A Functional Screen for the Type III (Hrp) Secretome of the Plant Pathogen Pseudomonas syringae

See allHide authors and affiliations

Science  01 Mar 2002:
Vol. 295, Issue 5560, pp. 1722-1726
DOI: 10.1126/science.295.5560.1722


Type III secreted “effector” proteins of bacterial pathogens play central roles in virulence, yet are notoriously difficult to identify. We used an in vivo genetic screen to identify 13 effectors secreted by the type III apparatus (called Hrp, for “hypersensitive response and pathogenicity”) of the plant pathogenPseudomonas syringae. Although sharing little overall homology, the amino-terminal regions of these effectors had strikingly similar amino acid compositions. This feature facilitated the bioinformatic prediction of 38 P. syringae effectors, including 15 previously unknown proteins. The secretion of two of these putative effectors was shown to be type III–dependent. Effectors showed high interstrain variation, supporting a role for some effectors in adaptation to different hosts.

The bacterial type III secretion system is responsible for some of the most devastating diseases of animals and plants. This remarkable system enables a bacterium to strategically inject proteins directly into the host cytoplasm or its extracellular milieu, and thereby subvert host cellular processes (1–3). The type III apparatus is required for pathogenesis and is highly conserved across a broad range of Gram-negative bacterial pathogens. Less conserved is the repertoire of proteins exported (the type III secretome) (1–3). Little is known about the function or mechanism of action of phytopathogenic effectors, although several are known to enhance the growth rate and transmission potential of the pathogen (4, 5). Avirulence (avr) genes are a class of phytopathogenic effectors that restrict host range (3). Most Avr proteins are thought to be secreted through the type III apparatus. Plants with the appropriate cognate resistance (R) genes recognize Avr proteins and mount a defense response characterized by a type of programmed cell death (PCD) called the hypersensitive response [HR (6)].

Variation in the constellation of effector genes among related pathogenic strains may facilitate adaptation to new hosts and permit the rapid evolution of novel pathogen specificities (4,7). To critically test this hypothesis, a comprehensive analysis of the identity and function of effector genes is necessary. To achieve this end, we devised an in vivo screen exploiting the modular nature of effectors (5, 811) and the well-characterized intracellular interaction, between the COOH-terminus of the AvrRpt2 effector and its cognate R protein, Rps2 (12). Because the COOH-terminal HR-inducing domain of AvrRpt2 (lacking the NH2-terminal secretion signal) is a good in vivo reporter for type III secretion (5, 13), we were able to devise a transposon containing the DNA coding for AvrRpt281-255 to capture type III secretion signals from unknown effector genes, also known as hop genes [hrp/hrc outer protein (14)] (Web fig. 1) (15). Insertions of the avrRpt281-255 transposon that create translational fusions with hop genes generateP. syringae strains that induce the HR in an Rps2-dependent manner upon infection of the model host plant, Arabidopsis thaliana. The screen relied on the type III secretion signal and the endogenous promoter of the hop gene and was thus highly specific. During infection at least one type III–utilizing pathogen delivers one set of effectors to the extracellular environment, and another set to the host cell interior (16). The precise mechanism governing the final location of an effector is largely unknown. On the basis of the data reported here, the final destination of the Hop::AvrRpt281-255 fusions, inside plant cells, appears to be driven largely by the AvrRpt281-255 moiety, although a specific translocation signal for AvrRpt2 is not known. This enabled the capture of type III delivery signals from a broad variety of effectors that are likely to have a number of different sites of action.

We chose to study effectors from P. syringae pv. maculicola strain ES4326 (PmaES4326) because theA. thaliana host responses elicited by this strain are very well characterized (17–19). Approximately 75,000 independent transposon insertion strains were screened (20). Twenty-five independent HR-inducing isolates with fusions to AvrRpt281-255 identified 13 differenthop genes (Table 1) that, aside from hopPmaA, were chromosomally located (21).

Table 1

Characterization of effectors identified by in vivo screening in PmaES4326 and functional genomic analysis of PtoDC3000. All hop genes were sequenced to completion, with the exception ofavrEPma . For this effector all characterizations were based on the PtoDC3000 homolog (AAF1499) except for the Blast analysis which used the incomplete (inc) sequence (40% missing). BLASTP and PSI-BLAST queries were done using the nr database at the National Center for Biotechnology Information (47). HopPtoO and HopPtoP were identified in the unfinished genome sequence of PtoDC3000 and confirmed to be secreted by creating fusions with AvrRpt281-255. holPmaN was amplified by PCR from a cosmid with primers designed on hopPmaL. Additional information is presented in Web table 1 and Web note 4 (15).

View this table:

Seven Hop proteins had widely varying degrees of similarity to known phytopathogenic effectors or P. syringae virulence factors (Table 1). These included a putative ortholog and an apparent paralog (HopPmaB) of AvrPphE (22, 23), the latter having a eukaryotic NH2-myristoylation motif at amino acid 2 that may facilitate targeting to host cell membranes. HopPmaD was a member of the Xanthomonas campestris AvrBsT and Yersinia pestis YopJ effector family present in many plant and animal pathogens. Members of this family induce PCD in both plant and animal hosts. Although the similarity between HopPmaD and AvrBsT is weak, all the amino acids implicated in inducing PCD were conserved (24). The first 20 amino acids of HopPmaD were very similar to those found in the AvrPto effector (25), suggesting recombination may generate new hop genes. HopPmaL was similar to VirPphA, a P. syringae virulence protein not previously known to be secreted (26). HolPmaN (Hop-like) was a truncated, presumably nonfunctional protein similar to HopPmaL that was found through polymerase chain reaction (PCR) analysis of PmaES4326. Orthologs of the harpin HrpW effector and the virulence protein AvrE were also recovered.

Five previously unknown Hops (including two identified using bioinformatics; see below) contained regions most similar to proteins found in animal pathogens or nonpathogenic bacteria. HopPmaA had similarity to a VT1- or VT2-Sakai prophage protein of unknown function in pathogenic E. coli (27, 28). HopPmaG and HopPtoP were similar to different transglycosylases. These enzymes act on the bacterial peptidylglycan layer; one such enzyme is required for flagellar formation (an assembly requiring a type III–like apparatus) in Salmonella enterica (29). HopPmaG and HopPtoP may similarly facilitate assembly of the type III apparatus. HopPmaH showed high similarity to a plant cell wall–degrading pectate lyase, an enzyme important for P. marginalis virulence (30). HopPmaI had COOH-terminal similarity to a DnaJ domain; such domains interact with the molecular chaperone Hsp70 and alter its substrate binding (31). HopPmaI also had an ATP- or GTP-binding site motif A (P-loop) at amino acid 118, and three full and one partial proline-rich tandemly repeated sequences (32). These latter repeats are present nearly five times in thePtoDC3000 homolog (32). The variable numbers of repeats could be important for adaptation to different hosts, similar to some Xanthomonas campestris effectors with repeated segments (33). ThehopPmaI1-315::avrRpt281-255 insertion strain had reduced virulence (Fig. 1, A and B). Finally, HopPtoO had a myristoylation motif at amino acid 2 and similarity to a region of an NAD(P)(+)-arginine ADP-ribosyltransferase 2 precursor that was also present in the P. aeruginosa effector ExoS (34).

Figure 1

Demonstration of reduced virulence and type III–dependent secretion of P. syringaeHop::AvrRpt281-255 fusions. (A) Virulence growth assay of PmaES4326hopPmaI1-315::avrRpt281-255 (squares) versus PmaDG6 [circles; PmaES4326recAΩavrRpt2 (5)] on A. thaliana rps2-101C. Leaves were inoculated at OD600(optical density at 600 nm) = 0.0002 as described (5). Six samples of each genotype were titered at each time point. *P < 0.003. Error bars are obscured by data symbols. Similar results were obtained in three separate experiments. (B) Leaves are shown at 3 days after infection from the virulence growth assay in (A). Note the reduced chlorosis on the leaf infiltrated with PmaES4326hopPmaI1-315::avrRpt281-255 . (C) HR (avirulence) assay demonstrating type III–dependent secretion of HopPmaA1-284::AvrRpt281-225 including a positive and a negative control [PmaDG6,PmaES4326 containing an integrated copy of full-lengthavrRpt2, and PmaDG3, PmaES4326 containing an integrated copy of a vector control, respectively (5)]. pHrcC is a PmaES4326 cosmid clone (3B3) containing the hrcC gene (20). All other Hop::AvrRpt281-255 fusions behaved similarly (Web table 1). (D) Western blot of total protein extracted from a PmaES4326 (WT) strain and aPmaES4326 hrcC strain both containing an integrated copy ofhopPmaA1-284::avrRpt281-255 (band corresponding to the fusion protein indicated by an asterisk) in order to verify that the HopPmaA1-284::AvrRpt281-225 fusion protein is expressed to similar levels in the HR+ and HR strains (48). A background band at 90 kD served as a loading control (Web fig. 2).

All hop::avrRpt281-255 strains showed RPS2-dependent HRs (Web table 1), indicating that the Hop-AvrRpt281-255 fusions were not generally cytotoxic. Type III–dependent delivery was confirmed for each of the Hop::AvrRpt281-255 fusion proteins (Fig. 1, C and D and Web table 1). Additionally, all but one hop gene (hopPmaJ) had upstream sequences called “hrp boxes” found in promoter regions of many effector genes (35). Finally, the synthesis of most of the Hop::AvrRpt281-255 fusion proteins was undetectable in rich medium and induced by minimal medium, similar to what has been found for other P. syringae effectors (Web fig. 2 and Web table 1) (36, 37). The only exceptions were HopPmaA1-284::AvrRpt281-255 which was synthesized in rich medium and was further induced in minimal medium, and HopPmaJ1-48::AvrRpt281-255 which was produced constitutively.

Pseudomonas syringae effectors have exceptionally high Ser content in the NH2-terminal 50 amino acid secretion regions compared to the rest of the protein (Tables 1 and2). A low Asp, Leu, and Lys content in the effectors' NH2-termini was also observed. These NH2-terminal amino acid biases resemble chloroplast and mitochondrial targeting sequences (38). Indeed, mostP. syringae effectors are predicted to localize to chloroplasts (Table 1 and Web table 3). This similarity could reflect a common mechanism used by the type III apparatus and organelle import complexes to recognize secretion signals and to facilitate secretion. Alternatively, the similarities could either indicate a common evolutionary origin for secretion and organelle targeting mechanisms, signify an analogous functional requirement for these sequences for proper targeting of effectors to host organelles [oneE. coli effector localizes to mitochondria (39)], or simply be coincidental. Salmonella enterica effectors also show NH2-terminal Ser and Asp biases (Table 2), suggesting a conserved function for this feature. The overall amino acid composition of P. syringaeeffectors was also biased, showing a high overall Ser and Asn content and a low overall Leu, Ile, and Val content (Web table 2). Some features of this amino acid bias were shared with P. syringae homologs of flagellar secreted proteins from S. enterica and with S. enterica effectors (Web table 2). Some amino acid biases may be important for the unique requirement of effectors and flagellar components to be unfolded during the delivery process and to refold upon reaching their final destinations (40).

Table 2

Analysis of NH2-terminal amino acid composition of type III effectors. The Hop group (Hop) includes the proteins shown to be secreted (Table 1) as well as the known type III–secreted proteins: HopPtoB (ORF1 EEL) ofPtoDC3000 (AAF71498), AvrRpt2 (CAA79815), AvrRpm1 (NP_114197), HopPsyA (AAF71481), AvrB (AAA25726), HrpZ (AAB00127), AvrPto (AAA25728), HrpA (AAB00126). The Flagella group (Fla) includesPtoDC3000 homologs (identified in the unfinished genome sequence at of flagellar type III–secreted proteins from S. enterica (n = 12). The control group (Con) includes a group of randomly selected proteins fromP. syringae (n = 23). TheSalmonella effector group (SalEff) includes S. enterica effectors (n = 24). TheSalmonella control (SalCon) group includes a group of randomly selected proteins from S. enterica(n = 32). All of these loci are presented in Web table 4. P values (P) were determined using the nonparametric Kolmogorov Smirnov test. P values in bold indicate significance (≤0.05).

View this table:

We estimated the total number of effectors in a single strain ofP. syringae (PtoDC3000) based upon the characteristic NH2-terminal amino acid biases (high Ser and low Asp) and the conserved hrp box promoter element (35) (Web table 3). Preliminary PtoDC3000 sequence data was obtained from The Institute for Genomic Research (TIGR; The proposed secretome includes 38 proteins, of which 15 are putative novel effectors. Two of these latter ORFs (HopPtoO and HopPtoP, Table 1) were tested and shown to be delivered to A. thaliana (Web table 1 and Web fig. 2). Pseudomonas syringae may contain additional effector genes induced in different environmental contexts and thus lack hrp boxes. Indeed,P. syringae may interact with hosts from at least two kingdoms, as it contains homologs of the Photorhabdus luminescens insectidical toxin complex genes tcaA-C,tcbA, tccA-C, and tcdA ( (41).

A core set of 10 of the 15 confirmed effectors was present in many commonly studied P. syringae strains causing diseases in different hosts (Table 3). Some of these effectors are also present in related Pseudomonads and other plant pathogens. A third of the effectors studied showed a high degree of variability, even among strains with similar host specificity. Several effectors also appear to be duplicated and some duplicatedhop genes may have been inactivated by base changes, transposon insertions or genetic rearrangements (see holPmaNexample above and the avrPphDPto ,hopPtoO2 and holPtoU2 examples in Web table 3).hop gene inactivation or loss may contribute to pathogen fitness by allowing the pathogen to evade the host surveillance mechanism. In summary, hop genes represent a highly dynamic set of genes in P. syringae populations.

Table 3

hop genes distribution in selectedPseudomonas syringae strains and other Gram-negative pathogens. Genomic DNA extraction, Eco RI digest and Southern blot followed standard protocols (45). Probes were generated by PCR and corresponded to full-length hop genes (except forhopPmaI, which was missing the region homologous to dnaJ) from Table 1. PmaES4326,PmaM5, PtoDC3000, PmaM3,PmaM1, PmaM2, PmaM6,Pto5034, Pc83-1, XccBP109, andXcc750 infect A. thaliana. PtoDC3000 andPto5034 also infect tomato. Pph3121,Psy61, and PsyB782 infect bean.

View this table:

We have shown that the type III secretome of P. syringae is likely to contain more effectors than in any other pathogen characterized so far (2). The principle of our in vivo functional screen can be adapted to any type III–utilizing bacterial species. This approach, coupled with the ability to predict effectors based on their biased amino acid composition and mode of regulation affords the possibility of making rapid progress in understanding the mechanisms of action and the functions of effectors in diverse pathogens.

  • * These authors contributed equally to this work.

  • To whom correspondence should be addressed. E-mail: guttman{at}


View Abstract

Navigate This Article