Current address: Max-Planck-Institut für Züchtungsforschung, Carl-von-Linné-Weg 10, D-50829 Köln, Germany.
These authors contributed equally to this work.
The Lycopersicon esculentum Bs4 resistance (R) gene specifies recognition of Xanthomonas campestris pv. vesicatoria (Xcv) strains that express the cognate AvrBs4 avirulence protein. Bs4 was isolated by positional cloning and is predicted to encode a nucleotide-binding leucine-rich repeat (NB-LRR) protein that is homologous to tobacco N and potato Y-1 resistance proteins. Xcv infection tests demonstrate that Bs4 confers perception of AvrBs4 but not the 97% identical AvrBs3 protein. However, when delivered via Agrobacterium T-DNA transfer, both, avrBs4 and avrBs3 trigger a Bs4-dependent hypersensitive response, indicating that naturally occurring AvrBs3-homologues provide a unique experimental platform for molecular dissection of recognition specificity. Transcript studies revealed intron retention in Bs4 transcripts. Yet, an intron-deprived Bs4 derivative still mediates AvrBs4 detection, suggesting that the identified splice variants are not crucial to resistance. The L. pennellii bs4 allele, which is >98% identical to L. esculentum Bs4, has a Bs4-like exon-intron structure with exception of a splice polymorphism in intron 2 that causes truncation of the predicted bs4 protein. To test if the receptor-ligand model is a valid molecular description of Bs4-mediated AvrBs4 perception, we conducted yeast two-hybrid studies. However, a direct interaction was not observed. Defense signaling of the Bs4-governed reaction was studied in Nicotiana benthamiana by virus-induced gene silencing and showed that Bs4-mediated resistance is EDS1- and SGT1-dependent.
Plants are a nutritious habitat for phytopathogenic microbes and thus have had to evolve countermeasures that minimize assimilate plundering. Cells of resistant plant genotypes challenged by pathogens frequently respond with a controlled suicide program, termed the hypersensitive response (HR; Klement and Goodman, 1967). Genetic analysis of plant–microbe interactions has shown that perception of microbial invaders is often determined by complementary pairs of plant resistance (R) and pathogen avirulence (avr) genes. The receptor-ligand model is one biochemical interpretation of gene-for-gene resistance and predicts that the Avr ligand binds directly to an R receptor that subsequently activates a defense reaction (Gabriel and Rolfe, 1990; Keen, 1990). Analysis of two gene-for-gene interactions have provided experimental support for this model (Jia et al., 2000; Kim et al., 2002; Scofield et al., 1996; Tang et al., 1996). An alternative biochemical interpretation of gene-for-gene resistance known as the guard model (Van der Biezen and Jones, 1998b) postulates that the R protein (the ‘guard’) detects Avr-triggered changes of a plant virulence target (the ‘guardee’). Identification and analysis of the Arabidopsis RIN4 protein provided experimental support for this model as RIN4 interacts physically with both the Arabidopsis RPM1 R protein and its cognate Avr proteins AvrRpm1 and AvrB (Mackey et al., 2002).
The majority of R genes cloned to date encode putatively cytoplasmic proteins with nucleotide-binding leucine-rich repeat (NB-LRR) domains (Dangl and Jones, 2001; Ellis et al., 2000b; Holub, 2001; Martin et al., 2003; Young, 2000), and this structure–function relationship has inspired homology-based approaches aimed at identification of potential candidate R genes (Pflieger et al., 2001). NB-LRR proteins differ structurally in their N-termini, comprising either a Toll/interleukin-1-receptor (TIR)-homologous region or a coiled-coil (CC) domain. Genetic dissection of defense signaling in Arabidopsis has shown that TIR-NB-LRR proteins signal through EDS1, whereas CC-NB-LRR proteins signal predominantly through NDR1 (reviewed in Feys and Parker, 2000; Glazebrook, 2001). By contrast, analysis of Arabidopsis Rar1 and SGT1 showed that these signaling components are engaged by both structural NB-LRR subtypes. Mutational studies of the Nicotiana benthamiana homologs of Rar1, EDS1 and SGT1, as well as the barley homologs of Rar1 and SGT1 indicate that these components are functionally conserved in Arabidopsis, N. benthamiana and barley (reviewed by Shirasu and Schulze-Lefert, 2003).
Genetic screens initially identified bacterial avr genes as mediators of gene-for-gene resistance (Staskawicz et al., 1984). The ability of many avirulent bacteria to elicit HR also depends upon the hypersensitive response and pathogenicity (hrp) genes. As the name implies, these genes are also required for bacteria to cause disease on susceptible plants. hrp genes encode a type III secretion system that injects bacterial Avr proteins into the host cell (Alfano and Collmer, 1997). This would allow these proteins to interact with NB-LRR type R proteins, which have an intracellular location. The existence of this sophisticated injection machinery implies that the primary function of Avr proteins is in virulence rather than avirulence. Indeed increasing evidence suggests a dual role of Avr proteins as recognition determinants in resistant and virulence determinants in susceptible hosts (Gabriel, 1999a; Luderer and Joosten, 2001; Van't Slot and Knogge, 2002).
We study tomato and pepper bacterial spot disease, which is caused by the Gram-negative bacterium Xanthomonas campestris pv. vesicatoria (Xcv). In particular, our interest is in the R gene-mediated perception of AvrBs3 and AvrBs4 – two members of the Xanthomonas AvrBs3 family that share 97% sequence identity (Bonas et al., 1993). AvrBs3 and AvrBs4 are recognized specifically by the corresponding pepper Bs3 and tomato Bs4 R genes, respectively (Ballvora et al., 2001a). We have shown that perception mediated by pepper Bs3 but not by tomato Bs4 depends on functional NLSs in the corresponding Avr proteins, which suggests different recognition principles for the detection of almost identical avirulence proteins in pepper and tomato (Ballvora et al., 2001a). The abundance and highly conserved structure of AvrBs3 family members make them a useful experimental system in which to study the specificity of R gene-mediated Avr protein perception. However, the molecular isolation of a corresponding R gene for any of the family members has not yet been reported.
Previously, we presented the genetic mapping and physical delimitation of the tomato Bs4 locus (Ballvora et al., 2001a,b). Here, we report the isolation and functional analysis of the tomato Bs4 gene. Bs4 encodes a TIR-NB-LRR protein with most similarity to the tobacco N (Whitham et al., 1994) and potato Y-1 proteins (Vidal et al., 2002). Transcript studies uncovered Bs4 splice variants, which, however, appear to be not required for Bs4-mediated HR. Analysis of recognition specificity indicates that Bs4 has the ability to mediate not only detection of AvrBs4 but also other AvrBs3-like proteins. Yeast two-hybrid (Y2H) studies suggest that Bs4 does not directly interact with AvrBs4. Furthermore, we demonstrate that Bs4 is functional in Solanum tuberosum and Nicotiana species and that its function is EDS1- and SGT1-dependent.
Bs4 mediates recognition of AvrBs4-deletion derivatives
Analysis of F2 segregants derived from a cross between Lycopersicon esculentum cultivar (cv.) Moneymaker (MM) and L. pennellii LA2963 has shown that recognition of Xcv strains that express AvrBs4 is mediated by the L. esculentum cv. MM Bs4 locus (Ballvora et al., 2001a). Earlier studies have shown that not only full-length AvrBs4 but also AvrBs4-deletion derivatives trigger an HR on L. esculentum cv. MM (Figure 1; Bonas et al., 1993). Yet, it remained unclear, if Bs4 or other unlinked R loci govern recognition of these AvrBs4-deletion derivatives. We therefore analyzed six different C-terminal AvrBs4-deletion derivatives on a set of 20 F2 progenies derived from a cross between L. esculentum cv. MM (Bs4) and L. pennellii LA2963 (bs4) (Figure 1). The allele configuration of these F2 segregants at the Bs4 locus was determined with molecular markers that flank Bs4 on either side of the locus (see Methods for details). Xcv infection tests showed that C-terminal AvrBs4-deletion derivatives, which contain 3.5 or more repeat units, triggered an HR in Bs4/− segregants and the L. esculentum cv. MM but not in bs4/bs4 segregants (Figure 1). On the contrary, an AvrBs4 derivative that contains 26 residues of the first repeat unit (AvrBs4 Δ233) did not induce an HR in any of the tested tomato genotypes. Notably, all tested Bs4/− segregants and L. esculentum cv. MM showed identical infection phenotypes with AvrBs4 and its deletion derivatives. These findings indicate that the previously determined avirulence activity of AvrBs4-deletion derivatives (Bonas et al., 1993) was because of Bs4-dependent recognition.
Degenerate PCR and high-resolution mapping uncovers a Bs4 candidate gene
A PCR-based strategy with degenerate primers targeting the conserved TIR motif of TIR-NB-LRR-encoding R genes was applied to isolate Bs4 candidate genes from tomato genomic DNA. The chromosomal positions of the cloned PCR products were assessed by RFLP mapping in a standard tomato mapping population (Tanksley et al., 1992) and placed one of the isolated fragments (T26 RGC) in the vicinity of the Bs4-linked RFLP marker TG432 (Figure 2; Ballvora et al., 2001b). T26 RGC was converted into a PCR-based RFLP marker and employed for the analysis of a mapping population that segregates for Bs4 resistance. Linkage analysis of 1972 meiotic events revealed that T26 RGC fragments diagnostic for the Bs4-parental genotype L. esculentum cv. MM co-segregated with an HR phenotype that was visible 48 h after infiltration of avrBs4-expressing Xcv. Furthermore, analysis of two yeast artificial chromosome (YAC) clones, that were shown previously to span the Bs4 locus (Ballvora et al., 2001b), demonstrated that both YACs contain the T26 RGC marker locus. In summary, high-resolution genetic and physical mapping supported the notion that T26 RGC was part of a potential Bs4 candidate gene.
A TIR-NB-LRR-encoding Bs4 candidate gene mediates specific recognition of AvrBs4
To determine if the YAC insert DNA that was used for physical delimitation of Bs4 could be employed for complementation studies, we performed infection tests with L. esculentum cv. VFNT Cherry, the DNA source for YAC library construction (Bonnema et al., 1996; Martin et al., 1992). Inoculation tests showed that avrBs4- but not avrBs3-expressing Xcv triggers an HR, indicating that the L. esculentum cv. VFNT Cherry contains the Bs4 gene.
In order to isolate T26 RGC-flanking sequences, we generated a cosmid library of YAC clone Y45, which was shown previously to span the Bs4 locus (Ballvora et al., 2001b). Sequence analysis of T26 RGC-containing cosmids revealed a putative TIR-NB-LRR-encoding candidate gene (termed Bs4c). Amplification and sequence analysis of the Bs4c homolog from L. esculentum cv. MM, the genotype that was used for high-resolution linkage mapping of Bs4, revealed that the nucleotide sequence of Bs4c is identical in VFNT Cherry and MM. Southern analysis of L. esculentum cv. MM and YAC Y45 DNA indicated that Bs4c is a single-copy gene (data not shown).
To determine whether Bs4c mediates AvrBs4 recognition, the previously established L. esculentum bs4 backcross line MMbs4–BC4 (Ballvora et al., 2001a) was transformed with a binary vector (pVTSB1) in which the Bs4c genomic fragment is under transcriptional control of the CaMV 35S promoter. Xcv infection tests showed that 16 out of 22 transgenic tomato plants displayed an AvrBs4-dependent HR (Figure 3). Analysis of several independent T1 and T2 lines showed that inheritance of AvrBs4-responsiveness was strictly dependent on the presence of Bs4c, indicating that Bs4c is indeed the Bs4 gene.
Recognition specificity of the Bs4c-transgenic plants was tested with Xcv strains that deliver (i) AvrBs4, (ii) AvrBs4 Δ227 (C-terminal deletion derivative of AvrBs4), (iii) AvrBs3, or (iv) AvrBs1 (no sequence homology with AvrBs4; Ronald and Staskawicz, 1988), respectively (Figure 3). Infiltration of the Bs4c-transgenic lines with xanthomonads that express avrBs4 or its deletion derivative avrBs4 Δ227 triggered a rapid HR. By contrast, avrBs3- and avrBs1-expressing Xcv did not induce an HR. Notably, all avr derivatives produced identical infection phenotypes in Bs4c-transgenic lines and the Bs4-containing genotype L. esculentum cv. MM and did not trigger HR in the bs4 genotype MMbs4–BC4 (Figure 3). Taken together, our complementation studies led us to conclude that Bs4c is indeed the tomato Bs4 gene.
Agrobacterium tumefaciens-mediated expression of avrBs3 triggers a Bs4-dependent HR
Recognition specificity of the L. esculentum cv. MM Bs4 allele was also studied by A. tumefaciens-mediated transient expression (agroinfiltration). In contrast to our Xcv infection tests we found that, when delivered by agroinfiltration, not only avrBs4 and its deletion derivative avrBs4 Δ227 but also avrBs3 triggered an HR in L. esculentum cv. MM and the Bs4 transgenic lines (Figure 4a). We did not observe an HR in the bs4 genotype MMbs4–BC4, indicating that agroinfiltrated avrBs3 was detected in a Bs4-dependent manner. It is worth noting that the HR was indistinguishable in the Bs4-transgenic lines and L. esculentum cv. MM although the Bs4 transgene is under control of the CaMV 35S promoter (Figure 4a).
Analysis of unchallenged tissue by real-time PCR revealed that the Bs4 transcript levels in our 35S::Bs4 transgenic tomato lines were approximately 100-fold higher when compared to L. esculentum cv. MM (Bs4) (data not shown). To examine whether Bs4 overexpression influences the infection phenotypes, we cloned a genomic fragment containing the Bs4 open-reading frame (ORF) and approximately 3.5-kbp upstream sequence into a promoterless binary vector (pVTSB3). Functionality of Bs4 under transcriptional control of (i) its putative native promoter or (ii) the CaMV 35S promoter was compared by Agrobacterium-mediated co-expression with avrBs4 (Figure 4b). In this assay, a bs4 plant genotype was infiltrated with a mixture of two Agrobacterium cultures, one expressing the respective avr gene, and the other expressing Bs4 under transcriptional control of either the CaMV 35S or the putative Bs4 native promoter. Phenotypic inspection of the reactions mediated by the CaMV 35S and the native promoter constructs showed no differences with respect to recognitional specificity, extent, timing, and intensity of the HR (Figure 4b).
The Bs4 transcript undergoes alternative splicing
Reverse transcriptase (RT)-PCR was used to determine the Bs4 exon-intron structure (summarized in Figure 5), because attempts to isolate a Bs4 full-length cDNA clone were unsuccessful. Based on rapid amplification of cDNA ends (RACE) analysis, the Bs4 transcript contains 25 and 500 bp of 5′- and 3′-untranslated region (UTR), respectively. A comparison of the Bs4 genomic sequence with the RACE products revealed the presence of a 90-bp intron in the 3′-UTR. Inspection of the Bs4 ORF by means of intron finder algorithms suggested the presence of three introns in the Bs4 ORF that could be confirmed by RT-PCR with primers flanking the predicted introns. In addition, we obtained Bs4-splice variants that still contained intron 2 or 3 but lacked intron 1. To determine the functional relevance of these splice variants, we removed introns 1–3 and used this intron-deprived Bs4 derivative for agroinfiltration of the bs4 genotype MMbs4–BC4. We found, that the intron-deprived Bs4 ORF mediates avrBs4 recognition, indicating that the introns are not crucial to Bs4 functionality. Likewise, we tested Bs4 constructs lacking the 5′- or 3′-UTRs and found that UTRs are functionally dispensable.
The predicted Bs4 protein is highly related to the virus resistance proteins potato Y-1 and tobacco N
The Bs4 transcript encodes a predicted protein of 1146 amino acids with a molecular weight of 131 kDa. Among proteins with a known function, Bs4 most closely resembles the TIR-NB-LRR proteins potato Y-1 (57% identity, 71% similarity; Vidal et al., 2002) and tobacco N (54% identity, 67% similarity; Whitham et al., 1994) (Figure 6). The Bs4 protein displays at its far N-terminus (aa residues 1–8) a characteristic sequence motif (MASSSSSS), which is also present in potato Y-1 (Vidal et al., 2002), tobacco N (Whitham et al., 1994), Arabidopsis RPP4 (Van der Biezen et al., 2002), Arabidopsis RPP5 (Parker et al., 1997), and other putative TIR-NB-LRR proteins (Hehl et al., 1999). Comparison of the Bs4 TIR (aa residues 17–162), NB (aa residues 224–500), and LRR (aa residues 626–987) regions with the corresponding domains of Y-1 and N shows that the TIR is the most conserved and the LRR the most divergent domain between these R proteins. The C-terminal region of Bs4 is composed of 15 repeat units with similarity to the cytoplasmic LRR consensus sequence (Jones and Jones, 1997). We found no apparent N-terminal signal sequences in Bs4, suggesting that it is a cytoplasmic protein. However, PROSITE motif search (http://www.expasy.ch/prosite) identified potential myristoylation sites in Bs4, which might mediate membrane anchoring of the protein.
bs4 transcripts encode a truncated bs4 protein
Sequence comparison of the L. pennellii LA2963 bs4 allele with the L. esculentum cv. MM Bs4 allele revealed in total 74 nucleotide polymorphisms (Figure 7). Analysis of putative exons predicts that 28 out of 60 nucleotide polymorphisms would lead to aa differences between the bs4- and Bs4-encoded proteins. The predicted bs4 exons do not contain an early stop codon that would unequivocally define bs4 as a null allele. Transcript analysis showed that bs4 and Bs4 share an almost perfect conservation of exon–intron junctions with the exception of the 5′ splice site of intron 2 (Figure 7). This modified exon–intron boundary is predicted to cause a frame-shift and hence to encode a truncated protein that lacks the LRR region. This is confirmed by sequence analysis of 32 cloned RT-PCR fragments, which revealed no bs4 transcripts that would encode an LRR region.
Bs4 and AvrBs4 do not show interaction in the Y2H system
To test whether AvrBs4 interacts physically with Bs4, we performed Y2H studies. An AvrBs4 bait construct (N-terminal fusion of AvrBs4 to the LexA DNA-binding domain) was found to activate both the leu2 and the lacZ reporter genes in the absence of the prey construct. To overcome this autoactivation problem, we constructed an N-terminal deletion derivative of AvrBs4 (AvrBs4 Δ152) lacking aa residues 1–152. AvrBs4 Δ152 no longer activates the yeast reporter genes but still triggers a Bs4-dependent HR when the corresponding gene is delivered by Agrobacterium. This finding suggests that the N-terminus of AvrBs4 is not required for Bs4-dependent recognition. Next, an intron-deprived Bs4 gene, that lacks introns 1–3, and that was shown to be functional (see above) was cloned into the prey vector. We also generated reciprocal bait and prey constructs (AvrBs4 Δ152, prey vector; Bs4, bait vector) and confirmed by Western blot analyses for all constructs expression of appropriate-sized proteins in yeast (data not shown). A repression assay (Brent and Ptashne, 1984) showed that the LexA-AvrBs4 Δ152 and LexA-Bs4 fusion proteins can bind to nuclear-localized operator sequences, indicating that both proteins are suitable for Y2H interaction studies. Pepper importin α1 (Caimpα1), previously identified as an interactor of AvrBs3 (Szurek et al., 2001), also interacts with AvrBs4 and was used as a positive control for our Y2H studies. The Drosophila bicoid protein was used as a control for non-specific interactions.
To study interaction of Bs4 and AvrBs4, we tested AvrBs4 Δ152 bait and Bs4 prey constructs and the reciprocal combination (AvrBs4 Δ152 prey; Bs4 bait). The lacZ reporter gene activity measured in AvrBs4 Δ152–Bs4 co-transformants in no case exceeded the levels of the negative controls (bicoid in combination with AvrBs4 Δ152 or Bs4) and was about 10 times lower than that for the positive control (Figure 8). The same yeast co-transformants lacked leucine prototrophy, indicating that AvrBs4 Δ152 and Bs4 proteins do not interact in yeast. We also cloned different Bs4 regions (TIR, NB-LRR, and LRR) into the prey vector and tested these for interaction with the AvrBs4 Δ152 bait protein. In accordance with the data observed for the full-length Bs4 protein, co-transformation of the separated Bs4 regions with AvrBs4 Δ152 (Figure 8) did not activate the leu2 or the lacZ reporter gene, suggesting that a simple receptor-ligand model does not hold true for the avrBs4–Bs4 gene-for-gene interaction.
Tomato Bs4 is functional in S. tuberosum and Nicotiana species
To examine whether species other than tomato have all accessory proteins that are needed for a Bs4-mediated defense response, we performed agroinfiltration of the solanaceous plant species N. tabacum, N. clevelandii, N. benthamiana, Capsicum annuum, C. frutescens, S. tuberosum, and Petunia hybrida. Agroinfiltration of a binary vector carrying the β-glucuronidase (GUS) gene induced reporter activity in the infiltrated tissues, thereby indicating that all tested plant species are transformable by Agrobacterium (data not shown). Subsequently, we performed agroinfiltration of these different plant species using Bs4 in combination with avrBs4, avrBs4 Δ227, avrBs3, and avrBs1. Co-expression of Bs4 with avrBs4, avrBs4 Δ227 and avrBs3, but not avrBs1 triggered an HR in N. tabacum, N. clevelandii, N. benthamiana, and S. tuberosum (Figure 9; data not shown). No HR was observed in P. hybrida, C. annuum, and C. frutescens. We noted in S. tuberosum that agroinfiltration of avrBs4, but not avrBs3, induced a Bs4-independent HR, which is most likely because of an intrinsic R gene. However, as the avrBs3-triggered HR in S. tuberosum was only observed when co-expressed with Bs4, we conclude that Bs4 is functional in S. tuberosum. In summary, our results indicate that S. tuberosum and Nicotiana species but not Capsicum or Petunia species contain the elements that are required for the Bs4-mediated HR.
Bs4-mediated HR is EDS1- and SGT1-dependent
As a first step towards elucidation of the signal components that mediate the Bs4-specified defense response we analysed known R gene pathway elements. To study the functional relevance of EDS1, NDR1, RAR1, and SGT1 in the Bs4-mediated HR, we employed virus-induced gene silencing (VIGS), a well-established tool for transient gene silencing in N. benthamiana (Baulcombe, 1999). As VIGS requires a high degree of sequence similarity between the recombinant virus and the target RNA (Mueller et al., 1995), it was necessary to use the N. benthamiana orthologs of EDS1 (NbEDS1), NDR1 (NbNDR1), Rar1 (NbRar1), and SGT1 (NbSGT1). The NbEDS1 and NbSGT1 genes have been described before (Peart et al., 2002a,b) and for isolation of NbNDR1 and NbRar1, we employed a homology-based approach (see Experimental procedures for details). VIGS was initiated by inoculation of tobacco rattle virus (TRV) vector derivatives carrying a fragment of NbEDS1, NbNDR1, NbRar1, or NbSGT1, respectively. Approximately 21 days after the first inoculation, Bs4 was agroinfiltrated in combination with avrBs4 or avrBs3, respectively. Plants that were infiltrated with an empty virus vector showed an HR in the leaf area, in which Bs4 was co-expressed with avrBs4 or avrBs3 (Figure 10). This indicates that the viral infection did not interfere with the Bs4-mediated HR. However, in NbEDS1- and NbSGT1-silenced plants co-expression of Bs4 with avrBs4 or avrBs3 (Figure 10) triggered no HR, indicating that both genes are crucial to Bs4-governed HR execution. On the contrary, Bs4-mediated HR was not affected in NbNDR1 and NbRar1-silenced plants.
A putatively cytoplasmic R protein mediates recognition of a nuclear-targeted Avr protein
A positional candidate approach was used to isolate the tomato Bs4 gene, which encodes a novel member of the TIR-NB-LRR class of R proteins. Tomato Bs4 was shown to mediate recognition of the Xanthomonas AvrBs4 protein, which is a member of the well-characterized AvrBs3 family (Lahaye and Bonas, 2001). AvrBs3-like proteins contain NLS sequences that are indicative for nuclear targeting (Gabriel, 1997; Leach and White, 1996; White et al., 2000), and recent immunocytochemical studies indeed demonstrated that AvrBs3 and AvrBs4 are localized to the plant nucleus (Szurek et al., 2002; Szurek and Bonas, unpublished results). Molecular analysis of several gene-for-gene interactions suggests that R-protein localization is generally dictated by the Avr protein destination (reviewed by Bonas and Lahaye, 2002; Martin et al., 2003) and thus Bs4 would be predicted to encode a nuclear protein. However, tomato Bs4 has no apparent sequence signatures that would indicate nuclear localization. This prediction is in agreement with the fact that Bs4 mediates recognition of NLS-deprived AvrBs4-deletion derivatives (this study and Ballvora et al., 2001a) and supports our previously proposed working model that Bs4-mediated perception of AvrBs4 occurs in the cytoplasm before AvrBs4 has reached its final destination, the nucleus (Ballvora et al., 2001a). However, it needs to be considered that computer-based predictions are error-prone as exemplified by the RPM1 R protein that was found to be membrane bound although it was predicted to be cytoplasmic (Boyes et al., 1998). Hence, localization studies are needed to clarify the subcellular localization of Bs4.
Agrobacterium-mediated avrBs3 expression triggers a Bs4-dependent HR
Xanthomonas infection tests demonstrated that Bs4 mediates recognition of AvrBs4 and its C-terminal deletion derivatives, but not the 97% identical AvrBs3 protein (Figure 3). However, when the corresponding genes are delivered via Agrobacterium, both AvrBs4 and AvrBs3 but not AvrBs1 trigger a Bs4-dependent HR (Figure 4). This finding indicates that Agrobacterium-mediated delivery causes a partial loss of recognition specificity. The reasons for this are not fully understood.
One possibility is that increased AvrBs3 expression levels generate the observed loss of Bs4 recognition specificity as transgene expression in the binary vector is under control of the CaMV 35S promoter. Thus, AvrBs3 levels in the Agrobacterium-mediated expression system might exceed the quantity that is delivered by the Xanthomonas-type III secretion system, possibly causing loss of Bs4 recognition specificity. However, in the converse experiment, the pepper Bs3 R gene that confers perception of AvrBs3 in an NLS-dependent fashion (Van den Ackerveken et al., 1996) does not mediate recognition of agroinfiltrated AvrBs4 (S. Schornack and T. Lahaye, unpublished results). The seemingly different degrees of specificity in Bs3- and Bs4-mediated resistance in agroinfiltration assays might be related to the fact that Bs3 confers only recognition of NLS-bearing AvrBs3 derivatives (Ballvora et al., 2001a; Van den Ackerveken et al., 1996). Previous studies have shown that AvrBs3 recruits the host's nuclear import machinery in order to reach the plant nucleus (Szurek et al., 2001). Assuming that the host's nucleocytoplasmic shuttle system has only a limited transfer capacity, it might be possible that in agroinfiltration assays, only a small fraction of NLS-bearing AvrBs3-like proteins are actually transferred into the nuclear compartment where it can interact with Bs3. In contrast, the putatively cytoplasmic tomato Bs4 might be confronted with high levels of AvrBs3-like proteins. Hence, the apparently different degrees of recognition specificity of Bs3 and Bs4 may reflect their subcellular localization rather than intrinsic properties of the proteins.
A second possibility is that ‘relaxed’ recognition specificity is an intrinsic property of Bs4. This hypothesis is supported by the identification of two novel AvrBs3-like proteins that are distinct from AvrBs4, which are both recognized in a Bs4-dependent manner when delivered by Xanthomonas (S. Kay and U. Bonas, unpublished results). Considering that several Xanthomonas strains contain multiple AvrBs3-like proteins (Leach et al., 2001; Van't Slot and Knogge, 2002) and given that genes encoding AvrBs3-like proteins can rapidly change their structure by intra- and inter-repeat recombination (Yang and Gabriel, 1995a), it seems economically sensible to employ one R protein with a ‘relaxed’ recognition specificity rather than expressing multiple, highly specific R proteins.
The Bs4 exon-intron structure is characteristic of an R gene transcript that encodes an TIR-NB-LRR protein (reviewed by Jordan et al., 2002) and has three conserved introns that are located between the TIR-, NB-, and LRR-encoding regions. RT-PCR uncovered in several Bs4 transcripts retention of introns 2 and 3, which has also been described for the Arabidopsis RPS4 (Gassmann et al., 1999) and the flax L6 gene (Ayliffe et al., 1999). Complementation analysis showed that intronless RPS4 derivatives mediate no or only a reduced defense response in comparison to genomic RPS4 constructs (Zhang and Gassmann, 2003). In contrast to the findings observed for RPS4, we did not observe obvious functional differences between genomic and intron-deprived constructs. Nonetheless, one cannot exclude the possibility that these Bs4 splice variants confer subtle biological effects.
Sequence analysis of the bs4 allele provides no unequivocal evidence for a complete-loss-of-function (null) allele
Analysis of 74 DNA polymorphisms between the L. pennellii LA2963 bs4 and L. esculentum cv. MM Bs4 genomic sequences (Figure 6) failed to uncover mutations that would unequivocally classify bs4 as a null allele. RT-PCR revealed that bs4 and Bs4 transcripts differ with respect to the 5′ splice site of intron 2 (Figure 4b). Usage of an alternative GT splice donor site in bs4 transcripts generates a frame-shift and is predicted to encode truncated bs4 proteins that lack the LRR region. However, we analyzed only 32 cloned RT-PCR fragments and it might well be possible that bs4-derived transcripts that encode a functional, full-length TIR-NB-LRR protein remained undetected. Phenotypic inspection of L. pennellii LA2963 and segregants that harbor the corresponding bs4 allele revealed a delayed HR in bs4 genotypes that appears about 10 days after Xanthomonas infection (data not shown; Bs4 mediates HR 48 h after infection). Notably, only avrBs4-expressing xanthomonads trigger this delayed HR, indicating that the L. pennellii LA2963 bs4 allele still has residual function with respect to AvrBs4 perception. In summary, the lack of mutations that would clearly classify L. pennellii bs4 as a null allele and the delayed AvrBs4-dependent HR in bs4 genotypes supports the hypothesis that bs4 is not a non-functional allele but rather an allele with reduced activity.
VIGS of NbEDS1 and NbSGT1 suppresses Bs4-mediated HR execution
Genetic dissection has shown that EDS1 and NDR1 define defense-signaling pathways that are differentially employed by TIR-NB-LRR and CC-NB-LRR proteins, respectively (reviewed by Feys and Parker, 2000; Glazebrook, 2001). In agreement with this postulate, we found that functionality of the TIR-NB-LRR protein Bs4 was compromised by EDS1- but not by NDR1-silencing.
Unlike EDS1 and NDR1, the pathway elements Rar1 and SGT1 were shown to be engaged by both TIR--NB-LRR and CC-NB-LRR subtypes (reviewed by Dodds and Schwechheimer, 2002). Noteworthy, activity of the tobacco N protein, which shares 54% sequence identity with Bs4, is impaired by VIGS of Rar1 and SGT1 (Liu et al., 2002a,b). By contrast, our studies show that Bs4-mediated HR execution is only impaired by SGT1- but not by Rar1-silencing. This is somewhat surprising, given that Bs4 and N share extensive sequence homology. However, previous studies of the barley MLA1, MLA6, and MLA12 mildew R proteins, which share approximately 90% sequence identity (Halterman et al., 2001; Shen et al., 2003; Zhou et al., 2001), have shown that, despite their pronounced sequence homology, only MLA6 and MLA12 require RAR1 and SGT1 for execution of a defense response (Halterman et al., 2001; Shen et al., 2003; Zhou et al., 2001). Domain swaps between the Rar1/SGT1-independent Mla1 and the Rar1/SGT1-dependent Mla6 alleles generated Rar1/SGT1-independent chimeras with Mla6 recognition specificity (Shen et al., 2003), thereby demonstrating that recognitional specificity and Rar1/SGT1 requirements are defined by distinct protein regions. Analysis of the barley Mla alleles demonstrates also that highly similar R proteins that recruit distinct signaling elements provide a functional tool for allocation of protein regions that specify dependency on certain downstream components. In analogy to the MLA variants, Bs4 and N represent highly similar R proteins, which however differ only with respect to their Rar1 but not with respect to their SGT1 dependency. Hence, domain swaps between Bs4 and N might allow allocation of domains that define Rar1-dependency.
However, there is a caveat to our VIGS-based Rar1 and NDR1 knockdown assays as gene-silencing does not generally facilitate complete elimination of the targeted transcripts. Thus, we cannot exclude that Bs4 acts in a NDR1- and Rar1-dependent manner and that residual amounts of the corresponding transcripts are sufficient for Bs4-mediated HR execution.
No direct interaction between Bs4 and AvrBs4?
Resistance proteins are thought to detect Avr proteins either directly by physical interaction or indirectly because of virulence-associated Avr actions (Bonas and Lahaye, 2002; Dangl and Jones, 2001; Martin et al., 2003; Van der Biezen and Jones, 1998b). Thus far, no virulence function has been assigned to AvrBs4. However, as AvrBs4 has been maintained in nature, despite the fact that it exerts negative selective pressure in the interaction with most tomato hosts, it is likely to have a virulence function. Conceivably, AvrBs4 employs molecular virulence strategies similar to those of the 97% identical AvrBs3 protein, and hence it seems likely that AvrBs4-deletion derivatives that lack C-terminal AAD and NLS domains have no or only reduced virulence activity. We showed by analysis of deletion constructs that an AvrBs4-derivative (AvrBs4 Δ230), which contains only 3.5 out of 17.5 repeat units and lacks its AAD and NLS regions was still able to trigger a Bs4-dependent HR. This might be interpreted as an argument against the guard model as recognition in the conceptual framework of the guard model requires biologically active, disease-promoting Avr effectors. Taken together, genetic analysis of the avrBs4–Bs4 gene-for-gene interaction favors a receptor-ligand rather than a guard model. However, AvrBs4 and Bs4 seem not to interact in yeast, despite the fact that Western analyses and repression assays showed that AvrBs4 and Bs4 derivatives are in principle suitable for Y2H studies. How can we explain these seemingly contradictory findings? Recent biochemical studies suggest that R proteins are part of multiprotein complexes (Bogdanove, 2002; Ellis et al., 2002; Leister and Katagiri, 2000), and it may well be possible that R proteins fulfill their Avr-receptor function only in the context of this multicomponent recognition complex. Further clarification of the molecular principles that govern the AvrBs4-Bs4 interplay will therefore foreseeable require techniques that allow analysis of proteins in their natural environment.
avr derivatives and Xcv inoculations
Plants were grown and inoculated as described previously by Bonas et al., (1989, 1993). Analysis of avrBs4-deletion derivatives in pLAFR6 (description of constructs by Bonas et al., 1993) was conducted with respective Xcv transconjugants. All other avr genes were assayed in Xcv transconjugants that carry pDSK602 constructs of avrBs1 (pDS100, Escolar et al., 2001), avrBs3 (pDS300F, Van den Ackerveken et al., 1996), avrBs4 (pDS200F, this study) and avrBs4 Δ227 (pDS227, this study). avrBs4 and avrBs4 Δ227 were cloned into pDSK602 (Murillo et al., 1994) as follows. An EcoRI/HindIII fragment from pUS200F, bearing avrBs4 (Ballvora et al., 2001a), was cloned into pDSK602, resulting in plasmid pDS200F. AvrBs4 Δ227 (pAT227) is a C-terminal avrBs4-deletion derivative that was generated by DNAse I digest of an EcoRI-linearized pAT200 (Bonas et al., 1993) and released from pAT227 by BamHI digest and replaced the corresponding avrBs3 fragment in pDSK602-36 giving rise to pDS227. pDSK602-36 was generated by cloning a BamHI–NdeI–BamHI linker into the BamHI site of pUXV1006 (Bonas et al., 1989) and ligating the NdeI–HindIII fragment with the avrBs3-ORF into EcoRI/HindIII sites of pDSK602.
Inoculation tests of avrBs4-deletion derivatives were performed on 10 bs4/bs4 and 10 Bs4/− F2 segregants of a cross between the L. esculentum cv. MM (Bs4) and L. pennellii LA 2963 (bs4) (Ballvora et al., 2001b). The genotype of F2 plants in the Bs4 locus was determined with the Bs4-flanking markers TG432 and P11M6 (Ballvora et al., 2001a,b).
PCR with degenerate oligonucleotide primers targeting the TIR motif
Primer pairs based on the conserved TIR motif of NB-LRR-encoding R genes were used for amplification of Bs4 candidate genes from VF36 tomato DNA. Their sequences were: RD5: GT(T/G)TT(T/C)TT(A/G)AGTTT(C/T)AG(A/G)GG; and RD10: GGATCCAC(A/C)(T/A)(C/T)ATA(G/A)AA(A/T)AT(A/C)GG. Amplification was carried out with the following program: 5 cycles of 94°C for 45 sec, 42°C for 45 sec, and 72°C for 90 sec; 30 cycles of 94°C for 45 sec, 50°C for 45 sec, and 72°C for 90 sec. PCR products were mapped on a population of 47 F2 plants from L. esculentum × L. pennellii (Tanksley et al., 1992) using mapmaker (Lander et al., 1987).
Genetic mapping of a Bs4 candidate gene
We developed two PCR-based RFLP markers within the Bs4 candidate gene; RGA2 (primers: Bs4-A02: CTACCATCATCTCTTCAGTACCAACTC; and Bs4-B02: GAAATTGGAGGAACCGAGCTCCAG; MspI-polymorphism); and RGA3 (primers: Bs4-A03: GGGTTGGAGTCCGAAGAGCAGG; and Bs4-B03: GACTAACCAACGCAAGTTATTGGACAGG; RsaI-polymorphism). Inheritance of AvrBs4 recognition in tomato was studied using a cross between the L. esculentum cv. MM (resistant parent) and L. pennellii LA 2963 (susceptible parent; Ballvora et al., 2001b).
Construction of binary vectors carrying Bs4
For construction of a cosmid library, YAC Y45 was partially digested with Sau3A, and cloned into the BamHI site of the binary vector pCLD04541 (Jones et al., 1992). The T26 RGC-bearing cosmid T2-2 was identified by colony hybridization and confirmed by PCR. A T2-2-derived genomic fragment containing Bs4 and 1.2 kbp upstream of the ATG was transferred into the SmaI site of pCP60 (kindly provided by C. Coronado and P. Ratet; Bs4 transgene is under control of CaMV 35S promoter) yielding pVTSB1. We removed the 35S* promoter from the binary vector pVB60 (Van den Ackerveken et al., 1996) yielding pVB61. Next, we cloned Bs4 and 3.5 kbp upstream sequence into the SalI–NotI sites of pVB61 yielding pVTSB3.
Construction of binary vectors containing avrBs4 derivatives
The binary vector pVB60 (Van den Ackerveken et al., 1996) was used for Agrobacterium-mediated transient expression of avr derivatives. We assayed avrBs4 (pVS200F, Ballvora et al., 2001a), avrBs4 Δ227 (pVS227, this study), and avrBs4 Δ152 (pVS256F, this study). pVS227 and pVS256F originate from pAT227 (avrBs4 Δ227 in pUC118; Bonas et al., 1993) and pUS256 (this study, see below for description), respectively. EcoRI/HindIII inserts from pAT227 and pUS256 were cloned into pBluescript KS yielding pBS227F and pBS256F, respectively. Subsequently EcoRI/XhoI fragments from pBS227F and pBS256F were transferred into pVB60 creating pVS227F and pVS256F, respectively.
The binary vector pVTSB1, containing Bs4 under transcriptional control of the CaMV 35S promoter, was transferred into A. tumefaciens strain LBA4404 (Hoekema et al., 1983) and transformed into the previously described bs4/bs4 L. esculentum genotype MMbs4–BC4 (Ballvora et al., 2001a). Transformation and plant regeneration was performed as described by Ling et al. (1998). Transgenic plants were confirmed by PCR with primers for the neomycin phosphotransferase (nptII) gene and primers that distinguished between the transgenic Bs4 and endogenous bs4 sequences.
Total RNA was isolated using the RNeasy Plant Mini kit (Qiagen, Hilden, Germany) according to the manufacturer's instructions and enriched for poly(A) RNA using the Oligotex mRNA Mini kit (Qiagen). RACE PCR was carried out using the SMART RACE cDNA Amplification Kit (Clontech, Heidelberg, Germany). Amplicons were cloned into pCR2.1 (Invitrogen, Karlsruhe, Germany) and sequenced using vector-specific primers.
Construction of an intron-deprived Bs4-derivative
An intron-deprived Bs4 derivative was generated by amplification of exonic regions. Restriction sites present at the end of Bs4 exons were used to fuse PCR-amplified exons. Briefly, an XbaI –NotI fragment from pVTSB1 containing the Bs4 genomic sequence was subcloned into pBluescript SK II (Stratagene, La Jolla, CA, USA) to yield pBlue:Bs4. A XbaI–KpnI fragment of pBlue:Bs4 that contained all introns to be removed was subcloned into pUC119 (Vieira and Messing, 1987) to yield pUCSB6. A Bs4 clone with pBluescript backbone lacking EcoRV, EcoRI, and PstI was generated by EcoRV–SmaI digest, religation, and XbaI–NotI introduction of Bs4 from pBlue:Bs4 into this vector to yield pBSB3. Primers I1fwd-EcoRV, I2rev-BstXI, and I3rev-BglII and the corresponding reverse primers derived from the genomic Bs4 sequence were used to amplify exonic regions from pBlue:Bs4 and introduce exon–exon junctions instead of introns. These amplicons replaced corresponding restriction fragments in pUCSB6 by EcoRV–SacII, SacII–BstXI, and BstXI–BglII digest, respectively, to create pUCSB6 Δ123. A XbaI–PstI fragment from pUCSB6 Δ123 was then cloned into pBSB3, thus replacing the corresponding genomic Bs4 fragment and creating pBSB5. pBSB5 represents an intron-deprived Bs4 fragment containing flanking genomic sequences. A XbaI–NotI fragment from pBSB5 was subsequently cloned into the binary vector pCP60 for in planta assays.
Bs4 pathway dissection by VIGS
Expressed sequence tag (EST) database searches using Arabidopsis NDR1 (AI776252.1) as input returned a closely related tomato EST (EST257352). Based on this EST, we designed primers in regions of highest homology to Arabidopsis NDR1 (JP NDR 2: TATAATCTCGTCGTAACGAACACCTTTGTC and JP NDR 1: ACTGCAGGCTTAACAGCTCTCTTTATCTGG) and performed RT-PCR on N. benthamiana cDNA. We observed a 198-bp PCR fragment that was cloned into pGEM-T Easy vector (Promega, Mannheim, Germany). The 5′ and 3′ ends of NbNDR1 were determined by RACE, using the MARATHON cDNA amplification kit (Clontech). A 480-bp NbNDR1 fragment used to produce TRV:NbNDR1 was generated by PCR on a plasmid containing NbNDR1 cDNA, using primers JP NDR 4: (TACCTGGCTTTTACCAAGGTCATGAC) and JP NDR 5: (CCAACTACAGTCAGTTTGCACCGA). We designed primers for amplification of NbRar1 based on the N. tabacum Rar1 mRNA sequence (AF480487). A 585-bp NbRar1 fragment used to produce TRV:NbRar1 was generated by RT-PCR on N. benthamiana cDNA, using primers Rar1-f2 (TGCACTTATCACGAATCCGG) and Rar1-r2 (TGGGCTGGCGTTGTGCCATCCTG). Construction of TRV:EDS1 and TRV:SGT1 has been described before (Peart et al., 2002a,b). Infection of plants with TRV derivatives was performed by agroinfiltration as described previously by Peart et al. (2002a).
Sequence analysis of the L. esculentum cv. MM Bs4 and L. pennellii LA2963 bs4 genes
The Bs4 sequence from L. esculentum cv. VFNT Cherry was obtained from cosmid T2-2. Corresponding sequences from L. esculentum cv. MM (Bs4), and L. pennellii LA2963 (bs4) were determined by sequencing of respective PCR products. Sequences were obtained using the BigDye Terminator Kit (PE Biosystems, Foster City, CA, USA) and analyzed using Sequencher 4.0 (Gene Codes Corp., Ann Arbor, MI, USA).
Y2H interaction studies
For Y2H studies, the yeast interaction trap was used (Gyuris et al., 1993). Bait and prey plasmids were co-transformed into yeast strain EGY48 (Estojak et al., 1995) containing the lacZ-reporter plasmid pSH18-34. Transformants grown on selective glucose medium were transferred on galactose medium containing 5-bromo-4-chloro-3-indolyl-β-d-galactopyranoside (X-Gal) to determine β-galactosidase (LacZ) activity, and on galactose plates lacking leucine to measure Leu2 expression. Quantitative assays were performed on liquid cultures. Three individual yeast transformant colonies for each construct were inoculated into liquid glucose medium and incubated overnight (o.n.) at 30°C. The culture was diluted in galactose medium inducing the expression of the prey and grown until an OD600 of 0.5–1.0. β-galactosidase activity was assayed using o-Nitrophenyl β-d-galactopyranoside as described previously by Ausubel et al. (1996). For Y2H analysis, avrBs4 and Bs4 were cloned into bait (pYB) and prey (pYP) vectors. Full-length AvrBs4 autoactivated transcription in yeast and to overcome this problem, we cloned the N-terminal deletion derivative AvrBs4 Δ152 (lacking aa residues 1–152) into bait (pYB256) and prey (pYP256) vectors, respectively. The N-terminal deletion was generated by replacing the StuI/HindIII fragment of pUS356 (contains N-terminal deletion of avrBs3, Szurek et al., 2001) with the corresponding fragment of pAT200 (contains avrBs4, Bonas et al., 1993), yielding pUS256 (avrBs4 Δ152 in pUC118). The EcoRI/XhoI fragment from pUS256 was transferred into pEG202 and pJG4-5 (OriGene Technologies Inc., Rockville, MD, USA) yielding pYB256 and pYP256. Bs4 domains were amplified from cDNA using primers annealing adjacent to TIR (Bs4-TIR-fwd, Bs4-TIR-rev), NB (Bs4-NB-fwd, Bs4-NB-rev) and LRR (Bs4-LRR-fwd, Bs4-LRR-rev) or combinations, all containing MunI/SalI sites that allowed cloning into EcoRI/XhoI of pEG202 and pJG4-5. The previously described Caimpα1 (Szurek et al., 2001) was used as a positive control for Y2H studies. In all cases, expression and stability of fusion proteins was confirmed by immunoblotting using a monoclonal anti-LexA antibody (Clontech Laboratories) and monoclonal anti-HA antibody (3F10, Roche, Mannheim, Germany), respectively.
We thank Daniela Büttner, Martin Parniske, John Page and Max Dow for helpful suggestions on the manuscript. We are also grateful to the two anonymous reviewers for their constructive comments on an earlier version of this manuscript. This work was in part funded by grants from the CNRS (ATIPE) and INRA (AIP Gènes de Résistance) to U.B. and grants of the Deutsche Forschungsgemeinschaft to U.B. and T.L. (SFB 363 and LA 1338/1–1). A.B. was supported by a long-term EMBO fellowship.
The GenBank accession number for the N. benthamiana NDR1 and Rar1 sequence is AY438029 and AY438026, respectively. The GenBank accession number for L. esculentum cv. MM Bs4 and L. pennelliiLA2963bs4 is AY438027 and AY438028, respectively.