is one of the preeminent model systems for the study of host specificity and virulence. This gram-negative plant-pathogenic bacterium is the causal agent of a variety of bacterial spot, speck, and blight diseases on a wide range of plant hosts, including (but not limited to) apples, beets, beans, cabbage, cucumbers, flowers, oats, olives, peas, tobacco, tomato, and rice (25
). Isolates of P. syringae
are taxonomically subdivided into pathogenic varieties known as pathovars, based largely on their host of isolation. The tremendous diversity of hosts and disease symptomatology found in this species presents a unique opportunity to investigate the factors that determine host specificity.
In the present study, we used functional, phenotypic assays to survey the ability of 95 P. syringae strains to produce toxins, nucleate ice, and resist antimicrobial agents. We also refined the MLST protocol so that it could be used on nearly any fluorescent pseudomonad. Furthermore, we reduced the number of MLST loci used for typing from seven to four, which dramatically increased the rate at which strains could be typed, without any significant reduction in the phylogenetic resolution. We then mapped the toxin and resistance phenotypes onto the phylogeny of the core genome and showed that most of these phenotypes were distributed in a manner that was not consistent with a clonal evolutionary process. The present study provides a robust, phylogenetic foundation for studies of host adaptation and virulence in P. syringae.
There has been a tremendous explosion of interest in the role played by virulence-associated molecules in P. syringae
host interactions. Most of the current studies focus on the widely conserved type III secretion system and its effector proteins. Nevertheless, strains of P. syringae
also produce an impressive array of toxins, which typically act in a non-host-specific manner. Although toxins are not required for pathogenesis, they have been found to enhance virulence by increasing the severity of lesions and by contributing to increased growth and movement of bacteria inside the plant tissue (5
). Pseudomonads in general are also widely recognized as being highly resistant to a broad range of medically and agriculturally important antimicrobial compounds. P. syringae
is no exception to this rule.
To date, no study of toxin production or antimicrobial resistance in P. syringae has used a precise phylogenetic framework. We have used an MLST approach to characterize the P. syringae core genome, which provides the most accurate reflection of the clonal evolutionary history of the species. By mapping the distribution of toxin production and antimicrobial resistance onto the MLST phylogeny, we can more precisely determine the evolutionary origin of these phenotypes.
Four conclusions emerge from our study of toxin production. First, toxin production is surprisingly rare in P. syringae. Of 95 strains assayed, at least 54 (56.8%) did not produce any of the four toxins tested. None of the six cucumber pathogens or three wheat pathogens produced any toxins. Only one of the eight pea pathogens produced a toxin, with the sole exception being a pea isolate that was not originally given a pathovar pisi designation (Psy 1212R, which produced syringomycin).
The conclusion that toxin production is relatively rare assumes (i) that our assays are robust in their ability to identify the four toxins in all strains, (ii) that strains do not lose their ability to produce toxins when stored in the laboratory, and (iii) that there are no other significant toxins produced by this species. We do not believe the first issue is significant since we used standard toxin assays that have been used widely in other studies. The second point is very important. Some strains of P. syringae have been known to lose toxin production when stored for long periods. These phenotypes can sometimes be recovered by passaging the strains in planta prior to the toxin assays. Unfortunately, this procedure was impossible in the present study given the very large diversity of hosts. We do not believe this issue is a significant problem since our date is consistent with published results in all cases except the loss of coronatine production in strain Pto Pt23. With respect to the last issue, it is very unlikely that there are significant numbers of undescribed toxins given their extensive study in P. syringae. Furthermore, most of the assays performed in the present study were of limited specificity, identifying any representative of a class of toxins. For example, the syringomycin assay would detect any lipodepsipeptide toxin. Given the lack of specificity, these assays should provide a conservative estimate of the frequency of toxin production.
The second conclusion comes from the observation of a surprising negative association among toxin. P. syringae strains are very unlikely to produce more than one toxin. Only 2 of the 95 strains produced two toxins, and only 1 strain produced three toxins (the tomato pathogen Pto KN10, which produced tabtoxin, phaseolotoxin, and coronatine). Of the 21 strains producing syringomycin, only one strain produced coronatine, and none produced phaseolotoxin. Of the 14 strains producing coronatine, only 1 produced phaseolotoxin, while 1 produced syringomycin. Of the six strains producing phaseolotoxin, only one produced coronatine, while none produced syringomycin. These negative associations are statistically significant as determined by Fisher exact test (P < 0.0001 [syringomycin-coronatine]; P < 0.0001 [syringomycin-phaseolotoxin]; and P = 0.003 [phaseolotoxin-coronatine]). In addition, three of the four tabtoxin-producing strains did not produce any other toxins. Is this negative association due to a cost associated with the production of multiple toxins or simply the by-product of a relatively low rate of toxin production in P. syringae? Based on our empirical determination of the frequency of each toxin, at least 7 strains of a collection of 95 should produce two or more toxins. Our observation of only three multiple-toxin-producing strains indicates that these strains are less frequent in the population than they should be if toxins were produced independently of each other. This suggests a cost to the production of multiple toxins in P. syringae.
Our third conclusion is that toxin production is very poorly associated with the host of isolation (Fig. 1
). The best evidence for host association comes from a cluster of tomato pathogens that produce coronatine. However, this case is just as readily explained by nonindependent evolutionary histories. Furthermore, only 7 of the 11 tomato pathogens produced coronatine. It is perhaps easier to make the case for a correlation between the lack of production of a particular toxin and a specific host. This appears to be the case with respect to the two divergent clades of pea pathogens (pv. pisi) in group 2. As discussed above, both of these clades lack syringomycin production, while this toxin is relatively common throughout the rest of the group 2 strains.
Finally, we conclude that toxin production is generally distributed in a manner inconsistent with clonal (vertical) evolution (Fig. 1
); therefore, the evolution of toxin producing genes is very likely driven by horizontal gene transfer. This conclusion is also supported by similarity analysis of the genes and proteins responsible for toxin production. For example, the Cma proteins, which are necessary for coronatine biosynthesis, are similar to proteins distributed throughout the bacterial and eukaryota domains, including other pseudomonads. However, the only similar nucleotide sequence found outside the highly conserved orthologs in the P. syringae
complex is in Burkholderia pseudomallei
(E value = 2e-37, NCBI discontiguous megablast). Either the P. syringae cma
genes were acquired vertically and diverged to such an extent that they no longer show any nucleotide similarity to homologs in other pseudomonads or, more likely, they were acquired horizontally from an as-yet-unsequenced organism.
The tabtoxin and phaseolotoxin biosynthetic genes have no homologs outside of the P. syringae
complex. The small tabtoxin biosynthetic cluster is sufficient for tabtoxin synthesis and was observed by Kinscherf et al. (28
) to excise from the chromosome at frequencies as high as 10−3
/CFU. The specific mechanism driving this process is not known.
The interpretation of the history of the syringomycin biosynthetic genes is less clear. Syringomycin production is heavily concentrated in group 2. It appears that a common ancestor of this group may have acquired the syringomycin regulon, and it has since passed vertically to the descendants of this ancient bacterium. In fact, syringomycin is the only toxin produced by the group 2 strains. Nevertheless, during the evolution of this group, the syringomycin system appears to have been lost or disabled in a substantial number of strains, most notably, the ancestor of the clade of six pathovar pisi strains. Is syringomycin production ancestral in the P. syringae complex? The most parsimonious answer to this question is no, since syringomycin production is very rare in the other four syringae groups. This is further supported by the fact that three of these four groups branched off basal to group 2.
This syringomycin biosynthesis operon is dominated by the enormous syrE
gene, which is 28.1 kb in length, encoding a protein of 1,039 kDa, making it the largest prokaryotic protein discovered (20
). The SyrE synthetase protein contains eight conserved modules that show high similarity (typically ca. 75% nucleotide identity, but often as high as 90%) to homologous modules in a wide range of related gram-negative bacteria and even in species as distant as high-GC gram-positive bacteria such as Streptomyces
spp. Given the common occurrence of syringomycin production in the group 2 strains and the wide distribution of the biosynthetic genes throughout the bacterial domain, further comparative analysis would have to be completed before we can conclusively determine whether the distribution of this gene cluster is due to horizontal genetic exchange or simply rampant gene loss after vertical descent.
Unlike toxin production, the ability to nucleate ice and detoxify copper appears to be an ancestral trait in P. syringae
). With 80% of strains being able to nucleate ice, and 75% of strains being copper resistant, perhaps the more interesting observations come from those strains that have lost this ability. Although all soybean and pea pathogens are INA positive, there was a negative association between INA activity and the bean pathogens. Alternatively, this pattern may simply be the result of nonindependent evolutionary histories, a circumstance analogous to the situation in the tomato pathogens with respect to coronatine production. One way to answer this question is to determine whether the three strains that are INA positive in the closely related bean pathogen clade gained (or retained) the phenotype independently. A much stronger case can be made for pathogens of brassicaceous crops, which are also negatively associated with INA activity. These strains are scattered throughout groups 1 and 5. It would be much more difficult to invoke nonindependent evolutionary histories as the cause for the correlation. No clear phylogenetic or host-specific pattern was seen with copper resistance, except for the observation that the group 3 bean pathogen clade that was deficient in its ability to nucleate ice nucleation also had an excess of copper-sensitive strains. The meaning of the correlation is unclear.
Resistance to ampicillin appears to be ancestral in P. syringae, although the relatively high rate of ampicillin sensitivity in groups 1 and 5 (39 and 33% resistant, respectively, versus 70% resistance among the three remaining groups) presents the possibility that ampicillin resistance was lost in the ancestral lineage that gave rise to groups 1 and 5 or acquired in the ancestral lineage that gave rise to groups 2, 3, and 4. It is impossible to judge whether one hypothesis is more valid than the other based on the current data, since the phylogenetic support for these basal branches is very weak. Resistance to the other five antibiotics is generally quite rate, and there is absolutely no association between antibiotic resistance, the core genome phylogeny, or host of isolation. These findings are not surprising given the well-known propensity of antibiotic resistance genes to be transferred horizontally.
The interaction between a pathogen and its host is complex and multifaceted. To date, only type III effectors have been conclusively shown to both qualitatively limit and enable pathogenesis on specific hosts (1
). We are unable to identify a similar role for toxins in the present study. No specific host association was found to be strictly associated with the production of a particular toxin or resistance to a specific antimicrobial agent. Despite the reputation of P. syringae
as a copious toxin producer and the well-established role toxins play in modulating virulence, the present study clearly shows that no toxin is common or definitively ancestral in this species.