The serovars chosen for this study included the host-specific serovars Choleraesuis (pigs) and Dublin (cattle), which, in nonendemic herds, cause invasive disease, with localization in the meninges, lungs, joints, and reproductive tract (21
). Both these serovars also cause invasive disease in other species, including humans. Serovar Virchow and serovar Enteritidis can cause invasive disease in humans but are not common causes of disease in chickens, the food source implicated in many human cases. Serovars Typhimurium and Heidelberg, and to a lesser extent serovars Derby and Infantis, are common causes of enteric salmonellosis in humans and other animals. Serovars Ohio, Hessarek, Zanzibar, Bovismorbificans, and Ratchaburi are relatively rare causes of disease and are principally associated with enteritis (13
). With the exception of serovar Typhimurium, the variation observed in the SPIs was conserved within a serovar. Most of the isolates used in this study were independent isolates obtained from sick animals, humans, or the environment of outbreaks. For example, one isolate of serovar Virchow was of human origin, isolated in Queensland, Australia, and another isolate originated from poultry in Victoria, Australia. The serovar Enteritidis strains were isolated from two different countries but had identical RFLP patterns in the SPI regions.
Although the bootstrap values obtained for some of the branches on the tree inferred from the restriction site data were relatively low, the phylogenetic relationships between seven of the serovars used in our study (serovars Enteritidis, Dublin, Infantis, Heidelberg, Typhimurium, Choleraesuis, and Derby) have been studied previously using multilocus enzyme electrophoresis (2
), and for these serovars, our phylogenetic analysis based on the restriction endonuclease maps in the five pathogenicity islands yielded a tree with a topology very similar to those obtained previously using multilocus enzyme electrophoresis. The other six serovars in our study have not been included in multilocus enzyme electrophoresis studies, so the inference of phylogenetic relationships is dependent on the restriction endonuclease cleavage data. The benefit of using restriction map data across all five islands, rather than restricted sequence data on one or a few genes, is that these data are less likely to be influenced by horizontal transfer of individual segments within the islands, a process that has been suggested by recent work (6
). Indeed, such horizontal transfer between serovars might be expected to result in low bootstrap values, no matter what data were used, since the consensus tree would represent a compromise between multiple regions within the same genome with differing evolutionary histories.
Major genetic variation was detected in SPI-1, SPI-3, and SPI-5 for some serovars examined. SPI-2 and SPI-4 were found to be relatively conserved, with only minor variation in the presence or absence of restriction endonuclease sites. The only significant variation within SPI-1 was found in the avrA
gene. The detection of the deletion of avrA
within serovars Choleraesuis, Ohio, and Ratchaburi confirmed and extended the previous studies (15
) that found the deletion of this gene in serovars Typhi, Choleraesuis, Ohio, Montevideo, Othmarschen, Nienstaedten, and Arizona and examined the sequence of this region in serovar Typhi and serovar Choleraesuis. While AvrA is secreted by the SPI-1 type III secretion system (15
), its expression is not regulated by HilA or InvF (8
). Although it is similar to YopJ of Yersinia pseudotuberculosis
, expression of AvrA in either serovar Dublin or Y. pseudotuberculosis
showed that it does not exert a YopJ-like activity on cytokine expression or in killing macrophages (29
While avrA is deleted in the host-adapted serovars Typhi and Choleraesuis, it is present in serovar Dublin, another host-specific serovar, and it is present in serovars from a range of hosts and with a range of pathogenicities. There is no clear phylogenetic relationship between the serovars containing a deletion in avrA (Fig. ), suggesting that an essentially identical deletion may have occurred several times during the evolution of different Salmonella serovars. This, and the similarity of the sequences that have replaced the deleted region in the different serovars, suggests that the deletion has been generated by insertion and imperfect excision of a site-specific, transmissible genetic element, such as a phage or insertion element, or alternatively by horizontal transfer of a discrete region of the SPI. The characterization of the full significance and genesis of this deletion will depend on the determination of a function for AvrA and on detection of the sequence that has been substituted for avrA in some other location.
The general conservation within the remainder of the SPI-1 region confirmed the proposition (26
) that the genes within SPI-1 were an early acquisition. It has been suggested (22
) that SPI-1 was acquired before the divergence of S. bongori
and S. enterica
, whereas SPI-2 was acquired by S. enterica
after its split from S. bongori
. Our studies found that all the genes within SPI-2 were conserved and that the structure of SPI-4 was also conserved in all the serovars examined, with the only variation identified being loss or gain of restriction sites, presumably as a result of single base changes.
An insertion encoding 58 amino acids with similarity to PagJ and PagK between PipC and PipB in SPI-5 was found in serovars Derby and Ohio (Fig. ). The genes pagJ
are PhoP-activated genes and are nearly identical to each other (14
). The PhoPQ regulatory system is necessary for activation of invasion genes in response to environmental signals. The distribution of pagJ
is limited; Southern blot analysis identified hybridizing sequences only in serovar Typhimurium and serovar Enteritidis and not in serovar Typhi or serovar Paratyphi or a range of other Enterobacteriaceae
). However, there has not been any investigation of the distribution of these genes in other serovars. The amino acid sequences within and around pagJ
are similar to those of proteins associated with mobile or extrachromosomal elements, including transposases, phage proteins, and proteins encoded on plasmids (10
). Deletion of either pagK
does not attenuate virulence in mice, but this may be because they have similar or identical functions. Thus far, no single mutation of a PhoP-activated gene has resulted in significant changes in the phenotype (14
). It is possible that the insertion in SPI-5 is a remnant of pagJ
from a previous recombination event or that it is a functional homologue of these genes. The phylogenetic analysis suggested that the event that led to the insertion in SPI-5 was likely to have been a single event, since the serovars carrying this insertion were clustered on the tree, and the topology of the tree in this area was strongly supported by the statistical analysis (Fig. ).
The position of the genetic variation seen in SPI-3 in our work was similar to that seen for different subspecies (3
), using specific regions of each gene as hybridization probes. However, the serovars examined in our studies were all pathogenic serovars within subspecies I. The high prevalence of genetic variation within SPI-3 in subspecies I and the similarity of some of the inserted sequences to other virulence-associated genes provides more information on Salmonella
diversity, since information to date is derived mainly from studies of serovars Typhimurium and Typhi.
Most genetic variation in SPI-3 occurred at the left-hand end next to selC
in the region containing genes sugR
. This region is the integration site for many pathogenicity islands of enteric bacteria, including E. coli
, and also the retron phage of E. coli
. In serovar Ohio there was a deletion of 0.5 kb in sugR
. The roles of SugR and RhuM have not yet been reported, but their position suggests that they may have been acquired independently of the remainder of SPI-3 and, even though the intergenic region between them is 580 bp, they are likely to form an operon because a lacZ
fusion to this intergenic region resulted in some β-galactosidase activity (3
). The link between these two genes is further supported by the deletion of both of them in many serovars.
The similarity of the variation at the left end of SPI-3 in serovars Bovismorbificans, Infantis, Virchow, and Zanzibar suggests a close relationship between these serovars, and this was supported by phylogenetic analysis (Fig. ).
The insertion that replaced sugR
in serovar Ratchaburi and that lay immediately adjacent to selC
had a high similarity to retron phage ΦR73, which is located in a similar position on the chromosome of E. coli
K-12. The P4-like ΦR73 prophage is 12.6 kb and is flanked by 29-bp direct repeats and integrated 3′ to selC
). Even though part of the prophage is similar to bacteriophage P4, it has a unique region at the right end that differs from P4. The retron-Ec73 region encodes the gene for RT and the genes msr
, and orf316
. The low similarity of RT-Ec73 to other bacterial RT genes, for example, 27% identity to RT-Ec67 of E. coli
(more than 47 amino acid residues), 25% identity to RT-Ec86 of E. coli
(more than 43 amino acid residues), and 26% identity to RT-Mx65 of Myxococcus xanthus
(more than 174 amino acid residues), indicates that it is equally distant from the other bacterial RTs (33
). The high level of DNA sequence similarity between RT-Ec73 and the insertion in serovar Ratchaburi implies a much closer relationship than that between RT-Ec73 and other bacterial RTs. However, the insertion in serovar Ratchaburi contains a significant deletion compared to the full-length phage, with more than 80% of the genome missing and the integrase gene disrupted. This suggests that the phage genome in serovar Ratchaburi is no longer fully competent, although at least two ORFs from the phage, including the RT gene, appear to be translationally competent. The presence of retron phage sequences within an SPI suggests that such phages may also have played a role in the evolution of SPIs. It is well established that retroviruses of eukaryotes can acquire host sequences. It may be that retron phages similarly acquire and transfer sequences from one host to another.
The 10-kb insertions adjacent to selC in serovars Derby and Hessarek were similar to each other but not to any SPI-3 genes. The majority of the insertion appeared to be composed of an operon for synthesis of fimbriae, although, surprisingly, this operon was most similar to an E. coli fimbrial operon. The inserted operon was bracketed by the remnants of two distinct transposases, suggesting the involvement of at least two different insertion elements in the evolution of this sequence. It is notable that both of the major insertions observed in SPI-3 were likely to have been derived from E. coli, implying that the evolution of pathogenicity islands is an ongoing process in pathogenic salmonellae and that the sources of the elements are likely to be other enteric bacteria. A surprising finding was that although serovars Hessarek and Derby do not appear to be closely related phylogenetically (Fig. ), they shared the same insertion, suggesting that this element has inserted independently into different serovars or that horizontal transfer of this region has occurred.
The possibility that a single event led to the deletion of sugR and rhuM in serovars Bovismorbificans, Infantis, Virchow, and Zanzibar is refuted by the observation that exactly the same apparent deletion was observed in serovar Ratchaburi, although in this case the deletion was replaced by an inserted sequence derived from a retron phage. Furthermore, a very similar deletion appears to have occurred during the insertion of the fimbrial operon into serovar Hessarek and serovar Derby. For serovar Typhi, sugR and rhuM have also been substantially deleted, with pseudogene remnants, pseudogenes of transposase genes, and two short ORFS with no significant similarity to other sequences in the databases. These findings suggest that there is a region within SPI-3, immediately adjacent to selC and 5′ to rmbA, that is particularly prone to deletion and/or insertion of transposable elements from a variety of sources and that the acquisition of the sugR/rhuM region in serovar Typhimurium is likely to have been a relatively recent event. It seems most probable that the ancestral SPI-3 sequence in this region was most similar to the sequence for serovars Bovismorbificans, Infantis, and Zanzibar, with multiple insertions, deletions, and short sequence duplications giving rise to the variations seen with the other serovars. It is clear that the region immediately following selC has been the target of a variety of insertion sequences, including at least two distinct transposable elements, similar to IS3 of E. herbicola and IS630 of S. sonnei, and a retron phage most similar to ΦR73 of E. coli. A fourth element is likely to have introduced the sugR/rhuM region into some serovars.
The high level of variability of DNA sequences adjacent to selC and the similarity of selC-related DNA sequences of other enteric bacteria to those in the serovars investigated here imply the occurrence of genetic transfer of sequences in this region between enteric bacteria and imply that in contrast to the other SPIs, SPI-3 may be still evolving through major sequence acquisitions.
Although the Salmonella serovars used in these studies cause a range of syndromes in both humans and animals, the major genetic variations identified were not significantly correlated with prominent differences in clinical features, host range, or levels of virulence. It may be that the variant genes are not important in pathogenesis or that there are other Salmonella genes that have the same action. Alternatively, the effect of these deletions and insertions may be reflected in less obvious benefits, for instance, in maintenance hosts that rarely suffer from disease or in survival in the environment. Understanding such issues will require a broader focus on the molecular biology of the ecology of these organisms.
This study has shown that even in relatively recent evolutionary history there have been a number of major genetic events shaping SPI-3, all of which appear to be focused on one specific region. Further studies of this region with a range of other serovars, using the techniques developed in this work, may reveal the range of transmissible elements that can contribute to the ongoing evolution of pathogenicity islands and thus provide a better understanding of the history of the more stable islands in both Salmonella and other enteric bacteria.
Additional information on this study, including oligonucleotide primers used, PCR conditions, maps of restriction endonuclease sites in different serovars, and aligned sequences, can be obtained from our website at ftp://jb:email@example.com/.