Redox-active cysteines in TGACG-BINDING FACTOR 1 (TGA1) do not play a role in salicylic acid- or pathogen-induced expression of TGA1-regulated target genes in Arabidopsis thaliana

Salicylic acid (SA) is an important signaling molecule of the plant immune system. SA biosynthesis is indirectly modulated by the closely related transcription factors TGA1 (TGACG-BINDING FACTOR 1) and TGA4. They activate expression of SARD1 (SYSTEMIC ACQUIRED RESISTANCE DEFICIENT1), the gene product of which regulates the key SA biosynthesis gene ICS1 (ISOCHORISMATE SYNTHASE 1). Since TGA1 interacts with the SA receptor NPR1 (NON EXPRESSOR OF PATHOGENESIS-RELATED GENES 1) in a redox-dependent manner and since the redox state of TGA1 is altered in SA-treated plants, TGA1 was assumed to play a role in the NPR1-dependent signaling cascade. Here we identified 193 out of 2090 SA-induced genes that require TGA1/TGA4 for maximal expression after SA treatment. One robustly TGA1/TGA4-dependent gene encodes for the SA hydroxylase DLO1 (DOWNY MILDEW RESISTANT 6-LIKE OXYGENASE 1) suggesting an additional regulatory role of TGA1/TGA4 in SA catabolism. Expression of TGA1/TGA4-dependent genes in mock/SA-treated or Pseudomonas-infected plants was rescued in the tga1 tga4 double mutant after introduction of a mutant genomic TGA1 fragment encoding a TGA1 protein without any cysteines. Thus, the functional significance of the observed redox modification of TGA1 in SA-treated tissues has remained enigmatic. SIGNIFICANCE STATEMENT Previous findings demonstrating a redox-dependent interaction between transcription factor TGA1 and NPR1 attracted considerable attention. Here we show that TGA1 can act in the NPR1- and SA-dependent signaling cascade, but that its SA-regulated redox-active cysteines do not affect its function in this process.


Introduction
Redox reactions drive all energy-converting processes in living organisms. To adjust metabolic and regulatory processes to the prevailing redox state, proteins possess reactive cysteines which can be subject to various oxidative modifications. Prominent examples of proteins regulated by these so called thiol switches are enzymes of the Calvin Cycle, which become inactivated during the night when less reducing power is available in the chloroplast (Michelet et al., 2013). Conversely, oxidation of yeast transcription factor yAP1 leads to its accumulation in the nucleus, where it activates genes of the anti-oxidative system (Delaunay et al., 2000).
Plant immune responses are associated with complex changes in the cellular redox state. The defence hormone salicylic acid (SA), for instance, promotes the production of reactive oxygen or nitrogen species, while on the other hand inducing genes of the anti-oxidative system, like e.g. oxidoreductases or glutathione biosynthesis genes (Herrera-Vasquez et al., 2015). Redox signals affect the activity of the important regulatory protein NON EXPRESSOR OF PATHOGENESIS-RELATED GENE1 (NPR1). NPR1 controls many processes that are induced by elevated SA levels. In SA-treated tissues, NPR1 becomes first nitrosylated, which is a prerequisite for the formation of intermolecular disulfide bonds. These force the protein into the inactive oligomeric form, which resides in the cytosol (Mou et al., 2003;Tada et al., 2008). On the other hand, transcription of the small oxidoreductase THIOREDOXIN h5 is activated, which in turn reduces the disulfide bonds resulting in monomerization and nuclear translocation of NPR1 (Spoel et al., 2009). In the nucleus, NPR1 protein levels are regulated by NPR3 and NPR4 in an SA-dependent manner (Fu et al., 2012). All three NPR1 proteins bind SA, which is essential for their regulatory function (Fu et al., 2012;Ding et al., 2018). NPR1 interacts with TGACGbinding (TGA) transcription factors TGA2, TGA3, TGA5 and TGA6 to induce the expression of defence genes (Zhang et al., 2003;Saleh et al., 2015), while NPR3 and NPR4 function as repressors (Ding et al., 2018). 4 TGA factors form a family of ten members which are grouped into five clades (Gatz, 2013). The partially redundant clade-II TGAs (TGA2, TGA5 and TGA6) function together with NPR1 in the context of the immune response "systemic acquired resistance" (SAR) (Zhang et al., 2003). TGA3 is -like NPR1 -required for basal resistance against the bacterial pathogen Pseudomonas syringae pv. maculicola ES4356 (Psm) (Kesarwani et al., 2007). Since NPR1 is sumoylated after SA treatment and since TGA3 only interacts with sumoylated NPR1, it has been concluded that TGA3 and NPR1 functionally interact in vivo (Saleh et al., 2015). The SA marker gene PR1 (PATHOGENESIS-RELATED 1) has been used as an example to provide evidence that the described NPR1/TGA interactions occur at TGA binding sites in SA-responsive promoters.
Initial studies also suggested that NPR1 and clade-I TGAsc (TGA1 and TGA4) act in the same pathway. First, TGA1/TGA and NPR1 are required for basal resistance against Psm; second, TGA1 interacts with NPR1 only if an inhibitory internal disulfide bridge between cysteine residues 260 and 266 of TGA1 is not formed; third, the interaction between NPR1 and TGA1 promotes its binding to DNA and fourth, TGA1 is partially oxidized in untreated leaves and becomes reduced after SA treatment. Based on these circumstantial pieces of evidence, models presenting redox-modulated TGA1 interacting with NPR1 at SA-responsive promoters were published in numerous reviews and book chapters (Eckardt, 2003;Pieterse & Van Loon, 2004;Li & Zachgo, 2009;Moore et al., 2011;Chi et al., 2013;Li & Loake, 2016;Gullner et al., 2017).
However, the functional significance of TGA1 for the expression of SA/NPR1regulated genes and the role of the often cited redox modulation has not yet been conclusively demonstrated. Using microarray analysis of SA-treated plants, Shearer et al. observed that expression of 584 of the 629 SA-induced NPR1-dependent genes were independent from TGA1/TGA4 and that basal levels of the remaining 45 genes including PR1 were up-regulated in tga1 tga4 (Shearer et al., 2012). This implicated that oxidized TGA1/TGA4, which has a low DNA binding activity at least in vitro, would repress these genes and repression would be released upon the interaction of reduced TGA1/TGA4 with NPR1. To explain the susceptibility of the tga1 tga4 mutant, an NPR1-independent defence mechanism was postulated and confirmed by the higher susceptibility of the npr1 tga1 tga4 mutant as compared to 5 npr1 and tga1 tga4 mutants. A very recent study explained the susceptibility of the tga1 tga4 mutant by lower SA and pipecolic acid levels after Psm infections. These are due to the reduced expression of the master regulator of SA and pipecolic acid biosynthesis, SARD1 (SAR DEFICIENT 1) in Psm-infected tga1 tga4 plants.
We decided to re-address the question whether the redox-regulated cysteines in TGA1 play a regulatory role. Since in our hands, basal expression of PR1 was not enhanced in tga1 tga4, we performed again transcriptome analysis to identify TGA1/TGA4-regulated genes. RNAseq analysis provided a number of SA-induced NPR1-dependent genes that were less expressed in tga1 tga4, with SA-3-

HYDROXYLASE (S3H)/DOWNY MILDEW RESISTANT 6-LIKE OXYGENASE1
(DLO1) (Zhang et al., 2013;Zeilmaker et al., 2015) being the most robust TGA1/TGA-dependent gene. Under the conditions tested so far, no evidence for a function of the previously postulated redox switch of TGA1 for the regulation of DLO1 and other genes was obtained.

Plant material and cultivation
All plants used in this study are in the Arabidopsis thaliana Columbia background.
Treatment was conducted 1 hour after the subjective dawn and samples were collected at eight hours after treatment.

Pathogen infection assays
Pseudomonas syringae pv. maculicola ES4356 (Psm) was cultivated at 28 o C in King's B medium. Overnight cultures were diluted in 10 mM MgCl 2 to the final optical density at 600 nm (OD 600 ) of 0.005. 10 mM MgCl 2 (mock) or the diluted bacteria were hand-infiltrated into three leaves of five-week-old plants. Two days after this primary infection, three younger upper leaves were infiltrated again with a Psm solution (OD 600 of 0.005) in 10 mM MgCl 2. These leaves were harvested for RNA extraction after eight hours post infection. Pathogen infiltrations were generally conducted at one hour after the subjective dawn.

Other methods
Construction of recombinant plasmids, transcriptome analysis, quantitative reverse transcription (qRT)-PCR, transient expression analysis in Arabidopsis protoplasts, Western blot analysis, and accession numbers can be found in Methods S1. Primer sequences are depicted in Table S1. Maps and sequences of plasmids can be found in Notes S1.

Results
TGA1/TGA4 positively regulate a subgroup of salicylic acid-induced genes In order to address the question whether the disulfide bridge-forming cysteines of TGA1 that become reduced in SA-treated plants indeed play a role for accurate 7 transcription of SA-responsive genes, we first tested SA-induced expression of SARD1, which has been recently identified as a target gene of TGA1/TGA4 (Sun et al., 2018). However, in contrast to Psm infections, spraying with 1 mM SA resulted in TGA1/TGA4-independent SARD1 expression ( Fig. 1). Under these conditions, SARD1 was controlled by the well-established SA-responsive regulatory module which consists of NPR1 and clade-II TGAs TGA2/TGA5/TGA6. Still, it has to be noted that basal levels of SARD1 were lower in the tga1 tga4 and the npr1 mutants than in wild-type plants suggesting that residual basal levels of TGA1/TGA4 and NPR1 stimulate basal SARD1 expression.
In contrast to previously published observations (Lindermayr et al., 2010;Shearer et al., 2012), basal PR1 expression was not enhanced in the tga1 tga4 mutant and SA-induced PR1 transcript levels were only slightly affected (see below,Fig. 4). Therefore, we performed transcriptome analysis of RNA harvested from leaves of mock-and SA-treated plants. We compared the expression pattern of sid2 and sid2 tga1 tga4 rather than that of wild-type and tga1 tga4 because we wanted to avoid any possible influence of TGA1/TGA4 on endogenous SA biosynthesis.
Moreover, we aimed to reduce fluctuations in gene expression due to environmental factors affecting endogenous SA levels in different experiments. Four-week-old plants were sprayed either with water or with 1 mM SA. Eight hours after treatment, three leaves of five individual plants were collected and total RNA was isolated. The experiment was repeated four times with batches of independently grown plants.
Thus, the RNA from 15 leaves of five plants served as one replicate and replicates originated from four independent experiments.
Principal component analysis (PCA) results in clusters of samples with a similar expression pattern and thus yields a first impression of the global structure of the data set. The clusters from sid2 and sid2 tga1 tga4 plants treated with water showed a clear separation (Fig. S1) indicating that the transcriptomes of both genotypes are different even in the absence of ICS1-derived metabolites. The clusters from SA-treated plants indicate that both genotypes respond to SA. Since our aim was to identify target genes of TGA1/TGA4 after SA treatment, we focused on those 2090 genes that were induced (log2 fold change (FC) >1) by SA in sid2 (Table S2). 8 Fourtyone % (864 genes) of the 2090 SA-induced genes showed a differential expression pattern in sid2 tga1 tga4. These 864 genes fall in two major groups (Fig. 2a). Genes with lower expression values in the sid2 tga1 tga4 plants as compared to sid2 establish the "green" group (346 genes). Three major subgroups were identified based on reduced gene expression in sid2 tga1 tga4 either after SA treatment (119), mock and SA treatment (71), or only after mock treatment (153). Only three genes were less expressed in SA-treated leaves while background levels were elevated.
The two major subgroups of the "red" group, which comprises 518 genes that are higher expressed in sid2 tga1 tga4 as compared to sid2, contain genes that have higher expression values only in the mock situation (401) and genes that had elevated transcript levels in mock-and SA-treated plants (114). Three genes were hyper-induced upon SA treatment and have wild-type transcript levels upon mock treatment. Figure 2b shows relative expression levels of representative genes of the green and the red group. DLO1 encodes for an SA hydroxylase that is involved in dampening the immune response by inactivating SA (Zhang et al., 2013;Zeilmaker et al., 2015). Its expression responded strongly to SA (30-fold) and we observed a 7fold reduction of expression in the tga1 tga4 mutant, both under basal conditions and after SA treatment. Thus, the induction factor after SA treatment was not changed, suggesting that TGA1/TGA4 act as amplifiers under both conditions. In contrast, induction factors were lower for other genes in sid2 tga1 tga4 as compared to sid2 (Table S3). ß-1,3 GLUCANASE (BGL2) for example was induced by a factor of 4.3 in SA-treated sid2, and by a factor of 1.4 in SA-treated sid2 tga1 tga4. Another example is GLUTATHIONE S-TRANSFERASE F 6 (GSTF6), which was induced by a factor of 52 in sid2 and by a factor of 4.3 in sid2 tga1 tga4. The well-known gene of the NPR1-dependent SA response gene PR1 barely missed the cut-off for being differentially expressed in sid2 tga1 tga4 versus sid2 in the RNAseq analysis, but its expression was still modulated by TGA1/TGA4. Two genes with elevated expression levels (ATAF1 and WKRY6) in sid2 tga1 tga4 as compared to sid2 are displayed as well. These genes code for transcription factors.

9
The TGACGTCA motif is specifically enriched in SA-induced genes that are positively regulated by TGA1/TGA4 The ideal binding site for TGA factors is the palindromic sequence TGAC/GTCA, which is an extended C-box (GAC/GTC) (Izawa et al., 1993;Qin et al., 1994).
However, the pentamer TGAC/G is sufficient for binding. Moreover, at least TGA1 binds to the A-box (TAC/GTA) in vivo (Wang et al., 2019) and tobacco TGA1a binds to A-and G-(CAC/GTG) boxes in vitro (Izawa et al., 1993). Therefore, we tested whether any of these potential binding sites is specifically enriched in promoters that are either lower or higher expressed in sid2 tga1 tga4 as compared to sid2. To this end, the 1-kb sequences upstream of the predicted transcriptional start sites were scanned using the Motif Mapper cis element analysis tool (Berendzen et al., 2012).
As displayed in Fig. 3, all potential TGA1 binding sites are enriched in the 2090 SAinducible promoters as compared to promoters arbitrarily selected from the whole genome. At least the enrichment of the extended C-boxes was expected since most of the 2090 genes are likely to be regulated by NPR1 acting in concert with TGA2/TGA5/TGA6 or TGA3 (Wang et al., 2006). However, when comparing the relative frequency of these motifs in TGA1/TGA4-regulated promoters with their relative frequency in the 2090 SA-regulated genes, an enrichment of the TGACGTCA motif was detected in the group of those 346 genes that required TGA1/TGA4 for maximal expression. DLO1, for instance, contains the TGACGTCA palindrome at position -72 bps with respect to the transcriptional start site. Likewise, SARD1, which was slightly but significantly activated by TGA1/TGA4 in the absence of SA ( Since the SA-induced redox-modification of TGA1 alters the interaction with NPR1, we tested, whether expression of the selected TGA1/TGA4-dependent genes ( Fig.   2b) requires NPR1. We also included the tga2 tga5 tga6 mutant, since NPR1 has been functionally associated with clade-II TGAs (Zhang et al., 2003). Col-0 wild-type and tga1 tga4 mutant plants were included in the experiment. As observed before for sid2 tga1 tga4 vs. sid2, the four marker genes that are positively regulated by TGA1/TGA4, were less expressed in tga1 tga4 vs. Col-0 (Fig. 4). However, the elevated background levels of WRKY6 and ATAF1 were not observed in the presence of a functional SID2 allele. Expression of all genes with the exception of ATAF1 was significantly regulated by NPR1 (Fig. 4). Since expression also depended on clade-II TGAs, a functional connection between TGA1/TGA4 and NPR1 could not be inferred.
The TGACGTCA motif in the DLO1 promoter is a target site for TGA1 and for TGA2 The DLO1 promoter contains only one of the typical TGA binding sites (TGACGTCA) and one A-box within 2000 bps upstream of the transcriptional start site. In order to investigate, whether representatives of both clades of TGAs might be accommodated at the promoter, we analyzed the effect of transiently expressed TGA1 and TGA2 on DLO1 promoter activity. The DLO1 regulatory region from -1777 bps (with respect to the transcriptional start site) to the ATG start codon was fused to the open reading frame of the firefly luciferase gene (fLUC), while TGA1 and TGA2 were expressed under the control of the UBIQUITIN10 (UBQ10) promoter. Proteins were tagged at their C-terminal ends with a triple HA and a streptavidin tag. An "empty" vector expressing only the triple HA tag under the control of the UBQ10 promoter was used as a control of background promoter activity and adjustment of equal amounts of DNA in the transfection mixture. Renilla luciferase (rLUC) served to normalize for transfection efficiency.
In order to avoid background activation by endogenous clade-I and clade-II TGAs, we used protoplasts of the tga1 tga2 tga4 tga5 tga6 mutant that was previously obtained by crossing the respective genotypes. In this assay, only TGA1, but not TGA2 activated the promoter. Co-expression of TGA1 with TGA2 slightly enhanced promoter activity ( Fig. 5a and Fig. S2). Since TGA2 -in contrast to TGA1 -does not 1 1 binding of TGA2 to the promoter might have been missed in this assay. SA treatment, which leads to association of the transcriptional co-activator NPR1 with clade-II TGAs in differentiated leaf cells, did not specifically increase TGA-enhanced expression of the Prom DLO1 :fLUC construct in protoplasts (Fig. S2). Thus, the SA signal transduction chain does not operate in the same manner in protoplasts as in mesophyll cells. To compensate for this, the coding regions of TGA1 and TGA2 were fused C-terminally to the activation domain of the Herpes simplex viral protein (VP) 16. Strong trans-activation of the DLO1 promoter by TGA1-VP and TGA2-VP ( Fig.   5b) allowed to address the importance of the TGA binding sites. Activation was abolished when the TGACGTCA motif at position -72 bps relative to the transcriptional start site was mutated while mutation of the A-box at position -1731 bps did not affect the responsiveness of the promoter to both TGAs. It is concluded that the single TGA binding site of the DLO1 promoter is the only site to recruit either clade-I or clade-II TGAs.

DLO1 and BGL2 expression depends on SARD1/CBP60g
On the one hand, the requirement of clade-I and clade-II TGAs for expression of a marker gene with only one TGACG motif in the promoter might be explained by a heterodimer being active at this promoter. This scenario seems feasible since heterodimerization between in vitro co-translated tobacco TGA1a and TGA2 has been shown before (Niggeweg et al., 2000). Alternatively, they might bind as homodimers not only at the final target gene, but also at genes encoding for regulators acting upstream in the SA-dependent signaling cascade. This hypothesis fits to the expression pattern of SARD1, which is regulated by the well-established NPR1-TGA2/TGA5/TGA6 module, but not by TGA1/TGA4 (Fig. 1). As shown in Fig.   6, expression of DLO1 and BGL2 was strongly reduced in the sard1 cbp60g double mutant, which lacks not only SARD1 but also the related and sometimes redundantly acting factor CBP60g (CALCIUM BINDING PROTEIN 60g (Wang et al., 2011)). This effect was less pronounced for PR1 and absent for GSTF6 expression. At least for DLO1 and BGL2, which are more affected by TGA1/TGA4 than PR1 and GSTF6, the concept of indirect regulation by clade-II TGA-activated SARD1 and direct regulation by clade-I TGAs seems plausible. Consistently, both promoters contain SARD1 binding sites. No SARD1 binding sites were found in the other two promoters.

2
Mutation of the redox-active cysteines does not alter the expression pattern of selected marker genes in SA-treated leaves As mentioned above, a disulfide bridge between C260 and C266 was detected in 50% of the TGA1 proteins in untreated tissue, while 100% of the TGA1 pool is reduced in SA-treated tissue resulting in a larger amount of TGA1 being able to interact with NPR1, which in turn leads to increased DNA binding (Despres et al., 2003). Having identified SA-induced TGA1/TGA4-dependent genes we were now able address the importance of these "SA-switchable" cysteines. Since the two flanking cysteines C172 and C287 are also prone to redox modifications (Lindermayr et al., 2010), all four cysteines were mutated (C172N C260N C266S C287S). The first three cysteines were changed into residues found in TGA2 at the corresponding positions, while the last cysteine was changed to serine, which is found at the corresponding positions in TGA3, TGA4, TGA7 and TGA9. Mutations were introduced into a genomic clone that consisted of 2671 bps upstream of the translational start site, exons and introns and 217 bps downstream of the transcribed region. These lines, along with a transgenic line transformed with the empty vector, were treated with water or with SA and the expression of TGA1/TGA4-dependent marker genes was monitored. As observed before, expression of DLO1, BGL2, GSTF6 and PR1 was reduced in the absence of TGA1/TGA4. Both, the wild-type and the mutated TGA1 protein rescued SA-induced expression of the four TGA1/TGA4dependent marker genes to the same degree ( Fig. 7a). Also, background levels were not differentially affected in plants expressing either TGA1 or mutated TGA1 (for background SARD1 transcript levels, see Fig. S3). Although the proteins accumulated to higher levels than endogenous TGA1 in the untransformed Col-0 wild-type plants (Fig. 7b), only partial complementation was observed for GSTF6 and PR1. This might be due to the absence of TGA4, to the N-terminal HA-tag that was introduced upstream of the ATG start codon, or the history of the untransformed Col-0 seeds. Since the non-mutated and the mutated proteins were equally effective, it is concluded that the SA-mediated redox switch in TGA1 does not contribute to the proper expression of TGA1/TGA4-dependent target genes at eight hours after mockor SA treatment.  et al., 2018), we questioned whether under these conditions TGA1/TGA4 might act in concert with NPR1 and SA. We tested SARD1 expression eight hours after infection with Psm in the tga1 tga4, tga2 tga5 tga6, npr1 and the sid2 mutants (Fig. 8a).
As observed in SA-treated tissue, SARD1/CBP60g, NPR1 and clade-I and clade-II TGAs were important for DLO1 and BGL2 expression in Psm-infected leaves, independent of whether they had been pre-treated with Psm or with MgCl 2 . Since clade-II TGAs were not required for SARD1 expression and thus did not influence ICS1 transcript levels ( Fig. S4), we consider it likely that SA levels were not reduced in tga2 tga5 tga6. It is concluded that these factors activate DLO1 directly, while the effect of the clade-I TGAs and NPR1 can be partially explained by reduced SA levels due to reduced SARD1 and ICS1 expression.
BGL2 transcript levels followed a similar trend. However, it has to be noted that in Psm-infected SAR leaves, SARD1 and BGL2 were not as stringently dependent on SA as in Psm-infected leaves from plants that had be pre-treated with MgCl 2 . NPR1 remained to be necessary even when SA levels were not as critical for induction. This suggests that a signaling molecule different from SA can activate the NPR1/TGA1/TGA4 regulatory module in Psm-infected SAR leaves. A similar phenomenon has been observed very recently in the auto-immune mutant camta123 showing that SARD1 transcript levels were higher in sid2 as compared to npr1 (Kim Having identified SARD1 as an SA/NPR1/TGA1/TGA4-dependent target gene after pathogen infection we analysed its expression in the complementation lines ( Fig. 9).
As expected, Psm-induced SARD1 expression was reduced in tga1 tga4 plants transformed with the "empty vector". Importantly, both TGA1 constructs (i.e. TGA1 and TGA1red) complemented the phenotype to the same extent. Elevated expression after Psm pre-infections as compared to mock pre-treatments and the contribution of TGA1/TGA4 to gene expression was more pronounced for DLO1 and BGL2. Again, TGA1 lacking all four cysteines complemented the phenotype to a similar extent as the wildtype protein, supporting the notion that the lack of potential oxidative modifications does not alter the regulatory properties of the protein under these conditions.

DISCUSSION
Arabidopsis TGA transcription factors TGA1 and TGA4 interact with the SA-activated transcriptional co-activator NPR1 in a redox-dependent manner (Despres et al., 2003;Lindermayr et al., 2010). Here, using TGA1 mutants with point mutations in all four cysteines, we show that these cysteines do not play a role in SA-or pathogeninduced NPR1-dependent expression of TGA1/TGA4-regulated marker genes. We identified TGA1/TGA4 as a positive regulator of the SA catabolizing gene DLO1.
Finally, we found that the relative influence of clade-I and clade-II TGAs on SARD1 expression depends on whether plants are treated with SA or with Psm.
In order to address the functional importance of redox-modulated cysteines in TGA1, we first identified SA-induced TGA1/TGA4-dependent genes by RNAseq analysis.
Since it was known that SA synthesis is controlled by TGA1/TGA4 (Sun et al., 2018), we performed the analysis in the SA biosynthesis mutant sid2. This strategy guaranteed that genes affected in the SA-treated sid2 tga1 tga4 mutant as compared to sid2 would require TGA1/TGA4 downstream of SA, while any effects upstream of SA were excluded. Only 193 out of the 2090 genes that were higher expressed at eight hours after SA treatment as compared to mock treatment showed reduced expression in SA-treated sid2 tga1 tga4 plants. It is likely that this number would be 1 5 even lower in the wild-type background since we observed more fluctuations in the presence of endogenous amounts of SA. The low frequency might be due to the expression pattern of TGA1/TGA4, the promoters of which are mainly active in the vascular tissue (Song et al., 2008;Wang et al., 2019). This correlates well with the expression pattern of the two robustly regulated target genes: DLO1 is expressed near the vascular tissue in Hpa-infected leaves (Zeilmaker et al., 2015), while BGL2 is expressed near the vascular tissue in SA-treated leaves (Spoel et al., 2009). We assume that the other regulatory components influencing DLO1 and BGL2 expression (NPR1, clade-II TGAs, SARD1) are also present in this tissue.
Still, the discrepancy to previously published gene expression patterns of the SAtreated tga1 tga4 mutant has to be pointed out. Shearer et al. Whatever the reason for these differences is, we were able to identify SA-induced NPR1-dependent genes that required TGA1/TGA4 for maximal expression. Due to the limited expression domain of TGA1/TGA4 we failed to prove direct binding to e.g the promoters of DLO1 or BGL2 by chromatin immunoprecipitation (ChIP) experiments. Similar problems were encountered before: binding of TGA1/TGA4 to the SARD1 promoter was only shown in protoplasts ectopically expressing TGA1 To answer our primary research question, whether the redox-modulated NPR1dependent DNA-binding activity of TGA1 influences the expression of SA-dependent target genes, we had to make sure that expression of the identified target genes are 1 6 regulated by the interplay between SA, TGA1 and NPR1. However, the analysis was complicated since SA-induced expression of all four tested TGA1/TGA4-dependent target genes also depended on clade-II TGAs, which can recognize the same binding site as TGA1/TGA4. Given the fact that at least the DLO1 promoter contains only one TGA binding site, we postulate for SA-treated tissues that SA activates NPR1 to stimulate expression of SARD1 in concert with clade-II TGAs. Subsequently, SARD1 acts at the DLO1 and the BGL2 promoters, the expression of which is further enhanced by TGA1/TGA4. Thus, in SA-treated tissues, we could not clearly establish that DLO1 or BGL2 are regulated by a mechanism that is controlled by TGA1/TGA4 interacting with NPR1.
Interestingly, the functions of clade-I and clade-II TGAs in the SA-dependent regulatory network were changed in Psm-infected leaves. Here, the SARD1 promoter remained to be responsive to NPR1, but was regulated by TGA1/TGA4 while TGA2/TGA5/TGA6 became dispensable. Thus, in this tissue, at least SARD1 was the candidate gene we were looking for to address the functional importance of the redox-regulated cysteines. However, the redox-regulated cysteines did not play a role for SARD1 expression, at least after eight hours after pathogen infection of naïve or SAR leaves. Under these conditions, endogenous SA levels might have already led to full reduction of the wild-type protein.
According to previously published data, interfering with the internal disulfide bridge formation should lead to a protein that constitutively interacts with NPR1 and subsequently binds to DNA with a higher affinity (Despres et al., 2003;Lindermayr et al., 2010). Thus, higher background activity of at least SARD1 and thus its downstream genes might have been the expected consequence of the mutations.
This was not observed, most likely due to other inhibitory mechanisms including the repressive effects of NPR3 and NPR4 (Ding et al., 2018). A phenotype might be expected if oxidation and thus inactivation of TGA1 would happen under certain conditions. Our complementation lines in combination with the TGA1/TGA4dependent marker genes might provide useful tools to analyse whether potential antagonistic effects of e.g. reactive oxygen species-generating abiotic stresses that interfere with the SA pathway are less pronounced in plants expressing a mutant TGA1 protein that cannot be oxidized.

Supporting Information
Additional Supporting Information may be found in the online version of this article.      Methods S1 Detailed description of methods.
Notes S1 Maps and sequences of plasmids used in this work.       qRT-PCR analysis of transcript levels of six TGA1/TGA4-modulated genes in wildtype (Col-0), tga1 tga4, tga2 tga5 tga6 and npr1 plants. Four-week-old plants were sprayed either with water (mock) or 1 mM salicylic acid (SA) and further incubated for 8 h. Transcript levels were normalized to transcript levels of UBQ5. Bars represent the average ± SEM of four to five biological replicates, each replicate representing three leaves from one plant. Statistical analysis was performed using one-way ANOVA followed by Tukey's post hoc test for mock-and SA-treated samples separately. Lowercase letters indicate significant differences (P < 0.05) between mock-treated samples; uppercase letters indicate significant differences (P < 0.05) between SA-treated samples. between various reporter constructs combined with the same type of effector plasmid; uppercase letters indicate significant differences (P < 0.05) between different effectors combined with the same reporter variant.   Statistical analysis was performed using one-way ANOVA followed by Tukey's post hoc test. Letters indicate significant differences (P < 0.05) between the different genotypes. Psm (OD 600 of 0.005). After 8 hours, these were harvested for RNA extraction.

Figure legends
Transcript levels were normalized to transcript levels of UBQ5. Bars represent the average ± SEM of five to six plants of each treatment. Statistical analysis was performed using one-way ANOVA followed by Tukey's post hoc test for mock-and Psm-pretreated samples separately. Lowercase letters indicate significant differences (P < 0.05) between mock-pretreated samples; uppercase letters indicate significant differences (P < 0.05) between Psm-pretreated samples. qRT-PCR analysis of SARD1 transcript levels in wild-type (Col-0), tga1 tga4, tga2 tga5 tga6 and npr1 plants.
Four-week-old soil-grown plants were sprayed with water (mock) or 1 mM SA and tissue was harvested after 8 hours. Transcript levels were normalized to the transcript levels of UBQ5. Bars represent the average ± SEM of four to five biological replicates, each replicate representing three leaves from one plant. Statistical analysis was performed using one-way ANOVA followed by Tukey's post hoc test for mock-and SA-treated samples separately. Lowercase letters indicate significant differences (P < 0.05) between mock-treated samples; uppercase letters indicate significant differences (P < 0.05) between SA-treated samples. (a) Euler diagram of 2090 SA-inducible genes identified at 8 h after spraying with 1 mM SA as compared to water (mock) treatment in sid2. Different square sizes represent the number of genes with significantly different (log2 fold change (FC) ≥1 or log2 FC ≤ -1, P < 0.05) transcript levels in sid2 tga1 tga4 under either mock, SAinduced-or under both conditions. Sketches are drawn to visualize the expression pattern in the respective groups.
(b) Relative expression of selected genes as identified by RNAseq analysis of four-week old Arabidopsis sid2 and sid2 tga1 tga4 plants treated with water (mock) or 1 mM SA. Bars represent the average of Transcripts Per Kilobase Million (TPM) ± SEM of four biological replicates of each genotype, with each replicate representing three leaves of five plants of one independent experiment. Statistical analysis was performed using unpaired Student's t-test (two-tailed) for mock-and SA-treated samples separately. Lowercase letters indicate significant differences (P < 0.05) between mock-treated samples; uppercase letters indicate significant differences (P < 0.05) between SA-treated samples. qRT-PCR analysis of transcript levels of six TGA1/TGA4-modulated genes in wild-type (Col-0), tga1 tga4, tga2 tga5 tga6 and npr1 plants. Four-week-old plants were sprayed either with water (mock) or 1 mM salicylic acid (SA) and further incubated for 8 h. Transcript levels were normalized to transcript levels of UBQ5. Bars represent the average ± SEM of four to five biological replicates, each replicate representing three leaves from one plant. Statistical analysis was performed using one-way ANOVA followed by Tukey's post hoc test for mockand SA-treated samples separately. Lowercase letters indicate significant differences (P < 0.05) between mocktreated samples; uppercase letters indicate significant differences (P < 0.05) between SA-treated samples. (a) Relative luciferase (LUC) activities yielded by the DLO1 promoter as a function of co-expressed TGA1, TGA2 or TGA1 together with TGA2. The Prom DLO1 :fLUC reporter plasmid was transformed into Arabidopsis tga1 tga2 tga4 tga5 tga6 mesophyll protoplasts with either an "empty" effector plasmid or effector plasmids encoding TGA1 and/or TGA2 under the control of the UBQ10 promoter.
(b) Relative LUC activities yielded by the TGA-VP-activated DLO1 promoter as a function of the presence of the TGACGTCA element or the A-box. Reporter plasmids were transformed into Arabidopsis Col-0 protoplasts with either an effector plasmid encoding TGA1 or TGA2 fused to the activation domain of viral protein 16 (VP) under the control of the UBQ10 promoter, respectively, or a control plasmid encoding non-fused VP. WTwild-type Prom DLO1 :fLUC reporter sequence, mTGAreporter plasmid with mutated TGACGTCA motif, mA-boxreporter plasmid with mutated A-box.
Firefly LUC activities were normalized to Renilla LUC activities. LUC activity obtained from the wild-type DLO1 promoter in the presence of the respective control vector plasmids was set to 1. Values are means of four independently transfected batches of protoplasts (+/-SEM). Different letters indicate significant differences at P < 0.05 (one-way ANOVA followed by Tukey's post hoc test) for the various transfections in (a). In (b), statistical analysis was done using two-way ANOVA followed by Bonferroni's post hoc test: lowercase letters indicate significant differences (P < 0.05) between various reporter constructs combined with the same type of effector plasmid; uppercase letters indicate significant differences (P < 0.05) between different effectors combined with the same reporter variant.
Budimir et al. qRT-PCR analysis of transcript levels of four representative TGA1/TGA4-dependent genes in wild-type (Col-0) and sard1 cbp60g plants. Four-week-old plants were sprayed either with water (mock) or 1 mM salicylic acid (SA) and further incubated for 8 h. The experiment is part of the experiment shown in Figure 4 with Col-0 used as a common control. Transcript levels were normalized to transcript levels of UBQ5. Bars represent the average ± SEM of four to five plants of each genotype. Statistical analysis was performed using unpaired Student's t-test (two-tailed) for mock-and SA-treated samples separately. Lowercase letters indicate significant differences (P < 0.05) between mock-treated samples; uppercase letters indicate significant differences (P < 0.05) between SA-treated samples. (a) qRT-PCR analysis of transcript levels of four TGA1/TGA4-dependent genes after SA treatment of wild-type (Col-0) and tga1 tga4 plants complemented either with a control vector (contr.), a wildtype TGA1 genomic construct (TGA1) or a mutated TGA1 genomic construct carrying mutations in four critical cysteine residues (TGA1red). Four-week-old plants were sprayed either with water (mock) or 1 mM SA at 1 h after the subjective dawn and further incubated for 8 h. Transcript levels were normalized to transcript levels of UBQ5. Bars represent the average ± SEM of four to six plants of each genotype. Statistical analysis was performed using one-way ANOVA followed by Tukey's post hoc test for mock-and SA-treated samples separately. Lowercase letters indicate significant differences (P < 0.05) between mock-treated samples; uppercase letters indicate significant differences (P < 0.05) between SA-treated samples.
(b) Western blot analysis of protein extracts obtained from roots of the different plant genotypes as indicated in (a). TGA1 protein levels were detected using an anti-TGA1 antibody. Coomassie blue staining served as a loading control.
qRT-PCR analysis of transcript levels of SARD1 and DLO1 in wild-type (Col-0), tga1 tga4, tga2 tga5 tga6, npr1, sid2 and sard1 cpb60g plants. Three leaves of five-week-old plants were either MgCl 2 (mock)-infiltrated (a) or infiltrated with Psm (OD 600 of 0.005) (b) at 1 h after the subjective dawn. Two days later, three younger upper leaves were infiltrated with Psm (OD 600 of 0.005). After 8 hours, these were harvested for RNA extraction. Transcript levels were normalized to transcript levels of UBQ5. Bars represent the average ± SEM of three to four plants of each treatment. Statistical analysis was performed using one-way ANOVA followed by Tukey's post hoc test. Letters indicate significant differences (P < 0.05) between the different genotypes.
Budimir et al.  qRT-PCR analysis of transcript levels of TGA1/TGA4 target genes in wild-type (Col-0) and tga1 tga4 plants complemented either with a control vector (contr.), a wildtype TGA1 genomic construct (TGA1) or a mutated TGA1 genomic construct carrying mutations in four critical cysteine residues (TGA1red). Three leaves of fiveweek-old plants were MgCl 2 (mock)-or Psm-infiltrated (OD 600 of 0.005) at 1 h after the subjective dawn. Two days later, three younger upper leaves were infiltrated with Psm (OD 600 of 0.005). After 8 hours, these were harvested for RNA extraction. Transcript levels were normalized to transcript levels of UBQ5. Bars represent the average ± SEM of five to six plants of each treatment. Statistical analysis was performed using one-way ANOVA followed by Tukey's post hoc test for mock-and Psm-pretreated samples separately. Lowercase letters indicate significant differences (P < 0.05) between mock-pretreated samples; uppercase letters indicate significant differences (P < 0.05) between Psm-pretreated samples. Budimir et al. Fig. 9