• Joel M. Kniskern,

    1. Department of Ecology and Evolution, University of Chicago, 1101 E. 57th Street, Chicago, Illinois 60637
    Search for more papers by this author
    • These authors contributed equally to this study.

    • Current address: Seminis Vegetable Seeds, Monsanto Vegetable Division, 37437 State Highway 16, Woodland, California 95695.

  • Luke G. Barrett,

    1. Department of Ecology and Evolution, University of Chicago, 1101 E. 57th Street, Chicago, Illinois 60637
    Search for more papers by this author
    • These authors contributed equally to this study.

    • Current address: CSIRO Plant Industry, GPO Box 1600, Canberra, Australian Capital Territory, 2601, Australia.

  • Joy Bergelson

    1. Department of Ecology and Evolution, University of Chicago, 1101 E. 57th Street, Chicago, Illinois 60637
    2. E-mail:
    Search for more papers by this author


Multihost pathogens occur widely on both natural and agriculturally managed hosts. Despite the importance of such generalists, evolutionary studies of host–pathogen interactions have largely focused on tightly coupled interactions between species pairs. We characterized resistance in a collection of Arabidopsis thaliana hosts, including 24 accessions collected from the Midwest USA and 24 from around the world, and patterns of virulence in a collection of Pseudomonas syringae strains, including 24 strains collected from wild Midwest populations of A. thaliana (residents) and 18 from an array of cultivated species (nonresidents). All of the nonresident strains and half of the resident strains elicited a resistance response on one or more A. thaliana accessions. The resident strains that failed to elicit any resistance response possessed an alternative type III secretion system (T3SS) that is unable to deliver effectors into plant host cells; as a result, these seemingly nonpathogenic strains are incapable of engaging in gene for gene interactions with A. thaliana. The remaining resident strains triggered greater resistance compared to nonresident strains, consistent with maladaptation of the resident bacterial population. We weigh the plausibility of two explanations: general maladaptation of pathogen strains and a more novel hypothesis whereby community level epidemiological dynamics result in adaptive dynamics favoring ephemeral hosts like A. thaliana.

The importance of pathogens as dynamic agents of selection on hosts is reflected in the persistence of genetic variation for resistance in wild host populations (Salvaudon et al. 2008; Lazzaro and Little 2009). With only a few exceptions (Gandon 2004; Thrall et al. 2005; Goss and Bergelson 2006; Woodhams et al. 2006) theoretical and empirical studies on the maintenance of genetic variation for resistance in hosts and virulence in pathogens have focused on tightly coupled interactions involving relatively specialized pathogens (Barrett et al. 2009). This is despite the fact that a large proportion of pathogens of plants and animals are generalists that infect multiple host species (Woolhouse et al. 2001; Barrett et al. 2009), and evidence that many emerging diseases are caused by generalist parasites infecting multiple host species (Cleaveland et al. 2001). There is also growing evidence that variation in host range and virulence can have a significant impact at the ecological level, widely influencing the potential for transmission, disease incidence, and impacts on host fitness (Power and Mitchell 2004; Colla et al. 2006; Barrett et al. 2009; Hellgren et al. 2009). However, the consequences of multihost–pathogen interactions for the evolution of host resistance and microbial pathogenesis remain largely unexplored (but see Gandon 2002; Spitzer 2006).

The ecological and evolutionary dynamics maintaining resistance and virulence variation in specialized, tightly coupled host–pathogen interactions have been well studied. For such systems, microbial pathogens are expected to have an adaptive advantage over their hosts due to their relatively large population sizes, fast generation times, and high migratory capacity (Hamilton et al. 1990; Kaltz and Shykoff 1998). However, the complexity generated by interactions with multiple host species calls into question the applicability of these predictions for generalist pathogens. Compared to specialist pathogens, generalists encounter hosts that vary widely in mechanisms of resistance and in the proportion of resources allocated to defense (Barrett et al. 2009). Encounter rates between individual host species and generalist pathogens within communities are also likely to be highly dynamic, depending on the diversity, spatial structure, and phenology of the host community (Carlsson-Graner and Thrall 2006; Mitchell and Power 2006; Barrett et al. 2008). Such heterogeneities in selection imposed by alternative host species are likely to limit the ability of pathogen species to adapt to any one particular host (Gandon 2002; Lajeunesse and Forbes 2002). Furthermore, evolutionary outcomes are likely to be highly context dependent, varying according to the host range of the pathogen concerned, the demographic structure of the host community, and the biology of the focal host (Wolinska et al. 2006; Poullain et al. 2008). For example, generalist pathogens may be expected to exhibit some degree of adaptation to one host species within a community, if that species represents a key resource for the pathogen (Woolhouse et al. 2001). In other cases, generalist pathogens may be maladapted to particular species, particularly if potential hosts are rarely encountered due to restricted availability in space or time. In such scenarios, we hypothesize that source–sink type evolutionary dynamics may favor adaptation of rare and short-lived hosts to generalist pathogens, rather than pathogens to hosts.

For both generalist and specialist pathogens, the potential for reciprocal evolutionary change largely depends upon a genotype-specific interaction between host and pathogen (e.g., Kaltz and Shykoff 1998; Lively and Dybdahl 2000; Thrall et al. 2002). One well-characterized, highly specific form of qualitative resistance common throughout eukaryotes is termed gene-for-gene (GFG) resistance (Thompson and Burdon 1992). The classic genetic definition of this form of resistance is that for every major resistance (R) gene in the host, there is a corresponding avirulence (Avr) gene in the pathogen (Flor 1956). When matching R and Avr proteins co-occur then a resistance response ensues, but when either protein is missing then the pathogen escapes R-gene-mediated recognition. This simple genetic interaction has been widely adopted as a general model for describing evolution between hosts and their pathogens in natural communities. However, to date, very little is known about the evolution of GFG interactions in interactions involving pathogens that can infect a broad range of host species.

In this study, we focus on GFG interactions between the herbaceous annual plant Arabidopsis thaliana and the generalist bacterial pathogen Pseudomonas syringae. Bacterial plant pathogens represent an excellent resource for the investigation of evolutionary dynamics in wild-host pathogen associations because the underlying mechanisms of pathogen virulence and host resistance have been well characterized in several model systems, allowing characterization and manipulation of genetic dynamics in an ecologically relevant setting. Bacterial pathogens inject pathogenicity effectors into plant cells using a needle-like apparatus referred to as the type III secretion system (T3SS) (Mudgett 2005). Avr genes code for effectors that have in some cases been demonstrated to promote pathogen growth (reviewed in Mudgett 2005; Kamoun 2006; Ellis et al. 2007), however, recognition of an effector gene induces a complex and highly effective defense response called the hypersensitive response (HR) that restricts or eliminates pathogen multiplication within the plant (reviewed in Heath 2000). The aim of this study was to investigate how pathogen adaptation to multiple host species may influence the evolution of virulence in pathogens and resistance in hosts. Specifically, we use a combination of molecular and phenotypic approaches to characterize the genetic structure and pathogenic potential of strains of P. syringae that are resident and nonresident on A. thaliana. We then examine patterns of host and pathogen adaptation with respect to the phylogenetic and geographic origin of host and pathogen genotypes. In particular, we assess to what extent our data are consistent with the hypothesis that generalist pathogens are maladapted to rare and short-lived hosts, because they specialize on other, more common hosts.



As an empirical model, we chose the interaction between the host plant A. thaliana and the pathogenic bacterium P. syringae. This system is well suited to investigating evolution in multihost pathogen interactions because P. syringae infects a wide phylogenetic range of plant species (including wild A. thaliana plants), and the genetic basis of the interaction between P. syringae and A. thaliana is well understood.

Pseudomonas syringae is a species complex encompassing a diverse group of gram-negative bacteria, many strains of which are pathogenic to plants (Sarkar and Guttman 2004). Pseudomonas syringae sensu lato has the capacity to cause disease on plant species within more than 100 families (Hirano and Upper 2000; Barrett et al. 2009), including A. thaliana (Jakob et al. 2002). The host ranges of subgroups and individual strains however are largely unknown, in part because the concept of what constitutes a host is largely undefined. Various strains of P. syringae have been recovered from the leaf surface and interior of a range of asymptomatic plants (Orser et al. 1985; Burr et al. 1996; Mohr et al. 2008), and in several cases, it has proven difficult to show that these strains cause disease on a range of tested plants, although the potential for disease on other unrepresented plants is largely impossible to rule out (Smith and Saddler 2001; Whipps 2001). Furthermore, Clarke et al. (2010) recently demonstrated that a clade of P. syringae strains that were isolated from various host species and that do not seem to cause disease on plants (Mohr et al. 2008) have an aberrant T3SS, seemingly acquired from an external source via lateral gene transfer. This xenologous T3SS is constitutively expressed, but is lacking in the capacity to deliver effectors into host cells (Clarke et al. 2010); in other words, these strains fail to elicit GFG resistance characterized by the HR phenotype. The characterization of strains that vary widely in pathogenicity adds weight to a growing recognition that P. syringae has a complex life history, including pathogenic, epiphytic, and saprophytic phases (Hirano and Upper 2000; Morris et al. 2008).

Arabidopsis thaliana is a short-lived winter annual that grows wild in agricultural fields and other disturbed areas. In the Midwest USA, seeds germinate in the late fall, and plants flower in April, completing their life-cycle by early May. Midwest A. thaliana normally has limited genetic variation within populations due to its recent introduction to North America (Nordborg et al. 2005) and to its high degree of self-fertilization [99%: (Platt et al. 2010)], although populations are typically polymorphic for R genes such as Rps5 and Rpm1 (Bakker et al. 2006), and adjacent populations are often genetically distinct (Bergelson et al. 1998; Bakker et al. 2006).

In the Midwest USA, P. syringae has been reported as a naturally occurring bacterial pathogen of A. thaliana (Jakob et al. 2002) that imposes significant fitness costs (Gao et al. 2009; Roux et al. 2010). The interaction between these two species is an important genetic and molecular model system for studying both microbial pathogenicity and plant responses to microbial infection. Multiple GFG interactions have been characterized in A. thaliana and P. syringae, including between the plant R protein Rps2 and bacterial effector AvrRpt2 (Dong et al. 1991; Whalen et al. 1991; Kunkel et al. 1993), Rps5 and AvrPphB (Jenner et al. 1991; Simonich and Innes 1995), and Rpm1, which recognizes both AvrRpm1 (Dangl et al. 1992) and AvrB (Bisgrove et al. 1994). However, all of these GFG interactions have been discovered in P. syringae strains derived from hosts other than A. thaliana, and it is not known whether these strains commonly infect A. thaliana, nor whether these specific effector genes are found in natural populations of P. syringae on wild A. thaliana plants.


Inbred lines of A. thaliana were obtained from the Arabidopsis Biological Resource Center (ABRC) or were generated from seeds collected from wild A. thaliana in northern Indiana and southwestern Michigan (see Table S1). For wild-collected seeds, an inbred line was generated by sowing a single seed in a 1:1 ratio of Metro-mix and Fafard in the University of Chicago greenhouse. Soil was watered to saturation and seeds were placed in a 4°C cold room for three days. Subsequently, seedlings were moved to a controlled growth chamber at a temperature of 20°C with a 16/8 h day/night cycle and light provided by 1:1 metal halide to HPS bulbs producing 400 μmoles m−2s−1 PAR, and selfed seeds were collected from the plants at maturity.

A collection of resident P. syringae strains was assembled from isolates derived from the leaf interior of wild A. thaliana growing in Michigan and Indiana (for collection details, see Jakob et al. 2002; Kniskern et al. 2007; Dunning 2008). Arabidopsis thaliana leaves were sterilized with either 70% ethanol or hydrogen peroxide and ground in 10 mM MgSO4 buffer. The resulting solution was plated onto solid KB growth media. Colonies were screened for a variety of morphological traits characteristic of Pseudomonas including color, size, shape, and texture (details described in Kniskern et al. 2007), but ultimately the identity of P. syringae strains was verified by sequencing a fragment of the 16S region and observing high sequence homology to published sequences (see Jakob et al. 2002; Kniskern et al. 2007). For example, the strains PNA29.1a (GenBank accession number AY574913) and RM29.1a (AY574914) were >99% similar to P. syringae pv. syringae (AB001443) and P. syringae pv. phaseolicola (AB001448), respectively (Jakob et al. 2002). The identity of these strains was further verified by developing an intraspecific phylogeny (see below). This collection of resident strains was compared to a group of nonresident strains of P. syringae collected from a variety of agricultural host species as described above (see Table S1).


We assembled 48 inbred lines of A. thaliana for the characterization of GFG resistance diversity and structure: 24 accessions were generated from seeds collected from wild populations in Midwest USA, and the remaining 24 accessions from the global distribution of A. thaliana were obtained from the ABRC (see above). Seeds of these 48 A. thaliana accessions were sown as described above but with seven days of stratification at 4°C to promote uniform germination.

For P. syringae, 24 strains were collected from the same Midwest populations as A. thaliana (resident strains), whereas the remaining 18 strains represent a worldwide collection of strains isolated from agricultural species (nonresident strains, see Table S1). We examined patterns of GFG resistance by screening all combinations of plant lines and pathogen strains for the HR, a simple resistance phenotype that involves leaf tissue collapse at the site of inoculation. This phenotype is associated with R protein-mediated recognition of effectors in A. thaliana–P. syringae interactions (Dong et al. 1991; Whalen et al. 1991; Dangl et al. 1992; Kunkel et al. 1993; Bisgrove et al. 1994; Simonich and Innes 1995). To screen for the HR resistance phenotype, plants were inoculated with bacterial strains 21–28 days following germination. To create the solution for inoculation, bacteria were grown overnight in liquid Kings Broth (KB) medium. The following morning, 5 mL of the growth media was diluted in 50 mL of fresh KB media, and then the resulting solution was grown for 4–5 h, centrifuged at 3000 rpm for 10 min., resuspended in sterile 10 mM MgSO4 buffer, and diluted to an OD600 of 0.2 in 10 mM MgSO4 buffer, or approximately 2.5 × 108 CFU/mL−1. Two leaves per plant, from four replicate plants per line, were inoculated with the bacterial solution by using a 1 mL blunt syringe. Plants were scored for the HR resistance phenotype approximately 20 h after inoculation. Prior work has shown that resistant plants express the HR within 20 h after inoculation by P. syringae (Aranzana et al. 2005; Van Poecke et al. 2007); susceptible accessions exhibit little if any disease symptoms that could be confused with the HR during this time frame. We scored each leaf as follows: no evidence of HR or mild disease symptoms = 0; severe leaf tissue collapse due to full HR = 1. Thus, each plant genotype had a total of eight leaves scored in binary format.


Because P. syringae is a diverse species complex, comprised of several discrete and divergent clades (Sarkar and Guttman 2004), we placed resident strains collected from A. thaliana within this broader framework by constructing a phylogeny of 22 resident and 22 nonresident P. syringae strains using the housekeeping gene gyrase B (gyrB). A partial fragment of gyrB was amplified from genomic P. syringae DNA extractions by using primers from Sawada et al. (1999). Forward and reverse DNA sequencing of amplicons was performed at the University of Chicago Cancer Research Center DNA Sequencing and Genotyping facility. Edited sequences were aligned using the software Bioedit (Hall 1999). Sequences used here have been deposited under Genbank accession numbers GQ199479-GQ199586. A minimum evolution phylogenetic tree was generated in MEGA version 4 (Tamura et al. 2007) by using a maximum composite likelihood model of sequence evolution and assuming uniform rates among sites. The tree was rooted with P. aeruginosa strain PA01 as the outgroup.


Many bacteria failed to elicit an HR resistance phenotype on any A. thaliana genotype (see Results). We thus tested all strains for the potential to induce the HR in tobacco as described elsewhere (Jakob et al. 2002); this test is important because it reliably distinguishes effector secreting, pathogenic Pseudomonads from nonpathogenic or saprophytic Pseudomonads (Klement 1963; Mohr et al. 2008). Failure to initiate an HR in tobacco would suggest a move away from a pathogenic life-history, while induction of the HR would leave open the possibility that these strains may be secreting novel effectors and acting as pathogens while escaping GFG recognition. Interestingly, the resident strains that failed to elicit the HR on A. thaliana also failed to elicit the HR on tobacco; these stains are hereafter referred to as HR−. Recently, Clarke et al. (2010) characterized a clade of P. syringae strains that have lost their ancestral T3SS and acquired a xenologous T3SS via lateral gene transfer. This xenologous T3SS is constitutively expressed, but seemingly lacks the capacity to elicit an HR resistance phenotype in host plants (including tobacco). The large proportion of the resident strains that were HR− led us to suspect that a similar phenomenon was occurring in our system.

To determine if our HR− strains lacked the ability to induce the HR resistance phenotype due to an inability to deliver pathogenicity effectors into the host cell, we first determined if they have the genes necessary to make the needle-like apparatus that injects effectors into plant cells. To do this, we used PCR to test for the presence or absence of conserved components of the ancestral and xenologous T3SS gene clusters in all of the HR− strains (Mohr et al. 2008; Clarke et al. 2010). For the ancestral P. syringae T3SS, we used the degenerate primer pairs developed by Mohr et al. (2008) to screen the hrp/hrc genes hrpK, hrpL, hrcC (present in all three fully sequenced reference strains of P. syringae: pv. tomato DC3000, pv. phaseolicola 1448A, and pv. syringae B728a), the CEL effector genes AvrE and hrpW, and the effector gene hop I1 (present in all tested pathogenic strains of P. syringae, including the three fully sequenced reference strains). For the xenologous T3SS described in Clarke et al. (2010), we designed primers specific for variants of the genes hrpL (F: GTCGCGATAACTCTCGATGT; R: ACCCAACAACAACTCCAGAA) and hrcC (F: TCGATATAACGCATCGGATT; R: CCAAGAACCAGATGATCCAG).

We next tested whether the HR− strains had lost the capacity to inject effectors into plants by engineering a known Avr gene into a subset of HR− strains (RMX815a, LP205a, and ME890.2a) and then determining if these engineered strains induce the HR resistance phenotype. To do this, we first created engineered strains of P. syringae with either a plasmid-borne copy of the bacterial Avr effector AvrPphB or an empty vector. To create the engineered strains, we amplified a 989 bp fragment containing the promoter and coding sequence of AvrPphB (provided by R. Innes), using high-fidelity polymerase Pfu (Stratagene) and two primers, one containing an XhoI restriction site (AvrPphB_67F_XhoI: TTTTCTCGAGCCCCTTCACAACCTCATAGC) and one containing an EcoRI site (AvrPphB_1055R_EcoRI: TTTTGAATTCAAATATTGCCGGCGTTACAG). The amplicon was then digested with XhoI and EcoRI and ligated into the similarly digested plasmid pME6010 (provided by B. Vinatzer). Escherichia coli DH5α was transformed via electroporation with either a plasmid containing the AvrPphB effector (pME6010:AvrPphB) or an empty plasmid (pME6010) and screened for tetracycline resistance (50 μg/mL). Sequencing of the entire insert with primers flanking the cloning site of pME6010 (pME6010MCSF2: GGGTGTTATGAGCCATATTCAA and pME6010MCSR2: ACTGAATCCGGTGAGAATGG) was used to confirm the absence of mutations in avrPphB that could have occurred during these manipulations. Plasmid extractions (Qiagen, Chatsworth, CA, USA) were subsequently transformed via electroporation into P. syringae strains. Colonies were selected for tetracycline resistance and PCR was used to verify the expected amplicon size for strains harboring either plasmid-borne copy of the effector AvrPphB or an empty vector.

These three engineered HR− strains now possessed a functional Avr effector that should induce the HR phenotype in a resistant host if the HR− strains possessed a functional T3SS. We next inoculated these engineered strains into the Col-0 plant ecotype, which possesses Rps5, an R gene that recognizes AvrPphB (Simonich and Innes 1995). As positive controls, we engineered with the same plasmids two strains that are known to have a functional T3SS but that do not normally elicit the HR on the Col-0 accession, because they lack effectors recognized by Col-0 (P. syringae pv. tomato DC3000 and P. syringae pv. maculicola ES4326).


To explore the influence of phylogeny and the T3SS on bacterial growth in planta, we measured growth of 40 P. syringae strains (including both resident and nonresident) in the Col-0 ecotype. PNA29.1a and Knox244a were excluded because they elicit the HR on Col-0 and exhibit low growth due to this resistance response. Col-0 was chosen to host the 40 P. syringae strains because it exhibited a low incidence of HR, and also for comparative purposes; Col-0 is the most widely used accession in P. syringae– A. thaliana studies. Each strain was inoculated into Col-0 plants at an initial average in planta density of 811.9 ± 149.9 CFU/cm2 leaf tissue (log10 value of 2.9; see Fig. 2). This initial density was estimated directly by enumerating bacterial population size for four to six replicate plants for two strains (PNA29.1a and P. syringae pv. tomato DC3000) immediately after inoculation in two separate experiments.

Figure 2.

Mean resistance of Midwest and globally distributed lines of the host A. thaliana when challenged with 12 strains of the pathogen P. syringae collected from Midwestern A. thaliana plants (resident strains) and 18 strains collected worldwide, from a largely unrelated group of 13 agricultural species (nonresident strains). (A) All P. syringae strains. (B) Only strains belonging to P. syringae clade 2 (see Fig. 1). See Table S1 for information on host and pathogen sampling. Error bars represent standard errors around the mean.

To prepare the solution for inoculation, bacteria were grown overnight in liquid KB media and prepared as described for HR tests, but were diluted to an OD600 of 0.0002 in 10 mM MgSO4 buffer, or approximately 2.5 × 105 CFU mL−1. Two leaves per plant on four replicate plants per accession were inoculated using a 1 mL blunt syringe. After three days, a hole punch was used to remove a disk of tissue from a single leaf per plant. The disk was sterilized in 70% ethanol and ground in 200 μl of 10 mM MgSO4, and then 10 μl of the resulting solution was plated in a dilution series (0, 10−1, 10−2, and 10−3 dilutions) on solid KB media. After three more days, colony number was counted to estimate bacterial population size.


We adopted a count-based approach for the analysis of the HR phenotypic data (Crawley 2007), summing for each bacterial × plant combination (n= 1432) the number of leaves exhibiting an HR (scored as 1) versus those without (scored as 0). These data were analyzed using a two-way analysis of deviance with quasibinomial errors, testing for the effects of host line and pathogen type (resident vs. nonresident). For the analysis of growth data, we compared among groups using analysis of variance (ANOVA) or t-tests in JMP 5.1 or R. All growth data were log10 transformed to meet assumptions of normality.



Resident strains were all contained within a single lineage (group 2 of Sarkar and Guttman (2004)), and are relatively closely related to P. syringae reference strain B728a (P. syringae pv. syringae B728). Within clade 2, strains were further subdivided into three discrete subclades (2a, b and c; Clarke et al. 2010). Nonresident strains were more diverse, representing at least three additional major cladistic groups (Fig. 1).

Figure 1.

Minimum Evolution tree based on a 570 bp fragment of the gyrase B gene showing evolutionary relationships among 42 strains of Pseudomonas syringae. Strains that elicit a host resistance response (HR+) are suffixed with +; those that carry a xenologous T3SS and do not elicit a host resistance response (HR−) are suffixed with −. Bootstrap values (>75%) from analysis of 10,000 replicates are shown above nodes. The tree was rooted with Pseudomonas aeuruginosa strain PA01 as the outgroup. Strains collected from Arabidopsis are marked with an asterisk. All strains collected from A. thaliana are within one of the group 2 clades, and HR− strains carrying a xenologous T3SS are found exclusively in clade 2c.


Twelve of the 24 resident P. syringae strains and all of the nonresident strains elicited the HR on at least one A. thaliana or tobacco line (although not all strains elicit the same pattern of HR; see next section). These strains are hereafter referred to as HR+. One of the HR+ strains used in this study, P. syringae pv. tomato JL1065, is known to possess the gene for the effector AvrRpt2 (Dong et al. 1991; Whalen et al. 1991). As expected, this strain induced a strong HR phenotype on A. thaliana lines that are known to possess the complementary resistance gene Rps2 (see Table S1). Several other HR+ strains, the effector complements of which are largely unknown, elicited similarly strong HR phenotypes across multiple hosts, indicating a high likelihood that these interactions are influenced by GFG resistance. The remaining 12 resident P. syringae strains failed to elicit the HR response on either A. thaliana or tobacco, suggesting the absence of GFG resistance to these strains. These strains are hereafter referred to as HR−.

When we investigated the pathogenic potential of HR− strains by engineering a known Avr gene (AvrPphB) into a subset of strains, we found that the positive control strains elicited a strong HR in Col-0 plants at 20 h after infection. In contrast, the engineered HR− strains containing a plasmid-borne copy of AvrPphB failed to elicit the HR on Col-0 plants within 20 h after infection, and also when checked at 48 h post infection. As expected, none of the engineered strains transformed with the empty control plasmid induced the HR. These data suggest that the HR− resident P. syringae strains in our study lack the ability to translocate AvrPphB at titers sufficient for induction of the HR in A. thaliana Col-0 plants, suggesting at least partial loss of T3SS function in these strains.

Consistent with the results of Clarke et al. (2010), all HR− strains fall exclusively in subclade 2c (Fig. 1). The results of PCR on conserved T3SS genes concur entirely with the phylogenetic, molecular, and phenotypic results. Specifically, primers designed to detect the ancestral T3SS in P. syringae successfully amplified conserved components of the T3SS only from the 30 (18 nonresident and 12 resident strains) HR+ strains. In contrast, primers designed to amplify the xenologous T3SS amplified fragments only from the 12 HR− strains in clade 2c. These PCR results indicate that the HR− strains have lost those genes that compose the ancestral T3SS, while gaining genes hypothesized to compose a xenologous version of the T3SS. Together, the PCR and HR test results suggest that the HR− resident P. syringae strains in our study lack a T3SS capable of delivering effectors at a concentration sufficient to induce a host resistance response.


For those HR+ strains possessing a functional T3SS and thus capable of engaging in GFG interactions with A. thaliana, there was a significantly higher frequency of resistance (HR) induced by resident bacterial strains (37%) relative to nonresidents (15%) (Fig. 2, Table 1; P < 0.00001). Because resident strains are all Clade 2 genotypes, there is a potential for phylogenetic nonindependence to influence this result. We therefore repeated the analysis comparing only strains from clade 2 (Fig. 1; n resident = 12, n nonresident = 9). Consistent with results obtained using the full dataset, resident strains induced a significantly higher frequency of resistance (37%) relative to nonresident strains (11%) (F= 134.3; P < 0.00001). Thus, we conclude that HR+ resident P. syringae strains carry a greater number of effectors recognized by A. thaliana relative to nonresident strains; that is, resident strains are maladapted compared to nonresident strains in terms of avoiding GFG recognition by co-occurring A. thaliana hosts.

Table 1.  Effect of pathogen origin (resident vs. nonresident) on patterns of gene-for-gene resistance in the interaction between Pseudomonas syringae and Arabidopsis thaliana. The analysis of deviance table was computed from a generalized linear model relating the frequency of major gene resistance to pathogen origin (resident vs. nonresident) and host origin (Midwest vs. globally distributed).
VariabledfDevianceResidual dfResidual DevianceF
  1. *P=2×10−16.

NULL  14318280.0 
Pathogen origin1677.114307602.9130.8*
Plant origin1  0.714297602.2  0.1431
Pathogen×plant1 14.114287588.1  2.72

In contrast, in testing for patterns of geographic adaptation in the hosts, we found no significant differences in the frequency of the HR phenotype among A. thaliana lines from the Midwest (23%) versus from the global distribution (24%), nor for the interaction between these factors (Fig. 2; Table 1). Consequently, we failed to find evidence that selective or demographic events have generated divergent R gene complements between A. thaliana lines collected from Midwest and global populations.


Three days after inoculation, many strains of P. syringae had increased in population size (Fig. 3). Several nonresident strains, including P. syringae pv. tomato DC3000 and P. syrinage pv. maculicola ES4328, were highly aggressive pathogens, growing to population sizes in excess of 106 CFU/cm2; these strains eventually induced disease symptoms on inoculated leaves. However, several other nonresident strains did not grow to large population sizes in A. thaliana. For example the strains P. syringae pv. syringae B728a and P. syringae pv. phaseolicola 1448A both grew to population sizes in the range of 103–104 CFU/cm2 and failed to induce disease symptoms. All resident strains of P. syringae consistently grew to population sizes in the range of 103–105 CFU/cm2. Thus, the resident strains are phenotypically quite distinct from the highly aggressive nonresident strains isolated from tomato (P. syringae pv. tomato DC3000) and Brassicaceae hosts such as radish and broccoli (P. syringae pv. maculicola ES4326; P. syringae pv. alisalensis BS136), but comparable to, or more aggressive than, nonresident pathogens isolated from other crops (e.g., P. syringae pv. syringae B728a and P. syringae pv. phaseolicola 1448A).

Figure 3.

Growth of 24 resident strains of the pathogen Pseudomonas syringae and 18 nonresident strains in the A. thaliana ecotype Col-0. Resident strains are further divided into strains that have a xenologous T3SS and do not trigger the hypersensitive response (HR) in the host (HR−), and strains that have an ancestral T3SS and do trigger HR (HR+). Bars reflect one standard error of the mean. Each strain is identified by a number found in Table S1.

Interestingly, there was no statistically significant effect of the type of T3SS (ancestral vs. xenologous) on in planta growth among 22 of the 24 resident strains (PNA29.1a and Knox244a were excluded from this analysis because they elicit the HR in Col-0 plants; F= 1.18, df = 1, 20; P= 0.28). Thus, at least under the experimental conditions used in this study, there appears to be little fitness penalty associated with the loss of a classical pathogenicity mechanism in P. syringae. This result suggests that strains of P. syringae do not require Avr effectors for successful colonization, growth and survival on A. thaliana. Finally, for strains capable of delivering effectors, there was no statistically significant effect of resident host (A. thaliana vs. crop) on in planta growth (F= 0.6151, df = 1, 28; P= 0.44).


Variation in both host resistance and pathogen virulence is well characterized in interactions between P. syringae and A. thaliana, yet we are lacking an ecologically relevant perspective on how this interaction evolves. This system is unlike pathogen–host systems that are typically modeled; it is neither obligate nor pairwise, and in fact, a small, ephemeral plant such as A. thaliana is probably not very important to P. syringae population biology. We discovered that GFG interactions are common between A. thaliana and resident P. syringae strains, and that A. thaliana plants are more likely to recognize HR+ resident P. syringae strains than nonresident strains. In other words, we found evidence that the raw genetic materials for a common form of coevolution are available, but that resident pathogens, despite faster generation times and higher migratory potential, appear to be maladapted compared to nonresident strains. Somewhat unexpectedly, we also discovered a relatively high prevalence of HR−strains that have lost the ability to interact with plants via the well-characterized mechanisms of the T3SS, which effectively removes them from participation in any characterized GFG-facilitated evolutionary interaction. These results suggest that evolution in multihost pathogen systems may proceed in ways that do not conform to expectations developed to explain evolution in more specialized, tightly coupled interactions.


In any host–pathogen interaction, the potential for evolutionary change is contingent upon genetic variation for host and pathogen traits that are important to the outcome of that interaction. Whenever resident P. syringae strains were capable of delivering effectors, at least some A. thaliana accessions exhibited a strong resistance phenotype referred to as the HR. The strength of this phenotype was identical to the HR induced by nonresident strains, and we interpret this response to reflect interactions between major genes for host resistance and pathogen virulence. All accessions of A. thaliana showed a resistance response to some HR+P. syringae strains. Moreover, no strain of P. syringae elicited a resistance response on all A. thaliana lines.

Arabidopsis thaliana plants displayed a high level of resistance to resident pathogen strains, compared to nonresident pathogens, suggesting that HR+ resident P. syringae strains carry a greater number of effectors recognized by A. thaliana relative to nonresident strains. Thus, we infer that resident strains are relatively maladapted in terms of avoiding GFG recognition by co-occurring A. thaliana hosts. As host species differ widely in resistance mechanisms, phenology, and environmental associations, variation in the level of resistance to resident and nonresident pathogens is not surprising. However, conventional wisdom holds that pathogens evolve more rapidly than their hosts and are therefore more likely to be ahead in any evolutionary arms race (Hamilton et al. 1990; Kaltz and Shykoff 1998). Given the expected adaptive advantages for P. syringae of short generation times and high migratory potential, the finding of pathogen maladaptation in this case (i.e., greater resistance of hosts to resident pathogen strains) warrants further explanation. Here, we discuss two broad explanations for the observation of maladaptation of resident P. syringae strains in the interaction with A. thaliana: (1) resident strains of P. syringae are maladapted specifically on A. thaliana, implying this host is ahead of P. syringae in their arms race of adaptive evolution; and (2) resident strains of P. syringae are generally maladapted to all hosts.

Demographic traits, such as phenology, density, and frequency, are typically highly variable among host species (Barrett et al. 2008). It is our hypothesis that variation in these traits may be a critical driver of patterns of adaptation in generalist pathogens. Specifically, we predict that uneven patterns of encounter between a pathogen and its hosts may favor the host in an evolutionary arms race, so that generalist pathogens should be maladapted to hosts that are rarely encountered due to restricted availability in space or time. A complete test of this hypothesis would require reciprocal cross-inoculation experiments investigating resistance to resident versus nonresident strains of P. syringae in other common, host species with which A. thaliana co-occurs. Our hypothesis would demand that P. syringae is ahead in the arms race against hosts that harbor substantial fractions of the pathogen population (i.e., resident strains induce resistance on fewer host genotypes than nonresidents), but behind in hosts that are ephemeral in time and space (i.e., resident strains induce resistance on more host genotypes than nonresidents). With this caveat in mind, our data do allow some insight into whether such dynamics occur.

The finding of relative maladaptation of resident strains on Midwest hosts is consistent with a scenario whereby A. thaliana hosts have an adaptive advantage over resident pathogens. In our Midwest study populations, A. thaliana is present only ephemerally and represents a relatively small resource base compared to the many other species with which it co-occurs (e.g., cultivated crop species). Indeed, bacterial populations can grow to large size in fields planted with susceptible agricultural plant varieties (Hirano and Upper 1990; Hirano and Upper 2000), and genetically similar P. syringae strains inhabit the leaves of a wide range of plants with which A. thaliana co-occurs, including the widely planted agricultural species soybean (Glycine max) and alfalfa (Medicago sativa) (L.G. Barrett, unpubl. data). Highly asynchronous epidemic dynamics can occur where agricultural and wild populations interact, so that generalist pathogens adapted to abundant agricultural species spill over into surrounding nonmanaged communities (Power and Mitchell 2004; Colla et al. 2006; Rand et al. 2006; Saleh et al. 2010). Under such a scenario, pathogens should exert selection on these incidental hosts, causing an evolutionary change in host resistance, but reciprocal selection on generalist pathogens by ephemeral weeds may be weak relative to the selection imposed by the more abundant hosts. Generalist pathogens may therefore show little evolutionary response to changes in the resistance structure of these incidental host populations, falling further and further behind in arms races involving GFG resistance.

However, results showing that Midwestern and global A. thaliana plants display a similar level of resistance to Midwestern pathogens raise questions regarding the likelihood of specific evolutionary responses in A. thaliana to infection by P. syringae. In particular, it might be expected that if the members of Midwestern plant communities that are important to P. syringae population biology differ from similarly important hosts elsewhere, then globally sampled accessions of A. thaliana will have been exposed to P. syringae adapted to a different suite of hosts. Countering this expectation would be the realization that many agricultural species that act as hosts have largely cosmopolitan distributions, thus interfering with any expectation of local adaptation. Nevertheless, if the community drivers of pathogen evolution were indeed different around the globe, and our expectations for pathogen spillover correct, then globally sampled A. thaliana should be less resistant on average to Midwestern P. syringae strains. It is thus important to entertain general maladaptation of the resident P. syringae strains as an alternative hypothesis. Such general maladaptation (in terms of triggering the HR across the majority of A. thaliana hosts) may reflect many possible scenarios. Nonadaptive dynamics in pathogen populations, such as demographic bottlenecking, limited genetic recombination or exchange, and restricted dispersal, are frequently cited as possible drivers of maladaptation (Parker 1991; Kaltz et al. 1999), although it is not obvious why nonadaptive dynamics should differentially influence resident and nonresident strains. Alternatively, adaptation to a broad range of environments, including multiple wild plant species, insect hosts (Stavrinides et al. 2009) and nonhost environments (Morris et al. 2008) may require the maintenance of a broad suite of effectors (i.e., potential Avr genes) compared to pathogens of crops, which may have a relatively small and finely tuned subset of effectors enabling growth on a handful of common crops (i.e., a cost of generalism; Poullain et al. [2008]). Rigorous tests of these competing hypotheses will require additional data on the frequency of resistance in other common, host species with which A. thaliana co-occurs. Regardless, it seems safe to conclude that adaptive constrains associated with generalism may limit the capacity of P. syringae to adapt specifically to A. thaliana.


A major life-history shift has seemingly occurred in one element of the P. syringae population resident on A. thaliana. One half (12 of 24) of resident P. syringae strains have apparently lost the ability to deliver pathogenicity effectors into cells of that host through a needle-like structure that is part of the bacterial T3SS. Mohr et al. (2008) recently described several wild P. syringae strains isolated from Primula and other plants that have likewise largely lost the ability to deliver effectors into host cells. Furthermore, these isolates form a monophyletic group and have acquired a seemingly xenologous T3SS via lateral gene transfer (Clarke et al. 2010). We used similar tests to demonstrate that our HR− resident strains lack conserved components of the classical T3SS, failed to elicit the HR in either A. thaliana or tobacco, and failed to elicit the HR in A. thaliana when transformed with an effector that the host recognizes. Despite being collected from unrelated hosts, our resident strains lacking a functional T3SS are closely related to those characterized by Mohr et al. (2008), falling into the same subclade 2c (Fig. 1).

The common presence of strains with this key polymorphism, and the finding that they do not, on average, show lower growth within the plant leaf in the A. thaliana ecotype Col-0, relative to those isolates that possess the ability to deliver effectors, raises important questions about the evolution of pathogenicity and virulence in a generalist such as P. syringae. The use of the T3SS to facilitate delivery of effectors into plant cells has hitherto been thought to be an important mechanism promoting pathogenicity and ecological fitness in P. syringae. For example, the experimental incapacitation of the T3SS in the A. thaliana HR+ resident P. syringae strain PNA29.1a causes an approximate 10-fold reduction in bacterial growth on A. thaliana (Barrett et al. 2009). Even larger reductions in pathogen growth in planta have been observed in P. syringae pv. tomato DC3000 (Chen et al. 2004) and P. syringae pv. syringae B728a (Mohr et al. 2008) when the T3SS was experimentally disabled. In contrast, experimentally disabling the xenologous T3SS does not appear to influence the growth ability of HR− strains in glasshouse experiments (Clarke et al. 2010). However, the impact of this polymorphism on the fitness of P. syringae under more natural conditions has yet to be investigated. For example, the T3SS and associated virulence genes have been demonstrated to significantly increase fitness on the surface of the leaf (Darsonval et al. 2008; Deng et al. 2009) and enhance transmission potential in the field (Hirano et al. 1999; Wichmann and Bergelson 2004; Darsonval et al. 2008).

What processes might have facilitated this major-life history transition within a diverse clade of plant pathogens? Certainly, the T3SS does not universally promote growth on all host plants. Rather, growth benefits conferred by the T3SS are typically specific to certain host–pathogen genotypic combinations (Clarke et al. 2010). Within this context, the strong maladaptation of HR+ strains resident on A. thaliana may provide some insight into the type of dynamics that potentially promote shifts along the symbiotic continuum. If, in some lineages, the benefits of the ancestral T3SS were reduced, or removed altogether, by host adaptation or pathogen maladaptation, then the benefits of maintaining the capacity to deliver effectors might be outweighed by the costs of maintaining this apparatus. Similarly, the xenologous T3SS may improve performance in alternative environments, with potentially little loss of performance in planta. These questions will provide a rich ground for further research.


The molecular and genetic basis of resistance to the bacterial plant pathogen P. syringae has been studied intensively in A. thaliana. However, previous studies have focused almost exclusively on interactions involving strains of P. syringae collected from hosts other than A. thaliana. As a result, very little is known about the virulence, phylogenetic relationships, or form of plant resistance induced by strains of P. syringae encountered by wild populations of A. thaliana. Our results suggest that this interaction can provide valuable insight into the evolutionary dynamics of interactions between hosts and multihost pathogens.

In particular, we argue that adaptive constrains associated with the capacity to infect a broad range of host species may limit the capacity of P. syringae to adapt specifically to A. thaliana, and suggest that pathogen maladaptation may be a common consequence of divergent selection pressures from alternative host species on generalist pathogens. More generally, our results suggest that dynamics in multihost pathogen systems may not conform to expectations developed to explain evolution in more specialized, tightly coupled interactions.

Associate Editor: A. Read


J. Greenberg generously provided many strains of P. syringae and B. Vinatzer graciously provided plasmids, protocols, and thoughts on this work. We thank E. B. Haney for help editing this manuscript. This research was funded by NSF grant MCB0603515 and NIH grant GM057994 to JB, and through generous support by the Dropkin Foundation.