INTRODUCTION

Top Abstract Introduction Materials and Methods Results and Discussion References

Photorhabdus luminescens is an insect-pathogenic gram-negative proteobacterium that forms a "symbiosis of pathogens" with insect-pathogenic nematodes (52). In this symbiosis the bacteria are carried in the guts of entomopathogenic nematodes belonging to the family Heterorhabditidae (members of a different group of bacteria, Xenorhabdus spp., are carried in the guts of members of a different group of nematodes, the Steinernematidae). Upon invasion of an insect host by a nematode, the bacteria are released from the gut directly into the open blood circulatory system of the insect, the hemocoel (52). Here the bacteria are thought to release a wide variety of potential virulence factors, including high-molecular-weight toxin complexes (Tc), lipopolysaccharide (LPS), proteases, lipases, and a range of different antibiotics (52). Inferences concerning the involvement of these factors in killing of the insect or in overcoming the insect immune system, however, often result merely from documentation of secretion of the factors into bacterial culture supernatants. Studies examining the precise role of virulence factors during the infection process in insects have not been performed, and studies of Photorhabdus mutants are rare. As a prelude to genetic analysis of potential virulence factors in P. luminescens, we were interested in obtaining a sample sequence of strain W14 in order to document the classes of genes present and to begin to design suitable experiments for analysis of the genes based on a likely idea of their functions.

The relative advantages of sample sequence analysis versus full-scale analysis of a finished bacterial genome have been discussed elsewhere (119). However, there are several points relevant to the current discussion, as discussed briefly below. First, a sample sequence can be completed at a fraction of the cost of completion of a full genome. Second, a surprisingly high percentage of the genome can be captured even with a 1× sample sequence. Given the current uncertainty concerning the exact genome size of P. luminescens, the percent coverage obtained in this study is hard to estimate; however, McClelland and Wilson (119) suggested that a 1× genome equivalent for the 4.78-Mbp Salmonella typhi genome would require only 12,000 reads of 400 bases. Such coverage would ensure that almost every cistron was represented in the sample sequence. The 2,000 reads reported here obviously do not give this level of coverage, but, as shown below, even the limited sample sequence obtained revealed ample evidence concerning the types of virulence systems that P. luminescens may employ in its complex life cycle.

Although few potential P. luminescens virulence factors have been examined in detail (either biochemically or genetically), we can attempt to predict the likely role of bacterial virulence systems in killing an insect, in overcoming an insect immune system, or in facilitating bacterial and/or nematode growth. It is thought that once P. luminescens is released from the nematode gut into the insect hemocoel, it plays multiple roles in helping the nematode overcome its host (52). To do this, the bacteria need to overcome both the cellular (hemocytic) and peptide-mediated (antibacterial polypeptide) components of the insect immune system. Furthermore, the bacteria stop the insect from feeding and probably render its tissues suitable for consumption by both the bacteria and the nematodes. Anti-insect virulence mechanisms might, therefore, include, but not be limited to, toxins active against the insect gut and/or hemocytes and enzymes (such as proteases) capable of both degrading insect tissue and disabling the antibacterial peptides also associated with the insect immune system. Equally important in its role in overcoming an insect host, P. luminescens must ensure that the insect cadaver does not act as a breeding ground for opportunistic soil bacteria, fungi, and/or other species of nematodes. We might, therefore, expect P. luminescens to secrete a wide range of antimicrobial, antifungal, and nematicidal compounds, as previously documented by other workers (52). The aim of the present study was, therefore, to identify genes that encode likely virulence factors as a prelude to a functional analysis of the genes via targeted knockout and assay of the resulting mutants in the host infection process. Not only should such an analysis allow us to elucidate how the virulence factors act on the insect, but the gene sequences may also provide an indication of the evolution and potential origins of the virulence factors.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results and Discussion References

Genomic library construction and sequencing. Genomic DNA from P. luminescens W14 was size selected to obtain 1- to 2-kb fragments and then cloned into M13 Janus as previously described (28, 115). DNA templates were purified from library clones and sequenced by using dye terminator-labeled fluorescent cycle sequencing (model ABI377 automated sequencer; Applied Biosystems Division, Perkin-Elmer). Single sequencing reads (average length, ~400 bp) were obtained for one end of 2,122 clones. Sequences were truncated to exclude the phage arms and multiple cloning site and were then submitted to the BLASTX servers at the National Center for Biotechnology Information. Clones giving hits to either Tc-, protease-, or Rtx-like-encoding genes were then sequenced from the other end or "flipped."

Comparison with Escherichia coli K-12. Trimmed (vector removed and high-quality trim with SeqManII) P. luminescens sequence reads were searched against the DNA and protein sequences of E. coli MG1655 by using BLASTN and BLASTX with a local server. The output was parsed and sorted to give three subsets of data with different levels of identity. No alignment length criteria were imposed on the output. The results, therefore, included short alignments and multiple hits for many sequences, all of which were legitimate similarities.

Nucleotide sequence accession number. The nucleotide sequence determined in this study has been deposited in the DDBJ/EMBL/GenBank database under accession no. AQ989457-AQ991805.

	RESULTS AND DISCUSSION

Top Abstract Introduction Materials and Methods Results and Discussion References

Comparison with E. coli K-12. The results of the comparison of P. luminescens sequences with E. coli K-12 sequences are shown in Fig. 1. Even though the number of query sequences was relatively small (<0.5× genome coverage), we clearly observed that there is significant conservation of sequences, particularly protein sequences, between the two genomes, which is consistent with the relatively close phylogenetic relationship of the two organisms (both are members of the family Enterobacteriaceae). Regions of the genome conserved in Escherichia and Photorhabdus strains begin to define the components of the putative ancestral chromosome of members of the gamma subdivision of the class Proteobacteria. The large excess of hits at the protein level compared to the DNA level at all stringencies suggests that the divergence between orthologous sequences is sufficient to obscure true matches at the nucleotide level. Even the number of protein hits changed extensively as we varied the criteria for a significant match. Thus, we were reluctant to choose an arbitrary cutoff for determining orthology (as has been done for other sample sequence comparisons) and instead describe three different levels of stringency below. This form of presentation provides not only a sense of the absolute number of sequences that are similar but also a sense of the strength of the similarities. It should be noted that although the hits are distributed, albeit unevenly, around the K-12 map, this does not necessarily indicate that there is colinearity, and indeed the sizes of the two genomes are probably different, which could account for some of the gaps and sparse regions in Fig. 1. In all, 1,133 W14 clones exhibited no significant matches in even the lowest-stringency analysis (E < E⁰⁵), and from this we inferred that approximately 53% (1,133 of the 2,122 clones examined) of the P. luminescens genome is clearly distinct from the genome of E. coli K-12.

View larger version (71K): [in this window] [in a new window]   FIG. 1.   Graphic display of sample sequence similarities to E. coli K-12 nucleotide and protein sequences, generated from a BLAST search of P. luminescens sequences performed with K-12. The three concentric sets of data show nucleotide (outer ring) and protein (inner ring) hits plotted at the coordinates of the K-12 target. From the outside, the three data sets show hits with BLAST expected value (E) limits of <10e05 (2,765 protein and 729 nucleotide hits), <10e20 (1,227 protein and 376 nucleotide hits), and <10e40 (664 protein and 234 nucleotide hits), respectively. The positions of the genetic markers ori, ter, and rrnA through rrnH are shown as landmarks to orient the circle. The figure was generated by using the program Genescene (DNASTAR).

Old and new toxin complex (tc) loci. We previously cloned and sequenced four Tc-encoding loci, tca, tcb, tcc, and tcd, from P. luminescens W14 (22); each of these loci encodes a different high-molecular-weight insecticidal Tc (Tca, Tcb, Tcc, and Tcd, respectively). The Tc proteins are secreted into the supernatant by P. luminescens grown in liquid culture (23). Despite the fact that the Tc toxins exhibit both oral and injectable activity against a range of insects (22, 23), their precise role as potential virulence factors in the infection process remains to be determined. However, one of the complexes, Tca, has highly specific histopathological effects on the lepidopteran midgut (18), suggesting that Tca proteins may be used by the bacterium to destroy the insect midgut and effectively stop feeding. In the sample sequence analysis, BLASTX searches gave 19 hits for the four known tc loci (22), but 27 additional sequences (Table 1) were also identified that could not be ascribed to the previously identified tc loci after careful examination of the sequence chromatographs (Fig. 2A). This suggests that there are other tc-like loci in the P. luminescens W14 genome in addition to those already reported. The matches with new tc-like loci were classified as tca-like (3 hits), tcc-like (13 hits), or tcb/tcd-like (11 hits; tcb and tcd are close homologs of one another).

View larger version (30K): [in this window] [in a new window]   FIG. 2.   Diagrams showing the relative locations of hits to known tc loci and new tc-like loci. (A) Locations of individual sequencing reads (arrows above the diagrams) and their associated contigs and BLASTX matches (boxes below the diagrams). Note that the predicted amino acid sequences for tcb and tcd are sufficiently similar that we could not distinguish matches with either locus. (B) One example of a difference in genomic organization of a new tc-like locus inferred from a small contig and adjacent flipped sequence. Note that two TccC-like BLASTX matches are located next to a TcaC-like ORF with a phage remnant in between (see text). aa, amino acid.

Antibiotics and antibiotic resistance. Having destroyed the insect gut, presumably by using the tc-encoded Tc (18), P. luminescens must then defend the insect cadaver from a wide range of other colonizing organisms, such as bacteria (including other strains of P. luminescens), fungi, and/or nematodes. It seems reasonable to assume that during this process P. luminescens W14 deploys a range of antimicrobial agents, such as antibiotics and antifungal agents, as documented for other Photorhabdus strains and also Xenorhabdus strains (52), in order to maintain a bacterial monoculture in the insect cadaver. Thus, in the sample sequence one of the largest classes of hits was hits for polyketide synthetase-like genes. This class of genes is responsible for nonribosomal synthesis of a diverse array of compounds involved in processes ranging from fatty acid synthesis to antibiotic production (including production of inhibitors of eukaryotic protein phosphatases [41]). Even if we took into account the effects of the large sizes of some of the polyketide synthetase loci (up to 28 kb of repeated subunits), these classes of hits were still some of the predominant classes of hits in the sample sequence, accounting for 3.7% (80 hits) of the total sequences. Of the matches with polyketide synthetase-like sequences, 31 were with a syringomycin synthetase from Pseudomonas syringae pv. syringae (Table 2). Interestingly, the syringomycin synthetase gene cluster is thought to provide a link between prokaryotic and eukaryotic peptide synthetases (62), while syringomycin itself has a wide range of antibacterial and antifungal properties. Deployment of similar antibiotics by P. luminescens W14 may, therefore, help maintain a bacterial monoculture in an insect cadaver. Furthermore, it is interesting to note that P. luminescens also contains a sequence that is similar to tolaasin (another lipodepsipeptide), which is used for self-protection in Pseudomonas tolaasii, which implies that W14 may employ this peptidoglycan-associated lipoprotein in self-protection against its own antibiotics. In addition to potentially deploying broad-spectrum antibiotics to repel other organisms that might colonize the insect cadaver, strain W14 also contains sequences similar to colicin activity proteins (CeaAB), colicin transport proteins (BtuB), and pyocin immunity proteins (S3). A sequence similar to colicin lysis protein was not found, although there was a match with a similar VlyS lysis protein S from lambda phage (BLASTX E value, 2e-16). Although the role of the colicin- or pyocin-like sequences in P. luminescens remains to be determined, they may be used to produce toxins and antitoxins designed to kill non-self bacteria.

Rtx-like homologs. Another large class of database matches comprises sequences similar to both Rtx-like and hemolysin A-like toxins and their associated export systems (Table 4). The RTX (repeats in toxin) toxins are cytolytic toxins that are virulence factors in many pathogenic gram-negative bacteria (182). The RTX elements of other gram-negative bacteria share certain aspects of genomic organization, including the presence of three elements: an exported protein (RtxA-like), an ATP-binding cassette ABC protein (RtxB-like), and a membrane fusion protein (RtxD-like). Figure 3 shows the sequences similar to each of these elements alongside the loci to which they are most similar as determined by BLASTX searches. This figure shows that the Photorhabdus sample sequence contains sequences similar to the sequences of RtxA and RtxB of Vibrio cholerae, ShlA and ShlB of Serratia marcescens, EthA and EthB of Erwinia tarda, and HecA and HecB of Erwinia chrysanthemi. We also discerned sequences similar to both HlyB and CvaA/CvaB of E. coli, which are involved in hemolysin secretion and colicin V secretion, respectively. Notably, even if we took into account the large predicted ORF size (size of rtxA, ~12 kb), there were still 24 hits with RtxA-like sequences alone, suggesting that more than one locus may be present. Furthermore, BLASTX E values were highly significant (e-25 to e-78), suggesting that there is a high level of amino acid conservation. We also observed that there is a sequence similar to TolC which is unlinked but is required both for hemolysin export and for colicin V export (58). We can only speculate as to the number of loci that these sequences correspond to and to the likely role of the encoded toxins in P. luminescens infection. However, given the propensity of Rtx-like toxins to attack host phagocytes (182), we postulate that the sequences may be important in attacking the insect cellular immune system, the hemocytes. This hypothesis could be tested by deleting the toxin loci or their export machinery and examining the infection process in their presence and absence.

View larger version (26K): [in this window] [in a new window]   FIG. 3.   Inferred genomic organization of different putative Rtx-like operons (rtx, shl, eth, hec, hly, and cva) from the sample sequence. The relative predicted positions of sequencing hits are shown below each predicted locus, and the range of percent identity values is shown. The putative operons are shaded in order to indicate their potential functions as either an exported protein, a activity regulator, an ATP-binding protein, or a membrane fusion protein.

Other putative virulence factors. In addition to the specific Tc-like and Rtx-like toxins discussed above, we also identified a wide range of other sequences related to a diverse array of genes that are potentially involved in infection and virulence. These genes include genes that encode factors involved in bioluminescence, other proteases, lipases, hemmaglutinins, chitinases, and other toxins, such as non-Rtx hemolysins and ADP-ribosyltransferases (Table 5). They also include genes involved in two-component sensor systems that have previously been implicated in regulation of virulence both in Photorhabdus strains and in other bacteria.

Locomotion, attachment, and invasion. During its complex life cycle, a Photorhabdus strain not only needs to detect which environment it is in (e.g., nematode gut versus insect hemocoel) but presumably also needs to recognize specific surfaces for attachment and potentially invasion. Although we have little understanding of when and where Photorhabdus strains go during the insect infection process, we do know that their titers in the hemolymph change rapidly (52) and that during the infection process the insect midgut is specifically destroyed (18). Thus, we do not know if the bacteria replicate in the insect hemocytes or if they invade the gut directly. However, even in the absence of substantial information concerning the basic biology of these organisms, we can make inferences about the likely infection process based on the array of genes that they carry which are putatively involved in locomotion and tissue-specific attachment and/or invasion. Most notable in the latter case are two hits to the attachment invasion locus (ail) found in Yersinia spp. (Table 6). In Y. enterocolitica this locus is responsible for the ability of the pathogen to cross the epithelium of the gut on its way to replicate in the reticuloendothelial system. Again, although we have no direct evidence that a putative homolog plays a similar role in Photorhabdus strains, we can test whether P. luminescens W14 can invade the gut (presumably from the hemocoel, not the lumen) and, if it can, whether deletion of the ail-like locus interferes with this ability. With respect to attachment, we also found a sequence similar to intimins, which are proteins homologous to the invasins of Yersinia spp. and which play a role in attachment and effacing of the brush border membrane (54).

Secretion and transport. One of the most striking features of P. luminescens grown in a liquid culture is the large number of proteins that are secreted into the supernatant. Some of these proteins have been well characterized, including the Tc toxins, proteases, and lipases discussed above. However, most of the secreted proteins are poorly characterized, and, perhaps equally importantly, their mechanisms of export are not known. Thus, for example, the mechanism and timing of secretion of the Tc toxins in the insect host remain obscure. Below we discuss sequences similar to different types of secretion machinery, notably type III-like secretion systems and ABC transporters. We observed a series of hits to the Yop type III secretion system of Yersinia species, including sequences similar to both Yop proteins, Ysc secretion proteins, and Syc Yop-specific chaperones (Table 7). The yop virulon enables Yersinia cells to survive and multiply in the lymphoid tissues of their hosts (36). The Yop proteins are encoded on the pYV plasmid at the low-calcium-response locus, and virulent Yersinia cells secrete these virulence determinants when they are incubated at 37°C in the absence of Ca²⁺ ions. The Yop proteins themselves are involved in contact-dependent delivery of toxins and effector molecules. Thus, in P. luminescens they could potentially be responsible for delivering toxins to either the gut or the insect hemocytes. As discussed above, the virF virulence regulon transcriptional regulator (BLASTX E value, 2e-44) (Table 5) regulates production of Yop proteins. This gene is, therefore, a very interesting candidate for knockout in P. luminescens, as its loss may alter the pathogenesis of Photorhabdus cells with different insect tissues and potentially ascribe a function to the presence of the Yop-like sequences in strain W14.

Polysaccharide biosynthesis. Another striking feature of the P. luminescens culture supernatant is the large amount of LPS present. LPS production has been directly implicated in virulence in P. luminescens, as it has been in a wide range of other bacteria. For example, in B. pertussis the LPS is biologically active and is both toxic and immunogenic (5). LPS can also act as a recognition or binding site for extracellular agents. Thus, core LPS can act as a binding site for bacteriocins (alongside the outer membrane proteins OmpA and OmpF, as discussed above), while the trsG operon (Table 8) is required for biosynthesis of the bacteriophage Phi R1-37 receptor structures (158). The lipid A-core component of LPS is synthesized by sequential addition of sugars and fatty acids, and several sequences similar those involved in LPS biosynthesis were found in the sample sequence. These include envA (lpxC), which encodes an enzyme necessary for synthesis of the lipid A moiety (94), and rfaC, which is required for LPS inner-core synthesis (31). We also found genes likely to encode polysaccharide export functions, such as rcsF, which confers a mucoid phenotype (57), and wza, which encodes an outer membrane lipoprotein probably responsible for colanic acid (extracellular polysaccharide) export (161). Finally, genes encoding pullulanaselike proteins (starch-debranching enzymes) are also present; these proteins may play a role in recycling of the cell wall.

Iron acquisition and transport. As iron is often a rate-limiting growth factor in the host, many pathogenic bacteria have high-affinity iron-binding systems which can capture iron from host iron chelators. Thus, P. luminescens W14 has sequences which predict proteins similar to those involved in biosynthesis, transport, and reception of the siderophore yersiniabactin. Yersiniabactin (Ybt) has a high affinity for ferric iron, and similar siderophore-dependent iron transport systems are found in Yersinia pestis, Yersinia pseudotuberculosis, and Y. enterocolitica (140). A similar system may, therefore, also be used by P. luminescens. The irp1 and irp2 genes are required for yersiniabactin synthesis, as is ybtE, which encodes yersiniabactin dihydroxybenzoate ligase (Table 9). Transport of the iron-yersiniabactin complex back into the cell requires the TonB-dependent surface receptor FyuA, which may also be present in P. luminescens W14. This receptor is highly conserved and is found in all pesticin-sensitive bacteria, including E. coli (146). The sample sequence also contained hits to an R4-like ferric siderophore receptor from E. coli, which may perform a similar function in P. luminescens, and a putative operon (pvcABCD) involved in synthesis of the chromophore moiety of the P. aeruginosa siderophore pyoverdine (162).

Extrachromosomal elements. Like the genomes of other bacteria, the P. luminescens sample sequence contains many sequences similar to sequences found in a wide range of phage and insertion sequence elements. These sequences are important because they may begin to explain how P. luminescens, an insect pathogen, acquired virulence factors previously associated only with vertebrate pathogenicity, such as sequences similar to the low-calcium-response stimulon from Yersinia discussed above. Numerous hits to tail proteins from P2-like bacteriophages (46) (P2, P4, 186, and HP1) and a range of other phage-related proteins were observed (n = 51). There were also 10 hits (BLASTX P values, 0.01 to 2e-86) to products of the integrase (int) gene, which controls phage site-specific integration. Notably, in the range of phage homologies there were hits (although with relatively low significance [BLASTX E values, 0.3 to 5e-15]) to three different ORFs (ORFs 16, 20, and 25) of the P. aeruginosa cytotoxin converting phage Phi CTX (cholera toxin). We note that the rtx gene cluster is physically linked to the CTX toxin element in the V. cholerae genome (112). Therefore, it will be interesting to investigate whether this element is linked with the rtxA-like sequences found in P. luminescens W14, suggesting that it could have been responsible for horizontal transfer of the toxin-encoding genes. Numerous transposon-like sequences were also found (n = 33), including 10 hits to a transposase from plasmid ColIb-P9 (BLASTX E values, 1e-04 to 4e-86). Again, although these sequences indicate that transfer events occurred, it is not known how long these transposons have been present and if any of them have retained functionality. Finally, the P. luminescens W14 genome contains numerous sequences related to sequences involved in plasmid maintenance and stability (Table 10). However, we cannot at this stage distinguish which of these sequences are plasmid encoded (plasmids have been found previously in Xenorhabdus spp. [103]) and which are chromosomal. The presence of these sequences, therefore, raises the possibility of plasmid maintenance in P. luminescens W14 but is not strictly indicative.

Conclusions. P. luminescens has a life cycle which introduces it into a diverse array of environments, and in only one of these environments, the insect environment, is the bacterium pathogenic. The sample sequence of strain W14 revealed sequences similar to the sequences of a diverse array of potential virulence factor-encoding genes, including the genes for several classes of toxins, proteases, lipases, and LPS. It also gave us some indication of the diversity of the transport and metabolic systems present. Furthermore, Photorhabdus spp. also seem to share potential virulence factors (Yops, a yersiniabactin-like siderophore, and the low-calcium-response stimulon) with distantly related vertebrate pathogens, such as members of the genus Yersinia. This hypothesis is supported by the presence of numerous phagelike and transposon-like sequences in the P. luminescens genome. The potential for horizontal transfer raises the intriguing possibility that the virulence factors present in invertebrate pathogens may also be present in vertebrate pathogens. Given the far greater diversity of invertebrates and, potentially, their associated pathogens, this raises interesting questions about the diversity and origins of potential vertebrate virulence factors. In relation to P. luminescens itself, complete elucidation of the genome sequence of strain W14 and other strains should allow us to begin to understand the roles of individual genes via targeted disruption and to begin to compare the diversity of virulence factors found in different invertebrate pathogens. Our findings are consistent with the hypothesis of Burland et al. (30), who hypothesized that all of the pathogenic genes shared by enteric bacteria form a pool or "pathosphere"; however, here we emphasize that the pool must be extended to include both invertebrate and vertebrate pathogens. Furthermore, as invertebrates evolved before vertebrates, this also raises the interesting possibility that pathogens such as P. luminescens include the progenitors of virulence factors in vertebrate pathogens.