Identification of PAD2 as a γ-glutamylcysteine synthetase highlights the importance of glutathione in disease resistance of Arabidopsis


(fax +41 26 300 9740; e-mail

The first two authors contributed equally to this paper.

Present address: UMR Plante Microbe Environnement, INRA 1088, CNRS 5184, Université de Bourgogne, Dijon, France.


The Arabidopsis pad2-1 mutant belongs to a series of non-allelic camalexin-deficient mutants. It was originally described as showing enhanced susceptibility to virulent strains of Pseudomonas syringae and was later shown to be hyper-susceptible to the oomycete pathogen Phytophthora brassicae (formerly P. porri). Surprisingly, in both pathosystems, the disease susceptibility of pad2-1 was not caused by the camalexin deficiency, suggesting additional roles of PAD2 in disease resistance. The susceptibility of pad2-1 to P. brassicae was used to map the mutation to the gene At4g23100, which encodes γ-glutamylcysteine synthetase (γ-ECS, GSH1). GSH1 catalyzes the first committed step of glutathione (GSH) biosynthesis. The pad2-1 mutation caused an S to N transition at amino acid position 298 close to the active center. The conclusion that PAD2 encodes GSH1 is supported by several lines of evidence: (i) pad2-1 mutants contained only about 22% of wild-type amounts of GSH, (ii) genetic complementation of pad2-1 with wild-type GSH1 cDNA restored GSH production, accumulation of camalexin in response to P. syringae and resistance to P. brassicae and P. syringae, (iii) another GSH1 mutant, cad2-1, showed pad2-like phenotypes, and (iv) feeding of GSH to excised leaves of pad2-1 restored camalexin production and resistance to P. brassicae. Inoculation of Col-0 with P. brassicae caused a coordinated increase in the transcript abundance of GSH1 and GSH2, the gene encoding the second enzyme in GSH biosynthesis, and resulted in enhanced foliar GSH accumulation. The pad2-1 mutant showed enhanced susceptibility to additional pathogens, suggesting an important general role of GSH in disease resistance of Arabidopsis.


The plant stress hormones salicylic acid (SA), jasmonic acid (JA) and ethylene (ET) play important roles in the establishment of disease resistance in many plant–pathogen interactions (Dong, 1998; Glazebrook, 2005). In contrast, the disease resistance of Arabidopsis to Phytophthora brassicae and of tomato (Lycopersicon esculentum) to Phytophthora infestans was found to be largely independent of known stress hormone signaling pathways (Roetschi et al., 2001; Si-Ammour et al., 2003; Smart et al., 2003). The disease resistance of the resistant Arabidopsis accession Col-0 to P. brassicae was maintained in mutants deficient in SA-, JA- or ET-dependent signaling. However, the phytoalexin-deficient mutant pad2-1 (Glazebrook and Ausubel, 1994) was found to be hyper-susceptible to P. brassicae. The pad2-1 mutant was originally described as being partially camalexin-deficient and showing increased susceptibility to the bacterial pathogen P. syringae (Glazebrook and Ausubel, 1994; Glazebrook et al., 1997). Interestingly, the camalexin deficiency of pad2-1 was not the cause of its enhanced disease susceptibility to either P. syringae or P. brassicae. A null allele of the cytochrome P450 monoxygenase gene CYP71B15, pad3-1 (Zhou et al., 1999), nearly abolished camalexin synthesis yet had little effect on resistance to these pathogens (Glazebrook and Ausubel, 1994; Roetschi et al., 2001). The pad2-1 mutant was later shown to be more susceptible to other pathogens as well (Ferrari et al., 2003; Van Wees et al., 2003). Apparently, PAD2 encodes a gene product with important but unknown functions in the general disease resistance of Arabidopsis.

The hyper-susceptibility of pad2-1 to P. brassicae was exploited for the positional cloning of PAD2. We present evidence that PAD2 encodes γ-glutamylcysteine synthetase (γ-ECS, γ-glutamylcysteine ligase, GSH1), the enzyme that catalyzes the first step of de novo GSH biosynthesis. GSH1 is encoded by a single-copy gene (At4g23100) in the Arabidopsis genome. The first GSH1 cDNA of plants cloned from Arabidopsis encoded a predicted 60 kDa protein that was structurally unrelated to GSH1 of mammals and yeast (May and Leaver, 1994). The tripeptide GSH (γ-l-glutamyl-l-cysteinylglycine) is synthesized in two ATP-dependent reactions. GSH1 (EC catalyzes the formation of a peptide bond between the γ-carboxyl of l-Glu and the α-amino group of l-Cys. GSH synthetase (GSH2; EC, which is also encoded by a single gene in Arabidopsis (At5g27380; Rawlins et al., 1995), catalyzes the addition of glycine to γ-glutamylcysteine (γ-EC). GSH1 and GSH2 both contain plastidic transit peptides. In Arabidopsis, multiple transcription initiation leads to an exclusive plastidial localization of GSH1, while GSH2 predominantly accumulates in the cytosol (Wachter et al., 2005). GSH synthesis is subject to complex regulation. The expression and enzymatic activity of GSH1 is under multiple controls at the transcriptional, post-transcriptional and post-translational levels (Jez et al., 2004; May et al., 1998b; Xiang and Oliver, 1998).

The identification of PAD2 as GSH1 suggested that adequate levels of GSH are important in Arabidopsis for limiting the spread of virulent P. syringae and for establishing disease resistance to P. brassicae. The ubiquitous thiol tripeptide GSH has been implicated in many different aspects of cellular biochemistry (reviewed in May et al., 1998a; Mullineaux and Rausch, 2005; Noctor, 2006; Noctor and Foyer, 1998; Noctor et al., 1998a,b, 2002). GSH is present in up to millimolar concentrations in plant cells, and functions as a major determinant of cellular redox homeostasis. GSH plays important roles in stress physiology by reducing reactive oxygen species (ROS) via the ascorbate–GSH cycle which consists of three interdependent redox couples: ascorbate/dehydroascorbate, GSH/GSSG and NADPH/NADP (Noctor and Foyer, 1998). As a result of a redox reaction, GSH is oxidized to GSSG, which is recycled back to GSH by a reaction catalyzed by GSH reductase using NAD(P)H as a reducing agent. GSH can function as a modulator of redox-controlled enzymatic reactions and thiol-based regulatory switches (Foyer and Noctor, 2005; Paget and Buttner, 2003), in protein modification via glutathionylation (Dixon et al., 2005; Klatt and Lamas, 2000), and as a co-substrate in conjugation and detoxification processes catalyzed by GSH transferases, GSH peroxidases and glyoxalases (Edwards et al., 2000). As a precursor of phytochelatins, the production of GSH is involved in heavy metal tolerance (Cobbett and Goldsbrough, 2002). In addition to its various roles in cellular protection, GSH serves as a storage and transport form of reduced sulfur (Kopriva and Rennenberg, 2004), participates in the regulation of cell division in root apical meristems (Sanchez-Fernandez et al., 1997), promotes flowering (Ogawa et al., 2004), and plays a role in the nodulation process (Frendo et al., 2005). GSH can engage in thiol–disulfide exchange reactions that may link the regulation of gene expression to the redox state of cells (Baier and Dietz, 2005; Pfannschmidt, 2003). In animals, GSH was demonstrated to be involved in redox-dependent activity changes of transcription (Mihm et al., 1995). Thiol–disulfide status also appears to be important in disease resistance signaling. The reduction of key cysteines on the regulatory protein NPR1 and on the transcription factors TGA1 and TGA4 was shown to be crucial in the SA-dependent activation of the PR1-gene encoding pathogenesis-related (PR) protein 1, and changes in the GSH level were suggested to play a role in this process (Després et al., 2003; Mou et al., 2003). Interestingly, GSH treatment has been reported to activate the expression of a number of stress and defense genes (Dron et al., 1988; Loyall et al., 2000; Wingate et al., 1988), including PR genes (Creissen et al., 1999; Gomez et al., 2004; Senda and Ogawa, 2004).

The effect of GSH deficiency on disease resistance has been analyzed previously. The cad2-1 mutant, which accumulated only about 30% of wild-type amounts of GSH, showed an unaltered disease resistance phenotype to virulent and avirulent strains of Hyaloperonospora parasitica and virulent and avirulent strains of P. syringae pv. tomato, respectively (May et al., 1996a). Thus, GSH was concluded to be of minor importance for plant disease resistance. In contrast, cad2-1 and rax1-1, an Arabidopsis GSH1 mutant with similar GSH content as cad2-1, were shown to be more susceptible to avirulent strains of P. syringae (Ball et al., 2004). The identification of pad2-1 as a GSH-deficient mutant demonstrates that adequate levels of GSH are important for the accumulation of resistance-related compounds (Glazebrook and Ausubel, 1994; Roetschi et al., 2001) and for the establishment of disease resistance to many pathogens.


Positional cloning identifies PAD2 as a γ-glutamylcysteine synthetase

The hyper-susceptibility of pad2-1 to P. brassicae was used for positional cloning of the PAD2 gene. The segregation of disease susceptibility was analyzed in reciprocal crosses of pad2-1 with the resistant Arabidopsis accession Wassilewskija (Ws). All plants of the F1 progeny were resistant, while 24% of the F2 progeny derived from self-pollination of the F1 progeny were susceptible. This is consistent with the 3:1 ratio expected for segregation of a recessive allele of a single nuclear gene (χ2 = 0.60, P < 0.05). Bulk segregant analysis confirmed the location of PAD2 on chromosome 4 (Glazebrook and Ausubel, 1994). P. brassicae susceptibility co-segregated with the SSLP marker ciw7 on the lower arm of chromosome 4 (data not shown). To refine the position of PAD2, a mapping population from the pad2-1 × Ws cross was established, and 412 F2 chromosomes were analyzed with SSLP and CAPS markers. The results are summarized in Figure 1(a). The markers Ig950/60 and CER428546, showing very low recombination frequencies, were found to flank the pad2-1 mutation. These markers defined a region of about 110 kb containing 23 annotated ORFs. The 23 candidate genes were PCR-amplified from pad2-1 DNA and sequenced. Comparison with the Col-0 wild-type sequence identified a single G to A nucleotide transition at position 1697 from the start codon of the gene At4g23100. This mutation resulted in replacement of a serine by an asparagine residue at position 298 in the 522 amino-acid protein (Figure 1b). At4g23100 encodes γ-glutamylcysteine synthetase (GSH1), which catalyzes the first dedicated step of GSH biosynthesis (May and Leaver, 1994). Mutant alleles of GSH1 have been characterized previously. They include cadmium-sensitive 2-1 (cad2-1; Cobbett et al., 1998), rootmeristemless 1 (rml1; Vernoux et al., 2000) and regulator ofAPX2 1−1 (rax1-1; Ball et al., 2004). Figure 1(b) shows the position and nature of the four mutations in a region considered to be the catalytic domain of GSH1. An additional PAD2 allele (eds47  = pad2-2) was identified based on complementation tests (Glazebrook et al., 1996). Sequencing of the GSH1 gene of the pad2-2 mutant revealed that the pad2-2 mutation was identical to that in pad2-1.

Figure 1.

 Positional cloning of PAD2.
(a) Genetic map of the PAD2 region at the bottom of chromosome 4 of Arabidopsis (left). CAPS and SSLP markers were used to map PAD2 to a region flanked by Ig950/60 and CER428546. The recombination frequency for each marker is indicated in brackets. The defined region included 23 predicted genes. Sequencing identified a point mutation in gene At4g23100 that encodes GSH1. Right side: gene structure of GSH1 and position and nature of the pad2-1 mutation in relation to other known GSH1 mutants. Mutations are underlined in the second line of each comparison. Deletions in cad2-1 are indicated by dashes. The rax1-1 and cad2-1 mutations are located in exon 6, rml1 in exon 7 and pad2-1 in exon 8.
(b) Partial amino acid sequence of GSH1 of Arabidopsis (At4g23100; positions 221–300) compared with other mutant alleles of GSH1. The line labeled ‘mutant’ shows the mutations of the known GSH1 mutants: rax1-1 (R228K; Ball et al., 2004), cad2-1 (deletion of P237, K238 and V239L; Cobbett et al., 1998), rml1 (D258N; Vernoux et al., 2000) and pad2-1 (S298N; this paper). The region includes the putative catalytic domain defined by Lueder and Phillips (1996). The cysteine residue highlighted in black at position 251 is thought to be part of the active site of GSH1.

The GSH1 mutation in the pad2-1 mutant causes reduced GSH and increased cysteine accumulation

The identification of PAD2 as GSH1 led to the question of whether pad2-1 is deficient in GSH. GSH and cysteine levels were determined in Col-0 and pad2-1 by HPLC using homoglutathione (which is not present in Arabidopsis) as an internal standard. Figure 2(b) shows that foliar GSH levels in pad2-1 were reduced to about 21% of the level in wild-type plants. However, pad2-1 contained about five times more of the GSH1 substrate cysteine. These results demonstrate that the mutation in GSH1 interfered with GSH accumulation in pad2-1. Real-time RT-PCR revealed that Col-0 and pad2-1 contained very similar transcript levels of GSH1 and GSH2 (Figure 2a), indicating that the GSH deficiency of the pad2-1 mutant was not based on differential accumulation of GSH1 or GSH2 transcripts.

Figure 2.

 Complementation of the glutathione deficiency of pad2-1 by overexpression of the wild-type GSH1 cDNA.
(a) GSH1 and GSH2 transcript levels in Col-0, pad2-1 and three complemented lines of pad2-1 (35S::GSH1). Transcript levels were determined by real-time RT-PCR and are given relative to Col-0. The mean and SE from two experiments are reported. The relative transcript levels of GSH1 in the three complemented lines were significantly higher than in Col-0 (P < 0.001, t-test).
(b) GSH and cysteine content of Col-0, pad2-1 and three complemented lines of pad2-1 (35S::GSH1). The values are given relative to Col-0 for GSH or pad2-1 for cysteine: 100% equals 304 nmol g−1 FW for GSH and 93 nmol g−1 FW for cysteine. Leaves of 7-week-old plants (four leaves of 6–8 plants) were analyzed. The results show means and SE of three independent experiments. The GSH content in the three complemented lines was not significantly higher than in Col-0.

To confirm that the multiple phenotypes of pad2-1 were caused by the point mutation in GSH1, the GSH1 wild-type cDNA was expressed under the control of the CaMV 35S promoter in transgenic pad2-1 plants. Figure 2(a) shows for three independent transgenic lines, that expression of the wild-type GSH1 cDNA in pad2-1 caused a greater than threefold increase in GSH1 transcripts. Genetic complementation restored GSH accumulation of pad2-1, but did not lead to overaccumulation of GSH. The three complemented lines contained GSH and cysteine levels similar to wild-type plants (Figure 2b). None of the 36 transgenic lines analyzed contained >1.5 times the wild-type amount of GSH. The expression of GSH2, encoding the second enzyme of GSH biosynthesis, was very similar in Col-0, pad2-1 and the complemented lines. The results from real-time RT-PCR experiments and RNA blot analysis were consistent (data not shown).

GSH content has been linked to several developmental roles in plants, so pad2-1 was analyzed for some growth parameters. Analysis of the leaf number and leaf size of 20 plants per genotype at 13, 25 and 44 days after germination did not reveal a significant difference between the pad2-1 mutant and Col-0. Seed production was identical in pad2-1 and Col-0. Apparently, the reduced GSH content of pad2-1 was sufficient for normal fitness under optimal growth conditions.

Complementation of camalexin deficiency and disease susceptibility of pad2-1

In order to establish a link between the mutation in GSH1, the GSH deficiency and the other pad2-1-related phenotypes, the complemented 35S::GSH1pad2-1 lines were tested for resistance to P. brassicae and P. syringae pv. maculicola ES4326 (Psm ES4326) and for restoration of camalexin production in response to Psm ES4326 infection. Figure 3(a) shows the result of an infection experiment with P. brassicae. Three-week-old seedlings of Col-0, pad2-1 and two complemented 35S::GSH1pad2-1 lines were inoculated with zoospores of P. brassicae isolate D and analyzed 8 days later. Col-0 proved to be fully resistant, while the pad2-1 mutant was highly susceptible to infection by P. brassicae. The first signs of susceptibility were observed 3 days post-inoculation (dpi) in the form of water-soaked lesions, and most pad2-1 plants were dead or dying 1-week post-inoculation. In contrast, the complemented 35S::GSH1pad2-1 lines were as resistant to P. brassicae as Col-0. Similarly, inoculation experiments with Psm ES4326 confirmed that pad2-1 showed enhanced susceptibility in comparison with wild-type, while the complemented lines accumulated a bacterial titer very similar to that in wild-type plants (Figure 3b). Finally, as shown in Figure 3(c), the complemented lines accumulated wild-type amounts of camalexin in response to infection with Psm ES4326, while camalexin accumulation in the pad2-1 mutant was reduced by about 90%. In summary, the constitutive expression of wild-type GSH1 transcripts complemented the GSH deficiency of pad2-1, restored resistance to P. brassicae, reduced the enhanced susceptibility to Psm ES4326 and restored wild-type levels of camalexin. It was concluded that PAD2 encodes GSH1, and that all observed phenotypes of pad2-1 plants are caused by the point mutation identified in GSH1.

Figure 3.

 Complementation of camalexin deficiency and disease resistance phenotypes of pad2-1.
(a) Disease resistance phenotype of Col-0, pad2-1 and complemented lines of pad2-1 (35S::GSH1) after zoospore inoculation with P. brassicase isolate D. Four-week-old plants were spray-inoculated with a zoospore suspension (150 000 zoospores ml−1) until run-off, and incubated in humid conditions at 18°C for 8 days. The experiment was repeated four times with similar results.
(b) Bacterial titer of Col-0, pad2-1 and complemented lines of pad2-1 (35S::GSH1). Leaves of 4.5-week-old plants were infiltrated with Psm ES4326 and the bacterial titer was determined 0 and 72 h post-inoculation (hpi). Bars represent means and SD of log-transformed data (four replicates at 0 h and 16–20 replicates are 72 h). At 72 hpi, the bacterial titer in pad2-1 is significantly higher than in the other three genotypes (P < 0.0001, Mann–Whitney U-test).
(c) Camalexin content of Col-0, pad2-1 and complemented lines of pad2-1 (35S::GSH1) at 48 hpi with Psm ES4326. Bars represent means and SD of eight replicates. Camalexin levels in the 35S::GSH1 lines were not significantly different from Col-0.

Physiological complementation of camalexin deficiency and disease susceptibility of pad2-1

The pad2-1 mutant has several biochemical phenotypes: it contains more cysteine, less γ-glutamylcysteine and GSH, and it has a reduced potential to accumulate phytochelatins. Experiments with excised leaves were performed to test whether feeding of GSH could reverse some of the pad2-1 phenotypes. Figure 4(a) shows that feeding reduced GSH (5 mm) to excised leaves of pad2-1 plants prior to inoculation resulted in increased protection against P. brassicae. In three independent experiments, a total of 28 out of 33 water-incubated pad2-1 leaves were susceptible (85%), while only six of 33 GSH-treated pad2-1 leaves showed signs of susceptibility (18%). No difference between water- and GSH-treated leaves was seen in uninoculated controls (data not shown). Feeding 1 mm cysteine to excised leaves of Col-0 did not interfere with their resistance to P. brassicae. Analysis of thiols demonstrated that feeding 5 mm GSH to cut leaves of pad2-1 led to a 40-fold increase of GSH at the time of inoculation, to a level 10-fold higher than in untreated Col-0. Feeding 1 mm cysteine to cut leaves of Col-0 led to a 50-fold increase in cysteine and a threefold increase in GSH at the time of inoculation. Figure 4(b) summarizes the effect of feeding GSH to excised leaves on camalexin accumulation. Feeding GSH to pad2-1 had no effect on camalexin accumulation. Feeding GSH to excised leaves of pad2-1 inoculated with Psm ES4326 enhanced camalexin production fourfold to levels similar to inoculated wild-type plants. Comparable results were obtained by feeding GSSG, while feeding NADPH or NADH was ineffective (data not shown). GSH treatment also led to a doubling of camalexin accumulation in Col-0 in response to Psm ES4326. Feeding of cysteine prior to inoculation with Psm ES4326 had no significant effect on the disease resistance of the leaves (data not shown). Finally, deficiency in phytochelatin was not the cause of the susceptibility of pad2-1. The phytochelatin-deficient mutant cad1-3 (Howden et al., 1995) was resistant to P. brassicae in experiments with zoospore inoculation of 4-week-old seedlings (data not shown). Together, these results support the idea that the pad2-1 phenotypes were caused by GSH deficiency rather than by elevated cysteine or deficiency in phytochelatin.

Figure 4.

 Physiological complementation of pad2-phenotypes.
(a) Physiological complementation of disease susceptibility of pad2-1 to P. brassicae. Wild-type Col-0 or pad2-1 leaves of 5-week-old plants were excised and incubated in either water, 1 mm cysteine or 5 mm GSH for 6 h. The leaves were then spray-inoculated with zoospores of P. brassicae isolate D (150 000 zoospores ml−1). The picture was taken at 7 dpi. The experiment was repeated twice with similar results.
(b) Physiological complementation of Psm-induced camalexin accumulation by feeding reduced GSH. Leaves of Col-0 and pad2-1 plants were excised and inoculated with Psm ES4326. The petioles were placed into 1.5 ml centrifuge tubes containing water or a solution of 1 mm GSH. Camalexin levels were determined at 40 h post-inoculation (hpi). Bars represent means and SD of eight replicates. The effects of GSH feeding on camalexin accumulation in Psm-inoculated Col-0 and Psm-inoculated pad2-1, respectively, were statistically significant (P < 0.001, t-test). Similar results were obtained in another replicate experiment.

Analysis of additional GSH-deficient mutants

Three mutant alleles of GSH1 of Arabidopsis are already known (Figure 1a). Among these, rml1 contains only about 3% of the wild-type amounts of GSH and shows severe developmental phenotypes that lead to lethality (Vernoux et al., 2000). No growth phenotypes were reported for pad2-1, cad2-1 or rax1-1. The GSH content of the three mutants was compared in order to place pad2-1 in this allelic series. The direct comparison shown in Figure 5(a) indicated that pad2-1 contained on average 22% of foliar GSH compared to Col-0, followed by cad2-1 with 30% and rax1-1 with 38%. The absolute and relative GSH contents of Col-0 and the GSH1 mutants were variable among different experiments depending on environmental conditions (see Discussion). Figure 5(a) also shows that pad2-1 accumulated about 5 times, cad2-1 about 3 times and rax1-1 1.7 times more cysteine than Col-0. Inoculation of the three GSH-deficient mutants with P. brassicae led to only minor increases in foliar GSH (Figure 6b; data not shown). The three mutants were used to test the effect of varying foliar GSH content on GSH-related phenotypes. Disease resistance against P. brassicae was tested using the zoospore inoculation method (Figure 5b, top) and the plug inoculation method (Figure 5b, bottom). In both tests, pad2-1 proved to be more susceptible than cad2-1, while the resistance of rax1-1 was hardly affected at all. Figure 5(c) shows the results of a comparative analysis of the mutants for their capacity to accumulate camalexin in response to Psm ES4326. Camalexin levels in pad2-1 were 16% of wild-type 48 h post-inoculation (hpi), while camalexin levels in cad2-1 and rax1-1 were not significantly different from wild-type. All three mutants showed significantly increased susceptibility to Psm ES4326 (Figure 5d).

Figure 5.

 Comparison of Col-0 and the GSH-deficient mutants pad2-1, cad2-1 and rax1-1 in terms of GSH content, camalexin accumulation and disease resistance to P. brassicae and P. syringae.
(a) Comparison of foliar GSH and cysteine content. The results represent means and SE of three independent experiments. Four leaves of 6–8 plants each were used per experiment. The 100% value for GSH of Col-0 corresponded to 319 nmol g−1 FW. The 100% value for cysteine for pad2-1 corresponded to 182 nmol g−1 FW. The GSH content of the three mutants was significantly different from that of Col-0 (P < 0.05, t-test). The GSH content of pad2-1 was significantly different from that of rax1-1 and cad2-1 (P < 0.005, t-test). The GSH content of cad2-1 was different from that of rax1-1 at a lower confidence level (P < 0.06, t-test). The cysteine contents of the three mutants were significantly different from that of Col-0 (P < 0.04, t-test) and from each other (P < 0.02, t-test) except for the difference between pad2-1 and cad2-1 (P < 0.08, t-test).
(b) Disease resistance to P. brassicae. Top: 3-week-old seedlings were spray-inoculated with zoospores of P. brassicae isolate D (150 000 zoospores ml−1) and incubated in a humid chamber for 8 days. The experiment was repeated three times with similar results. Bottom: 7-week-old plants were plug-inoculated with P. brassicae isolate HH and incubated in a humid chamber for 6 days. The results represent means and SE of two repetitions. A minimum of five leaves from six plants were used for each experiment. Resistance scores were determined as described in Experimental procedures. The resistance scores of pad2-1 and cad2-1 (P < 0.0001, t-test) and of rax1-1 (P = 0.05, t-test) were significantly different from Col-0.
(c) Comparison of camalexin levels 24 and 48 h post-inoculation (hpi) with Psm ES4326. Bars represent means and SD of six replicates. Only pad2-1 is significantly different from Col-0 (P < 0.0001, t-test). The experiment was repeated with similar results.
(d) Disease resistance to P. syringae. Leaves were inoculated with Psm ES4326 and the bacterial titer was determined 0, 1, 2 and 3 days post-inoculation (dpi). Data points represent means and SD of log-transformed data from four (0 dpi), eight (1 and 2 dpi) or 16 (3 dpi) replicates. The bacterial titer in Col-0 is significantly lower than for the other three genotypes at 1, 2 and 3 dpi (P < 0.01, Mann–Whitney U-test). Similar differences at 3 dpi were observed in two more independent experiments.

Figure 6.

 Accumulation of GSH1 and GSH2 transcripts and of GSH following inoculation with P. brassicae.
Leaves of 7-week-old Col-0 and pad2-1 plants were plug-inoculated with P. brassicae isolate HH. Plugs not containing P. brassicae were used as a control treatment.
(a) Relative transcript levels of GSH1 and GSH2 at 24 hpi determined by real-time RT-PCR. The transcript level of uninoculated Col-0 was set to 1. Means and SE from two experiments are shown. The increase in transcript levels of GSH1 and GSH2 was significant in Col-0 and pad2-1 (P < 0.005, t-test).
(b) Relative GSH content determined by HPLC 48 and 72 hpi. The GSH content of Col-0 (176 nmol g−1 FW) was set to 100%. The results represent means and SE of two experiments. Three to five leaves from at least six plants were used for each time point. Increases in GSH content were statistically significant for Col-0 only at 72 hpi (P < 0.05, t-test).

Accumulation of GSH1 and GSH2 transcripts and GSH in response to P. brassicae

GSH seems to play an important role in the establishment of disease resistance to P. brassicae and other pathogens. Figure 6 analyses GSH1 and GSH2 transcript accumulation and the level of foliar thiols in Col-0 and pad2-1 in response to P. brassicae. The abundance of both transcripts increased in Col-0 about 2.5-fold and in pad2-1 more than sixfold at 24 hpi (Figure 6a). Despite this strong increase in transcript abundance, the GSH levels remained low and the cysteine levels high (data not shown) in inoculated pad2-1 over a period of 72 hpi. In contrast, the foliar GSH content of Col-0 was increased 1.7-fold at 72 hpi (Figure 6b).

Reduced GSH levels cause susceptibility to other pathogens

The GSH deficiency suggested that the disease susceptibility of pad2-1 was unlikely to be restricted to P. brassicae and virulent strains of P. syringae. The disease resistance phenotype of pad2-1 has been reported in a number of publications (see Table S1 for overview). The pad2-1 mutant was resistant to Leptosphaeria maculans, Erysiphe orontii and avirulent isolates of P. syringae. Depending on the accession–isolate combination, the pad2-1 mutation had no or only a weak effect on resistance to H. parasitica. In contrast, pad2-1 was more susceptible to Botrytis cinerea (Ferrari et al., 2003) and Alternaria brassicicola (Van Wees et al., 2003). To confirm and extend these results, we tested the disease resistance of pad2-1 to two necrotrophic pathogens. Figure 7 shows that pad2-1 was significantly more susceptible than Col-0 to B. cinerea or Plectospaerella cucumerina. Three days post-inoculation, the lesion size was about 2.7 times larger in B. cinerea-inoculated pad2-1 compared to Col-0. The lesions caused by P. cucumerina were about nine times larger in pad2-1 at 6 dpi. These results confirmed that the GSH deficiency of pad2-1 interferes with disease resistance against a variety of pathogens.

Figure 7.

 GSH deficiency causes susceptibility to other pathogens.
Disease resistance of Col-0 and pad2-1 was tested against (a) B. cinerea isolate BMM and (b) P. cucumerina. Leaves of 7-week-old plants were inoculated as described in Experimental procedures. Disease resistance was scored by measuring the diameter of lesions (3 dpi for B. cinerea and 6 dpi for P. cucumerina). The results show mean values and SE from two independent experiments. A minimum of three leaves from six plants were used for each treatment. The lesion diameter of pad2-1 was in both cases significantly different from that of Col-0 (P < 0.0001, t-test).


We have used the susceptibility of pad2-1 to P. brassicae to clone the PAD2 gene based on its map position (Figure 1). Our results indicate that PAD2 encodes the known metabolic enzyme GSH1, which catalyzes the first dedicated step of GSH biosynthesis (May and Leaver, 1994). Our conclusion is supported by several lines of evidence. First, pad2-1 contained reduced levels of GSH, and genetic complementation of pad2-1 with the wild-type GSH1 cDNA restored GSH content and complemented all pad2-1-associated phenotypes (Figures 2 and 3). Second, another GSH-deficient allele of GSH1, cad2-1, was also more susceptible to P. brassicae and Psm ES4326 (Figure 5). Finally, feeding of GSH to excised leaves of pad2-1 led to the restoration of pathogen-induced camalexin synthesis and to increased resistance to P. brassicae (Figure 4). The physiological complementation experiments indicated that the pad2-1 phenotypes were a direct result of GSH deficiency.

Comparison of pad2-1 with other GSH1 mutants

Three other GSH1 mutants of Arabidopsis have been described previously. Of these, rml1 (Vernoux et al., 2000) accumulated about 3% of the wild-type amounts of GSH and showed severe developmental defects. The cad2-1 mutant was identified as a cadmium-sensitive mutant that accumulated only 15–45% of the wild-type amounts of GSH (Cobbett et al., 1998; Howden et al., 1995). The rax1-1 mutant, which accumulated 20–50% of wild-type GSH, was identified based on its constitutive expression of photo-oxidative stress-inducible ascorbate peroxidase 2 (Ball et al., 2004). The GSH deficiency apparently did not negatively affect these mutants’ growth and fitness under normal growth conditions.

The pad2-1 mutant contained less GSH than cad2-1 or rax1-1, and showed the most severe phenotypes (Figure 5). It was the most susceptible mutant with regard to infection with P. brassicae, and it accumulated much less camalexin in response to Psm ES4326 than did cad2-1 and rax1-1. All mutants were similarly susceptible to Psm ES4326. Surprisingly, rax1-1 showed wild-type accumulation of camalexin and resistance to P. brassicae. Under our experimental conditions, rax1-1 contained more GSH (38% of Col-0) than cad2-1 (30%) and pad2-1 (22%). The levels of the GSH1 substrate cysteine were highest in pad2-1 (4.8-fold increase compared to Col-0) and lowest in rax1-1 (1.7-fold increase compared to Col-0). Cysteine feeding experiments with Col-0 excluded increased cysteine levels as a cause of the observed phenotypes (Figure 4; data not shown). The phenotypic differences also appeared not to be caused by the different capacity of the mutants to accumulate GSH in response to pathogens. All three mutants showed at best only a small increase in GSH in response to inoculation with P. brassicae (Figure 6b, data not shown). In conclusion, relatively small differences in foliar GSH appeared to have dramatic but variable effects on the various phenotypes, indicating that the range of GSH concentrations in the various mutants is close to a threshold below which the phenotypic effects become more severe.

Although pad2-1 consistently showed the lowest GSH content, the absolute as well as the relative level was variable between experiments depending on environmental parameters. Large variations in relative GSH content have been reported for cad2-1 (15–45%; Cobbett et al., 1998; Howden et al., 1995) and rax1-1 (20–50%; Ball et al., 2004). GSH biosynthesis is positively regulated by increasing light intensity (Karpinski et al., 2003; Noctor et al., 1998b; Ogawa et al., 2004), while high humidity has a negative effect on GSH accumulation (data not shown; May et al., 1996b). Interestingly, high humidity was also reported to negatively affect hypersensitive cell death and disease resistance (May et al., 1996b; Wang et al., 2005; Zhou et al., 2004).

Regulation of GSH accumulation in response to P. brassicae

Col-0 reacted to inoculation with P. brassicae with a coordinated 2.5-fold increase in GSH1 and GSH2 transcripts (Figure 6a). Similar results were reported by microarray analysis of the interaction of Arabidopsis with other pathogens (Figure S1; Zimmermann et al., 2004). The transcriptional activation of both biosynthetic genes resulted in a 1.7-fold increase in foliar GSH at 3 dpi with P. brassicae. Overexpression of the wild-type GSH1 cDNA in pad2-1 transgenics led to a three to fourfold increase in GSH1 transcript abundance but caused the transgenic plants to accumulate at best 1.5 times more GSH. GSH1 transcript abundance and GSH accumulation were apparently only loosely connected. Overexpression of GSH1 in various plant species rarely resulted in more than a twofold increase in foliar GSH (Cobbett et al., 1998; Creissen et al., 1999; Gomez et al., 2004; Noctor et al., 1996, 1998a,b; Xiang et al., 2001).

The low concentration of GSH in unstressed pad2-1 did not affect the transcript abundance of the GSH biosynthetic genes, indicating that low GSH is not sufficient to cause transcript accumulation (Figure 6a). The expression of GSH1 and GSH2 was, however, much more strongly induced in pad2-1 compared with Col-0. This potentiated induction is unlikely to be caused by a reduced negative feedback of GSH on the rate of transcription of these genes. Feeding of GSH to Arabidopsis had no inhibitory effect on GSH1 or GSH2 transcript abundance (Xiang and Oliver, 1998). The biosynthesis of GSH is under complex control involving regulatory steps at the transcriptional and post-transcriptional level (Jez et al., 2004; May et al., 1998b; Xiang and Oliver, 1998). Interestingly, treatment with JA increased the transcript levels of GSH1 and GSH2 in Arabidopsis without affecting GSH levels (Xiang and Oliver, 1998). It was postulated that full activation of GSH1 depended on at least two signals, one that triggers increased transcript accumulation and a second that leads to the post-translational activation of the GSH1 enzyme. A candidate for this second signal is hydrogen peroxide which, when externally applied, led to the accumulation of GSH without an increase in GSH1 and GSH2 transcript abundance (May and Leaver, 1993; Xiang and Oliver, 1998). GSH1 was shown to undergo a redox-regulated reversible conformational change that modulates enzyme activity (Jez et al., 2004). Developing strategies to make plants more stress- and disease-resistant by increasing their GSH pool would therefore be quite challenging.

GSH plays an important role in numerous plant–pathogen interactions

The accumulation of GSH in response to P. brassicae suggests an increased need by the host for GSH, presumably to counteract disturbances of redox status. GSH was found to accumulate in response to various pathogens (Fodor et al., 1997; May et al., 1996a,b; Mou et al., 2003; Vanacker et al., 2000). Interestingly, significant increases in GSH were measured in incompatible but not compatible interactions (May et al., 1996b; Vanacker et al., 2000). A significant inverse correlation between GSH content and disease incidence was also observed in chemically induced plants (Bolter et al., 1993). GSH content was found to decrease in tomato leaves infected with B. cinerea (Kuzniak and Sklodowska, 1999). Considering the negative effect of GSH deficiency on Arabidopsis disease resistance against B. cinerea (Figure 7), one might speculate that GSH biosynthesis could be a virulence target for this pathogen.

The importance of GSH in abiotic stress resistance is much better documented than its role in biotic stress resistance (May et al., 1998a; Mullineaux and Rausch, 2005; Noctor et al., 1998a,b). In a test of its disease resistance phenotype, the cad2-1 mutant did not show increased susceptibility to virulent and avirulent isolates of either the biotrophic oomycete H. parasitica or the bacterial pathogen P. syringae pv. tomato DC3000 (May et al., 1996a). It was concluded that GSH was of minor importance for disease resistance of Arabidopsis. However, impaired pathogen defense against avirulent isolates of P. syringae pv. tomato was reported for rax1-1 and cad2-1 (Ball et al., 2004), and virus-induced gene silencing of GSH2 compromised GSH accumulation and the disease resistance of Nicotiana megalosiphon to bluemold (Borras-Hidalgo et al., 2006). The reported resistance of cad2-1 to some pathogens might be explained by its slightly higher GSH content than pad2-1. In support of this conclusion, disease resistance to P. brassicae was less affected in cad2-1 than in pad2-1 (Figure 5b). The pad2-1 mutant was previously tested for altered disease resistance and found to be more susceptible to a number of pathogens (Table S1). Clearly, GSH deficiency in Arabidopsis interferes with disease resistance against many pathogens. The reason for the differential effect of GSH deficiency on disease resistance to different pathogens is not clear. A possible explanation is that GSH only becomes important when the first layers of defense are breached by the pathogen. This does not seem to be the case in the interaction of Arabidopsis with E. orontii, L. maculans, and some avirulent isolates of H. parasitica (Bohman et al., 2004; Glazebrook et al., 1997; Reuber et al., 1998). Although Col-0 is resistant, P. brassicae managed to penetrate into the leaf and then triggered a hypersensitive response. This second phase of the interaction appears to be disturbed in pad2-1 (Roetschi et al., 2001), suggesting that GSH deficiency may interfere with pathogen-triggered oxidative signaling (Foyer and Noctor, 2005; Noctor et al., 2002).

How GSH homeostasis might influence plant disease resistance

Why does GSH deficiency lead to increased disease susceptibility? An obvious hypothesis is that GSH deficiency leads to disturbances of cellular homeostasis that eventually result in compromised plant fitness. Although this possibility cannot be ruled out, there is increasing evidence for a role of GSH in plant defense beyond that of a mere redox buffer (Foyer and Noctor, 2005; Noctor, 2006). The level of S-nitrosothiol (S-nitrosoglutathione, GSNO) was recently shown to affect plant disease resistance (Feechan et al., 2005). The GSH-deficient mutants pad2-1, cad2-1 and rax1-1 look phenotypically normal and are still capable of mounting a successful defense against some pathogens. Application of GSH can activate the expression of stress and defense genes in a number of plant species (Dron et al., 1988; Gomez et al., 2004; Loyall et al., 2000; Wingate et al., 1988; Zhang et al., 1997). Our results suggest a positive role for GSH in pathogen-induced camalexin production. Increased camalexin synthesis in response to pathogens has been shown to be independent of SA-, JA- and ET-signaling but linked to oxidative stress (Thomma et al., 1999; Zhao et al., 1998). Perturbations in redox signaling could form the link between GSH deficiency and reduced camalexin accumulation. As an unexplored alternative possibility, GSH could be the cysteine-related intermediate Cys-R that was proposed to function as an S-donor in formation of the camalexin precursor S-dihydro-camalexic acid (Schuhegger et al., 2006).

There is evidence for a connection between GSH and the accumulation of PR proteins. PR1 expression was induced by GSH1 overexpression and by GSH feeding, while inhibition of GSH synthesis suppressed PR gene expression in tobacco (Creissen et al., 1999; Gomez et al., 2004). Similarly, PR1 transcript accumulation was suppressed in the GSH-deficient cad2-1 mutant and was promoted in GSH1-overexpressing Arabidopsis (Senda and Ogawa, 2004). Finally, in contrast to Col-0, pad2-1 showed no detectable accumulation of PR1 transcripts in response to P. brassicae (Roetschi et al., 2001). The mechanisms of how GSH levels influence PR gene expression are not known. The expression of PR1 and other SA-regulated genes is under the control of the regulatory protein NPR1 which interacts with the TGA1 and TGA4 transcription factors to activate PR gene expression. The reduction of specific disulfides of NPR1 and TGA1/TGA4 is crucial for the induction process (Després et al., 2003; Mou et al., 2003). GSH deficiency could possibly interfere with the reduction of NPR1 and/or the TGA transcription factors, thus preventing the transcriptional activation of PR genes. Analysis of global gene expression supports a specific effect of GSH on the expression of defense-related genes (Ball et al., 2004). The expression of defense-related genes, but surprisingly not genes encoding anti-oxidant proteins, was affected by GSH deficiency in cad2-1 and rax1-1. Feeding of GSH had a positive effect on defense gene expression but transcript levels of anti-oxidative enzymes were not much affected (Gomez et al., 2004).

Similar to the results with GSH-deficient mutants (Ball et al., 2004), the most striking changes in transcript abundance in ascorbate-deficient mutants were not observed for genes encoding anti-oxidative enzymes but for genes involved in responses to biotic stress (Pastori et al., 2003). GSH and ascorbate cooperate directly via the GSH–ascorbate cycle in anti-oxidant defense (Noctor and Foyer, 1998). It is therefore surprising that ascorbate and GSH deficiency have opposite effects on plant disease resistance: high ascorbate levels were found to suppress the expression of PR proteins, and ascorbate-deficient mutants showed increased SA content, constitutive accumulation of defense gene transcripts and increased resistance to virulent strains of P. syringae and H. parasitica (Barth et al., 2004; Pastori et al., 2003; Pavet et al., 2005). The opposite effects of changes in the level of GSH and ascorbate are difficult to reconcile with their well-described cooperative function in anti-oxidant defense, and suggest additional more specific roles of these molecules in plant disease resistance.

Experimental procedures

Biological material and plant inoculation with pathogens

After stratification for 3 days at 4°C, Arabidopsis thaliana seeds of accession Col-0 and the various Arabidopsis mutants were grown in jiffy-7 peat pellets (42 mm, Samen Mauser AG, Winterthur, Switzerland) in a growth chamber with a 10/14 h day/night photoperiod at 19/17°C. Light intensity varied between 80 and 120 μE m−2 sec−1. The following mutant lines (all in the Col background) were used: pad2-1 and pad2-2 (Glazebrook and Ausubel, 1994; Glazebrook et al., 1996), cad2-1 and cad1-3 (Howden et al., 1995), and rax1-1 (Ball et al., 2004).

P. brassicae isolate HH and P. brassicae isolate D (CBS179.89) were cultivated as described previously (Roetschi et al., 2001). Plants were inoculated in two ways. Unless otherwise indicated, 6–7-week-old plants were inoculated by placing agar plugs with growing mycelia upside-down on the leaf surface (Roetschi et al., 2001). Empty plugs were used for control treatments. In some experiments, 3-week-old plants were spray-inoculated at low pressure with zoospore suspensions at a concentration of 1–2 × 105 zoospores ml−1. Zoospores were produced as described by Roetschi et al. (2001). The inoculated plants were incubated at 100% relative humidity at 17–19°C. Plants were inoculated shortly before the onset of the 14 h dark period. Plug-inoculated plants were scored for disease resistance on a scale of 0–4 based on symptom development. Zero corresponds to a completely susceptible leaf and 4 to a fully resistant leaf, while scores of 1, 2 or 3 represent leaves with lesions on about 75%, 50% or 25% of their surface, respectively.

P. syringae pv. maculicola strain ES4326 was cultured in King's B medium at ambient temperature (22–25°C). Plants used for Psm ES4326 infection were grown in a controlled-environment chamber at 22°C, 75% relative humidity and 100 μmol m−2 sec−1 fluorescent illumination on a 12 h light/12 h dark cycle. All plants were 4–5 weeks old at the time of infection. For assay of camalexin and bacterial growth, plants were infected at a concentration of 4 × 104 or 1 × 103 CFU cm−2 of leaf, respectively. Bacterial infection and determination of bacterial titer was performed as described previously (Glazebrook and Ausubel, 1994), except that each sample consisted of two leaf discs, and tissue was ground in deep 96-well plates using ball bearings and a paint shaker. P. cucumerina was grown in the dark at 18°C on potato dextrose agar (PDA) plates. B. cinerea was grown on PDA plates at room temperature. Plants were inoculated by applying 4 μl droplets of spore suspensions to the upper surface of the leaves. The spore concentration was 105 spores ml−1 for P. cucumerina. The spores of B. cinerea were applied in quarter-strength potato dextrose broth (PDB) at a concentration of 3 × 104 conidia ml−1. The inoculated plants were incubated at 100% relative humidity at 20–24°C with a 12 h photoperiod. Symptom development was scored by measuring lesion diameter. Control inoculations were performed with sterile water or quarter-strength PDB, respectively.

Genetic complementation of pad2-1

The cDNA from GSH1 of Arabidopsis (May and Leaver, 1994) was PCR-amplified using the primers F-GSH1-BamHI/GW (5′-CACCGGATCCTATACCATGGCGCTCTTGTCTC-3′) and R-GSH1-SacI (5′-GCGAGCTCCCGGAGACTCGAATTCTTCAG-3′). The resulting 1653 bp product was sequence-verified and mobilized into the Gateway pENTR donor vector by directional TOPO cloning to generate the pENTR:GSH1 cDNA entry clone (Invitrogen, Basel, Switzerland). GSH1 cDNA insertion was verified by PCR, and the cDNA of a positive pENTR:GSH1 cDNA entry clone was mobilized into the Gateway destination vector pB2GW7 (Karimi et al., 2002). In the resulting pB2GW7-GSH1 plasmid, expression of the GSH1 cDNA is under the control of the constitutive CaMV 35S promoter. Agrobacterium tumefaciens GV3101 was transformed by electroporation. Positive colonies were selected on LB plates containing spectinomycin (50 μg ml−1), rifampicin (25 μμl−1) and gentamycin (25 μμl−1), and tested for the presence of the GSH1 cDNA by PCR. One positive Agrobacterium colony was used for in planta transformation of Arabidopsis. Transformed Arabidopsis were selected based on their resistance to basta. Thirty-six primary transformants were tested for their GSH content, and three with a single T-DNA insertion were selected for further analysis.

Physiological complementation of pad2-1

For physiological complementation of camalexin deficiency, leaves were excised from 5-week-old plants and the petioles were placed in 1.5 ml microcentrifuge tubes containing various solutions. Leaves were infected with Psm ES4326 at a concentration of 4 × 104 CFU cm−2 of leaf and incubated in a controlled-environment chamber as described above. For physiological complementation of disease resistance against P. brassicae, excised leaves from 6-week-old Arabidopsis plants were incubated in test solutions 6 h prior to inoculation with a zoospore suspension of isolate D. The leaves were incubated under the conditions described above and analyzed at 7 dpi. The foliar thiol content was determined at the time of inoculation.

Biochemical analysis

GSH and cysteine were extracted and quantified by HPLC as described by Harms et al. (2000). The samples were analyzed on a reverse-phase HPLC column (C18, 250 × 4 mm, 5 μm particle size; Macherey-Nagel, Oensingen, Switzerland). A solvent system consisting of 10% v/v methanol, 0.25% v/v acetic acid, pH 3.9 (NaOH) and 90% v/v methanol with a flow rate of 1.5 ml min−1 was used. Chromatography was followed by fluorescence detection (excitation: 380 nm, emission: 480 nm, SFM 25 fluorescence detector; Kontron, Zürich, Switzerland). Chromatograms were quantified by integration of peak areas. An internal standard (homoglutathione, 50 μg g−1 plant fresh weight) was added to the plant material before extraction. The amounts of GSH and cysteine were calculated relative to homoglutathione using a calibration curve. The measurement of camalexin was performed as described previously (Glazebrook and Ausubel, 1994).

Analysis of gene expression

Total RNA was extracted from Arabidopsis leaves using the RNeasy Plant Mini Kit following the manufacturer's instructions, including a treatment with RNase-free DNase I (Qiagen, Hombrechtikon, Switzerland). Total RNA (2 μg) was reverse-transcribed into cDNA using the Omniscript RT Kit (Qiagen). Real-time PCR with a Rotor-Gene 2000 apparatus (Corbett Research, Sydney, Australia) was performed using SYBR-Green as fluorescent reporter dye (SYBR Green PCR Master Mix; Applied Biosystems, Foster City, CA, USA) with the following primers: F-GSH1 (5′-GTGTGCCCTGCCAGCTTTCT-3′), R-GSH1 (5′-CAGGACATCTTCAGCGACATGC-3′), F-GSH2 (5′-CTTGGTGCGAGAAGGCGTTTA-3′), R-GSH2 (5′-ACTCCAAAACCAGCTGCAACG-3′), F-actin (5′-gacctttaactctcccgctatg-3′) and R-actin (5′-CACCAGAATCCAGCACAATAC-3′).

The respective primer pairs amplified a 183 bp region of GSH1 and 170 bp region of GSH2 from the 3′ end of the target transcript. Each reaction was performed in triplicate. The specificity of the amplification was confirmed by melting curve analysis and agarose gel electrophoresis. The relative abundance of GSH1 and GSH2 transcripts was calculated and normalized to At-Actin2 (At3g18780) transcript levels based on analysis with the Rotor-Gene 4.4 software package. Results represent the means and standard errors of two experiments. The real-time RT-PCR data of GSH1 expression were qualitatively confirmed by RNA blot analysis using 32P-radiolabeled GSH1 cDNA.


We thank Christopher Cobbett, University of Melbourne, Australia, for providing us with cad1-3 and cad2-1 seeds, and Phillip Mullineaux, University of Essex, Colchester, UK, for seeds of rax1-1. Doris Rentsch, Marianne Suter and Elisabeth Kuslys, University of Berne, Switzerland, are thanked for their initial help with thiol determination. This work was supported by grant 3100–67038 from the Swiss National Science Foundation to F.M., and grant IOB-0419648 from the US National Science Foundation to J.G.