The ubiquitin pathway is required for innate immunity in Arabidopsis

Authors

  • Sandra Goritschnig,

    1. Michael Smith Laboratories, Room 301, 2185 East Mall, University of British Columbia, Vancouver, BC V6T 1Z4, Canada
    2. Department of Botany, Room 3529, 6270 University Blvd., University of British Columbia, Vancouver, BC V6T 1 Z4, Canada
    Search for more papers by this author
  • Yuelin Zhang,

    1. Michael Smith Laboratories, Room 301, 2185 East Mall, University of British Columbia, Vancouver, BC V6T 1Z4, Canada
    2. National Institute of Biological Sciences, Zhongguancun Life Science Park, 7 Science Park Road, Beijing 102206, China
    Search for more papers by this author
  • Xin Li

    Corresponding author
    1. Michael Smith Laboratories, Room 301, 2185 East Mall, University of British Columbia, Vancouver, BC V6T 1Z4, Canada
    2. Department of Botany, Room 3529, 6270 University Blvd., University of British Columbia, Vancouver, BC V6T 1 Z4, Canada
      (fax +1 604 822 2144; e-mail xinli@interchange.ubc.ca).
    Search for more papers by this author

(fax +1 604 822 2144; e-mail xinli@interchange.ubc.ca).

Summary

Plant defences require a multitude of tightly regulated resistance responses. In Arabidopsis, the unique gain-of-function mutant suppressor of npr1-1 constitutive 1 (snc1) carries a point mutation in a Resistance (R)-gene, resulting in constitutive activation of defence responses without interaction with pathogens. This has allowed us to identify various downstream signalling components essential in multiple defence pathways. One mutant that suppresses snc1-mediated constitutive resistance is modifier of snc1 5 (mos5), which carries a 15-bp deletion in UBA1, one of two ubiquitin-activating enzyme genes in Arabidopsis. A mutation in UBA2 does not suppress snc1, suggesting that these two genes are not equally required in Arabidopsis disease resistance. On the other hand, a mos5 uba2 double mutant is lethal, implying partial redundancy of the two homologues. Apart from affecting snc1-mediated resistance, mos5 also exhibits enhanced disease susceptibility to a virulent pathogen and is impaired in response to infection with avirulent bacteria carrying the protease elicitor AvrRpt2. The mos5 mutation in the C-terminus of UBA1 might affect binding affinity of the downstream ubiquitin-conjugating enzymes, thus perturbing ubiquitination of target proteins. Furthermore, SGT1b and RAR1, which are necessary for resistance conferred by the SNC1-related R-genes RPP4 and RPP5, are dispensable in snc1-mediated resistance. Our data reveal the definite requirement for the ubiquitination pathway in the activation and downstream signalling of several R-proteins.

Introduction

Plants are constantly being challenged by a variety of biotic and abiotic stresses, and have evolved a range of sophisticated mechanisms to cope with them. Attacking pathogens often encounter an unfavourable environment or pre-formed defences in non-host plants. On host plants, pathogens either cause disease or are recognized and rapidly contained at the site of infection by a mechanism usually involving the programmed death of the cells surrounding the infection, referred to as the hypersensitive response (HR; Hammond-Kosack and Jones, 1996; Nimchuk et al., 2003). This specific recognition of pathogens is mediated by Resistance (R) proteins, which directly or indirectly interact with pathogen-derived avirulence (Avr) gene products to induce a resistance response. Resistance proteins are proposed to associate into multimolecular complexes and most likely sense the presence of the pathogen indirectly through the action of its elicitors in the host cell (Belkhadir et al., 2004).

Most R-genes encode proteins with highly conserved structural domains, which are also frequently found in mammalian immune system modules (Ausubel, 2005; Dangl and Jones, 2001). For example, the majority of R-proteins share a central nucleotide binding (NB) domain, and carboxy-terminal leucine-rich repeats (LRR), similar to mammalian NOD proteins involved in innate immunity (Inohara et al., 2005). The amino-terminal portions discern two classes of R-proteins: those containing the Toll/Interleukin1-receptor-like (TIR) domain with high homology to the cytoplasmic signalling domain of mammalian and Drosophila Toll-like receptors, involved in recognizing microbe-specific antigens in innate immunity (Athman and Philpott, 2004), or proteins with a coiled coil (CC) motif potentially involved in protein–protein interactions (Belkhadir et al., 2004; Burkhard et al., 2001; Martin et al., 2003).

Upon recognition of the cognate avirulence elicitor, TIR-NB-LRR and CC-NB-LRR R-proteins activate a range of defence responses. Their requirement for downstream regulators is sometimes, but not always, overlapping. Generally, signalling downstream of TIR-NB-LRR R-protein activation is dependent on the lipase-like protein EDS1, whereas CC-NB-LRR R-protein signalling depends on NDR1 (Aarts et al., 1998). Other signalling components, such as RAR1, SGT1b and Hsp90, are shared by TIR and CC class R-protein pathways (Hubert et al., 2003; Muskett and Parker, 2003; Takahashi et al., 2003).

RAR1 and SGT1b were independently identified in screens for suppressors of various R-protein-mediated responses (Austin et al., 2002; Shirasu et al., 1999; Tor et al., 2002; Tornero et al., 2002). In Arabidopsis, both are required for resistance conferred by RPP5 (Resistance to Peronospora parasitica 5; Austin et al., 2002). Recent studies have revealed the direct interaction between several R-proteins and SGT1b (Bieri et al., 2004; Leister et al., 2005), as well as Hsp90 (Liu et al., 2004). No direct interaction has been reported between RAR1 and R-proteins, but RAR1 interacts with both Hsp90 and SGT1b and might thus be indirectly involved in the assembly and stability of R-protein complexes (Azevedo et al., 2006; Bieri et al., 2004). RAR1 and SGT1b have also been shown to act antagonistically as positive and negative regulators, respectively, on accumulation of R-protein prior to infection (Holt et al., 2005).

A gain-of-function mutation in a close homologue of RPP5, suppressor of npr1-1 constitutive 1 (snc1), results in the constitutive activation of basal defence responses manifested as resistance to the virulent pathogens Pseudomonas syringae pv. maculicola (P.s.m.) ES4326 and Peronospora parasitica (P.p.) Noco2, elevated levels of the endogenous signalling molecule salicylic acid (SA) and dwarf morphology (Li et al., 2001; Zhang et al., 2003). As opposed to other constitutively resistant mutants, snc1 does not exhibit spontaneous lesions. However, like other TIR-NB-LRR R-genes, snc1-mediated resistance fully depends on EDS1 and PAD4, but is only partially dependent on accumulation of SA (Li et al., 2001; Zhang et al., 2003).

To further understand the signalling downstream of TIR-NB-LRR R-proteins, a suppressor screen was performed in the snc1 and snc1 npr1-1 backgrounds. This screen identified a number of modifier of snc1 (mos) mutants, including several alleles of pad4. The identities of MOS3 (a putative nucleoporin 96; Zhang and Li, 2005) and MOS6 (an importin alpha 3 homologue; Palma et al., 2005) reveal an essential role for nucleo-cytoplasmic trafficking in resistance signalling. MOS2, a nuclear protein with putative RNA-binding motifs, highlights the importance of RNA processing in plant innate immunity (Zhang et al., 2005).

Here, we report on the identification and cloning of mos5, a modifier of constitutive disease resistance in snc1, which suppresses snc1-associated phenotypes and shows differential responses to avirulent bacteria. Cloning of MOS5 revealed that it encodes an essential component of the ubiquitin pathway and implicates a requirement for ubiquitination in R-protein signalling. We also show that the resistance regulators RAR1 and SGT1b are not required for snc1-mediated constitutive resistance.

Results

Isolation and genetic analysis of mos5 snc1 npr1-1

The screen for suppressors of snc1-mediated resistance has been previously described (Zhang and Li, 2005). mos5 was identified in the snc1 npr1-1 double-mutant background based on its ability to revert snc1 morphology to almost wild type and to abolish constitutive expression of the pBGL2–GUS reporter transgene (Figure 1a,b). Lack of snc1-induced constitutive expression of pathogenesis related (PR) genes in mos5 snc1 npr1-1 was further confirmed by semiquantitative RT-PCR (Figure 1c).

Figure 1.

 Phenotypic characterization of mos5 snc1 npr1-1.
(a) Morphology of soil-grown plants. The picture was taken 5 weeks after planting.
(b) Expression of the pBGL2–GUS reporter gene. Twenty-day-old seedlings grown on MS plates were stained for GUS activity.
(c) Semi-quantitative RT-PCR of pathogenesis-related genes. Ribonucleic acid was extracted from twenty-day-old seedlings grown on MS plates and reverse transcribed to obtain total cDNA. The cDNA samples were normalized using the actin probe (Zhang et al., 2003). PR1, PR2 and Actin were PCR-amplified in 30 cycles using equal amounts of cDNA.

When mos5 snc1 npr1-1 was backcrossed with snc1 npr1-1, F1 progeny displayed the characteristic snc1 morphology indicating that the mutation is recessive, and pBGL2-GUS expression in the F2 progeny segregated 75:21 (staining:non-staining) as determined by GUS staining, demonstrating that the phenotype of mos5 is caused by a single recessive mutation (expected 3:1, χ2 = 0.5, P = 0.48).

Characterization of defence related phenotypes of mos5 snc1 npr1-1

Salicylic acid is an important signalling molecule in plant defence responses and is associated with systemic resistance (Durrant and Dong, 2004; Ryals et al., 1996). snc1 npr1-1 mutant plants have high endogenous levels of SA (Li et al., 2001). The mos5 snc1 npr1-1 mutant exhibits an approximate 12-fold reduction in endogenous levels of both free and total SA as compared with snc1 npr1-1 (Figure 2a,b). However, the mutant still displays ten- and fivefold higher levels of SA than either Col-0 or the npr1-1 single mutant, respectively. This indicates that suppression of snc1 by mos5 is not complete regarding SA levels.

Figure 2.

 Suppression of constitutive resistance in mos5 snc1 npr1-1.
(a,b) Levels of endogenous SA are reduced in mos5 snc1 npr1-1. Free (a) and total (b) SA was extracted from 5-week-old soil-grown plants and analysed with HPLC.
(c,d) mos5 suppresses resistance against virulent oomycete and bacterial pathogens. (c) Two-week-old seedlings were infected with P. parasitica Noco2 conidiospores (104 ml−1) and disease symptoms were assessed and rated 7 days post-inoculation (dpi) as follows: 0, no conidiophores on entire plant; 1, at least one leaf with 1–5 conidiophores; 2, some infected leaves with 6–20 conidiophores, but most with 1–5; 3, most infected leaves with 6–20 conidiophores; 4, all infected leaves with >5 conidiophores, or most infected leaves with >20; 5, all infected leaves with >20 conidiophores.
(d) Four-week-old plants were infected with P. syringae pv. maculicola ES4326 (OD600 = 0.0001) and bacterial growth was measured by quantifying colony forming units (cfu) at day 0 and day 3. Experiments were repeated at least twice with similar results.

The snc1 npr1-1 double mutant exhibits enhanced resistance to the virulent pathogens P.p. Noco2 and P.s.m. ES4326 compared with Col-0 wild type (Li et al., 2001). To investigate the role of MOS5 in disease resistance, plants were inoculated with either pathogen. The mos5 mutation suppresses snc1-mediated constitutive resistance to P.p. Noco2, resulting in wild-type-like susceptibility (Figure 2c). Furthermore, mos5 snc1 npr1-1 supports wild-type-like levels of P.s.m. ES4326 growth (Figure 2d). However, the mutation is not sufficient to restore npr1-like susceptibility against the bacterial pathogen, again indicating that suppression of snc1 by mos5 is incomplete.

Taken together, these results demonstrate that the mutation in MOS5 strongly impairs defence signalling in snc1 npr1-1, affecting responses to bacterial and oomycete pathogens as well as the accumulation of SA.

Map-based cloning of mos5

In order to map the recessive modifier of snc1 in the Col-0 ecotype, mos5 snc1 npr1-1 was crossed with Ler-snc1, in which snc1 had been introgressed into the Landsberg erecta (Ler) ecotype by repeated backcrossing as described previously (Zhang and Li, 2005). Using 68 plants homozygous for mos5 from the F2 progeny of the mapping cross, the approximate position of mos5 was determined to be between markers T8O18 (12.28 Mb) and T16B12 (13.25 Mb) on the lower arm of chromosome 2 (Figure 3a). The progeny of F2 plants heterozygous for mos5 in this region and homozygous for the pBGL2–GUS transgene was used for fine mapping. Out of 747 F3 plants, 46 recombinants between markers T8O18 and T16B12 were identified. The phenotypes of these recombinants were confirmed by following segregation of the F4 progeny on soil as well as with GUS staining. The mos5 mutation was ultimately mapped to a 95 kb region between T27E13-2 and T9D9, with one and four remaining recombinants, respectively (Figure 3a). Open reading frames (ORFs) covering this region were sequenced in mos5 snc1 npr1-1 and a 15 bp deletion was found in the coding region of At2g30110. Sequence analysis indicated that the deletion is located in the last exon of At2g30110, leading to an amino acid substitution (arginine to serine) and the deletion of the following five amino acids (Figure 3b,c).

Figure 3.

 Map-based cloning of mos5.
(a) Mapping of mos5 to the bottom of chromosome 2. Markers used for crude and fine mapping are indicated with the number of recombinants below. Open reading frames between the final markers T27E13-2 and T9D9 were sequenced and a deletion was identified in At2g30110 (indicated by *).
(b) Intron–exon structure of At2g30110. The mutation in mos5 (*) lies close to the stop codon in exon 7.
(c) Deoxyribonucleic acid and amino acid sequence alignment of mos5 and UBA1. The 15 bp deletion in At2g30110 results in an amino acid substitution and deletion of five amino acids.
(d) Complementation of mos5 by AtUBA1 genomic DNA. mos5 snc1 npr1-1 plants carrying the 7 kb transgene of the genomic sequence of AtUBA1 show typical snc1 morphology. Representative plants are shown, photographed at 5 weeks.

At2g30110 encodes one of two ubiquitin-activating (E1) enzymes in Arabidopsis (AtUBA1). AtUBA1 is very similar to AtUBA2 (81% amino acid sequence identity) and E1 enzymes in other organisms, stressing the evolutionary conservation of this essential enzyme (Hatfield et al., 1997). Ubiquitin-activating enzymes from all kingdoms share regions of high homology that are involved in binding of the ubiquitin molecule, including the catalytic cysteine residue, and binding regions for the nucleotides that provide the energy for ubiquitin activation. The mutation in mos5 is located in the C-terminal domain of the E1, outside this highly conserved nucleotide and ubiquitin-binding regions. Interestingly, the C-termini of ubiquitin E1s from different kingdoms are also very similar, suggesting a conserved function of this region in the activity of the enzyme (Figure 4).

Figure 4.

 Alignment of the C-terminal domains of ubiquitin-activating enzymes from different species.
Protein sequences were aligned using clustalx and shaded using boxshade (http://www.ch.embnet.org/software/BOX_form.html). Conserved residues are shaded in black and similar residues are shaded in grey. The location of the deletion in mos5 is indicated with a bar. AtUBA1, A. thaliana UBA1 (accession AAC16961); AtUBA2, A. thaliana UBA2 (accession BAB08968); Ta_E1, Triticum aestivum (accession A38373); Os_E1, Oryza sativa (accession NP_910456); Nt_E1, Nicotiana tabacum (accession BAD00983); Hs_E1, Homo sapiens (accession CAA40296); Mm_E1, Mus musculus (accession AAF00149); Xl_E1, Xenopus laevis (accession BAB19357); Sc_Uba1p, Saccharomyces cerevisiae (accession NP_012712).

To confirm that the deletion in mos5 is responsible for the suppression of the snc1 phenotype, a 7 kb genomic clone including the complete At2g30110 ORF as well as 2 kb of the 5′ promoter region and the 3′ untranslated region (UTR) were cloned into pGreen229 (Hellens et al., 2000) and transformed into the mos5 snc1 npr1-1 mutant. Thirty-one out of 33 T1 transformants restored snc1 morphology (Figure 3d), indicating that At2g30110 can complement the mutation in mos5 snc1 npr1-1 and that MOS5 is AtUBA1.

The mos5 single mutant displays enhanced disease susceptibility

The mos5 single mutant was generated by crossing mos5 snc1 npr1-1 with Col-0 (carrying the pBGL2–GUS transgene) and identified in the F2 progeny by genotyping using PCR. mos5 single-mutant plants are phenotypically indistinguishable from mos5 snc1 npr1-1 (data not shown). To see whether mos5 affects basal disease resistance, plants were infected with a low dose of the virulent bacterial pathogen P.s.m. ES4326 (OD600 = 0.0001). Compared to the Col-0 wild type, mos5 supports fivefold more bacterial growth (Figure 5a). These data indicate a minor involvement of mos5 in basal resistance against virulent bacteria.

Figure 5.

 Requirement for MOS5 in basal and R-protein-mediated resistance.
(a) Basal resistance is affected by mos5. Four-week-old plants were infected with P.s.m. ES4326 (OD600 = 0.0001) and bacterial growth measured as described.
(b–d) Response of mos5 to avirulent bacteria. Plants were infected with P.s.t. DC3000 AvrRpt2 (b), P.s.m. ES4326 AvrB (c) or P.s.t. DC3000 AvrRps4 (d) at OD600 = 0.001. Bars represent the average of four replicates, error bars represent standard deviation. Experiments were repeated at least twice with similar results.

mos5 exhibits differential susceptibility to avirulent pathogens

To investigate whether mos5 affects resistance mediated by R-proteins other than snc1, the mos5 single mutant was infected with bacteria carrying Avr-determinants. mos5 plants support wild-type-like growth of bacteria carrying AvrB or AvrRps4, determinants recognized by the CC-type R-protein RPM1 and the TIR-type R-protein RPS4, respectively (Figure 5c,d). Only when infected with bacteria carrying AvrRpt2, the avirulence determinant recognized by the CC-type R-protein RPS2, was a reproducible 10-fold increase in bacterial growth observed (Figure 5b). These data suggest that the activation or downstream signalling of certain R-proteins depends on a functional ubiquitination machinery.

UBA2 is not required for resistance

Arabidopsis contains two E1 paralogues. To investigate whether AtUBA2 (At5g06460) is also involved in disease resistance we obtained a T-DNA insertion line from the ABRC (Salk_108047; Alonso et al., 2003), containing an insertion in the fifth exon of UBA2 (Figure 6a), henceforth referred to as uba2. Reverse transcriptase-PCR analysis showed that uba2 does not accumulate UBA2 mRNA (Figure 6b), indicating that the mutation causes a loss of function. In contrast to mos5, uba2 mutant plants were phenotypically indistinguishable from Col-0 wild-type plants (Figure 6c), suggesting that the loss of UBA2 function has no major effect on development. In bacterial infection assays, uba2 plants did not exhibit increased susceptibility to virulent P.s.m. ES4326 (Figure 6d). We crossed uba2 with snc1 to obtain the double mutant, and uba2 snc1 plants displayed the typical snc1 stunted morphology (not shown). This indicates that a loss of UBA2 function is unable to suppress snc1, in contrast to the mos5 mutation in UBA1. These data suggest that UBA2 activity is not required in resistance responses.

Figure 6.

 UBA2 is not essential for plant innate immunity.
(a) Exon–intron structure of UBA2. The position of the T-DNA insertion in SALK_108047 is indicated by a triangle.
(b) Salk_108047 does not express UBA2 mRNA. Polymerase chain reaction was performed using cDNA-specific primers.
(c) uba2 plants are morphologically indistinguishable from Col-0 wild-type plants. The picture is of 5-week-old soil-grown plants.
(c) Resistance to virulent pathogens is not affected in uba2. Four-week-old plants were infected with P.s.m. ES4326 (OD600 = 0.0001) and bacterial growth was measured as described. Experiments were repeated at least twice with similar results.

A mos5 uba2 double mutant is lethal

Since uba2 does not have an obvious phenotype whereas mos5 is defective in innate immunity and responses to the plant hormone auxin (SG, unpublished data), we investigated whether mos5 could be a peculiar allele of UBA1 by two genetic approaches. We first attempted to find insertion alleles that knock out UBA1 function from ABRC seed stocks. Unfortunately, all available alleles with putative T-DNAs in exons of UBA1 did not actually carry an insertion in the gene (for a list of tested T-DNA insertion lines, see Supplementary Table S1). Furthermore, T-DNA insertions in the promoter and 5′-UTR regions of UBA1 did not cause any discernible phenotype and transcription of the gene was unaffected as determined by RT-PCR (data not shown). We could thus not distinguish whether mos5 is a complete or only a partial loss-of-function allele of UBA1. We then attempted to generate a mos5 uba2 double mutant. Out of 183 randomly chosen F2 plants, none were homozygous for both mutations, indicating that a combination of mos5 and uba2 is lethal. Furthermore, plants homozygous for one mutation and heterozygous for the other (thus only containing one fully functional copy of E1) are statistically underrepresented, whereas all other classes are represented close to expected values (15 and 10 observed plants of MOS5/mos5 uba2/uba2 and mos5/mos5 UBA2/uba2, respectively, versus 23 expected for either combination). These data suggest that the two copies of UBA present in Arabidopsis are partially redundant and that the loss of both genes causes lethality. Thus mos5 is most likely a complete loss-of-function allele of UBA1, unless a null mutation in UBA1 alone is lethal.

Resistance in snc1 is independent of SGT1b and RAR1

RAR1 and SGT1b were previously identified as necessary components in RPP5-mediated resistance responses and are also fully required in RPP4-mediated resistance (Austin et al., 2002; Muskett et al., 2002). SGT1b was shown to interact with components of ubiquitin E3 ligases of the Skp/cullin/F-box (SCF) type in yeast and Arabidopsis (Gray et al., 2003; Kitagawa et al., 1999), indicating an additional function in protein degradation. Since MOS5 encodes an essential component of the plant's protein degradation machinery, and given the fact that snc1 encodes an RPP5 homologue that is located in the RPP4 cluster (Zhang et al., 2003), snc1 resistance signalling might also be dependent on either or both proteins.

To investigate a potential role for SGT1b and RAR1 in snc1 signalling, sgt1b-1 (in the Ler ecotype) and rar1-21 (in Col-0) were crossed with snc1 npr1-1. As expected, the F1 progeny of both crosses looked phenotypically like wild type. In the F2 generation, 85 out of 353 plants of the rar1-21 × snc1 npr1-1 cross showed typical snc1-like morphology, indicating that rar1 does not suppress the snc1 growth phenotype (expected ratio 1:3, χ2 = 0.16, P = 0.69). The snc1 rar1-21 double mutant was then isolated using genotype-specific markers. SGT1b is closely linked to SNC1 on chromosome 4, resulting in a skewed ratio in F2 progeny of the snc1 × sgt1b-1 cross. The snc1 sgt1b-1 double mutant was isolated by genotyping, and the presence of both mutations was subsequently confirmed by sequencing.

Both the snc1 sgt1b-1 and the snc1 rar1-21 double mutant plants display similar morphological phenotypes as snc1, i.e. small stature, dark green colour and curly leaves (Figure 7a). To test whether the rar1 and sgt1b mutations affect snc1-mediated constitutive resistance against P.p. Noco2, a pathogen to which rar1 and sgt1b are hyper-susceptible (Austin et al., 2002), two-week-old seedlings were sprayed with a conidiospore suspension. As shown in Figure 7(b,c), the snc1 rar1-21 double mutant and snc1 had a few infected leaves whereas snc1 sgt1b-1 is completely resistant against P.p. Noco2. We also investigated the effect of either mutation on snc1-mediated enhanced basal resistance towards virulent P.s.m. ES4326. As expected, both the rar1-21 and the sgt1b-1 single mutant showed much higher bacterial growth than snc1 (Figure 7d). In the snc1 sgt1b-1 double mutant, snc1-like resistance was completely restored, whereas the snc1 rar1-21 double mutant could partially restore resistance. These data indicate that constitutive resistance against P.p. Noco2 and P.s.m. ES4326 in snc1 is not mediated by these two regulators.

Figure 7.

 SGT1b and RAR1 are not required for snc1-mediated resistance.
(a) Morphology of single and double mutants. Pictures are taken of 5-week-old soil-grown plants.
(b–d) sgt1b and rar1 do not suppress constitutive resistance towards virulent pathogens in snc1. (b,c) Plants were infected with P.p. Noco2 conidiospores and disease ratings assessed 7 dpi as described in Figure 2. (d) Plants were infected with P.s.m. ES4326 (OD600 = 0.0001) and bacterial growth measured as described.
(e,f) sgt1b and rar1 do not suppress elevated endogenous SA levels in snc1. Free (e) and total (f) SA was extracted from 5-week-old plants and analysed by HPLC as described. All experiments were repeated at least twice with similar results.

To fully evaluate the involvement of RAR1 and SGT1b, the levels of endogenous SA in the single and double mutants were also measured. Both double mutants exhibited elevated levels of SA, similar to those found in the snc1 mutant (Figure 7e,f). Thus, our results show that neither sgt1b nor rar1 suppress snc1-mediated phenotypes including stunted morphology, constitutive pathogen resistance and elevated endogenous SA levels.

Discussion

In a screen for suppressors of constitutive resistance responses in snc1 npr1-1, we identified mos5, a mutant that restores wild-type morphology and susceptibility to virulent bacterial and oomycete pathogens. The mutation abolishes constitutive expression of PR-genes and mutant plants accumulate reduced levels of endogenous SA, an important signalling molecule in R-protein-mediated resistance. Using a map-based approach, the mos5 mutation was identified in AtUBA1, one of two ubiquitin-activating E1 enzymes in Arabidopsis. Both E1s have previously been shown to bind ubiquitin and to transfer it to various ubiquitin-conjugating E2 enzymes (Hatfield et al., 1997). UBA1 is expressed in all parts of the plant, predominantly in young cells and dividing tissue, as determined by promoter–GUS fusion experiments (Hatfield et al., 1997).

The identification of mos5 reveals an essential role for ubiquitination in plant defence signalling. In animal systems, ubiquitination has been shown to have a conserved role in different immunity pathways. In the mammalian immune system, ubiquitination is associated with processing of the transcription factor NF-κB precursors into functional products and their activation through the degradation of the inhibitory protein IκB (Ben-Neriah, 2002). Apart from these proteolysis-associated ubiquitination events, a number of regulatory functions for ubiquitination in animal immune signalling have been identified. These include signal termination via ubiquitin-associated receptor endocytosis, inhibition of T-cell activation, modulation of ubiquitination through the activity of deubiquitinating enzymes and activation of IκB kinase (IKK) via ubiquitination of the upstream E3 ligase TRAF6 (Ben-Neriah, 2002). TRAF6 ubiquitination is mediated by a hetero-dimeric ubiquitin-conjugating enzyme complex consisting of Ubc13 and UEV1a, and does not lead to its degradation by the proteasome (Deng et al., 2000). The Drosophila homologues of Ubc13 and UEV1a are similarly required for IKK activation and induction of an immune response (Zhou et al., 2005), revealing a strong evolutionary conservation in eukaryotic immune systems. In plants, the role of ubiquitination in R-protein signalling has been elusive, although ubiquitin-dependent protein degradation has previously been implicated in plant disease resistance responses (Devoto et al., 2003). Tobacco plants expressing a ubiquitin variant unable to form the polyubiquitin chains necessary for recognition by the 26S proteasome show altered responses to infection with tobacco mosaic virus (Becker et al., 1993). Jasmonate-dependent responses to wounding and necrotrophic pathogens have been shown to require the action of the SCFCOI1 ubiquitin ligase complex (Xie et al., 1998; Xu et al., 2002). Two RING ubiquitin ligases, RIN2 and RIN3, which were initially identified as interactors of the R-protein RPM1, affect the HR mediated by RPM1 and RPS2 without having an effect on pathogen proliferation (Kawasaki et al., 2005). In tomato, a set of U-Box ubiquitin ligases that are induced upon elicitor treatment appear to regulate the HR, adding further evidence for the role of ubiquitination in specific defence signalling (Gonzalez-Lamothe et al., 2006; Yang et al., 2006).

Interestingly, the mos5 mutation affects the resistance responses conferred by only a subset of R-proteins. This might reflect divergent signalling pathways employed by different R-gene products. Indeed, studies of RIN4, a protein involved in the activation of both RPM1- and RPS2-mediated resistance, showed that these R-proteins are differentially regulated by RIN4. Induction of RPM1-mediated resistance involves phosphorylation of RIN4 (Mackey et al., 2002), whereas the proteolytic processing of RIN4 is necessary for RPS2 activation (Axtell et al., 2003; Chisholm et al., 2005), where MOS5 is also required (Figure 5b).

Since snc1 was originally identified as a gain-of-function R-gene rather than from traditional gene-for-gene interactions, the cognate Avr-gene product recognized by wild-type SNC1 is not known. We are therefore not able to speculate on the requirement of MOS5 in wild-type SNC1-mediated resistance signalling.

Apart from its effect on resistance, mos5 also showed slightly enhanced resistance towards several natural and synthetic auxins (SG, unpublished data). These findings are not surprising, since ubiquitination and targeted degradation of Aux/IAA proteins are involved in auxin signalling (Dharmasiri and Estelle, 2004). The auxin-resistant mutant axr1 contains a mutation in the RUB-activating enzyme, analogous to UBA1 (Leyser et al., 1993). We therefore conclude that the defects observed in mos5 are likely to be due to alterations in ubiquitination of target proteins.

mos5 has a deletion of 15 bp very close to the C-terminus of UBA1, resulting in a substitution of arginine to serine and the deletion of five amino acids. The mutation lies in a region with high homology among ubiquitin E1s from different organisms (Figure 4), and some recent structural studies of E1s of ubiquitin-like proteins point to a possible function of the C-terminus of UBA1. Walden and co-workers reported the three-dimensional structure of the E1 for the human RUB1 orthologue Nedd8, a heterodimer composed of APPBP1 and UBA3 (Walden et al., 2003a,b). Interestingly, the C-terminus of UBA3 adopts a ubiquitin-like fold, which plays a role as an adapter domain for the docking of the cognate conjugating enzyme UBC12 (Huang et al., 2005). The human SUMO1 E1 heterodimer Sae1/Sae2 adopts a very similar structure to APPBP1/UBA3 and also features a C-terminal ubiquitin-fold domain (Lois and Lima, 2005). In both cases the ubiquitin fold is necessary for binding of the cognate E2s as determined by mutational analyses. Given the strong evolutionary conservation among E1s of different ubiquitin-like proteins over all kingdoms (Walden et al., 2003b), it seems plausible that UBA1 has a similar structure. It might be speculated that the mos5 mutation somehow disrupts the putative ubiquitin-fold domain in UBA1, resulting in reduced binding affinity of some, if not all, conjugating enzymes. This disruption or alteration of the ubiquitination cascade may result in increased stability of negative regulatory proteins, the degradation of which might be necessary during snc1-mediated resistance responses. Alternatively, activation of positive regulators by ubiquitination could be affected by the mos5 mutation.

Our data support a model in which pathogen elicitors target a population of host proteins, some of which might be involved in basal defence. Modification of these proteins is perceived by the corresponding R-proteins, which are then activated to initiate defence signalling. The ubiquitination of those target proteins might be impaired in the mos5 mutant, affecting activation of the corresponding R-proteins. We cannot exclude the possibility, however, that R-protein pathways which are unaffected by mos5 involve signalling via pathways independent of ubiquitination or, alternatively, that UBA2 might be partially redundant in these pathways.

We have identified a T-DNA insertion mutant line for UBA2 (SALK_108047), containing the T-DNA in an exon and probably resulting in a complete loss-of-function phenotype. The uba2 mutant plants, however, do not exhibit any morphological phenotype different from Col-0 wild type and are unaffected in disease susceptibility (Figure 6), indicating that UBA2 is not essential. In addition, no T-DNA insertion mutants for UBA1 could be identified and a mos5 uba2 double mutant is lethal. This could hint at a primary requirement for UBA1, and the mutation in mos5 reduces the activity of UBA1. Most species contain only one copy of the E1 enzyme, and the lack of phenotypic effect in the uba2 insertion mutant indicates that there might be a preferential recruitment of UBA1 in the ubiquitination process.

Ultimately, however, the specificity of protein ubiquitination is probably dependent on the action of ubiquitin-conjugating enzymes (E2) and ubiquitin ligases (E3), rather than on E1. Furthermore, ubiquitination might be a strategy of the pathogen to eliminate host proteins involved in basal defence in order to facilitate infection. The Pseudomonas syringae protein AvrPtoB has recently been shown to possess ubiquitin ligase activity and to inhibit the HR in susceptible tomato plants (Janjusevic et al., 2006). The pathogenicity of the human pathogen Shigella flexneri relies on OspG, a protein that specifically binds ubiquitinated E2s to inhibit a host immune response (Kim et al., 2005). Pathogens are thus able to exploit the host's ubiquitination machinery to suppress basal defences.

SGT1b and RAR1 are essential signalling components of a number of R-gene products (Muskett and Parker, 2003). SGT1 was first identified as an interactor of SCF ubiquitin ligases in yeast (Kitagawa et al., 1999), and its plant homologue SGT1b has been shown to be essential for ubiquitin-dependent responses to auxin and jasmonic acid (Gray et al., 2003). These data suggested an involvement of SGT1b in ubiquitination-dependent plant defence responses. Our data, however, indicate that SGT1b and RAR1 act independently from the ubiquitin-proteasome pathway in resistance signalling mediated by snc1, because, unlike mos5, sgt1b and rar1 do not suppress constitutive defence responses against virulent pathogens in snc1. Our data support other reports positioning RAR1 and SGT1 upstream of R-protein activation, where they are potentially involved as co-chaperones in the assembly and stability of putative R-protein recognition complexes (Azevedo et al., 2006; Bieri et al., 2004). RAR1 was shown to affect the steady-state levels of the R-proteins Mla1 and Mla6 in barley, although no direct interaction was observed (Bieri et al., 2004). In the same study, SGT1 was identified as an interactor of the LRR C-terminal region of Mla1 in a yeast two-hybrid screen, which was not observed when the bait also contained the N-terminal NBS domain. This suggests a steric hindrance of the R-protein structure and might reflect an intramolecular switch (Bieri et al., 2004). In another study, intramolecular association of the R-protein Bs2 was shown to be dependent on SGT1b (Leister et al., 2005), and accumulation of the R-proteins Rx and N in tobacco is dependent on the presence of SGT1 (Azevedo et al., 2006). These reports indicate that SGT1b and RAR1 directly or indirectly interact with several NB-LRR R-proteins, controlling their abundance on the one hand, and their intra- and intermolecular interactions on the other. Interestingly, SGT1b was also shown to act as a RAR1 antagonist in negatively regulating R-protein accumulation (Holt et al., 2005). Thus, SGT1b and RAR1 presumably act as co-chaperones in the formation of R-protein recognition complexes and may be necessary for effector recognition but are probably dispensable once the R-protein is activated. This would explain why the constitutively active R-protein snc1 is unaffected by mutations in RAR1 and SGT1b. We cannot exclude, however, the possibility that those co-chaperones are important in stabilizing the wild-type SNC1 complex.

Our results suggest that the ubiquitination pathway is essential for the activation of some, but not all, R-protein-mediated resistance responses, as well as for basal defence. Ubiquitination appears to act both positively, by promoting defence responses in the plant, and negatively, by suppressing defences when employed by the attacking pathogen. The balance of these positive and negative aspects may determine the outcome of a plant–pathogen interaction.

Experimental procedures

Plant growth and mutant phenotypic characterization

All plants were grown at 22°C under 16-h light/8-h dark cycles. The screen for suppressors of snc1 npr1-1 double mutants has been described elsewhere (Zhang and Li, 2005). pBGL2–GUS reporter gene expression was tested using 20-day-old plants grown on MS plates as described previously (Zhang et al., 2003). Infection experiments with P. syringae and P. parasitica were performed as described in Li et al. (2001). Endogenous SA was extracted from 4-week-old soil-grown plants and determined by HPLC as described previously (Li et al., 1999). Ribonucleic acid was extracted from 20-day-old seedlings grown on MS plates using the Totally RNA kit (Ambion, Austin, TX, USA) and reverse transcribed to cDNA using the RT-for-PCR kit (Clontech, Palo Alto, CA, USA). Expression levels of the pathogenesis-related genes PR1 and PR2 were determined as described previously (Zhang et al., 2003). The expression of uba2 was determined using the cDNA-specific primers UBA2-RT-F (5′-tccagtttgaaaaggacgatg-3′) and UBA2-RT-R (5′-tcaccactttaggtggaacc-3′).

Map-based cloning of mos5

The markers used to map mos5 corresponding to the respective bacterial artificial chromosome clones were derived from insertion–deletion (InDel) and single sequence polymorphisms between the Col-0 and Ler Arabidopsis ecotypes, identified by mining the available genomic sequences of both ecotypes as well as a database provided by Monsanto on the TAIR homepage (Jander et al., 2002; http://www.arabidopsis.org/Cereon/). Marker T8O18 was amplified using primers T8O18-NF (5′-tgtgatgtgaaccaagattg-3′) and T8O18-R (5′-agcttcgagtggattctac-3′) yielding PCR fragments of 728 and 401 bp in Col-0 and Ler, respectively. Marker T16B12 was amplified using primers T16B12-F (5′-atactattaccgtactcatg-3′) and T16B12-R (5′-acgcatgcattagacaacg-3′) yielding PCR fragments of 542 and 238 bp in Col-0 and Ler, respectively. Marker F23F1-1 was amplified using primers F23F1-1F (5′-ctctgtttccagcttgtatg-3′) and F23F1-1R (5′-gtgacgtacactactttctc-3′) yielding PCR fragments of 231 and 206 bp in Col-0 and Ler, respectively. Marker T9D9 was amplified using primers T9D9-F (5′-tatgtttagtcaacgcctcc-3′) and T9D9-R (5′-cattaccatactaacgtacg-3′) yielding PCR fragments of 267 and 222 bp in Col-0 and Ler, respectively. The marker T27E13-2 was amplified using primers T27E13-2F (5′-gattagtgtcacaagttcttg-3′) and T27E13-R (5′-tctaaagtcagaaccaactag-3′) yielding 291 bp PCR fragments in both ecotypes, but only the Col-0 derived product was digested with HinP1I.

Complementation of mos5

A genomic clone of At2g30110 encompassing 7 kb was amplified by PCR using Platinum Pfx DNA polymerase (Invitrogen, Carlsbad, CA, USA) in a two-step reaction using primers with modified restriction enzyme cleavage sites (underlined). The N-terminal fragment UBA1-N was amplified using primers mos5-KpnI (5′- cttggtaccaggtttcaactgcatc-3′) and mos5-intR (5′-ggctgtttcacttgagtgac-3′) and the C-terminal fragment UBA1-C was amplified using primers mos5-intF (5′-catacaggcatcattgcgtc-3′) and mos5-NotI (5′-ttttccttttgcggccgcgaaacaccacctgcaag-3′). Both fragments were first cloned into pBluescript (Alting-Mees et al., 1992) using the restriction enzymes KpnI/SalI and SalI/NotI, respectively, and sequenced to ensure that no mutations were introduced. UBA1-C was then subcloned into pBS-UBA1-N to create pBS-UBA1g. UBA1g was subsequently cloned into the binary vector pGreen229 (Hellens et al., 2000) to create pG229-UBA1g. mos5 snc1 npr1-1 plants were transformed with Agrobacterium containing pSoup and pG229-UBA1g using the floral dip method and T1 plants containing the UBA1g transgene were selected by spraying with glufosinate.

Creating the mos5 single and mos5 uba2 double mutants

The mos5 single mutant was obtained by crossing mos5 snc1 npr1-1 with Col-0 carrying the pBGL2-GUS transgene. F1 progeny of the cross displayed wild-type morphology and were allowed to self-pollinate and set seed. mos5 single mutants were identified among the F2 progeny using genotype-specific markers. The presence of the mos5 deletion was determined by PCR using the primers mos5del-F (5′-aactcttcgtgaggtgttgc-3′) and mos5del-R (5′-actcgactttcgcaacatcc-3′), which amplify fragments of 181 and 166 bp in Col-0 and mos5, respectively.

The uba2 homozygous T-DNA insertion mutants were identified in SALK_108047 using the gene-specific primers 1972900F (5′-ctcacctcactgagaactatg-3′) and 1974050R (5′-tcaccactttaggtggaacc-3′). One uba2 line was crossed with mos5 to yield F1 progeny with wild-type phenotype. Segregating progeny in F2 were genotyped using the PCR markers described above. The presence of the T-DNA was determined using the combination of T-DNA-specific LBa1 (5′-tggttcacgtagtgggccatcg-3′) and 1974050R.

Creating the snc1 sgt1b-1 and snc1 rar1-21 double mutants

In order to create the snc1 sgt1b-1 double mutant, snc1 (in the Col-0 background) was crossed with sgt1b-1 (in the Ler background). F2 plants that were homozygous Col-0 at the SNC1 locus, but heterozygous Col-0/Ler at the SGT1b locus were selfed and their progeny screened for lines homozygous Ler for sgt1b-1. The obtained double-mutant line was confirmed by sequencing to be homozygous for both snc1 and sgt1b-1 and used for further analysis.

To create the snc1 rar1 double mutant, snc1 npr1-1 was crossed with rar1-21 (both in the Col-0 background). Selfed F2 progeny segregated 1:3 snc1-like:wild type and plants displaying the characteristic snc1-like phenotype were screened with genetic markers specific for the npr1-1 and rar1-21 mutations. Nine plants out of 32 snc1-like plants were homozygous for rar1-21. Among those, two plants were homozygous for the segregating npr1-1 mutation, three plants were heterozygous and four plants were homozygous for wild-type NPR1-1. The presence of the rar1-21 mutation in those four plants was confirmed by sequencing the locus and one line was used for further characterization.

Acknowledgements

We thank Sarah Westelmajer and Yu-ti Cheng for excellent technical assistance, Dr Jeff Dangl for rar1-21 seeds and Youssef Belkhadir for dCAPS-marker information for rar1-21 genotyping, Dr Jane Parker for sgt1b-1 seeds and helpful comments on the manuscript and Dr Ljerka Kunst, Kristoffer Palma and Dr Marcel Wiermer for critical reading of the manuscript. We are grateful for financial support to SG by a doctoral fellowship of the Austrian Academy of Sciences (DOC) and a UBC Graduate Fellowship (UGF), and to XL from the Natural Sciences and Engineering Research Council of Canada (NSERC), the Canadian Foundation for Innovation (CFI), the British Columbia Knowledge Development Fund (BCKDF), the UBC Blusson Fund and the UBC Michael Smith Laboratories.

Ancillary