MBE Advance Access originally published online on September 1, 2006
Molecular Biology and Evolution 2006 23(12):2342-2354; doi:10.1093/molbev/msl103
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Research Articles |
Diversifying Selection Drives the Evolution of the Type III Secretion System Pilus of Pseudomonas syringae
Department of Botany, Centre for the Analysis of Genome Evolution and Function, University of Toronto, Toronto, Ontario, Canada
E-mail: david.guttman{at}utoronto.ca.
 |
Abstract |
|---|
 TOP Abstract IntroductionMaterial and MethodsResultsDiscussionAcknowledgementsReferences |
|---|
The plant pathogenic bacterium Pseudomonas syringae uses a type III secretion system to inject virulence proteins directly into the cytoplasm of its hosts. The P. syringae type III secretion apparatus is encoded, in part, by the HrpZ operon, which carries the hrpA gene encoding the pilin subunit of the pilus, various components of the structural apparatus, and the HrpZ harpin protein that is believed to produce pores in the host cell membrane. The pilus of the type III system comes into direct contact with the host cell and is, therefore, a likely target of the host's pathogen surveillance systems. We sequenced and analyzed 22 HrpZ operons from P. syringae strains spanning the diversity of the species. Selection analyses, including Ka/Ks tests and Tajima's D, revealed strong diversifying selection acting on the hrpA gene. This form of selection enables pathogens to maintain genetic diversity within their populations and is often driven by selection imposed by host defense systems. The HrpZ operon also revealed a single significant recombination event that dramatically changed the evolutionary relationships among P. syringae strains from 2 quite distinct phylogroups. This recombination event appears to have introduced genetic diversity into a clade of strains that may now be undergoing positive selection. The identification of diversifying selection acting on the Hrp pilus across the whole population sample and positive selection within one P. syringae lineage supports a trench warfare coevolutionary model between P. syringae and its plant hosts.
Key Words: type III secretion system (T3SS) • diversifying selection • pilus • recombination • Pseudomonas syringae • HrpA
|
Introduction |
|---|
|
TOPAbstractIntroductionMaterial and MethodsResultsDiscussionAcknowledgementsReferences |
|---|
The type III secretion system (T3SS) is one of the most notorious virulence systems carried by pathogenic bacteria. This specialized protein secretion system is conserved among a wide range of both animal pathogens (e.g., Escherichia coli, Salmonella enterica, Shigella spp., and Yersinia spp.) and plant pathogens (e.g., Pseudomonas syringae, Xanthomonas spp., and Erwinia spp.) and is required for the full bacterial virulence (Hueck 1998
; Zaharik et al. 2002; Bretz and Hutcheson 2004; He et al. 2004). The T3SS is typically only expressed and assembled to appreciable levels after the bacterium comes into direct contact with a eukaryotic host. The assembly of the secretion apparatus includes the production of the T3SS pilus, which forms a direct conduit between the pathogen and its host. This pilus is used to inject T3SS effector proteins directly from the bacterial cell into the cytoplasm of its host, where they target host proteins and modulate the host defense response. T3SS effectors are known to suppress the defense response by interfering with signal transduction, causing cytoskeletal changes, or by having direct cytotoxic effects.
The T3SS is encoded by a regulon of approximately 20 genes clustered in a classic pathogenicity island (fig. 1) (Dobrindt et al. 2004). In the plant pathogen P. syringae, this island is called the hrp cluster because the system is required for both the hypersensitive response (an important component of the plant defense response) and pathogenesis. Genes in the hrp cluster that are conserved among a wide range of pathogens are called hrc genes, for hypersensitive response and conserved. The genes that actually encode the T3SS apparatus are encoded on 4 operons in the hrp cluster. One of these operons is the HrpZ operon and is particularly interesting because it contains genes that encode the pore-forming hairpin HrpZ; the pilus subunit HrpA; a conserved HrcJ protein that is associated with both inner and outer bacterial membranes by its C-terminal hydrophobic domain and its N-terminal lipid moiety, respectively (Holmes et al. 1999); and 3 other proteins of unknown function, HrpB, HrpD, and HrpE.
|
The type III secreted HrpZ harpin binds to lipid bilayers and is believed to form ion-conducting pores in the host cell membrane. This protein is a potent inducer of the plant defense response and, in fact, is the only P. syringae T3SS effector that can induce the host defense response from the outside of the host cell.
HrpA proteins self-assemble into the extracellular T3SS pilus (Roine, Saarinen, et al. 1997; Roine, Wei, et al. 1997) and form the conduit through which all T3SS effectors, including HrpZ and HrpA itself, must pass through on their way from the bacterial cytoplasm to the host cell (Li et al. 2002). Mutational analysis of HrpA revealed that insertions in the C terminus of the protein created dominant negative mutations that are neither able to cause disease nor able to induce the host defense response (Taira et al. 1999; Lee et al. 2005). These mutations presumably act by disrupting pilus formation, thereby blocking the secretion of T3SS effectors.
The necessity of the T3SS and its intimate interaction with the host cell exposes it to intense and often competing selective pressures. On one hand, the system must vary enough so that it does not provide an obvious target for the pathogen surveillance systems of the host, whereas on the other hand, it must be conserved enough to retain its essential function. Conserved bacterial factors (pathogen-associated molecular patterns, PAMPs) are extremely important inducers of the host defense response (Parker 2003; Nurnberger et al. 2004; Zipfel and Felix 2005). Because the T3SS apparatus and, most particularly, its pilus are at the very front lines of bacterial pathogenesis, it would not be surprising if they were PAMPs. However, the P. syringae T3SS apparatus, is not known to be a PAMP and does not induce a defense response. In fact, the only hrp-encoded protein that is a classic PAMP is the secreted HrpZ harpin (Brown et al. 2001; Lee et al. 2001; Nurnberger and Lipka 2005). Interestingly, the YscF protein that forms the base of the Yersinia T3SS apparatus does induce an antibody response when injected into mice (Matson et al. 2005), implying that mammalian immune systems may target the structural (pilus) needle component of T3SS.
Why do hosts not evolve mechanisms to recognize essential pathogenicity factors such as the T3SS? Two models have been proposed to explain the coevolutionary process that drives pathogen–host interactions: the arms race model and the trench warfare model. The arms race model is based on the assumption that pathogens and their hosts are constantly engaged in a deadly game of one-upmanship. Each is evolving counter measures to the defenses of the other. From a population genetic perspective, the arms race model predicts that positive selection will act on both the bacterial virulence factors and the host defense systems. Positive selection may result in repeated selective sweeps of favorable alleles through populations. As these favorable alleles increase in frequency in the population, there will be a corresponding loss of other alleles, resulting in relatively little variation at the selected loci, and other linked loci, and an excess of alleles carrying the selected amino acid substitution. Positive selection is most easily identified by an increased rate of amino acid–modifying substitutions (nonsynonymous changes) relative to the rate of synonymous or silent changes. From a coalescent perspective, positive selection results in a star genealogy, where all lineages radiate out from a common ancestor. The relatively low rate of recombination among some bacterial species can link a favorable allele to the rest of the genome, and this genetic hitchhiking may result in the repeated rise and eventual demise of clonal groups, a process called periodic selection (Atwood et al. 1951) or epidemic population dynamics. These evolutionary dynamics have been identified in species such as Neisseria meningitidis (Zhu et al. 2001) and Pseudomonas aeruginosa (Wang PW, Morgan RL, Guttman DS, unpublished data) (Lomholt et al. 2001; Pirnay et al. 2002; Salunkhe et al. 2005).
The trench warfare coevolutionary model posits that evolutionary pressures select for the maintenance of polymorphism within pathogen and host populations. Conceptually, this maintenance comes about because individuals of different genotypes are selectively favored under different circumstances. Selection that maintains genetic variation within or among populations has classically been referred to as balancing or diversifying selection. A number of different mechanisms have been invoked to explain this evolutionary model, including frequency-dependent selection, heterozygote advantage in diploids, and population structure. Regardless of the underlying evolutionary mechanism, the population genetic consequences are fairly clear. Diversifying selection will result in an excess of polymorphism at the selected locus (and linked sites) or, put in a coalescent context, the time to most recent common ancestry will be greater than expected under neutrality. It is important to make clear that diversifying selection does not exclude positive selection, so long as the positively selected allele is not able to displace all other alleles in the population. Diversifying selection has been shown to be extremely important in driving the evolution of animal immune systems, such as the human leukocyte antigen or mammalian major histocompatibility complex (Potts and Slev 1995; Prugnolle et al. 2005), the plant innate immune system (McDowell et al. 1998; Stahl et al. 1999; Mauricio et al. 2003; Allen et al. 2004), and most importantly for the current study, antigenic diversity in pathogens (Smith et al. 1995; Boyd and Hartl 1998; McGraw et al. 1999; Polley and Conway 2001; Baum et al. 2003; Wang et al. 2003; Kalia and Bessen 2004; Schurch et al. 2004; Wildschutte et al. 2004).
Very little evolutionary analysis has been performed on the genes that encode the T3SS. Sawada et al. (1999) sequenced 2 regulatory genes located on opposite ends of the P. syringae hrp cluster and 2 housekeeping genes among 56 strains. Their phylogenetic analysis indicated that the 2 hrp genes share a common evolutionary history with the housekeeping genes, indicating that the P. syringae T3SS has been maintained since the origin of the P. syringae species complex. One of the T3SS regulatory genes in the study, hrpS, is located immediately upstream of the HrpZ operon. Weber and Koebnik (2006) recently analyzed the Xanthomonas campestris hrpE locus, which encodes its T3SS pilus. This study identified the footprint of positive selection through an analysis of synonymous and nonsynonymous substitutions.
Here we report a population genetic and phylogenetic analysis of the P. syringae HrpZ operon. We find that the gene encoding the T3SS pilus, hrpA, is under diversifying selection, presumably to maintain genetic diversity in order to avoid detection by the host innate immune system. Although the evolutionary history of genes in the hrp cluster have been reported to be congruent with the evolutionary history of the P. syringae core genome (Sawada et al. 1999), we find clear evidence of recombination between 2 of the major P. syringae phylogroups in the HrpZ operon. The implications of these findings are discussed from the perspective of pathogen–host interactions.
|
Material and Methods |
|---|
|
TOPAbstractIntroductionMaterial and MethodsResultsDiscussionAcknowledgementsReferences |
|---|
Bacterial Strains and Growth Conditions
Bacterial strains (table 1) were grown overnight with continuous shaking in 5 ml of King's medium B (King et al. 1954
) at 30 °C for P. syringae or in 5 ml of Luria-Bertani medium (Sambrook et al. 1989) at 37 °C for E. coli. All strains used are listed in table 1.
|
Genomic DNA Extraction and HrpZ Operon Amplification
Genomic DNA was prepared using the PureGene DNA purification kit (Gentra Systems, Minneapolis, MN), according to the manufacturer's instructions. The HrpZ operon was polymerase chain reaction (PCR) amplified from the P. syringae strains using a universal hrcC-19 primer (5'-CCCGATCAACAATAAAGGCAACCA-3') and a degenerate hrpS + 506dgn primer (5'-GMARTTTCGDCGSGAYCTGTAYTTTCGCCT-3'). The exceptions are strains PmaH7311, which used a strain-specific hrpS primer (hrpS + 21 Psm; 5'-GGATATCTGCGTGATCGCCTCCG-3'), and strains PlaN7512, PmoM301020, and PmpFTRS_U7, which used a second specific primer (hrpS + 87 Psmml; 5'-GTCGGGATCTGTACTTTCGCCTTAACG-3'). PCR amplification of the operon was performed on 650–800 ng of template DNA using Platinum Pfx DNA polymerase (Invitrogen, Carlsbad, CA), following the manufacturer's protocol. Reactions included 30 µM dNTPs and 10 µM of each primer. Cycling conditions were performed on a Hybaid PCR Express thermal cycler as follows: an initial denaturation at 94 °C for 5 min, then 35 cycles of amplification with denaturation at 94 °C for 15 s, annealing at 60 °C for 30 s, and extension at 68 °C for 5 min.
Shotgun Cloning, DNA Sequencing, and Assembly
A minimum of 3 independent PCR reactions were performed on the HrpZ operon for each strain to reduce PCR-induced sequencing errors. The reactions were then pooled and independently restriction digested with Bsh1236I, HincII, or RsaI (Fermentas, Hanover, MD) overnight at 37 °C. Fragments were purified using the QIAQuick PCR purification kit (Qiagen, Valencia, CA) according to the manufacturer's instructions and were shotgun cloned into the pBluescript SK vector (Stratagene, La Jolla, CA). The resulting ligation mixtures were transformed into chemically competent E. coli DH5
cells.
Individual colonies were picked and dipped into a standard 25-µl PCR mixture containing 0.1 µM M13 forward and reverse primers. The cycling conditions were 1 cycle of 94 °C for 5 min; 35 cycles of 94 °C for 30 s, 55 °C for 30 s, and 72 °C for 2.5 min; and 1 cycle of 72 °C for 5 min. Products larger than 600 bp were cleaned enzymatically by the addition of 0.2 µl of calf intestinal phosphatase and 0.2 µl of exonuclease I (New England Biolabs, Beverly, MA), and incubated at 37 °C for 30 min, followed by inactivation at 85 °C for 15 min.
Sequencing was performed with the CEQ Sequencing Quick Start kit (Beckman Coulter, Fullerton, CA) on a Beckman Coulter CEQ8000XL DNA sequencer. A single 10-µL cycle sequencing reaction contained 2.5 µl of cleaned PCR product, 2.5 µl Beckman dye terminator cycle sequencing mix, 1.5 µl of 1x PCR buffer with MgCl2, 3 µl of water, and 0.5 µl (0.5 µM) of either T3 or T7 primer. The cycling conditions were 96 °C for 20 s, 55 °C for 20 s, and 60 °C for 4 min, repeated 55 times. After ethanol precipitation, the products were resuspended in Beckman sample-loading solution. Chromatograms were reviewed and edited with BioEdit (v. 7.0.4.1
[EC]
, http://www.mbio.ncsu.edu/BioEdit/bioedit.html) and were subsequently imported into either the contig assembly program Sequencher (Gene Codes Corp., Ann Arbor, MI) or the STADEN package (Staden 1996). Gaps were closed by PCR amplification from contig ends (primer sequences available from authors upon request). A minimum of 6x coverage was obtained for all the strains.
Data Analyses
The data set was supplemented with full-length HrpZ operons obtained from the National Center for Biotechnology Information (NCBI) sequence databank via Entrez and BlastP searches (table 1). We also attempted to obtain the source strain for each of these records for multilocus sequence typing (MLST). Alignments were performed on the amino acid sequence with ClustalX and back-translated using the Tranalign module of EMBOSS (Sarachu and Colet 2005). Analyses were performed on individual genes as well as on the concatenated data set. Neighbor-Joining (NJ) trees were generated in MEGA ver3 (Kumar et al. 1994) using the K2P evolutionary model with gamma correction of 0.2 (determined via PAUP*) and 1,000 bootstrap replicates. Maximum likelihood (ML) and maximum parsimony trees were generated in PAUP* ver4.0b10 for UNIX (Swofford 1993) and PHYLIP ver3.6 (Felsenstein 1993). Phylogenetic congruence between ML gene trees was tested using the Shimodaira–Hasegawa (SH) test (Shimodaira and Hasegawa 1999), which determines the likelihood of a data set given alternative trees. The SH test was implemented via the PHYLIP program DNAML using ML-generated trees.
Population genetic statistics and tests of selection were calculated with DnaSP ver4.10 (Rozas J and Rozas R 1999) and DAMBE (Xia 2000). Genetic distances were calculated using a Kimura 2-parameter model with 
correction of 0.2 using MEGA ver3 (Kumar et al. 1994). Recombination rates were also measured using a coalescent-based method as implemented in LDhat (McVean et al. 2002). The CodeML module of PAML ver3.14 (Yang 1997) was used to identify codons and lineages under positive selection. HKA (Hudson et al. 1987) and McDonald–Kreitman (McDonald and Kreitman 1991) tests could not be performed on hrpA because there are no homologs outside of P. syringae, making it impossible to get divergence data.
For analytical reasons, we only used NCBI records that contained full HrpZ operons, although many more accessions were available that contained individual gene sequences. A preliminary population genetic analysis performed on over 50 hrpA sequences obtained from NCBI revealed absolutely no significant differences when compared with the full analysis done on the complete 22-strain data set.
|
Results |
|---|
|
TOPAbstractIntroductionMaterial and MethodsResultsDiscussionAcknowledgementsReferences |
|---|
Phylogenetics
We sequenced the complete T3SS HrpZ operon from 15 strains of P. syringae and downloaded the DNA sequence data for full HrpZ operons for an additional 7 strains from NCBI (table 1). We used MLST (sequence data from 4 housekeeping genes: sigma factor 70, rpoD; DNA gyrase B, gyrB; citrate synthase, gltA; and glyceraldehyde-3-phosphate dehydrogenase, gapA) to characterize the "core genome" for 21 of the 22 strains (Sarkar and Guttman 2004
; Hwang et al. 2005), but could not obtain strain Pcm HL1 for MLST typing. All population genetic analyses were performed on the full 22-strain data set (table 1), whereas phylogenetic analyses were performed on the 21-strain data set in order to permit comparisons with the core genome MLST phylogenetic analysis. Major phylogenetic groups identified in the HrpZ operon will be referred to simply as clades, whereas major phylogenetic groups identified in the MLST analysis of the P. syringae core genome will be referred to as phylogroups.
Independent phylogenetic analyses of the loci encoded within the HrpZ operon revealed a very high level of congruence indicative of a common evolutionary history (fig. 2). In all cases, the same 5 major clades were recovered, and an SH test (Shimodaira and Hasegawa 1999), which determines if the gene tree derived directly from that gene's sequence is a better fit to the data than a gene tree derived from another locus, clearly demonstrates that the gene genealogies from all of the loci except hrpA are congruent with each other and with the genealogy built from the entire operon (table 2). Although the hrpA genealogy has the same 5 clades as the other loci, the SH test determined that its genealogy did not fit as well with the data from the other genes or the full operon (with the exception of the hrpB data).
|
|
The most notable difference between the hrpA gene genealogy and the other gene genealogies is the depth of the tree, which is best measured by the pairwise nucleotide diversity (
) among strains. The per nucleotide pairwise diversity (using both synonymous and nonsynonymous sites) for hrpA is 0.364 ± 0.038 (standard deviation), whereas the average for the 5 remaining loci in the operon is 0.203 ± 0.017 (fig. 3), indicating that hrpA is significantly more diverse than any other locus in the operon. We confirmed these results using the coalescent simulation feature of DnaSP to calculate the 95% confidence limits on given the observed number of segregating sites (data not shown). Interestingly, although hrpA is substantially more diverse than the other loci, the average diversity found within the major clades at the hrpA locus is no larger than that found at the other loci. The average, within-clade for the 5 major clades is 0.025 for hrpA and 0.021 for the other loci in the operon (fig. 3 and table 3). Consequently, the primary difference between the hrpA tree and the other trees is the length of the internal branches connecting the 5 major clades. Additionally, the hrpA tree has an extremely short branch leading to the clade that contains the P. syringae phylogroup 3 strains (primarily bean pathogens; see the phylogroup notation on the MLST tree in fig. 3).
|
|
Although all of the loci in the HrpZ operon exhibit identical clade structure, this large-scale branching pattern does not entirely match that of the MLST core genome tree. An alignment of the MLST core genome tree and a gene tree of the entire HrpZ operon (fig. 3) clearly shows that 2 strains (Pla N7512 and Pmp FTRS_U7) that should tightly cluster with phylogroup 3 strains instead cluster with highly divergent phylogroup 5 strains (Pma ES4326, Pma YM7930, and Pci 0788_9). The pairwise nucleotide diversity among the 2 strains that should cluster in phylogroup 3 (based on the MLST phylogenetic analysis) and the 3 phylogroup 5 strains is 0.060 for the MLST tree but only 0.026 when the same strains are examined in the full HrpZ operon tree. This is particularly notable given that the depth of the HrpZ operon tree is nearly 4 times that of the MLST tree. The only possible explanation for these data is a recombination event that mobilized a phylogroup 5–like HrpZ operon (or perhaps more of the hrp cluster) into an ancestral phylogroup 3 strain that gave rise to Pla N7512 and Pmp FTRS_U7.
Recombination
The unusual relationships among the phylogroup 3 and 5 strains brought about by a putative recombination event at the HrpZ operon are surprising because Sawada et al. (1999) did not find evidence that horizontal gene transfer influenced the P. syringae hrp cluster based on their observation of congruence between phylogenies of 2 loci from the hrp cluster (hrpL and hrpS) and 2 housekeeping genes (rpoD and gyrB). The HrpZ operon is less than 100 bp downstream from the end of hrpS, whereas the hrpL locus is less than 15 kb upstream of the end of the HrpZ operon. The most likely explanation for the discrepancy between our data and that of Sawada et al. (1999) is simply a lack of sampling of the relevant strains in the Sawada study. They only uncovered 3 of the 5 major phylogroups in the P. syringae species complex. Alternatively, it is possible that the recombination event that transported the phylogroup 5 HrpZ operon into the phylogroup 3 strains actually occurred within the operon itself. A recombination analysis of the loci in the HrpZ operon reveals that this latter explanation is less likely than the former. We used LDhat to estimate 
, the per nucleotide rate of recombination, and DnaSP to estimate Watterson's 
, a measure of the polymorphism based on the total number of segregating sites at a locus (table 4). The LDhat estimate of is a coalescent approach based on the composite likelihood estimator first developed by Hudson (2001). We can determine the relative importance of mutation versus recombination in generating diversity at a locus by taking the ratio of to (Guttman and Dykhuizen 1994; Maynard Smith and Smith 1998; Feil et al. 1999). / for the first 3 loci in the HrpZ operon (hrpA, hrpZ, and hrpB) is approximately 17 and decreases to approximately 7 for the last 3 loci (hrcJ, hrpD, and hrpE). In contrast, Sarkar and Guttman (2004) showed that / = 4 for the 7 housekeeping genes used in the original MLST study. Therefore, recombination appears to actually be suppressed in the HrpZ operon, particularly so in the first 3 genes.
|
We also used Hudson's estimator (Hudson and Kaplan 1985
) to determine the minimum number of recombination events (Rm) that occurred in our sample, a method that is based on the 4-gamete test. In contrast to the coalescent recombination approach, Rm was actually substantially higher for hrpA than for any other locus in the operon (table 4). An alternative and perhaps more likely explanation for the relatively high Rm is that the high level of polymorphism at these loci has introduced a significant number of homoplasies that would inflate the calculation of Rm.
Selection
An examination of the pattern of polymorphism in the HrpZ operon strongly supports the contention that diversifying selection is acting on the hrpA locus. Figure 4 shows a sliding window analysis of polymorphism () of the individual loci in the HrpZ operon using both synonymous and nonsynonymous sites. There is a high level of variation throughout the hrpA locus not seen in other loci, which is consistent with the phylogenetic analysis. The overall value of for the entire locus is 0.219 (table 4), nearly twice as high as any other locus in the operon. The pairwise nucleotide diversity () is also higher for the hrpA locus, with the most dramatic difference being seen when analyzing only nonsynonymous substitutions. hrpA has twice the number of nonsynonymous substitutions than that of the locus with the next highest number in the operon and 3 times the average for the 5 other loci in the operon. These observations support the hypothesis of diversifying selection at the hrpA locus but could also be consistent with a simple relaxation of selective constraints.
|
We performed a number of selection tests to clarify if the patterns of polymorphism seen in hrpA are due to diversifying selection or simply a relaxation of purifying selection. Tajima's D evaluates discrepancies between measures of polymorphism based on the number of segregating sites (
) and the pairwise nucleotide diversity () to identify genes under positive or diversifying selection. Significant positive Tajima's D values indicate that there is an excess of polymorphisms at intermediate frequency and is typically associated with diversifying selection. No locus in the HrpZ operon deviated significantly from neutrality; however, when Tajima's D scores were calculated using a sliding window approach, a number of internal regions were identified as having significantly positive values (identified by boxes along the bottom of fig. 4), most notably, small stretches at the 3' end of hrpA and 5' end of hrpZ. The region identified at the 3' end of hrpA corresponds to that part of the gene that is believed to be involved in HrpA subunit polymerization for the formation of the pilus (Taira et al. 1999; Lee et al. 2005). The region identified in hrpZ corresponds to the most conserved region of this gene and the only part of the gene with similarity to a conserved domain. The NCBI conserved domain database identified a low significance hit for this locus (E value = 0.003) to a predicted membrane protein (COG4907).
To further corroborate our findings, we also conducted the tests of selection of Fu and Li's (1993) (D* and F*). These tests use the number of singletons in a data set to infer the number of changes at the tips of a phylogeny relative to the total number of changes in the entire tree. An excess of changes at the tips indicates positive selection. The hrpA, hrpZ, and hrpB loci were all found to give significantly positive Fu and Li D* statistics, although none were significant by the F* test. A sliding window analysis found significant D* throughout the hrpA locus, clustered near the 5' region of hrpZ that corresponded to a region of reduced polymorphism and scattered throughout the hrpB locus (identified by "x" along the bottom of fig. 4). There were also numerous subregions of the other 3 loci that showed windows of significance by these tests. Despite these intriguing results, it is not clear that the Fu and Li tests are very powerful for the HrpZ operon data, mainly because these tests assume that singletons represent recent mutations that occurred at the tips of a phylogeny. This assumption is likely violated in highly polymorphic locus such as those found in the HrpZ operon.
Another powerful approach to identifying selection relies on comparing the rate of nonsynonymous substitutions (Ka) with the rate of synonymous substitutions (Ks). Ka/Ks ratios greater than 1.0 indicate positive selection. A Ka/Ks test performed using the Nei–Gojobori estimator (Nei and Gojobori 1986) as implemented in DnaSP (table 4) confirmed that the hrpA locus was least constrained, with Ka/Ks = 0.42, whereas the conserved structural gene hrcJ was the most highly constrained, with Ka/Ks = 0.07. The other genes in the operon had Ka/Ks ratios around 0.15–0.2. All of these values are well within the range of moderately conserved genes. A limitation of this approach is that selection acting on individual lineages or a small number of amino acids may be missed. We used PAML to perform maximum likelihood analysis of selection at the codon level (identified by "+" along the bottom of fig. 4). Likelihood ratio tests (LRTs) for positive selection were performed between the M3 (discrete) and M0 (single Ka/Ks) models, the M2 (positive selection) and M1 (neutral) models, and the M8 (beta distribution + positive selection) and M7 (beta distribution) models (Yang 1997, 2000). All loci except hrpB showed highly significant (P < 0.0001) LRTs for the M3–M0 comparison, but none of these loci had a maximum Ka/Ks class above 1.0 (table 5). The most interesting result from this analysis was the finding of a positive selection at hrpA amino acid 34 (relative to the Pto DC3000 sequence), which encodes a serine (Ka/Ks = 1.403, CodeML model 2) corresponding to a protein-processing site identified by Roine, Saarinen, et al. (1997). We were also surprised to find that the generally conserved hrcJ structural gene had 5 positively selected codons near its 3' end—a region of significantly higher polymorphism. This region corresponds to a hydrophobic domain that is putatively associated with the bacterial inner membrane.
|
We can refine this selection analysis by examining Ka, Ks, and Ka/Ks ratios for each of the major clades to determine if selection has been favoring the allele(s) of one lineage or another. Figure 5 shows Ka, Ks, and Ka/Ks ratios (calculated using a branch model analysis of the CodeML module of PAML) (Yang 1998
; Yang and Nielsen 1998) for each clade and gene in the HrpZ operon. First, given the Ka and Ks for each clade and gene, it is clear that the unusual pattern of polymorphism seen in hrpA are not simply due to a relaxation of selective constraints on this locus. If this were the case, we would expect a general increase in the nonsynonymous substitution rate for all clades at this locus. Instead, we see an average to below average rate of nonsynonymous substitutions for clades 1 through 4 but a dramatic increase in the rate for clade 5. The Ka/Ks ratio for the clade 5 strains at the hrpA locus is twice that of any other clade in any of the loci in the HrpZ operon. Second, the unusual Ka measure in this clade cannot be due to a greater time to common ancestry for these alleles because the phylogenetic congruence analysis clearly indicates that all of the loci within the operon have shared a common evolutionary history. Therefore, the hrpA alleles carried by strains in clade 5 must be experiencing stronger positive selection than the rest of the alleles in the population, perhaps indicating that these alleles are locally favored. The specific site of this selection can be inferred by looking at the distribution of nonsynonymous substitutions across the protein (fig. 7). Of the 9 nonsynonymous substitutions that occur among the clade 5 alleles, 6 occur within the first 18 amino acids, with 4 being clustered between amino acids 14 and 18 (fig. 6). All of the substitutions except one occur in putatively unstructured regions of the protein (Lee et al. 2005). It may be noteworthy that the single significant recombination event we identified in the HrpZ operon modified the evolutionary relationships of 2 strains that should cluster in phylogroup 3, according to the MLST analysis, to their observed position clustering with strains of phylogroup 5. It is not clear if this recombination event and the elevated Ka/Ks ratios observed in clade 5 are related.
|
|
|
Codon bias can also give an indication of whether greater genetic diversity is due to relaxed selective constraints or diversifying selection. A locus experiencing relaxed constraints is expected to have lower codon bias because codon bias is positively associated with expression level. We used the codon adaptation index (CAI) to measure codon bias on all of the loci in the HrpZ operon (table 4). Contrary to the expectations assuming relaxed selective constraints, hrpA actually has the highest level of codon bias of all of the loci in the operon, with a CAI of 0.75. The other loci range from a low of 0.56 for hrpZ to a high of 0.72 for hrcJ and hrpE. This analysis further supports the action of diversifying selection on the hrpA locus.
|
Discussion |
|---|
|
TOPAbstractIntroductionMaterial and MethodsResultsDiscussionAcknowledgementsReferences |
|---|
The P. syringae HrpZ operon is an excellent forum for studying the evolutionary forces that drive pathogen–host interactions. The tight linkage among loci that encode proteins that are both required for virulence and obvious targets for pathogen surveillance systems provides insight into how competing evolutionary pressures shape genetic variation and evolutionary history. By all rights, the type III pilus, composed of HrpA subunits, should be an obvious target for the pathogen surveillance systems of hosts. This is reinforced by the direct analogy between the HrpA subunits that make up the T3SS pilus and the flagellin subunits that make up the flagella, with the former being evolutionarily derived from the latter (Van Gijsegem et al. 1995
; Aizawa 2001). Nevertheless, although flagellin is the prototypical PAMP (Felix et al. 1999; Hayashi et al. 2001), HrpA has never been so identified. This begs the question of how HrpA stays ahead of the pathogen surveillance systems.
Our population genetic analysis of hrpA finds strong evidence supporting a role for diversifying selection at this locus. The evolutionary forces acting on hrpA are favoring the maintenance of genetic diversity, which apparently enables the pilus and, by extension the bacterium, to avoid recognition by the host. These analyses are supported by the hrpA phylogenetic data, which revealed long internal branches that are typically associated with either population subdivision or diversifying selection (Nordborg 2000). Population subdivision influences organisms and genomes as a whole and is therefore not a viable explanation because only 1 of the 6 loci in the HrpZ operon shows this pattern. Diversifying selection, in which there is selection for the maintenance of genetic diversity, can act on individual loci and is therefore a possible explanation of the pattern seen in the hrpA genealogy. Furthermore, diversifying selection has been shown to play an important role in the maintenance of genetic diversity of both the immune system and in parasites and pathogens that are trying to avoid a host defense response.
Weber and Koebnik (2006) recently studied the X. campestris hrpE locus, which encodes the T3SS pilus. They used an analysis of synonymous and nonsynonymous substitutions to conclude that this locus is under positive selection. Although the hrpE alleles clustered in 2 quite distinct clades, the analysis did not extend far enough to determine if there was any indication of diversifying selection at this locus.
The strongest signature of selection is found within the clade 5 alleles. This group is notable because it includes both strains that cluster in P. syringae phylogroup 5 and 2 additional strains that should cluster in P. syringae phylogroup 3 based on the MLST analysis. This raises the interesting possibility that a recombination event mobilized a genetic variant from a phylogroup 5 strain into a phylogroup 3 strain, where it apparently provides a selective advantage. It appears that selection may be acting on the unstructured region between the first 2 predicted alpha helices as depicted in figure 6, but the precise nature of this selective advantage would require significant structure–function analysis that is beyond the scope of the current work.
In contrast to hrpA, the hairpin hrpZ does not appear to be under strong selective pressures, despite its intimate interaction with the host cell. The only indication that diversifying selection is acting on this locus is the window of sequence that shows significantly positive Tajima's D starting around amino acid 70. This window corresponds to the region of lowest overall polymorphism and predicted membrane localization, which may be significant because HrpZ is believed to form ion pores in host membranes.
Although we do observe one significant recombination event in this data, overall, most of the apparently significant variation has been introduced by mutation. This is in contrast to what is observed among most of the effector proteins that are secreted through the T3SS, where the most obvious evolutionary changes are governed by horizontal gene transfer. The prominent role of allelic variation makes this study more consistent with the pathoadaptive changes observed in the fimH locus of E. coli (Sokurenko et al. 1998; Weissman et al. 2003; Sokurenko et al. 2004) than with other evolutionary studies of the T3SS, which generally find a dominant role for horizontal gene transfer (Prager et al. 2000; Guttman et al. 2002; Hansen-Wester et al. 2002; Gophna et al. 2003; Rohmer et al. 2004). This could simply be a consequence of the fact that the T3SS of P. syringae was acquired prior to the diversification of the species complex. Nevertheless, the question of whether the ancient status of the T3SS in P. syringae really imposes a strong constraint on the action of horizontal transfer now needs to be questioned because we clearly observe a significant recombination event between phylogroups 3 and 5 in the HrpZ operon. An alternative explanation is that genes that are essential components of intricate structures and embedded within complex regulons, such as the T3SS, are less likely to undergo recombination simply due to the higher likelihood of disrupting the function of the whole system.
Evolutionary studies such as this not only provide insight into fundamental evolutionary processes but can also be useful for identifying functionally significant regions in proteins. It will be interesting to see if the region under strong selection in HrpA corresponds to a potential PAMP once the protein structure has been solved. Even more interesting will be identifying the host pattern recognition receptors that may be targeting the HrpA locus and driving diversifying selection in this system.
|
Acknowledgements |
|---|
|
TOPAbstractIntroductionMaterial and MethodsResultsDiscussionAcknowledgementsReferences |
|---|
We would like to acknowledge the individuals who generously provided strains to the Guttman laboratory. We thank J. Stavrinides and R. Ness for careful review of the manuscript and members of the Guttman laboratory for their valuable input. D.S.G. is supported by the Canada Research Chair program and grants from the Natural Science and Engineering Research Counsel of Canada and Performance Plants Inc., Kingston, Ontario.
|
Footnotes |
|---|
John H McDonald, Associate Editor
|
References |
|---|
|
TOPAbstractIntroductionMaterial and MethodsResultsDiscussionAcknowledgementsReferences |
|---|
Aizawa S. (2001) Bacterial flagella and type III secretion systems. FEMS Microbiol Lett 202:157–164.[CrossRef][ISI][Medline]
Alfano JR, Charkowski AO, Deng WL, Badel JL, Petnicki-Ocwieja T, van Dijk K, Collmer A. (2000) The Pseudomonas syringae Hrp pathogenicity island has a tripartite mosaic structure composed of a cluster of type III secretion genes bounded by exchangeable effector and conserved effector loci that contribute to parasitic fitness and pathogenicity in plants. Proc Natl Acad Sci USA 97:4856–4861.
Allen RL, Bittner-Eddy PD, Grenville-Briggs LJ, Meitz JC, Rehmany AP, Rose LE, Beynon JL. (2004) Host-parasite coevolutionary conflict between Arabidopsis and downy mildew. Science 306:1957–1960.
Atwood KC, Schneider LK, Ryan TJ. (1951) Periodic selection in Escherichia coli. Proc Natl Acad Sci USA 37:146–155.
Baum J, Thomas AW, Conway DJ. (2003) Evidence for diversifying selection on erythrocyte-binding antigens of Plasmodium falciparum and P. vivax. Genetics 163:1327–1336.
Boyd EF and Hartl DL. (1998) Diversifying selection governs sequence polymorphism in the major adhesin proteins fimA, papA, and sfaA of Escherichia coli. J Mol Evol 47:258–267.[CrossRef][ISI][Medline]
Bretz JR and Hutcheson SW. (2004) Role of type III effector secretion during bacterial pathogenesis in another kingdom. Infect Immun 72:3697–3705.
Brown IR, Mansfield JW, Taira S, Roine E, Romantschuk M. (2001) Immunocytochemical localization of HrpA and HrpZ supports a role for the Hrp pilus in the transfer of effector proteins from Pseudomonas syringae pv. tomato across the host plant cell wall. Mol Plant-Microbe Interact 14:394–404.[ISI][Medline]
Dobrindt U, Hochhut B, Hentschel U, Hacker J. (2004) Genomic islands in pathogenic and environmental microorganisms. Nat Rev Microbiol 2:414–424.[CrossRef][ISI][Medline]
Feil EJ, Maiden MC, Achtman M, Spratt BG. (1999) The relative contributions of recombination and mutation to the divergence of clones of Neisseria meningitidis. Mol Biol Evol 16:1496–1502.[Abstract]
Felix G, Duran JD, Volko S, Boller T. (1999) Plants have a sensitive perception system for the most conserved domain of bacterial flagellin. Plant J 18:265–276.[CrossRef][ISI][Medline]
Felsenstein J. (1993) PHYLIP (phylogeny inference package). Distributed by the author(Department of Genetics, University of Washington, Seattle (WA)).
Fu Y-X and Li W-H. (1993) Statistical tests of neutrality of mutations. Genetics 133:693–709.[Abstract]
Gophna U, Ron EZ, Graur D. (2003) Bacterial type III secretion systems are ancient and evolved by multiple horizontal-transfer events. Gene 312:151–163.[CrossRef][ISI][Medline]
Guttman DS and Dykhuizen DE. (1994) Clonal divergence in Escherichia coli as a result of recombination, not mutation. Science 266:1380–1383.
Guttman DS, Vinatzer BA, Sarkar SF, Ranall M, Greenberg JT. (2002) A functional screen for the type III (Hrp) secretome of the plant pathogen Pseudomonas syringae. Science 295:1722–1726.
Hansen-Wester I, Stecher B, Hensel M. (2002) Analyses of the evolutionary distribution of salmonella translocated effectors. Infect Immun 70:1619–1622.
Hayashi F, Smith KD, Ozinsky A, Hawn TR, Yi EC, Goodlett DR, Eng JK, Akira S, Underhill DM, Aderem A. (2001) The innate immune response to bacterial flagellin is mediated by Toll-like receptor 5. Nature 410:1099–1103.[CrossRef][Medline]
He SY, Nomura K, Whittam TS. (2004) Type III protein secretion mechanism in mammalian and plant pathogens. Biochim Biophys Acta 1694:181–206.[Medline]
Holmes EC, Worobey M, Rambaut A. (1999) Phylogenetic evidence for recombination in dengue virus. Mol Biol Evol 16:405–409.[Abstract]
Hudson RR. (2001) Two-locus sampling distributions and their application. Genetics 159:1805–1817.
Hudson RR and Kaplan NL. (1985) Statistical properties of the number of recombination events in the history of a sample of DNA sequences. Genetics 111:147–164.
Hudson RR, Kreitman M, Aguadé M. (1987) A test of neutral molecular evolution based on nucleotide data. Genetics 116:153–159.
Hueck C. (1998) Type III protein secretion systems in bacterial pathogens of animals and plants. Microbiol Mol Biol Rev 62:379–433.
Hwang MSH, Morgan RL, Sarkar SF, Wang PW, Guttman DS. (2005) Phylogenetic characterization of virulence and resistance phenotypes in Pseudomonas syringae. Appl Environ Microbiol 71:5182–5191.
Kalia A and Bessen DE. (2004) Natural selection and evolution of streptococcal virulence genes involved in tissue-specific adaptations. J Bacteriol 186:110–121.
King EO, Ward MK, Raney DE. (1954) Two simple media for the demonstration of phycocyanin and fluorescin. J Lab Clin Med 44:301–307.[Medline]
Kumar S, Tamura K, Nei M. (1994) MEGA—molecular evolutionary genetics analysis software for microcomputers. Comput Appl Biosci 10:189–191.
Lee J, Klusener B, Tsiamis G, Stevens C, et al. (12 co-authors). (2001) HrpZ(Psph) from the plant pathogen Pseudomonas syringae pv. phaseolicola binds to lipid bilayers and forms an ion-conducting pore in vitro. Proc Natl Acad Sci USA 98:289–294.
Lee YH, Kolade OO, Nomura K, Arvidson DN, He SY. (2005) Use of dominant-negative HrpA mutants to dissect Hrp pilus assembly and type III secretion in Pseudomonas syringae pv. tomato. J Biol Chem 280:21409–21417.
Li CM, Brown I, Mansfield J, Stevens C, Boureau T, Romantschuk M, Taira S. (2002) The Hrp pilus of Pseudomonas syringae elongates from its tip and acts as a conduit for translocation of the effector protein HrpZ. EMBO J 21:1909–1915.[CrossRef][ISI][Medline]
Lomholt JA, Poulsen K, Kilian M. (2001) Epidemic population structure of Pseudomonas aeruginosa: evidence for a clone that is pathogenic to the eye and that has a distinct combination of virulence factors. Infect Immun 69:6284–6295.
Matson JS, Durick KA, Bradley DS, Nilles ML. (2005) Immunization of mice with YscF provides protection from Yersinia pestis infections. BMC Microbiol 5:38.[CrossRef][Medline]
Mauricio R, Stahl EA, Korves T, Tian D, Kreitman M, Bergelson J. (2003) Natural selection for polymorphism in the disease resistance gene RPS2 of Arabidopsis thaliana. Genetics 163:735–746.
Maynard Smith J and Smith NH. (1998) Detecting recombination from gene trees. Mol Biol Evol 15:590–599.[Abstract]
McDonald JH and Kreitman M. (1991) Adaptive protein evolution at the adh locus in Drosophila. Nature 351:642–654.[CrossRef]
McDowell JM, Dhandaydham M, Long TA, Aarts MG, Goff S, Holub EB, Dangl JL. (1998) Intragenic recombination and diversifying selection contribute to the evolution of downy mildew resistance at the RPP8 locus of Arabidopsis. Plant Cell 10:1861–1874.
McGraw EA, Li J, Selander RK, Whittam TS. (1999) Molecular evolution and mosaic structure of alpha, beta, and gamma intimins of pathogenic Escherichia coli. Mol Biol Evol 16:12–22.[Medline]
McVean G, Awadalla P, Fearnhead P. (2002) A coalescent-based method for detecting and estimating recombination from gene sequences. Genetics 160:1231–1241.
Nei M and Gojobori T. (1986) Simple methods for estimating the numbers of synonymous and nonsynonymous nucleotide substitutions. Mol Biol Evol 3:418–426.[Abstract]
Nordborg M. (2000) Coalescent theory. In Balding DJ, Bishop M, Cannings C (Eds.). Handbook of statistical genetics(John Wiley & Sons, Ltd, New York) pp. 179–212.
Nurnberger T, Brunner F, Kemmerling B, Piater L. (2004) Innate immunity in plants and animals: striking similarities and obvious differences. Immunol Rev 198:249–266.[CrossRef][ISI][Medline]
Nurnberger T and Lipka V. (2005) Non-host resistance in plants: new insights into an old phenomenon. Mol Plant Pathol 6:335–345.[CrossRef]
Parker JE. (2003) Plant recognition of microbial patterns. Trends Plant Sci 8:245–247.[CrossRef][ISI][Medline]
Pirnay JP, De Vos D, Cochez C, Bilocq F, Vanderkelen A, Zizi M, Ghysels B, Cornelis P. (2002) Pseudomonas aeruginosa displays an epidemic population structure. Environ Microbiol 4:898–911.[CrossRef][Medline]
Polley SD and Conway DJ. (2001) Strong diversifying selection on domains of the Plasmodium falciparum apical membrane antigen 1 gene. Genetics 158:1505–1512.
Potts WK and Slev PR. (1995) Pathogen-based models favoring MHC genetic diversity. Immunol Rev 143:181–197.[CrossRef][ISI][Medline]
Prager R, Mirold S, Tietze E, Strutz U, Knuppel B, Rabsch W, Hardt WD, Tschape H. (2000) Prevalence and polymorphism of genes encoding translocated effector proteins among clinical isolates of Salmonella enterica. Int J Med Microbiol 290:605–617.[ISI][Medline]
Prugnolle F, Manica A, Charpentier M, Guegan JF, Guernier V, Balloux F. (2005) Pathogen-driven selection and worldwide HLA class I diversity. Curr Biol 15:1022–1027.[CrossRef][ISI][Medline]
Rohmer L, Guttman DS, Dangl JL. (2004) Diverse evolutionary mechanisms shape the type III effector virulence factor repertoire in plant pathogenic Pseudomonas syringae. Genetics 167:1341–1360.
Roine E, Saarinen J, Kalkkinen N, Romantschuk M. (1997) Purified HrpA of Pseudomonas syringae pv. tomato DC3000 reassembles into pili. FEBS Lett 417:168–172.[CrossRef][ISI][Medline]
Roine E, Wei W, Yuan J, Nurmiaho-Lassila EL, Kalkkinen N, Romantschuk M, He SY. (1997) Hrp pilus: an hrp-dependent bacterial surface appendage produced by Pseudomonas syringae pv. tomato DC3000. Proc Natl Acad Sci USA 94:3459–3464.
R