Eukaryotic features of the Xanthomonas type III effector AvrBs3: protein domains involved in transcriptional activation and the interaction with nuclear import receptors from pepper

Authors


For correspondence (fax +49 345 55 27277; e-mail bonas@genetik.uni-halle.de).

Summary

The AvrBs3 protein of the phytopathogenic bacterium Xanthomonas campestris pv. vesicatoria is targeted to host-plant cells by the bacterial Hrp type III secretion system. In pepper plants containing the Bs3 resistance gene, AvrBs3 induces the hypersensitive response (HR). AvrBs3 recognition is thought to occur in the plant cell nucleus as HR induction is dependent on nuclear localization signals (NLSs) and an acidic transcription activation domain (AAD). In a search for AvrBs3-interacting pepper proteins using the yeast two-hybrid system, we have isolated eight different classes of cDNA inserts including two genes for importin α proteins. Importin α is part of the nuclear import machinery and interacts with AvrBs3 through an NLS in the carboxy-terminus of the protein, both in yeast and in vitro. The mechanism of AvrBs3 recognition was further studied by analysis of the C-terminal AAD. This putative transcription-activation domain was shown to be required for AvrBs3 HR-inducing activity, and could be functionally replaced with the VP16 AAD from the Herpes simplex virus. Our data support the model in which the AvrBs3 effector localizes to the nucleus, where the Bs3-mediated surveillance system of resistant plants detects AvrBs3 through its interference with host gene transcription.

Introduction

Specific recognition of bacteria, fungi, viruses and nematodes is, in many cases, based on a gene-for-gene interaction in which a resistance (R) gene in the plant mediates the detection of a pathogen expressing a matching avirulence (avr) gene. Recognition usually triggers the hypersensitive response (HR), a rapid localized cell death that is part of an active plant defence response (Morel and Dangl, 1997). More than 30 avr genes have been isolated from gram-negative plant pathogenic bacteria (Leach and White, 1996). The avirulence phenotype and the ability to colonize the plant depend on an additional set of bacterial genes, the hrp (hypersensitive response and pathogenicity) genes (Alfano and Collmer, 1997). These genes encode a type III secretion system which is highly conserved between plant and animal bacterial pathogens, and is believed to target bacterial proteins to the host-cell cytoplasm (for a comprehensive review on type III secretion see Cornelis and Van Gijsegem, 2000). Translocation of bacterial effector proteins into eukaryotic cells has first been demonstrated for the YopE protein of the mammalian pathogen Yersinia pseudotuberculosis (Rosqvist et al., 1991). There is strong indirect evidence that also plant pathogenic bacteria translocate effector proteins into host cells. In particular, expression of a bacterial avr gene in plant cells containing the cognate resistance gene generally triggers the HR (Bonas and Van den Ackerveken, 1999). Bacterial Avr proteins therefore appear to act as signal molecules inside plant cells that have evolved to specifically recognize them.

The model of intracellular Avr protein recognition is supported by the finding that most R genes effective against bacterial pathogens encode proteins that are predicted to be cytoplasmic (Parker and Coleman, 1997). The product of the tomato R gene Pto, a serine-threonine kinase, has been shown to interact with the corresponding bacterial avirulence protein AvrPto from Pseudomonas syringae in yeast (Scofield et al., 1996; Tang et al., 1996). A second example of an interaction between the products of corresponding avr and R genes has recently been described for AVR-Pita from the fungal pathogen Magnaporthe grisea and the rice resistance protein Pi-ta (Jia et al., 2000). For most other Avr proteins, however, a direct interaction with the corresponding R protein has not been demonstrated.

In our laboratory, we study the pepper and tomato pathogen Xanthomonas campestris pv. vesicatoria (Xcv). A number of avirulence genes have been cloned from Xcv (Minsavage et al., 1990), one of which is avrBs3. Pepper plants that contain the resistance gene Bs3 specifically recognize Xcv strains expressing avrBs3 (Bonas et al., 1989). Natural strains virulent on pepper Bs3 plants lack the avrBs3-containing self-transmissible plasmid. avrBs3 encodes a 122 kDa protein with a central domain composed of 17.5 nearly identical repeats of 34 amino acids (Bonas et al., 1989). The number and order of the repeats determine R gene specificity (Herbers et al., 1992). Recognition of avrBs3 requires hrp genes (Knoop et al., 1991), and it was recently shown that AvrBs3 is secreted in an Hrp (type III)-dependent way in vitro (Rossier et al., 1999). AvrBs3 is probably translocated from Xcv into host cells, as avrBs3 expression in resistant pepper cells triggers the HR (Van den Ackerveken et al., 1996). The induction of the HR depends on the presence of at least one nuclear localization signal (NLS), located in the C-terminal region of the AvrBs3 protein (Van den Ackerveken et al., 1996). The fact that NLSs are required for the induction of the HR suggests that AvrBs3 acts in the nucleus.

avrBs3 belongs to a family of avirulence and pathogenicity genes widespread in the genus Xanthomonas (White et al., 2000). All AvrBs3 family members carry functional NLSs (Yang and Gabriel, 1995), but differ in the number and order of the 34-amino-acid repeats. In addition, they all contain a putative transcriptional activation domain in the C-terminus, which was shown to be required for the avirulence activity of AvrXa10 and AvrXa7 from Xanthomonas oryzae pv. oryzae (Xoo), and AvrBs3 (Zhu et al., 1998; Zhu et al., 1999). A double-stranded DNA-binding activity was demonstrated for AvrXa7, suggesting that the protein interacts with host DNA (Yang et al., 2000).

In this study we have searched for host proteins interacting with the Xcv AvrBs3 protein, using the yeast interaction trap system. We describe the identification and analysis of the interacting proteins importin α-1 and importin α-2 from pepper, which bind to an NLS motif in AvrBs3. The potential role of AvrBs3 in modulating host-gene expression was further studied by analysis of the transcription activation domain which is required for HR induction on pepper Bs3 plants. Our results suggest that AvrBs3 induces the HR via activation of host-gene expression in resistant plant cells.

Results

AvrBs3 contains a transcription activation domain essential for HR-inducing activity

The C-terminal region of the AvrBs3 protein contains a domain that is rich in acidic amino-acid residues interspersed with bulky hydrophobic residues, a property typical of eukaryotic acidic transcription activation domains (AADs). This motif is present in all AvrBs3 family members and was reported to be functional and required for avirulence activity of the AvrXa10 protein in rice (Zhu et al., 1998; Zhu et al., 1999). Swapping experiments between avrXa10 and avrBs3 have suggested that the AAD region is also required for AvrBs3 activity (Zhu et al., 1998). To confirm that the putative AAD is important for AvrBs3 recognition in pepper plants carrying the resistance gene Bs3, a series of deletion constructs was tested for HR-inducing activity (Figure 1). A 25-amino-acid deletion in the C-terminal region of AvrBs3 starting from NLS3 (M3-Δ25, Figure 1) does not affect HR induction, nor does the modification of three amino acids through introduction of a SacI site in pDSF340 (construct BXS). In contrast, the deletion derivative ΔAAD, from which the putative AAD was removed by SacI digestion of construct BXS (Figure 1), was no longer able to induce the HR (Figure 2a). The stability and type III-dependent secretion of AvrBs3 were not affected, as demonstrated by immunoblot analysis of Xcv expressing ΔAAD (construct pDSF341, Figure 2b).

Figure 1.

Mutations in the AvrBs3 C-terminal region and their effect on HR induction in pepper Bs3 plants.

Mutations were created by PCR and combined by exchange of restriction fragments in the wild-type avrBs3 gene. The nomenclature on the left is used throughout the manuscript. Deleted amino-acid residues are indicated by dashes. The reactions on pepper ECW-30R (Bs3) are shown on the right. HR = normal HR [tissue collapse 24 h post-inoculation (pi), necrosis within 3 days]; h = delayed HR (tissue collapse 48 hpi, necrosis 4–6 dpi); hr– = weak and very delayed HR (first visible 72 hpi); – = no HR. Identical results were obtained in five independent inoculations.

Figure 2.

Role of the AvrBs3 acidic activation domain (AAD) in HR induction.

(a) Xcv transconjugants were inoculated at 5 × 108 cfu ml−1 into leaves of the pepper cultivar ECW-30R. The bacteria contained the following plasmids: pDSK602 (empty vector); pDSF340 (avrBs3 wild type); pDSF341 (AAD deletion); pDSF342 (AAD replaced with the VP16 AAD); and pDSF343 (minimal VP16 AAD). The photograph was taken 4 days after inoculation. Inoculations were repeated five times with identical results.

(b) Expression and secretion of AvrBs3 and derivatives in Xcv. The following strains were analysed by immunoblot analysis. Lanes: 1, 85*(pDSK602); 2, 85*ΔhrcV(pDSF340); 3, 85*(pDSF340); 4, 85*(pDSF341); 5, 85*(pDSF342); 6, 85*(pDSF343). Total protein extracts (Total) and culture supernatants (SN) were loaded on an 8% SDS–polyacrylamide gel and analysed by immunoblotting using AvrBs3-specific antibody.

To test if a foreign AAD could functionally complement AvrBs3ΔAAD, we replaced the AvrBs3 AAD with that from the VP16 protein of the Herpes simplex virus (ΔAAD-VP16, Figure 1). As shown in Figure 2(a), this heterologous AAD could partially complement HR-inducing activity. In an attempt to fully complement AvrBs3 activity, we used only the second subdomain of the VP16 AAD (ΔAAD-mVP16, Figure 1) which carries autonomous transcriptional activity (Sullivan et al., 1998) and was successfully used in the study of AvrXa10 in Xoo (Zhu et al., 1999). In contrast to the results for AvrXa10, the mVP16 domain construct only weakly complemented the ΔAAD mutation (Figure 2a). Both the ΔAAD-VP16 and ΔAAD-mVP16 proteins were stable in Xcv and were secreted by the type III system in vitro(Figure 2b).

AvrBs3 activates transcription in yeast

To characterize the AvrBs3 transcription activation domain, and to prepare a bait for the yeast two-hybrid screen, the protein was fused to the LexA DNA-binding domain and tested in a yeast one-hybrid system. The LexA-AvrBs3 fusion (in pYB300) did activate the lacZ reporter gene (Figure 3b). Surprisingly, deletion of the C-terminal AAD from LexA-AvrBs3 (in pYB341) did not result in loss of transcriptional activation. In contrast, deleting the AvrBs3 N-terminal 152 amino acids (in pYB356) or 300 amino acids (in pYB310) completely abolished reporter gene activation (Figure 3). The transcription-stimulating activity of the C-terminal AAD became apparent only when the avrBs3 repeat domain was deleted from the bait construct pYB356. Construct pYB311, containing only 1.5 repeats, strongly induced reporter gene expression in yeast. This activity was mediated by the C-terminal AAD, as additional deletion of this domain (construct pYB312) no longer induced transcription of lacZ(Figure 3). In all cases, expression and stability of the fusion protein in yeast were confirmed by immunoblot analysis (data not shown).

Figure 3.

LexA-AvrBs3 fusions and their effect on transcriptional activation of the lacZ reporter gene in yeast strain EGY48 (pSH18-34).

(a) Schematic overview of the LexA-AvrBs3 fusion proteins which were investigated for transcription activation in yeast.

(b) Yeast transformants were streaked onto glucose medium containing X-Gal and grown for 48 h at 30°C to assay for β-galactosidase activity. This assay was repeated four times with identical results.

Two-hybrid screen for AvrBs3-interacting proteins

To screen for AvrBs3-interacting proteins we used pYB356 (wild-type AvrBs3 without the first 152 aa) as a bait in the interaction trap system. As mentioned above, this construct does not auto-activate transcription in yeast. In addition, Agrobacterium tumefaciens-mediated expression of the N-terminal deletion derivative still induced the HR in Bs3 pepper plants (Rossieret al., unpublished results), suggesting that the N-terminal 152 amino acids are dispensable for AvrBs3 activity in resistant plants. The prey cDNA library was made from mRNA isolated from leaves of pepper line ECW-123. Approximately 0.5 × 106 independent cDNA clones were transformed into yeast and tested for leucine prototrophy and LacZ activity. Approximately 500 colonies were isolated, of which approximately 50% showed galactose-dependent LacZ activity. Plasmids from these yeast clones were grouped into 19 distinct classes based on restriction analysis, two of which contained the majority of clones. For one clone of each class, reproducibility of the interaction with AvrBs3 was tested by re-transforming the library plasmid into yeast. For eight of these different cDNA inserts, the interaction with AvrBs3 could be reproduced in yeast. LexA-bicoid and LexA-AvrBs1 were used as controls for non-specific interactions. AvrBs1 is an Xcv Avr protein (Ronald and Staskawicz, 1988) unrelated in sequence to AvrBs3. As the repeat region determines the avr specificity of the avrBs3 gene, LexA-372 was used to assess whether proteins interacted with AvrBs3 in a repeat-specific manner. This construct is derived from AvrBs3Δrep16, which lacks four repeats and is not recognized by the Bs3 gene but by bs3 in pepper (Herbers et al., 1992). All of the eight different cDNAs identified encode proteins that interact with both AvrBs3 and AvrBs3Δrep16. The analysis of two interactors is presented here.

Characterization of two pepper genes for importin α

cDNA clones corresponding to two different importin α genes were identified. The cDNA for pepper importin α-1, designated Caimpα1 (Capsicum annuum importin α-1) was isolated 25 times. The longest clone contained a 1.9 kb insert with an open reading frame (ORF) of 1605 bp. Based on sequence comparison with Arabidopsis thaliana importin α genes, this clone is probably full-length and encodes a protein of 535 aa with a molecular mass of 59.3 kDa. Another cDNA clone is 65% identical to Caimpα1 at the DNA level, and thus represents a different importin α gene designated Caimpα2. As the latter lacked the 5′ end containing the putative translation initiation codon, the complete sequence of this ORF was characterized by 5′ RACE. The predicted CaIMPα2 protein has 529 aa and a molecular mass of 58.5 kDa.

Sequence comparison of the putative pepper importin α-1 and importin α-2 revealed 76% identity (85% similarity) at the amino-acid level. The proteins are highly related to importin α from yeast SRP1 (52 and 57% identity, respectively) (Loeb et al., 1995) and A. thaliana (70 and 84% identity) (Ballas and Citovsky, 1997; Hicks et al., 1996), and display characteristics that are found in all members of the importin α family: an N-terminal stretch of basic amino acids which binds to importin-β (Görlich et al., 1996; Weis et al., 1996); eight armadillo repeats (Peifer et al., 1994); and an acidic C-terminal domain (Köhler et al., 1997) (Figure 4a).

Figure 4.

Amino-acid sequence comparison of the pepper importin α proteins with homologues from other plants and yeast SRP1.

(a) Multiple alignment of yeast SRP1, A. thaliana At-IMPα, pepper CaIMPα1 and CaIMPα2. Identical residues are shaded according to the degree of conservation: black, 100%; dark grey, 80–100%; light grey, 60–80% conserved positions. The IBB (interaction with importin β) domain and arm (armadillo) repeats are indicated above the alignment. GenBank accession numbers: M75849for SRP1, AF077528for At-IMPα, AF369706 for CaIMPα1 and AF369707 for CaIMPα2.

(b) Phylogenetic tree of plant importin α proteins.

A tIMPa1–4 are from A. thaliana (Y14615, Y14616, Y15224, Y15225); OsIMP1 is from rice (AB004660); and LeKAPα1 is from tomato (AF017252). GenBank accession numbers are given in brackets. Protein sequence alignments and tree construction were performed using the dnastar clustal algorithm.

A phylogenetic tree of plant importin α proteins (Figure 4b) shows that CaIMPα1 is most similar to AtIMPa4, and CaIMPα2 to AtIMPa2 of A. thaliana (Schledz et al., 1998). The latter protein is highly homologous to the well studied At-IMPα (Hicks et al., 1996). CaIMPα2 groups together with importin α from tomato (Kunik et al., 1999) and rice (Shoji et al., 1998). Southern blot hybridization using the pepper importin α-1 cDNA as a probe detected multiple fragments in pepper genomic DNA at low stringency, suggesting the presence of a small gene family (Figure 5). Similarly, there is also a small importin α-2 gene family (data not shown). RT–PCR and Northern blot analysis indicated that both Caimpα1 and Caimpα2 are constitutively expressed in pepper leaves independently of infection with Xcv (data not shown).

Figure 5.

Southern blot analysis of the Caimpα1 homologues in pepper.

Genomic DNA (20 µg) from pepper ECW-30R, digested with the restriction enzymes EcoRV, KpnI, DraI and EcoRI, was separated on an agarose gel, transferred to a nylon membrane, and hybridized at high stringency to a full-length CaImpα1 cDNA probe. Arrows on the right indicate the presence of two faint bands in the DraI digestion.

AvrBs3 interacts with CaIMPα1 and CaIMPα2 through NLS2

The interaction of the pepper importin α proteins with AvrBs3 was analysed in more detail. Deletion of the NLS region (83 aa) of AvrBs3, in construct pYB356ΔNLS, completely abolished the interaction with importin α. AvrBs3 carries two functional NLSs (see Figure 1, NLS2 and NLS3) that are required for HR induction (Van den Ackerveken et al., 1996). NLS2 and NLS3 were deleted individually, generating bait plasmids pYB356Δ2 and pYB356Δ3, respectively, as well as pYB356Δ2Δ3 in which both deletions are combined. As shown in Figure 6, CaIMPα1 and CaIMPα2 interact with AvrBs3 only when NLS2 is present. Curiously, replacement of the NLS region of AvrBs3 by the simian virus 40 (SV40) large T-antigen NLS, in construct pYB356ΔNLS-SV40, did not restore the importin α interaction, although the SV40 NLS does restore HR-inducing activity of avrBs3ΔNLS when delivered by Xcv (Van den Ackerveken et al., 1996). In all cases, expression and stability of the fusion proteins in yeast were confirmed by immunoblot analysis (data not shown).

Figure 6.

AvrBs3–importin α interaction is dependent on NLS2.

AvrBs3 and derivatives in the bait vector are depicted on the left. Plate assays were performed with yeast transformants spotted at equal density on galactose medium containing X-Gal. The plate was incubated for 48 h at 30°C, and photographed. This assay was repeated four times with identical results. The result for CaIMPα1 is shown; identical results were obtained with CaIMPα2. Interaction levels were quantified in liquid-culture assays using ONPG as substrate. β-galactosidase activity is expressed in Miller units. Measurements are mean (with standard deviation) of six independent assays. The avrBs3 activity in resistant pepper (HR induction) is shown on the right (see Figure 2 and Experimental procedures for details).

To measure the strength of the protein interactions, β-galactosidase activities were determined. When AvrBs3 was co-expressed with either CaIMPα1 or CaIMPα2, β-galactosidase activity was approximately 200 times higher than with bicoid (negative control). β-galactosidase activity was 10-fold lower for AvrBs3Δ3 compared with the wild type (pYB356), which might be due to a conformational change in the C-terminus in AvrBs3Δ3. However, Xcv expressing the NLS3 deletion derivative avrBs3Δ3 induced the HR in pepper ECW-30R as efficiently as the wild-type avrBs3 gene. Similarly, avrBs3Δ2 had wild-type HR-inducing activity. In contrast, avrBs3Δ2Δ3 failed to induce the HR, indicating that the presence of NLS2 or NLS3 is required for AvrBs3 to be fully active (Figures 1 and 6), and that NLS1, located 46 residues upstream of NLS2, is not sufficient for AvrBs3 activity as was observed previously (Van den Ackerveken et al., 1996).

AvrBs3 binds to importin αin vitro

The interaction between AvrBs3 and importin α was analysed further in an in vitro GST pull-down assay. AvrBs3 clearly bound GST-CaIMPα1 and not GST alone, as shown by immunoblot analysis (Figure 7). In addition, AvrBs3Δ3 also bound GST-CaIMPα1 efficiently, whereas no binding was observed with AvrBs3Δ2 and AvrBs3Δ2Δ3. Therefore these results confirm the data obtained for yeast, and suggest a direct interaction between AvrBs3 and pepper importin α-1 through NLS2.

Figure 7.

In vitro binding of importin α to AvrBs3.

The interaction between AvrBs3 and importin α was tested by a GST pull-down assay (see Experimental procedures). Lysates of E. coli expressing avrBs3 or NLS deletion derivatives were incubated with GST (G) or GST-CaIMPα1I) immobilized on sepharose beads. Total protein extracts of the different E. coli strains are shown to estimate the relative expression level of each avrBs3 construct. 10 µl eluates (G or I) and 5 µl of total protein extracts of E. coli (T) were loaded on an 8% SDS–polyacrylamide gel and analysed by immunoblotting using AvrBs3-specific antibody.

Discussion

Knowledge of the molecular mechanisms of host–pathogen biology in the Xcv–pepper interaction mainly derives from studies on the bacterial side. In order to identify plant cellular components involved in the interaction with Xcv, we screened for pepper proteins interacting with the avirulence protein AvrBs3. Using the yeast interaction trap we have isolated two nuclear import receptors from pepper, CaIMPα1 and CaIMPα2. The deduced amino-acid sequence of these proteins shows high similarity to the yeast NLS-binding protein SRP1 (Loeb et al., 1995) and other importin α (also designated karyopherin α) proteins from animals and plants. Several A. thaliana and rice importin α genes are known, and the corresponding proteins specifically bind to NLS-conjugated BSA or NLS-containing cargo in vitro (Hicks et al., 1996; Shoji et al., 1998; Smith et al., 1997).

Importin α acts as a subunit of a heterodimeric NLS receptor complex involved in the nuclear import pathway of proteins containing basic NLSs (Mattaj and Englmeier, 1998). Importin α binds to the NLSs of the import substrate in the cytoplasm (Görlich et al., 1994) and then associates with importin β, which mediates docking at the nuclear pore complex (Görlich et al., 1995). The CaIMPα1 and CaIMPα2 proteins contain the conserved importin β interaction domain and therefore probably use the corresponding nuclear import pathway in pepper. Thus AvrBs3 is the first known example of a type III-secreted effector recruiting the host's nuclear import machinery.

The only other example of a bacterial protein interacting with the eukaryotic nuclear import machinery is VirD2 from A. tumefaciens, which binds to importin α AtKAPα from A. thaliana (Ballas and Citovsky, 1997). VirD2 is covalently bound to the T-DNA and is required for efficient transport of the T-DNA into the plant cell nucleus (Zupan et al., 2000).

We have demonstrated interaction of both CaIMPα1 and CaIMPα2 with AvrBs3 in yeast and in vitro by GST pull-down assays. In both approaches, protein interaction was dependent on the presence of an NLS in the C-terminal region of AvrBs3, confirming the predicted NLS receptor function of CaIMPα1 and CaIMPα2. Moreover, the sequence responsible for binding AvrBs3 was mapped to a few amino acids encompassing NLS2. It is intriguing that both importin α proteins identified here bind to NLS2, and apparently not to NLS3, which was confirmed by in vitro assays. No interaction was observed with AvrBs3 containing only NLS3 or the SV40 NLS (in AvrBs3ΔNLS::SV40), although fusion proteins containing either NLS were previously shown to target a GUS fusion protein to the plant nucleus (Van den Ackerveken et al., 1996). Taken together, these data suggest that NLS receptors other than CaIMPα1 and CaIMPα2 in pepper could recognize AvrBs3 proteins containing NLS3 or the SV40 NLS. There are more members of the importin α gene family in pepper (Figure 5).

Previous analysis of the AvrBs3 NLS region showed that combined mutations in NLS2 and NLS3 (M23) reduce and delay HR induction in Bs3 pepper plants, pointing to a minor role of NLS1 (Van den Ackerveken et al., 1996). Here we show that the simultaneous deletion of NLS2 and NLS3 completely abolished HR induction on resistant plants, confirming that NLS1 is not sufficient for AvrBs3 activity. The weak activity of construct M23 is therefore due not to NLS1, but rather to residual activity of the mutated NLSs. Xanthomonas oryzae pv. oryzae (Xoo) strains expressing the AvrBs3 homologue AvrXa10 with the corresponding NLS (nlsA) alone retained some avirulence activity (Zhu et al., 1998). This discrepancy might be due to a single amino-acid difference between NLS1 of AvrBs3 (KRAK) and nlsA of AvrXa10 (KRVK).

The cumulative evidence that signalling in Bs3-mediated recognition of AvrBs3 takes place in the nucleus is supported by the finding that AvrBs3 and all family members contain an acidic transcription-activation domain (AAD) (Zhu et al., 1998). The AvrBs3-AAD deletion mutant could be partially restored for HR induction in pepper by the AAD from the H. simplex VP16 protein. The VP16 domain does not share any primary amino-acid sequence homology with AvrBs3, suggesting that complementation is due to the ability to recruit the transcription machinery, rather than to a structural role in recognition. The observation that the VP16 domain only partially complements could be due to a reduced ability of the viral AAD to activate transcription in pepper. The VP16 domain is composed of two subdomains, each possessing activation activity (Sullivan et al., 1998). Our construct harbouring only the second domain (mVP16) had a strongly reduced HR-eliciting activity compared to the full-length VP16 construct, suggesting that the two domains might co-operate in elicitation of the AvrBs3::VP16 HR. In contrast to our results, the AvrXa10 AAD deletion mutant could be fully complemented for HR induction on rice by the mVP16 AAD (Zhu et al., 1999). This difference in results could be due to the cloning strategies used leading to slightly different protein fusions, or to differences between the respective host plants, pepper and rice.

While the NLSs and AAD in AvrBs3 are important for eliciting the plant defence response, they would not have arisen in evolution having such a counter-selective purpose. Presence of these motifs in a bacterial type III-secreted protein implies a role in susceptible plants, where manipulation of host genes would be of benefit to the bacteria. Indeed, the AvrBs3 homologues PthA from X. citri and Avrb6 from X. campestris pv. malvacearum not only elicit the HR on resistant plants, but also confer a selective advantage to the bacteria on susceptible plants through canker elicitation on citrus and increased watersoaking on cotton, respectively (Swarup et al., 1991; Swarup et al., 1992; Yang et al., 1994; Yang et al., 1996). Recent field studies on AvrXa7, an AvrBs3 family member from Xoo, demonstrated its contribution to pathogen aggressiveness on susceptible rice lines (Bai et al., 2000). This avrXa7 effect correlates with an NLS- and AAD-dependent increase in lesion length (Bai et al., 2000; Yang et al., 2000). We have observed that AvrBs3 triggers cell hypertrophy on susceptible plants in an NLS- and AAD-dependent manner (E.M. and co-workers, unpublished results). This supports a model in which importin α is an intermediate target of AvrBs3 necessary for nuclear localization and subsequent gene activation. We therefore hypothesize that resistant pepper plants have evolved a recognition system for AvrBs3 in which signalling occurs downstream of nuclear localization and gene activation. This is in contrast to our model of recognition of the highly related protein AvrBs4 (previously named AvrBs3-2) in tomato, in which recognition is proposed to occur in the plant cell cytoplasm (Ballvora et al., 2001).

Together, our results are consistent with the model in which the AvrBs3 protein is injected into the host cell via the Hrp secretion system (Rossier et al., 1999) and transported into the nucleus by importin α, where AvrBs3 then alters host-gene expression. In susceptible plants, the function of AvrBs3 is probably to manipulate host gene transcription. In the presence of the corresponding resistance gene (Bs3), AvrBs3 recognition depends on both the NLSs and the AAD, suggesting that some components of the Bs3-mediated signalling pathway resulting in cell death require transcription activation. The recent observation that the AvrBs3 homologue AvrXa7 (Yang et al., 2000) shows double-stranded DNA-binding activity in vitro suggests that recognition might take place via a direct DNA interaction, as proposed for AvrXa10 (Zhu et al., 1998). Alternatively, AvrBs3 could first interact with host protein(s) to form a protein complex, which then binds DNA and activates gene transcription. Identification of host genes that are activated by AvrBs3, and subsequent analyses of their promoter regions and DNA-binding proteins, are the next challenge in understanding the molecular basis of AvrBs3-triggered resistance.

Experimental procedures

Bacterial strains and growth conditions

Bacterial strains used in this study were Escherichia coli strains DH5α (Bethesda Research Laboratories, MD, USA), XL1-MRF′ (Stratagene, La Jolla, CA, USA) and KC8 (Ausubel et al., 1996); Agrobacterium tumefaciens strain C58C1 (Van Larebeke et al., 1974); and Xanthomonas campestris pv. vesicatoria (Xcv) strains 85-10 (Minsavage et al., 1990), 85* and 85*ΔhrcV (Rossier et al., 2000). Escherichia coli cells were cultivated at 37°C in Luria–Bertani medium, Xcv at 30°C in NYG broth (Daniels et al., 1984), and A. tumefaciens at 30°C in YEB medium. Plasmids were introduced into E. coli by electroporation and into Xcv and A. tumefaciens by conjugation, using pRK2013 as a helper plasmid in triparental mating (Figurski and Helinski, 1979).

Plant material and inoculations

Pepper (Capsicum annuum) plants of cultivar Early Cal Wonder (ECW), and the near-isogenic line ECW-30R containing the resistance gene Bs3 (Minsavage et al., 1990), were grown and inoculated with Xcv as described (Bonas et al., 1989). Pepper cultivar ECW-123 contains the genes Bs1, Bs2 and Bs3, conferring resistance to Xcv strains carrying the corresponding avirulence gene avrBs1, avrBs2 and avrBs3, respectively. Agrobacterium-mediated transient-expression assays were performed as described previously (Van den Ackerveken et al., 1996).

Construction of AvrBs3 C-terminal deletion derivatives

To construct M3-Δ25, the 3′ end of avrBs3 was amplified from pBS300 using Pfu and primers CT-AAD1A (GCACGCGTGTCATGAGGGAACAAGATGAGGAC) and T3. The PCR product was digested with MluI (underlined in primer sequence) / HindIII and ligated into pBS300-M3 (in which NLS3 had been mutated and a MluI restriction site introduced; Van den Ackerveken et al., 1996), digested with the same enzymes. To create an XhoI and SacI site at the C-terminal BamHI site in avrBs3, a fragment was amplified from pBS300 using primers MUT-B-XS (GAAGATCTCGAGCTCCCCACGGCTGCCGAC) and T3. The PCR product was digested with BglII (underlined in primer sequence) and HindIII and ligated into BamHI (partial) / HindIII-digested pUS300 to obtain pUS300-BXS. The EcoRI/HindIII inserts from all constructs were cloned into pDSK602 (used for gene expression in Xcv;Murillo et al., 1994), resulting in clones pDSF300-M3-Δ25 and pDSF300-BXS, respectively. The AAD was deleted from pDSF300-BXS by SacI digestion and re-ligation to give pDSF341.

Replacement of the AvrBs3 AAD with that of the Herpes simplex protein VP16 was achieved by inserting an XhoI/SacI-digested amplification product obtained from plasmid pRG50 (Cousens et al., 1989) using primers VP16-up-Xho (GGGGCTCGAGCTGCACTTAGACGG) and VP16-low-Sac (GATAAGCTTAGAGCTCCCCACCGTACTCGTCAATTC) into XhoI/SacI-restricted pUS300-BXS. This operation introduced 67 codons from VP16 followed by one extra codon (glutamic acid) into the 36-codon AAD deletion derivative of AvrBs3 (Figure 1). The same strategy was used to produce the VP16 minimal domain (mVP16) fusion, using primer mVP16-up-Xho (GAGACTCGAGGCCGACTTCGAGTTTG) and VP16-low-Sac, resulting in the introduction of 23 codons from VP16 and an extra codon (glutamic acid) (Figure 1). The EcoRI/HindIII fragments carrying avrBs3::VP16 and avrBs3::mVP16 were cloned into pDSK602, yielding pDSF342 and pDSF343, respectively. NLS2 and NLS3 were deleted from AvrBs3 by site-elimination mutagenesis. By using the oligonucleotide del-nls2 (GAGGGAGATCAGACGCGTGCTGTCACCGGTC), 33 nucleotides were removed from pBS300, creating a MluI restriction site. For deletion of NLS3, the oligonucleotide del-nls3 (CTGCATTTGCCCCTCTCGATCGGGGGCGGCCTC) was used, creating a BsiEI restriction site. A double mutant was created by releasing a fragment from pBS300Δ3 by digestion with AgeI and HindIII and cloning it into AgeI/HindIII-digested pBS300Δ2. pBS300Δ2, pBS300Δ3 and pBS300Δ2Δ3 were digested with EcoRI/HindIII and cloned into pDSK602.

LexA-AvrBs3 and derivatives

Constructs pUS356 (O. Rossier and U. Bonas, unpublished results) and pUS310, respectively, contain AvrBs3 carrying N-terminal deletions of 152 and 300 aa. For this, sequences were deleted from pUSF300 (Van den Ackerveken et al., 1996) starting from the BamHI site, immediately downstream of the avrBs3 translation start codon. To remove the repeat domain, pUS310 was digested with NcoI and re-ligated, leaving 1.5 of the original 17.5 repeats (to obtain construct pUS311). To construct pUS312, the AgeI/HindIII fragment from pUS300-BXS was ligated into pUS311 (AgeI/HindIII). For AvrBs3 AAD analysis in yeast, EcoRI/XhoI or EcoRI/SalI fragments were released from pUS300, pUS341, pUS356, pUS310, pUS311 and pUS312, and cloned into EcoRI/XhoI-digested pEG202 to generate pYB300, pYB341, pYB356, pYB310, pYB311 and pYB312, respectively (see also Figure 3).

For interaction-trap analysis, StuI/HindIII fragments were released from the pBS300 series and cloned in StuI/HindIII-digested pBS356 to obtain pBS356Δ2, pBS356Δ3 and pBS356Δ2Δ3. The same procedure was applied to obtain pBS356ΔNLS and pBS356ΔNLS::SV40 (Van den Ackerveken et al., 1996). The EcoRI/XhoI inserts were cloned from the pBS356 series into pEG202 to generate pYB356, pYB356Δ2, pYB356Δ3, pYB356Δ2Δ3, pYB356ΔNLS and pYB356ΔNLS::SV40. In all cases, expression and stability of the fusion protein were confirmed by immunoblotting analysis using polyclonal anti-AvrBs3 antibody (Knoop et al., 1991) and a monoclonal anti-LexA antibody (Clontech, Palo Alto, CA, USA).

Secretion assays

Secretion assays were performed as described (Rossier et al., 1999). To verify that detection of AvrBs3 in the culture supernatant was due to secretion and not to bacterial lysis, the intracellular protein HrcN served as lysis control. Detection was with polyclonal anti-HrcN antibody (Rossier et al., 1999).

Construction of an interaction-trap cDNA library

Tissue was harvested from healthy young leaves of pepper line ECW-123. RNA was isolated using GuanidineHCl and LiCl precipitation (Logemann et al., 1987), and poly(A)+ RNA was prepared using oligo(dT)-cellulose columns (Pharmacia, Uppsala, Sweden). cDNA was synthesized using a cDNA kit in which the first strand synthesis is primed with a poly(dT)-XhoI linker primer (Stratagene). cDNAs were size-fractionated on a sepharose column, and fractions containing cDNA fragments of 0.5–4 kb were ligated into the EcoRI–XhoI sites of pJG4-5 (prey vector). The ligation mixture was transformed into E. coli XL1-Blue MRF′. Approximately 0.5 × 106 independent cDNA clones were obtained with an average cDNA insert size of 1.5 kb.

AvrBs3 interactor screening

For the interactor screening, the yeast-interaction trap was used (Gyuris et al., 1993) following standard protocols (Ausubel et al., 1996). The plasmid library was transformed en masse into yeast strain EGY48 containing the lacZ-reporter plasmid pSH18-34 and the bait plasmid pYB356, using the PEG–lithium acetate method (http://www.umanitoba.ca/faculties/medicine/units/human_genetics/gietz/). 2 × 106 yeast transformants were recovered on selective glucose medium. Approximately 4 × 107 yeast cells were plated on galactose medium containing X-Gal (5-bromo-4-chloro-3-indolyl-β-d-galactopyranoside). Fast-growing colonies (appearing within 3–6 days) with high β-galactosidase (LacZ) activity were subjected to a second round of reporter gene analysis. cDNA inserts from yeast cells showing galactose-dependent LacZ activity were amplified by PCR, using pJG4-5 primers BCO1 and BCO2 which flank the cDNA cloning sites. PCR products were digested with AluI and HaeIII, and separated on 1.5% agarose gels, leading to the identification of different classes of clones. The prey plasmid was transformed into E. coli KC8 or DH5α. The reproducibility and specificity of the interaction were tested by re-transformation of the prey plasmid into yeast strains containing pYB356 and a panel of different bait plasmids, respectively. Alternative bait plasmids were pRFHM-1 (LexA-bicoid; Ausubel et al., 1996); pYB100 (LexA-AvrBs1); and pYB372 (LexA-372, which derives from AvrBs3Δrep16 and carries a 152 aa deletion in the N-terminus).

Transcriptional activation assays

For β-galactosidase assays on plates, four independent yeast transformants were spotted at equal density onto glucose or galactose plates containing X-Gal and incubated for 48 h at 30°C. β-galactosidase activities were quantified after growth of yeast strains in liquid culture using ONPG (o-nitrophenyl β-d-galactopyranoside) as substrate (Ausubel et al., 1996). For each interaction test the mean and standard deviation of six measurements are shown.

Rapid amplification of cDNA ends (RACE)

5′ RACE PCR was performed on Poly(A)+ RNA from pepper line ECW-123, using a 5′ RACE kit (Life Technology, Gibco, Rockville, MD, USA) according to the manufacturer's instructions. The gene-specific primers used were GSP1 (AGAGTGGTGATTGACC) and GSP2 (TGCAACTGGAGGCGAACACC). Amplicons obtained from independent PCR reactions were sequenced and analysed using the dnastar package (DNASTAR Inc., Madison, WI, USA).

Southern blot analysis

Pepper genomic DNA was isolated from leaf material using extraction buffer (0.2 m Tris–HCl pH 7.5, 0.25 m NaCl, 25 mm EDTA, 0.5% SDS). After incubation at 60°C for 30 min and extraction with chloroform, DNA was precipitated by isopropanol, washed with 70% (v/v) ethanol, and re-suspended in distilled water. For Southern blot analysis, approximately 20 µg DNA was digested with EcoRV, KpnI, DraI and EcoRI, respectively, separated on a 1% agarose gel and transferred to Hybond N+ (Amersham, Uppsala, Sweden) according to the manufacturer's instructions. The Caimpα1 cDNA was labelled with 32P-dCTP (ICN Biomedicals, Irvine, CA, USA) using the Megaprime DNA Labelling Kit (Amersham). Filters were washed twice with 2 × SSC/0.1% SDS (10 min, 65°C), once with 0.5 × SSC/0.1% SDS (15 min, 65°C), and twice with 0.1 × SSC/0.1% SDS (10 min, 65°C) and exposed to X-ray film.

GST pull-down assay

Importin α-1 was expressed as a glutathione S-transferase (GST) fusion protein in E. coli BL21. For this, an EcoRI/XhoI (partial) fragment from Caimpα1 cDNA clone No. 26 was ligated in-frame into pGEX-2TK (Amersham Pharmacia Biotech), in which the MCS was changed to the sequence GGATCCGAATTCCCGGAGCTCGAGCTAGCAATTC. Bacteria were re-suspended in PBS (1/10 vol) containing 1 mm DTT, and broken using a French Press. Soluble proteins were immobilized on a glutathione resin and non-bound E. coli proteins were removed by washing of the resin three times with PBS buffer, as described in the manufacturer's instructions (Amersham Pharmacia Biotech). AvrBs3 deletion derivatives were expressed in E. coli BL21 (DE3), under the control of the T7 promoter using the pBS356 series (see above). 600 µl of bacterial protein extracts were applied to 40 µl of prepared glutathione resin. After incubation for 1 h at 22°C room temperature, the resin was washed five times with 1 ml ice-cold PBS buffer to eliminate non-specific binding proteins. Elution was performed at RT with 40 µl 10 mm reduced glutathione, and eluates were subjected to SDS–PAGE and immunoblot analysis using anti-AvrBs3 and anti-GST (Amersham Pharmacia Biotech) antibodies to verify that GST constructs were correctly expressed.

Acknowledgements

We thank Angelika Landgraf for excellent technical assistance. We are grateful to Dr Roger Brent for providing the yeast interaction-trap plasmids and strains, and to Ombeline Rossier for providing pUS356 and pBS356. We appreciated the technical hints on the yeast two-hybrid experiments by Jeff Leung and the group of Gunter Reuter, and the scientific discussions with all members of our group. This work was funded in part by an EC grant (BIO4-CT97-2244) and the Deutsche Forschungsgemeinschaft (SFB 363) to U.B. and E.M. was supported by a Marie Curie fellowship from the European Community.

GenBank accession number sequences: Caimpα1: AF369706; Caimpα2: AF369707.

Ancillary