sequenced a region of the chromosome farther downstream from the cdt locus. Here we report the presence and characterization of a classical genomic island, designated GIY4-1, and discuss the implications of its location proximal to a major genotoxin locus.

The nucleotide sequence of GIY4-1 was obtained using a combination of chromosome-walking strategies by employing oligonucleotide primers based on data from previous sequencing runs, restriction endonuclease sites and homologous sequences in the Haemophilus ducreyi 35000HP genomic sequence (GenBank accession no. AE017143) and genomic DNA from A. actinomycetemcomitans Y4 (GenBank accession no. AF006830; Mayer et al., 1999). The sequence was obtained by walking along both DNA strands. Automated-cycle sequencing reactions were conducted by the Genetics Core Facility at the University of Pennsylvania using an Applied Biosystems 377 sequencer with dye primer chemistry.

The sequences were compiled and assembled with LaserGene (DNASTAR). ORFs were identified and examined with the European Molecular Biology Open Software Suite (EMBOSS, release 3.0; ). The Lipman–Pearson algorithm was used to make amino acid sequence alignments (Altschul et al., 1997). The DNA and protein sequence databases were searched for homologous sequences using the TBLASTX and BLASTP algorithms (Gish & States, 1993) accessed through the National Center for Biotechnology Information (NCBI) website (). Conserved domain matches were obtained from the Conserved Domain Database (CDD; Marchler-Bauer et al., 2005) using the Conserved Domain Architecture Retrieval Tool (CDART; Geer et al., 2002). Multiple sequence alignments were performed with CLUSTAL_X version 1.83 (Thompson et al., 1997) using the default settings. Phylogenetic trees were constructed using DRAWGRAM in Phylogeny Inference Package (PHYLIP) version 3.65 (Felsenstein, 1989; ).

The genetic organization of GIY4-1 and adjacent coding regions surrounding the cdt locus is shown in Fig. 1. The previously sequenced cdt locus (Mayer et al., 1999) is included to orientate the genomic island and to provide updated annotation for several of the ORFs. GIY4-1 (bp 6859–28 537) is flanked by bacteriophage integration or attachment sequences (att) which delineate the boundaries of the genomic island. The sequence of the left att site (bp 6859–7018) has been reported previously (Mayer et al., 1999) and is included in GenBank under the accession number AF006830. The sequence of the right att site (bp 28 353–28 537) is identical to that of the left att site. A total of 22 ORFs, each containing 50 or more codons and matching sequences in the GenBank database, reside between the att sites (Table 1). These ORFs include both partial and complete gene sequences, determined by interpretation of the results of TBLASTX, BLASTP and CDART search matches with sequences in the databases. Five of the ORFs could be assigned putative functions based on deduced amino acid matches in the CDD. These include genes for an integrase/resolvase (xerD; bp 8069–7218), DNA primase (prm; bp 14 651–13 386), single-strand DNA-binding protein (ssb; bp 19 721–19 290), nuclease (parB; bp 26 104–24 320) and a combined DNA helicase–plasmid partitioning protein (dnaB–parA; bp 28 268–26 055). All of these gene products have putative roles in the insertion, replication and partitioning of extrachromosomal DNA. The predicted functions of each CDD match are detailed in Table 1. It is interesting that although the dnaB–parA sequence contains a single ORF, it represents two full-length in-frame genes. It is not clear at this time whether both genes are expressed. The remaining 17 ORFs have significant matches to hypothetical protein gene sequences deposited in GenBank and have no conserved domains. One ORF, designated hae, matched to a Haemophilus-specific protein. The other ORFs are designated ORF1–16. Three ORF cluster groups could be arbitrarily assigned based on visual inspection of ORF positions. These are labelled integrase cluster, primase cluster and par cluster in Fig. 1. ORF3 and ssb are also present in the A. actinomycetemcomitans plasmid pVT745 (GenBank accession no. NC002579). Sequence matches to the A. actinomycetemcomitans temperate bacteriophage phi 23 were not found (Resch et al., 2004). Many of the ORF sequences had a mol% G+C content well below that of the A. actinomycetemcomitans genome. This characteristic is indicative that the DNA is from a foreign source (Table 1).

(35K): Fig. 1. Genetic organization of GIY4-1 and the upstream cdt locus region. GIY4-1 is bounded by the att site (magenta). Genomic island ORFs are in black except for xerD (magenta) and ORF3 and ssb (green). Green ORFs have homologues on the A. actinomycetemcomitans plasmid pVT745. The cdt ORFs are in red and ORFs that reside outside of GIY4-1 are in grey (Mayer et al., 1999). The 74 bp direct repeat is in yellow. Three ORF clusters are shown. The arrows indicate the direction of transcription.

Table 1. Features of GIY4-1 and upstream cdt gene locus region

The properties of GIY4-1 just described are typical features of genomic islands. These features include (i) association with tRNA-encoding genes (att site), (ii) mol% G+C content differing from that of the host genome, (iii) flanking repeat structures (att site), (iv) a mosaic-like structure comprising a multitude of functional, truncated and non-functional putative ORFs with known or unknown functions and (v) the presence of many fragments of mobile genetic elements (Hacker & Kaper, 2000). GIY4-1 is clearly different from AAI-1 of A. actinomycetemcomitans HK1651 (Chen et al., 2005) and the atypical pathogenicity island in Porphyromonas gingivalis (rag locus; Curtis et al., 1999) reported previously. We searched the complete genomes of A. actinomycetemcomitans HK1651 (the only strain of this species for which there are complete genome sequencing data; ), H. ducreyi 35000HP, Haemophilus influenzae 86-028NP and Haemophilus somnus 129PT for deduced amino acid sequence homologues to the GIY4-1 ORFs. Sequences matching those of the GIY4-1 ORFs were then mapped on each of the genomes with some very interesting results (Fig. 2). All maps are drawn to relative scale. A. actinomycetemcomitans HK1651 does not have a genomic island related to GIY4-1, lacks the xerD ORF and has only a single copy of the att sequence. The single copy of the att sequence is located immediately upstream of the cdt locus. There are two or more att copies in H. ducreyi 35000HP, H. influenzae 86-028NP and H. somnus 129PT. A genomic island very similar to GIY4-1 was found in H. ducreyi 35000HP (GI35000HP) and H. influenzae 86-028NP (GI86-028NP). To the best of our knowledge, information about these two genomic islands has not been published. GI35000HP and GI86-028NP are flanked by the att sequence as in GIY4-1. The cdt locus is located within the boundaries of GI35000HP. The H. influenzae 86-028NP and H. somnus 129PT genomes do not contain cdt genes. The cdt gene order is reversed in A. actinomycetemcomitans HK1651 relative to that in GIY4-1 and GI35000HP. The organization of the integrase, primase and par ORF clusters in GIY4-1 is remarkably similar to those in H. ducreyi 35000HP (GI35000HP), H. influenzae 86-028NP (GI86-028NP) and H. somnus 129PT. There is one copy of the ssb ORF in A. actinomycetemcomitans HK1651 and it is relatively close to the cdt locus. There are two copies of ssb in H. ducreyi 35000HP, H. influenzae 86-028NP and H. somnus 129PT. One of the two copies of ssb in H. ducreyi 35000HP is in GI35000HP and one of the copies in H. influenzae 86-028NP is in GI86-028NP.

(19K): Fig. 2. GIY4-1 ORF deduced amino acid sequence homologues in A. actinomycetemcomitans HK1651, H. ducreyi 35000HP, H. influenzae 86-028NP and H. somnus 129PT. Chromosome maps are in kilobase pairs. Genomic island maps (GIY4-1, GI35000HP and GI86-028NP) are in base pairs. All maps are drawn to relative scale. The colour scheme of the ORFs is the same as in the legend to Fig. 1. The three ORF clusters (integrase, primase and par) are labelled. ORF coordinate data are from GenBank [accession numbers: AF006830 and EF196803 (cdt locus and GIY4-1), AE017143 (H. ducreyi 35000HP), CP000057 (H. influenzae 86-028NP) and CP000436 (H. somnus 129PT)]. A. actinomycetemcomitans HK1651 sequence information was obtained from the Actinobacillus Genome Sequencing Project ().

Based on the observations from the genetic map comparisons, it is very tempting to speculate about the origins of the cdt locus in A. actinomycetemcomitans Y4. It seems reasonable to suggest that a prototype genomic island originated from a region of the H. somnus 129PT chromosome that contained the integrase, primase and par ORF clusters. The xerD ORF, in proximity to these ORF clusters, in H. somnus 129PT could have contributed to the horizontal gene transfer of this region. This region could have been acquired by H. influenzae 86-028NP, where it became a true genomic island (flanked by an att sequence). The ORF clusters and xerD are bounded by the att sequence in GI86-028NP. A recombination event could have inserted the cdt locus in GI86-028NP either prior to or after transfer to H. ducreyi 35000HP, resulting in GI35000HP. This modified genomic island could then have been passed to A. actinomycetemcomitans. This theory is supported by results from an examination of the phylogenetic relationships among the xerD gene products produced by A. actinomycetemcomitans Y4, H. ducreyi 35000HP, H. influenzae 86-028NP and H. somnus 129PT. The deduced amino acid sequences of these xerD genes, including one found on the H. influenzae plasmid ICEhin1056 (Mohd-Zain et al., 2004), are approximately 40 % identical (Fig. 3a). The XerD proteins comprise three distinct cluster groups that fit the predicted pattern of horizontal gene transfer of the genomic islands (Fig. 3b). It is important to note that our conclusions are limited by the number of currently available genome sequences. These interpretations may be subject to change as more genomes are sequenced.

(46K): Fig. 3. Comparison of XerD homologues. (a) CLUSTAL_X alignment of the xerD deduced amino acid sequences from GIY4-1, H. ducreyi 35000HP, H. influenzae 86-028NP and H. somnus 129PT. Identical and similar amino acids are marked with asterisks and colons, respectively. Sequence data were obtained from GenBank. Accession numbers are the same as those in the legend to Fig. 2 except for xerD from ICEhin1056 (GenBank accession no. AJ627386). (b) Rooted phylogenetic tree showing the relationships of the XerD homologues aligned in (a). The tree was constructed using PHYLIP 3.6.

Another supporting argument for the proposed order of genomic island passage among the species examined here is that prokaryotic genomes, especially those of pathogens, tend not to increase in size. Prokaryotic genomes are continuously evolving (Nilsson et al., 2005) with the changes a consequence of horizontal gene transfer, duplications that lead to paralogous genes, and deletions (Mira et al., 2001; Ochman, 2001; Rocha, 2004). Even though the size of genomes may increase due to some of these processes, bacterial genomes remain relatively small and generally lack non-functional sequences. The most likely explanation for this occurrence is a process known as deletional bias in which genetic change is skewed towards deletions rather than insertions. The current view is that evolution of pathogens differs from that of non-pathogenic bacteria. Interaction with a host leads to smaller effective bacterial population sizes, low genetic diversity and infrequent recombination (Lawrence, 2005; Lawrence & Hendrickson, 2005). Calculated gene loss rates indicate that genome size can be extensively reduced during a relatively short evolutionary time period (Nilsson et al., 2005). It appears that various genetic rearrangements have occurred in A. actinomycetemcomitans, since most of the genomic island sequence appears to have been lost in strain HK1651 and reduced in size in strain Y4. In addition, restriction fragment length polymorphism analysis of the region of the chromosome containing the cdt locus revealed deletions, of various lengths, in members of a collection of clinical strains (DiRienzo & McKay, 1994; Mayer et al., 1999). Evidence that the cdt locus in strain Y4 was once part of GIY4-1 is supported by the presence of a 74 bp repeat sequence immediately upstream of the cdt locus (bp 2552–2625) and within GIY4-1 (bp 11 333–11 406) (Table 1 and Fig. 1). Interestingly, there are no significant ORFs immediately downstream of the copy of the repeat sequence in GIY4-1. It will be interesting to see how widely GIY4-1 is distributed in A. actinomycetemcomitans and the extent of sequence variation.

Although possible, it does not seem likely that the A. actinomycetemcomitans Y4 cdt locus was obtained by a plasmid or bacteriophage recombination event. ORF3 and ssb sequences are present on the A. actinomycetemcomitans plasmid pVT745 (Galli et al., 2001) but are also widespread since they are found in each of the bacterial genomes shown in Fig. 2. Except for the xerD deduced amino acid sequence, no other sequences were shared by ICEhin1056 and GIY4-1. In addition, a complete att sequence, containing the right and left conserved and middle non-conserved regions, was found only in GIY4-1, GI35000HP, GI86-028NP and the H. influenzae 86-028NP genome (Fig. 4). A related att sequence was also found in the H. influenzae bacteriophage HP2 (Williams et al., 2002). However, only the left and right conserved regions were present.

(30K): Fig. 4. Alignment of att nucleotide sequences from GIY4-1, H. ducreyi 35000HP, H. influenzae 86-028NP and H. somnus 129PT. (a) Sequences that contain only the left conserved region. (b) Sequences that contain the complete att site. (c) Sequences that contain only the right conserved region. Multiple att sequences in a genome are designated by position in bp. Identical nucleotides are marked with an asterisk. The att sequences in GIY4 and A. actinomycetemcomitans HK1651 (457263) are identical.

Thus we have identified a 22 kb genomic island (GIY4-1) in the periodontal pathogen A. actinomycetemcomitans Y4. GIY-4 is located immediately downstream from a cdt locus. Although no virulence genes were found in GIY4-1, the functions of 17 ORFs that encode hypothetical proteins have not yet been identified. Indirect evidence supports the possibility that the cdt locus was once part of the genomic island. We thank Sam Cao and Jon Korostoff for critical reading of the manuscript and Paul Sniegowski (Department of Biology) for valuable discussions. This study was supported in part by USPHS grant DE012593 from the National Institute of Dental and Craniofacial Research.

Footnotes

†Present address: Department of Physiology and Biochemistry, Faculty of Dentistry, Mahidol University, Bangkok, Thailand.

The GenBank/EMBL/DDBJ accession number for the GIY4-1 sequence of Aggregatibacter actinomycetemcomitans is EF196803.