The discovery 45 years ago that many Pseudomonas syringae pathovars elicit the hypersensitive response in plant species other than their hosts fostered the use of these bacteria as experimental models. However, the basis for host specificity and the corresponding resistance of nonhosts remain unclear. Pseudomonas syringae is now known to inject into the host cytoplasm, via the type III secretion system, effector proteins that suppress basal innate immunity, but may be recognized by cognate resistance (R) proteins in a second level of defence. The identification and manipulation of complete repertoires of type III effectors have revealed the highly polymorphic nature of effector repertoires and their potential to limit the host range. However, the maintenance of compatible effector repertoires may be driven by adaptations to life in a given plant species involving many factors. Tools are now available to test several hypotheses for the nature and evolution of P. syringae host specificity and nonhost resistance.
Pseudomonas syringae is a host-specific, hemibiotrophic pathogen whose interactions with plants are both archetypal and diverse. Research with a few model strains of P. syringae, notably P. syringae pv. tomato DC3000, has contributed significantly to the paradigm of effector-mediated pathogenesis, which holds that typical (hemi)biotrophic bacteria, fungi and oomycetes succeed as parasites mainly because the ‘effector’ proteins they translocate into host cells subvert defences that would otherwise be activated by microbial signatures (Abramovitch et al., 2006; Alfano and Collmer, 2004; Boller and Felix, 2009; Hogenhout et al., 2009; Nomura et al., 2005). Widening the study of effector-mediated pathogenesis to multiple strains of P. syringae has revealed a remarkable diversity in effector repertoires that scales with the diversity in hosts and disease symptoms associated with these bacteria (Cunnac et al., 2009; Hirano and Upper, 1990). The host specificity of the c. 50 pathovars of P. syringae for different plant species is a particularly striking aspect of this variability that has been investigated for more than 45 years but is still poorly understood.
Pseudomonas syringae and many other Proteobacteria causing diseases in plants and animals inject effector proteins into host cells via the type III secretion system (T3SS) (Cornelis and Van Gijsegem, 2000). The T3SS in plant pathogens is encoded by hrp genes, which are required for the bacteria to elicit the hypersensitive response (HR; a rapid, localized programmed cell death) in nonhost or resistant plants or to be pathogenic in susceptible host plants (Lindgren et al., 1986). The strong phenotype of hrp mutants highlights the collective importance of type III effectors (T3Es) in P. syringae interactions with plants. The host specificity of P. syringae is observed at two levels (with two corresponding types of resistance): pathovar–host species (‘nonhost resistance’) and race–host cultivar (‘host resistance’). For example, at the first level of specificity, soybean is a nonhost for pv. tomato but a host for pv. glycinea; at the second level, certain cultivars of soybean are resistant to particular races of pv. glycinea. As first demonstrated by the avirulence conferred by heterologous expression in P. syringae pv. glycinea race 5 of the avrA T3E gene cloned from race 6, the presence of a single betraying T3E can entirely explain the specificity at the race–cultivar level (Staskawicz et al., 1984). This gain-of-function approach was subsequently applied to several other P. syringae pathovars and crop species, most notably P. syringae pv. phaseolicola on bean, which revealed the general role of T3Es in the control of specificity at the race–cultivar level (Keen, 1990).
The earlier discovery that many P. syringae pathovars elicit HR in nonhosts was seminal in the development of these bacteria as important experimental models (Klement, 1963; Klement et al., 1964), and the common association of HR with avirulence suggested that T3E repertoires similarly control the specificity of interactions at the pathovar–host species level as well as the race–host cultivar level (Kobayashi et al., 1989). Indeed, the heterologous expression of pv. tomato genes in pv. glycinea revealed three T3Es that could collectively account for the avirulence of pv. tomato on all tested cultivars of soybean (Kobayashi et al., 1989). However, the construction of a polymutant in pv. tomato lacking all three of these T3E genes failed to extend the host range of the mutant to soybean (Lorang et al., 1994). These classic observations are based largely on qualitative differences in P. syringae interactions with plants. Since they were published, we have learned far more about P. syringae T3E repertoires and their function in defeating or evading two levels of plant defence, and the quantitative contributions that effectors and other factors can make to virulence. Despite this progress, the basis for the host specificity of P. syringae pathovars for different plant species and the corresponding resistance of nonhosts remains unclear. This mini-review addresses the dynamics of T3E repertoire evolution in the context of plant defences and the potential role of dysfunctional repertoires in determining pathovar–host species specificity.
TWO LEVELS OF PLANT DEFENCE AND VARIETIES OF NONHOST RESISTANCE
Successful effector-mediated pathogenesis involves the defeat of two levels of plant defence (Boller and Felix, 2009; Gohre and Robatzek, 2008; Jones and Dangl, 2006). In the first level, plants detect P. syringae and other microbes in the apoplast via pathogen (or microbe)-associated molecular patterns (PAMPs; e.g. flagellin, lipopolysaccharide, peptidoglycan and elongation factor TU) using surface-arrayed pattern recognition receptors (PRRs) that elicit PAMP-triggered immunity (PTI) (Boller and Felix, 2009; Jones and Dangl, 2006). Pseudomonas syringae defeats this defence largely by injecting T3Es that suppress PTI. The second level involves the recognition of the activity of one or more of these T3Es inside plant cells via resistance (R) proteins, which are typically nucleotide-binding leucine-rich repeat (NB-LRR) proteins. So far, it appears that R proteins recognize all of their cognate P. syringae T3Es indirectly by detecting effector activity on other plant proteins that are ‘guarded’ by NB-LRR proteins. According to an emerging model, ‘guardees’, such as the Pto kinase, are decoys for primary effector targets, such as PRR kinases (van der Hoorn and Kamoun, 2008; Zhou and Chai, 2008).
Resistance (R) gene-dependent effector-triggered immunity (ETI) confers qualitative resistance that has been exploited by plant breeders and pathologists because of the relative ease with which single R genes can be introduced into elite cultivars of many crop species (Poland et al., 2009). However, such resistance is typically not durable in the field because of the multiple mechanisms by which pathogens inactivate a betraying T3E gene, thus producing a new race that is virulent (Jones and Dangl, 2006; Ma and Guttman, 2008). This dynamic works for the pathogen because the co-evolutionary battle between plant R gene surveillance systems and pathogen T3Es has produced effector repertoires with internal redundancy (discussed further below), and consequently individual T3Es are dispensable. Another mechanism for evading ETI is through disruption of ETI itself. For example, many of the P. syringae pv. tomato DC3000 T3Es are capable of suppressing the defence response elicited by the T3E HopA1 (Guo et al., 2009) and, remarkably, AvrPtoB can suppress its own detection using its C-terminal E3 ligase domain to facilitate the degradation of Fen kinase, a host protein capable of activating ETI in response to the AvrPtoB N-terminal region (Rosebrock et al., 2007). The dispensability of individual effectors in a repertoire has been observed repeatedly through studies involving bacterial mutants in the laboratory or through R gene ‘breakdown’ in the field, and similar observations suggest that this property of repertoires is widespread among bacterial, fungal and oomycete pathogens of plants (Hogenhout et al., 2009; Jones and Dangl, 2006; Kvitko et al., 2009). Thus, the internal redundancy of P. syringae T3E repertoires appears to be archetypal for a wide collection of pathogens that defeat plants through effector-mediated pathogenesis.
In contrast with host resistance, nonhost resistance appears to be highly stable in the field (Mysore and Ryu, 2004). Nonhost resistance is canonically shown by all members of a plant species against a pathogen species or pathotype (Heath, 2000), and experience with various crops suggests that a given plant species is durably resistant in the field to all but a few of the P. syringae pathovars. For example, common bean is susceptible to pv. phaseolicola (halo blight) and syringae (brown spot), but not to pv. tomato or glycinea (Schwartz et al., 2005). However, many P. syringae strains can cause disease in more than one plant species (or family), and some plants can show nonhost resistance to nonadapted strains that have been nominally assigned to a pathovar that also contains adapted strains. For example, pv. tomato strain DC3000 can cause disease in tomato and Arabidopsis, whereas pv. tomato strain T1 causes disease only in tomato; Arabidopsis is a nonhost for this strain (Whalen et al., 1991). Importantly, multilocus sequence typing of close relatives of DC3000, which occurs in pv. tomato, maculicola and antirrhini (Hendson et al., 1992), reveals that strains in the T1 clade, but not the DC3000 clade, are consistently isolated from tomato plants with bacterial speck in the field (Yan et al., 2008). As discussed below, T1 is demonstrably nonadapted for Arabidopsis. Why DC3000-like strains are virulent on tomato in the laboratory but rarely isolated from tomato in the field is unknown.
Two types of nonhost resistance have previously been distinguished on the basis of interaction phenotypes: type I occurs without apparent host cell death, whereas type II is accompanied by HR (Mysore and Ryu, 2004). Arabidopsis displays type I nonhost resistance against P. syringae pv. phaseolicola and type II nonhost resistance against P. syringae pv. tomato T1. Although ETI has typically been distinguished from PTI by the elicitation of HR, it now appears that some PAMPs may elicit plant cell death and some T3E/R protein combinations, such as AvrB/TAO1 and HopA1Psy61/RPS6, may contribute to ETI without HR (Eitas et al., 2008; Kim et al., 2009; Naito et al., 2008; Shimizu et al., 2003), which suggests that there is overlap in the threshold for the elicitation of HR (Jones and Dangl, 2006). Similarly, there is overlap in the profiles of plant genes expressed in response to elicitors of PTI and ETI (Navarro et al., 2004; Tao et al., 2003; Thilmony et al., 2006). Thus, the threshold model for HR elicitation raises the possibility that there is no fundamental difference between type I and type II nonhost resistance, and both could result from T3E repertoires that are incompatible with nonhost ETI systems.
POLYMORPHISMS IN THE MOLECULAR INTERACTIONS OF P. SYRINGAE WITH PLANT DEFENCES
Polymorphisms in the repertoires of T3Es and R genes underlie race–cultivar specificity, but polymorphisms in multiple systems could underlie pathovar–host specificity, as summarized in Fig. 1. The model depicted postulates that all P. syringae strains share a basic ability to grow in the apoplast and that, despite potential adaptations to differences in nutrients and antimicrobials associated with different plant species, the polymorphisms that qualitatively limit host range involve defence recognition. The model further postulates that polymorphisms involving the recognition of PAMPs or the suppression of PTI by T3Es are not qualitative limiters of host range, although such polymorphisms do exist. For example, FLS2, the prototypical PRR, is lacking in some Arabidopsis ecotypes, and EFR, which mediates the recognition of the elongation factor Tu PAMP, appears to be unique to the Brassicaceae (Boller and Felix, 2009). However, FLS2 is found in all plant groups, plants appear to recognize multiple PAMPs and there has been no report of a P. syringae strain that defeats PTI by simply evading detection by the PAMP perception system of its host species (Boller and Felix, 2009), although calcium chelation by bacterial polysaccharides may partially reduce PAMP signalling for all strains (Aslam et al., 2008). Similarly, it appears less likely that polymorphisms in plant factors involved in defence signalling or antimicrobial deployment systems are qualitative limiters of host range. Pseudomonas syringae T3Es that suppress PTI and have been well characterized, such as AvrPto and HopM1, have so far been found to target apparently universal plant proteins (Cunnac et al., 2009; Nomura et al., 2006; Shan et al., 2008; Xiang et al., 2008). Furthermore, an analysis of 10 Arabidopsis genes encoding defence signal transduction proteins revealed that most of them lacked the signatures of positive selection that would be a hallmark of co-evolution with pathogen effectors (Caldwell and Michelmore, 2009). In contrast, a single-nucleotide analysis of 20 Arabidopsis accessions revealed that, among all protein families, the NB-LRR gene family was the most polymorphic (Clark et al., 2007).
The current data continue to support the originally proposed central role in P. syringae pathogenesis of T3SS and the effectors it delivers into the host cytoplasm (Alfano and Collmer, 1996), and highlight the importance of two aspects of relative timing in P. syringae–plant interactions and their evolution. First, the P. syringae T3SS appears to be capable of delivering effectors rapidly on inoculation (Huynh et al., 1989; Puri et al., 1997), and this enables PTI suppression to trump PAMP detection. Second, the highly dynamic genetics of T3E repertoires suggest that P. syringae can rapidly and repeatedly trump ETI in game-like interactions of populations of bacteria and plants, which underpins the perpetual importance of co-evolving T3E and R gene repertoires in determining the host range at all levels. Thus, the molecular dialogue between P. syringae and plants can involve a variety of factors, but ETI will always have the last word.
THE DYNAMIC GENETICS UNDERLYING T3E REPERTOIRES
The evolution of T3E repertoires has been addressed in recent reviews (Ma and Guttman, 2008; Stavrinides et al., 2008), but it is useful to highlight a few key points and examples here. The acquisition of new effector capabilities occurs through a combination of horizontal transfer and recombination, the latter process facilitating the re-assortment of domains for the creation of T3Es having novel functions (Stavrinides et al., 2006; Yan et al., 2008). Genome sequencing has provided extensive support for the horizontal transfer origins of many T3Es, revealing the atypical sequence profiles and proximity to mobile genetic elements that are hallmarks of this process (Lindeberg et al., 2008). For example, avrPto and hopW1 are found to be associated with phage genes in multiple P. syringae genomes and diverse sets of T3E genes are embedded within integrative and conjugative elements (ICElands) (Lindeberg et al., 2008; Pitman et al., 2005). Intra-genome recombination also plays a role in the generation of new T3Es, most importantly in uniting new functional domains with the regulatory elements and translocation motifs necessary for their co-expression with and translocation by T3SS.
Pseudomonas syringae genomes are also littered with effector pseudogenes with coding sequences or regulatory regions disrupted by frame shifts, internal stop codons or mobile elements (Lindeberg et al., 2006). Pseudomonas syringae genomes have an unusually high proportion of mobile elements (Buell et al., 2003), and efficient inactivation of betraying T3E genes is expected to be important for the evasion of ETI. Indeed, such inactivation was observed with the P. syringae pv. phaseolicola hopAR1 T3E gene during exposure to HR in a resistant bean plant (Pitman et al., 2005). In addition to simple loss of T3E gene function, P. syringae possesses other mechanisms for evading ETI, for example through the generation of allelic variations. Members of the AvrB family (formerly AvrB and AvrC) originate from a common ancestor, but trigger a resistance response in different soybean cultivars (Tamaki et al., 1991). Similar differences in host recognition are observed among alleles of the HopX1 and HopZ1 subfamilies, in some cases with only a single residue change accounting for the difference (Ma et al., 2006; Stevens et al., 1998). In general, P. syringae strains appear to be capable of rapid adaptation of T3E repertoires in response to plant ETI surveillance.
THE DIVERSITY OF EFFECTOR REPERTOIRES
Comprehensive T3E repertoires have been identified for model strains that represent different phylogenetic groups within P. syringae and that have different host specificities. Phylogenetic analyses show that the P. syringae pathovars fall into three major clades, plus a fourth that is enriched in pathogens of monocots (Sarkar and Guttman, 2004; Sawada et al., 1999). Complete genome sequences are available for strains in each of the first three clades: P. syringae pv. tomato DC3000 (Buell et al., 2003), P. syringae pv. syringae B728a (Feil et al., 2005) and P. syringae pv. phaseolicola 1448A (Joardar et al., 2005). Genome-enabled approaches exploiting promoter and T3SS targeting patterns associated with T3E genes, coupled with functional screens and extensive experimental validation, have produced comprehensive inventories of the T3E repertoires for these strains (Chang et al., 2005; Cunnac et al., 2009; Lindeberg et al., 2006; Vinatzer and Greenberg, 2007). These approaches have also been used to identify T3E genes in a draft genome of P. syringae pv. tomato T1 (Almeida et al., 2009). Although B728a and 1448A are both pathogens of bean, and DC3000 and T1 are both pathogens of tomato, the T3E repertoires of these strains are remarkably different, beyond a small core of apparently universal T3Es (Cunnac et al., 2009). Despite the high overall similarity of DC3000 and T1, only 14 of the T3Es are shared between them, whereas as many as 15 T3E genes are present only in the DC3000 genome and 11 only in the T1 genome (Almeida et al., 2009). Thus, there appears to be no pattern in the composition of these four repertoires that would predict their hosts, and there are multiple combinations of effectors that can confer P. syringae virulence in the same plant.
Combinatorial deletions involving 20 of the active T3E genes in P. syringae pv. tomato DC3000 have revealed a redundancy-based structure in the effector repertoire, such that some deletions diminish growth in planta only in combination with other deletions (Kvitko et al., 2009). It was found that two redundant effector groups are particularly important in promoting DC3000 growth in planta and, based on the known activities of some of the members, these internally redundant groups have been proposed to target different high-level processes in PTI: perception of PAMPs and vesicle trafficking of antimicrobial factors. These observations suggest that a few defence processes are particularly important for DC3000 to block, and each process is targeted redundantly. Two of these effectors, AvrPto and HopM1, as mentioned above, have potentially universal plant targets, an observation that further supports the hypothesis that T3E repertoires may productively suppress PTI in a wide range of plants, although they may be incompatible with the ETI systems of all but a few plant species.
TESTING THE ROLE OF T3Es IN NONHOST RESISTANCE
Experiments that address the role of T3Es and plant defences in nonhost resistance have provided new insights but no comprehensive model. A microarray analysis of Arabidopsis genes expressed at different levels in response to bacterial challenge revealed similar patterns for P. syringae pv. phaseolicola and P. syringae pv. tomato DC3000 heterologously expressing T3Es with HR-eliciting avirulence activity, although the amplitude of the expression changes was lower for pv. phaseolicola (Tao et al., 2003). There is additional evidence that P. syringae pv. phaseolicola elicits weak ETI in Arabidopsis. Wild-type bacteria elicit the formation of larger callose deposits and stronger defence gene expression than do T3SS– mutants (Ham et al., 2007; Mishina and Zeier, 2007). Importantly, only small callose deposits are elicited by the wild-type in an ETI-compromised rar1 mutant. Any T3Es in the P. syringae pv. phaseolicola repertoire that elicit weak ETI in Arabidopsis await identification, but the type I nonhost resistance of Arabidopsis to pv. phaseolicola appears to be ETI based.
There is now evidence from solanaceous plants that the presence of betraying T3Es can indeed limit the host range at the pathovar–species level and that interference with the recognition of these T3Es can overcome type II nonhost resistance. For example, HopQ1-1 acts as a typical avirulence determinant for P. syringae pv. tomato DC3000 in Nicotiana benthamiana, and DC3000 mutants lacking this one T3E gain the ability to grow from low levels of inoculum to high population levels and to produce typical bacterial speck lesions in this apparent nonhost species (Wei et al., 2007). Another, broadly important example of this phenomenon involves tomato Pto/Prf-mediated recognition of AvrPto and AvrPtoB (Lin and Martin, 2007). Both of these T3Es are widespread among nonadapted pathovars, which cause disease in plants other than tomato. However, 10 of these nonadapted pathovars, e.g. P. syringae pv. glycinea F117, were able to cause pathovar-unique symptoms in tomato lacking Pto or Prf. Although bacterial growth and symptom severity were reduced relative to those observed with P. syringae pv. tomato DC3000, this finding highlights the potential importance of Pto and Prf in the nonhost resistance of tomato against multiple P. syringae pathovars, as well as the importance of pathogen factors in the production of pathovar-unique symptoms (Lin and Martin, 2007). The role of ETI in the nonhost resistance of solanaceous plants is also supported by the observation that compromising ETI in N. benthamiana plants through virus-induced gene silencing of NbSGT1 results in a 100-fold higher growth of nonadapted P. syringae pv. maculicola (Peart et al., 2002). Similarly, multiple plant species in different families produce necrotic responses suggestive of ETI during Agrobacterium-mediated transient expression of T3Es from P. syringae strains nonadapted for the test plant more frequently than with T3Es from adapted strains (Wroblewski et al., 2009).
However, a simple model of nonhost resistance based on ‘correctable’ incompatibilities at the ETI level may not be universally applicable. The nonhost resistance of Arabidopsis to P. syringae pv. tomato T1 was not defeated despite the use of Arabidopsis mutants compromised in ETI. Among the T3Es present in T1, but not DC3000, is AvrRpt2, which elicits an RPS2-dependent HR in Arabidopsis (Almeida et al., 2009). Strain T1 no longer elicited HR in an rps2-101c mutant line of Arabidopsis, but it also did not grow, thus revealing an underlying capacity for type I nonhost resistance against T1. Interestingly, Arabidopsis lines with mutations in RAR1 and SGT1, which are needed for ETI mediated by many T3E–R gene combinations, also retained resistance to T1. This raises the possibility that Arabidopsis possesses a novel mechanism for type I nonhost resistance against T1 that is independent of ETI.
The multifactorial nature of nonhost resistance is particularly evident with the type I resistance of Arabidopsis to P. syringae pv. phaseolicola, because growth of the bacteria can be partially promoted by a variety of manipulations. These include the mutation of a single plant defence gene, such as NHO1 (Lu et al., 2001), or inoculation of P. syringae pv. phaseolicola strains heterologously expressing HopM1 (a known suppressor of PTI normally not produced by this pathovar) into Arabidopsis mutants with multiple defence gene mutations, such as pmr4/pad4 or pmr4/npr1 (Ham et al., 2007). Importantly, co-inoculation of P. syringae pv. tomato DC3000 suppressed defences elicited by P. syringae pv. phaseolicola and promoted strong growth of the latter strain (Ham et al., 2007). The simplest explanation for this observation is that T3Es in the DC3000 repertoire suppress the weak ETI elicited by the P. syringae pv. phaseolicola repertoire. The availability of a growing collection of Arabidopsis defence mutants and comprehensive T3E repertoires for DC3000 and two contrasting nonadapted pathogens, P. syringae pv. phaseolicola 1448A and P. syringae pv. tomato T1, provides an excellent toolkit for exploring the basis for nonhost resistance in a model plant. Similarly, strains closely related to P. syringae pv. tomato DC3000 but differing in host specificity and performance in the field provide tools for exploring host specialization in the field using a model pathogen (Yan et al., 2008).
NONHOST RESISTANCE EVOLUTION AND DURABLE DISEASE RESISTANCE
It is paradoxical that host specificity at the pathovar–species level appears so stable given the flexible genetics of P. syringae T3E repertoires and the demonstration that the loss of a single effector can enable a model laboratory strain to cause disease in a new plant species (Wei et al., 2007). The explanation for this paradox may reside in the ubiquity of P. syringae in the environment and the diverse interactions of P. syringae with plants in the field (Hirano and Upper, 2000; Morris et al., 2008). Because of these ecological factors, clonal groups within P. syringae are likely to have ongoing opportunities to interact with many plants. Importantly, competition among such groups may drive host specialization through subtle adaptations for life in different plant species. Strains adapted to a particular species would then be under ongoing selection pressure to maintain T3E repertoires that are compatible with their host R gene surveillance systems, whereas less adapted (and therefore less competitive) strains could accumulate incompatibilities in their T3E repertoires without relevant penalty. As discussed above, this model suggests that the defeat of PTI is ancient and that any given T3E repertoire will have a basic ability to suppress PTI in a variety of plants. Therefore, host specialization may be driven primarily by the accumulation of many factors that contribute quantitatively to fitness, but a consequence of this process will be the evolution of T3E repertoires that may trigger qualitative nonhost resistance in a myriad of other species.
The quantitative factors driving host adaptation could involve T3E-mediated PTI suppression and/or factors unrelated to T3Es. Importantly, no single T3SS-delivered P. syringae effector has been shown to be sufficient to confer substantial growth and symptom production to a nonpathogen (Huang et al., 1988; Mohr et al., 2008), and neither the minimal set of T3Es needed for pathogenesis nor the basis for needing multiple T3Es has been determined. One key to the exploration of quantitative factors underlying host adaptation is to eliminate qualitative ETI from the experimental system, which otherwise masks weaker interaction phenotypes. One promising approach involves the exploitation of natural variation in the susceptibility of Arabidopsis ecotypes to P. syringae pv. tomato DC3000 to identify quantitative trait loci underlying basal defence (Fan et al., 2008). An understanding of these factors may have practical benefits. Quantitative disease resistance in crops shares with nonhost resistance the properties of being multifactorial and usefully durable, but the underlying basis for quantitative resistance is poorly understood (Poland et al., 2009). Our increasing capacity to deconvolute the diverse interactions of P. syringae with model plants and crops may provide new tools for improving quantitative resistance and for breeding crops with durable resistance.
Research in our programmes is supported by National Science Foundation grant DBI-0605059.