Dual-Localized WHIRLY1 Affects Salicylic Acid Biosynthesis via Coordination of ISOCHORISMATE SYNTHASE1, PHENYLALANINE AMMONIA LYASE1, and S -ADENOSYL-L-METHIONINE-DEPENDENT METHYLTRANSFERASE1 1[OPEN]

Salicylic acid (SA) in ﬂ uences developmental senescence and is spatiotemporally controlled by various mechanisms, including biosynthesis, transport, and conjugate formation. Altered localization of Arabidopsis WHIRLY1 (WHY1), a repressor of leaf natural senescence, in the nucleus or chloroplast causes a perturbation in SA homeostasis, resulting in adverse plant senescence phenotypes. WHY1 loss-of-function mutation resulted in SA peaking 5 d earlier compared to wild-type plants, which accumulated SA at 42 d after germination. SA accumulation coincided with an early leaf-senescence phenotype, which could be prevented by ectopic expression of the nuclear WHY1 isoform (nWHY1). However, expressing the plastid WHY1 isoform (pWHY1) greatly enhanced cellular SA levels. Transcriptome analysis in the WHY1 loss-of-function mutant background following expression of either pWHY1 or nWHY1 indicated that hormone metabolism-related genes were most signi ﬁ cantly altered. The pWHY1 isoform predominantly affected stress-related gene expression, whereas nWHY1 primarily controlled developmental gene expression. Chromatin immunoprecipitation-quantitative PCR assays indicated that nWHY1 directly binds to the promoter region of isochorismate synthase1 ( ICS1 ), thus activating its expression at later developmental stages, but that it indirectly activates S-adenosyl- L -Met-dependent methyltransferase1 ( BSMT1 ) expression via ethylene response factor 109 (ERF109). Moreover, nWHY1 repressed expression of Phe ammonia lyase -encoding gene ( PAL1 ) via R2R3-MYB member 15 (MYB15) during the early stages of development. Interestingly, rising SA levels exerted a feedback effect by inducing nWHY1 modi ﬁ cation and pWHY1 accumulation. Thus, the alteration of WHY1 organelle isoforms and the feedback of SA are involved in a circularly integrated regulatory network during developmental or stress-induced senescence in Arabidopsis.

Salicylic acid (SA) is crucial for plant growth, responses to pathogens, programmed cell death, and environmental responses. Its content is temporally and spatially controlled by various mechanisms, including biosynthesis, transport, and conjugate formation. For example, leaf development in Arabidopsis (Arabidopsis thaliana) is regulated by SA biosynthetic/signaling genes. Early leaf senescence is a result of SA overproduction in mutant lines overexpressing isochorismate synthase (ICS1) and Phe ammonia lyase (PAL; Love et al., 2008;Rivas-San Vicente and Plasencia, 2011), whereas the hypersensitive response (a fast form of programmed cell death) has been intensively investigated in the S-adenosyl-L-Met-dependent methyltransferase mutant (bsmt1; Vlot et al., 2009). There are two main SA biosynthetic pathways in plants: the PAL pathway and the isochorismate (IC) pathway, which both depend on the primary metabolite chorismate (Dempsey et al., 2011). In the PAL pathway, the chorismate-derived 1 This work was supported by the National Natural Science Foundation of China (grant nos. 31770318 and 31470383), the National Science Foundation Fujian Provincial (grant no. 2016J01103), and the international exchange program of Fujian Agriculture and Forestry University (grant no. KXB16009A).
2 Senior authors. 3 Author for contact: ymiao@fafu.edu.cn. The author responsible for distribution of materials integral to the findings presented in this article in accordance with the policy described in the Instructions for Authors is: Ying Miao (ymiao@fafu.edu.cn. Y.M. designed the study; W.L. performed salicylic acid measurements, immunoblots, phenotyping, and reverse transcription quantitative PCR; D.H. performed chromatin immunoprecipitation sequencing and quantitative PCR; H.Z. generated plasmid constructs and promoter activation in addition to mutant screening; B.W. performed microarray data analyses; W.L. and Y.M. analyzed the data; Y.M. wrote the article; and D.S. and D.C. critically read the manuscript. [OPEN] Articles can be viewed without a subscription. www.plantphysiol.org/cgi/doi/10.1104/pp.20.00964 L-Phe is converted into SA via either benzoate intermediates or coumaric acid through a series of enzymatic reactions involving PAL, benzoic acid 2-hydroxylase (BA2H), and other uncharacterized enzymes (León et al., 1995b). The cytosolic PAL pathway produces ;10% of defense-related SA and in Arabidopsis, four PAL enzymes have been identified. In the IC pathway, chorismate is converted to SA via isochorismate in a two-step process involving isochorismate synthase (ICS) and isochorismate pyruvate lyase (IPL). In Arabidopsis, two ICS enzymes have been proposed to convert chorismate to ICS, but in recent studies another ICS was identified (Rekhter et al., 2019;Torrens-Spence et al., 2019). This pathway accounts for ;90% of the SA production generated by the plastid-localized ICS1 inducible by pathogens and UV light (Wildermuth et al., 2001). Endogenous SA undergoes a series of chemical modifications including hydroxylation, glycosylation, methylation, and amino acid conjugation. These modifications directly affect the biochemical properties of the SA derivatives and play a pivotal role in SA catabolism and homeostasis to regulate leaf senescence (Zhang et al., 2013). It has been shown that SA affects regulation of gene expression during leaf senescence (Morris et al., 2000;Vogelmann et al., 2012;Zhang et al., 2013) and plays a role in advancing flowering time in Arabidopsis (Martínez et al., 2004) as well as in inhibiting seed germination (Alonso-Ramírez et al., 2009;Lee et al., 2010). Although SA biosynthesis and its function in both local and systemic acquired resistance (SAR) against microbial pathogens and in plant development are well understood (Park et al., 2007;An and Mou, 2011), the underlying molecular mechanism of free SA homeostasis in cells is less clear. WHIRLY family proteins are localized to both the nucleus and organelles and perform numerous cellular functions in both locations (Krause et al., 2005;Grabowski et al., 2008). In the nucleus, WHIRLY1 (WHY1) protein was found to regulate the expression of genes related to defense and senescence by binding to their respective promoters (Desveaux et al., 2000(Desveaux et al., , 2004Xiong et al., 2009;Krupinska et al., 2013;Miao et al., 2013). For example, it binds to the promoter of WRKY53 and represses WRKY53 and WRKY33 expression in a developmental-dependent manner during early senescence in Arabidopsis (Miao et al., 2013;Ren et al., 2017), whereas it activates the HvS40 gene during natural and stress-related senescence in barley (Hordeum vulgare; Krupinska et al., 2013) and the PsbA gene in response to chilling treatment in tomato (Solanum lycopersicum; Kong et al., 2018). In the nucleus, WHY1 protein also modulates telomere length by binding to their AT-rich region (Yoo et al., 2007) and affects microRNA synthesis ( Swida- Barteczka et al., 2018). Moreover, in chloroplasts, WHY1 plays a role in organelle genome stability, facilitating accurate DNA repair (Cappadocia et al., 2010(Cappadocia et al., , 2012Lepage et al., 2013) and affecting RNA editing/splicing (Prikryl et al., 2008;Melonek et al., 2010). The intracellular localization of WHY1 and/or the developmental stage of the plant may contribute to its various functions . Furthermore, WHY1 has been reported to be involved in (a)biotic stress signaling pathways, e.g. in response to chilling (Kong et al., 2018), high light (Kucharewicz et al., 2017), nitrogen deficiency (Comadira et al., 2015), reactive oxygen species (Lepage et al., 2013;Lin et al., 2019), hormones such as SA and abscisic acid (Xiong et al., 2009;Isemer et al., 2012), and defense signaling, and it is required for SA-and pathogen-induced PR1 expression (Desveaux et al., 2005).
In this study, we extend the roles of the dual-localized WHY1 protein to SA biosynthesis, via regulation of PAL1 and ICS1 expression, and SA modification, by affecting BSMT1 gene expression, in a developmental-dependent manner. Moreover, cellular SA level affected the distribution and status of WHY1 protein in the nucleus and in plastids, suggesting a feedback mechanism to regulate SA content. Further, global analysis of gene expression in WHY1 loss-of-function mutants and a pWHY1 or nWHY1 gain-of-function mutant indicated that the levels of hormone metabolism-related genes were significantly altered. Our results provide evidence that the dual-localized WHY1 protein functions in both the nucleus and chloroplasts to finetune SA dynamics affecting plant development in Arabidopsis.

WHY1
Changes the Gene Expression Level of PAL, ICS, and BSMT1, and Alters SA Contents during Plant Development To explore how WHY1 is involved in the SA metabolism pathways (Fig. 1A), we used the why1-1 mutant previously deployed in several of our studies (Miao et al., 2013;Ren et al., 2017;Lin et al., 2019). Similar to the bsmt1 mutant (Vlot et al., 2009) and the SA 3-hydroxylase mutant (s3h; Zhang et al., 2013), this why1-1 mutant displays an early-senescence phenotype (Miao et al., 2013). We analyzed the expression levels of ICS, PAL, and BSMT1, encoding proteins with both benzoic acid (BA) and SA carboxyl methyltransferase activity and genes for SA glycoside/Glc ester modification enzymes, such as UGT71B1, UGT89B1, or UGT74F2 (Dempsey et al., 2011), in the why1 mutant compared to the wild type during plant development 28 to 42 d after germination (dag). Interestingly, transcript levels of PAL1 and PAL2 were increased at 37 dag, while those of BSMT1 were greatly decreased at 35 and 37 dag and those of ICS1 at 42 dag in response to WHY1 loss-of-function mutation (Fig. 1B), whereas transcript levels of UGT71B1, UGT74F2, UGT89B1, and S3H were not altered in the why1 mutant during plant development (Supplemental Fig. S1).
Thus, we tested whether SA contents also changed in the why1 mutant during plant development. The SA contents, including conjugated and free types of SA, of the why1 and wild-type plants were measured by HPLC assay during the period from 28 to 58 dag. Our results indicate that WHY1 loss-of-function mutation made both conjugated and free SA peak 5 d earlier (at 37 dag) than in the wild type (at 42 dag; Fig. 1, C and D).
In order to genetically confirm this hypothesis, we produced the why1pal1, why1sid2, why1pal1sid2, and why1bsmt1 double and triple mutants (Supplemental Fig. S2) and measured the SA contents in these mutants during plant development (Fig. 1E). Interestingly, the early SA peak disappeared in the why1pal1 line at 37 dag, showing a SA profile similar to that of the wild type, whereas SA accumulation in why1 mutants combined with bsmt1 mutation was not that strongly affected, displaying the same early-senescence phenotype as the why1 line. However, SA accumulation in why1 combined with sid2 (ics1) was inhibited at 42 dag. The whylpal1sid2 triple mutant showed a delayedsenescence phenotype and again had no earlier SA peak, with SA maintained at low levels even at 37 and 42 dag, suggesting that PAL activity is crucially important for SA accumulation during the early stages of plant development. Thus, we genetically confirmed that SA content in cells is affected by WHY1, predominantly via its effect on PAL1. Figure 1. Transcript levels of genes encoding key enzymes related to the SA metabolism pathway and SA content in the why1 line during plant development. A, The SA metabolism pathway in the cell. B, Transcript levels of genes encoding key enzymes related to SA metabolism in the why1 line during plant development. C, Content of conjugated (C-SA) and free (F-SA) SA in the wild-type (WT) and why1 mutant plants during the period 28 to 55 dag. D, Changes of conjugated and free SA content in a series of double mutants at 37 and 42 dag. Error bars represent standard deviation of triplicate experiments and statistical significance was checked by a two-way ANOVA. The different lowercase letters indicated the significance. E, Senescence phenotype of 37-d-old double mutants. The relative expression level was normalized to GAPC, and the wild type at 28 dag (B) was set as 1. The error bars represent the mean 6 SD of triplicate biological replicates and triplicate technical replicates. Asterisks in B and C indicate significant differences relative to the wildtype line according to pair-wide multiple t tests (*P , 0.05 and **P , 0.01). nWHY1/pWHY1 Affects the Gene Expression Levels of PAL1, ICS1, and BSMT1 As Well As SA Content during Plant Development WHY1 is known to be dual-localized to the nucleus and plastids (Grabowski et al., 2008). To clarify which isoform of WHY1 affects SA metabolism and its dynamics, we complemented the why1 mutant line with pWHY1, nWHY1, or pnWHY1 under 35S promoter control (Lin et al., 2019) and analyzed the subsequent transcript levels of PAL, ICS, and BSMT1 at 28 to 42 dag. Complementation with nWHY1 or full-length WHY1 (pnWHY1) restored transcript levels of PAL1, PAL2, and BSMT1 to near wild-type levels, whereas the nWHY1/why1 line had an even lower PAL1 expression level at 37 and 42 dag compared to the wild type. Surprisingly, complementation with pWHY1 resulted in a 2-fold increase in the transcript level of PAL1 and repressed the transcript level of BSMT1 at 37 dag, and also significantly increased the transcript level of ICS1 at 42 dag ( Fig. 2A). Measurement of SA contents in the complemented why1 mutant background from 28 to 42 dag showed that, until 37 dag, both nWHY1/why1 and pnWHY1/ why1 lines exhibited significantly restored wild-type SA accumulation compared to the why1 line, and that the SA content in the nWHY1/why1 mutant was even lower at 42 dag. However, pWHY1 significantly enhanced SA accumulation during the whole period of development ( Fig. 2B), indicating that nWHY1 somehow repressed SA accumulation via suppression of PAL1 expression. On the other hand, pWHY1 might enhance SA accumulation by repressing BSMT1 during early developmental stages and promoting ICS1 at later developmental stages.

WHY1" Transgenic Plants
In order to globally understand the differences and similarities in the transcriptome response following pWHY1 and nWHY1 expression, a microarray sequencing analysis was deployed. Phenotypic differences were observed in the short-term response and thus, to avoid long-term secondary artifacts caused by continuous expression, an estradiol-inducible promoter was used to generate "inducible compartmental WHY1" transgenic plants (VEX:pWHY1/why1 and VEX:nWHY1/why1) as described in Ren et al. (2017). We found that the WHY1 protein level increased about 14-fold following a 2-h induction with 20 mM estradiol . The total RNA isolated from the 35-d-old rosette leaves of inducible VEX:pWHY1/why1 and VEX:nWHY1/why1 plants before (0 h) and after estradiol application (2 h), as well as total RNA from why1 and wild-type plants, was used for transcriptome analysis by ATH1 Arabidopsis GeneChip microarrays with two biological replicates. Comparison of the transcriptome of inducible pWHY1 plants to that of noninducible pWHY1 plants revealed a complex genetic reprogramming with 1,165 and 4,560 transcripts being at least 2-fold up-and downregulated, respectively. Comparison of inducible nWHY1 plants to noninducible nWHY1 plants also revealed a complex genetic reprogramming with 920 and 3,965 transcripts up-and downregulated, respectively. Transcriptomic comparison of the why1 mutant to wild-type plants identified 4,432 and 1,190 transcripts up-and downregulated, respectively (Supplemental Fig. S3).
To visualize gene expression reprogramming in the VEX:pWHY1 VEX:nWHY1 and the why1 plants, their entire nuclear transcriptome was subjected to MapMan analysis, allowing identification of biological processes with significant alterations (Thimm et al., 2004). The hormone metabolism pathways were significantly overrepresented after induction of pWHY1 or nWHY1, or by WHY1 loss-of-function mutation, with auxin, jasmonic acid (JA), and ethylene metabolism, as well as SA metabolism, particularly affected ( Fig. 3; Supplemental Datasets S1-S4). The regulation of secondary metabolism and stress were also significantly enriched after induction of pWHY1 expression (Fig. 3A). These stresses were associated with biotic-and abiotic-stress responses related to redox imbalance, which were mostly upregulated by pWHY1 (Fig. 3A). By contrast, regulation of RNA, development, and signaling terms was significantly enriched after induction of nWHY1 expression. Given that the opposite regulation of signaling, development, RNA, and transport terms was observed in WHY1 loss-of-function mutant plants (Fig. 3A), these changes could be attributed to the inducible expression of pWHY1 or nWHY1 (Fig. 3A). Globally, a net enrichment of biological processes linked to hormone metabolism was found within the most significantly differentially expressed genes after induction of pWHY1 or nWHY1 or deletion of WHY1 (Fig. 3B). A net enrichment of biological processes linked to hormone metabolism, secondary metabolism, and photosynthetic stress was found within the most differentially expressed genes in the inducible pWHY1 line (Fig. 3B). A net enrichment of biological processes linked to RNA regulation, development, or signaling was found within the most differentially expressed genes in the inducible nWHY1 line (Fig. 3B). Finally, a net enrichment of biological processes linked to photosynthesis and signaling, development, or RNA regulation was found within the most differentially expressed genes in the why1 line (Fig. 3B).
Among the differentially expressed genes, 153 were common to the inducible pWHY1 and nWHY1 lines. Among these, 42 hormone-related genes exhibited upor downregulated expression in the pWHY1 or why1 lines, including metabolism-and signaling-related genes for SA, JA, IAA, and ethylene ( Fig. 3; Supplemental Datasets S2 and S4). The 24 most expressed or suppressed genes in pWHY1, nWHY1, or why1 plants, which encode key components of the SA metabolism pathway including ICS1, ICS2, PAL1, PAL2, UGT71B1, UGT89B1, UGT74F2, and BSMT1, as well as SA signalingrelated genes, or senescence/cell death-related genes, are shown in the heatmap (Fig. 3C).
WHY1 Directly Binds at the Promoter Region of ICS1 and Indirectly Affects PAL1 and BSMT1 Expression in a Developmental-Dependent Manner WHY1 was first reported as a transcription factor in the nucleus (Desveaux et al., 2000). To investigate whether WHY1 directly regulates ICS1, PAL1/PAL2, and BSMT1 gene expression, we analyzed our previous chromatin immunoprecipitation sequencing (ChIP-seq) dataset and the above microarray dataset and found that ICS1, MYB15, and ERF109 are direct targets of WHY1 ( Fig. 4A; Miao et al., 2013), whereas PAL1 and BSMT1 are not. A search for transcription factor binding motifs in promoter regions of ICS1, MYB15, ERF109, PAL1, and BSMT1 genes was conducted with PlantCARE (Lescot et al., 2002), which identified two w-boxes, six MYC elements, and four MYB motifs in the promoter of PAL1; 63ERE elements in the BSMT1 promoter ( Fig. 4B; Supplemental Fig. S3); and several GTNNNNAAT and AT-rich motifs in the ICS1, MYB15, and ERF109 promoters (Supplemental Dataset 5). In order to clarify the relationship among these motifs, we first confirmed WHY1 binding at the target genes by ChIP quantitative PCR (ChIP-qPCR) using leaf material collected 37 and 42 dag from plants expressing hemagglutinin (HA)-tagged WHY1 under its native promoter (P why1 :WHY1-HA), as described in previous work (Miao et al., 2013). The putative cis-elements found in WRKY53, ICS1, MYB15, ERF109, and WRKY33 promoters included several GTNNNNAAT or AT-rich motifs ( Fig. 4B; Supplemental Dataset S5) and were enriched 5-to 20fold (Fig. 4C). The regions containing GTNNNNAAT and AT-rich motives of MYB15, ERF109, and WRKY53 were enriched 10-to 15-fold at 37 dag, whereas fragments of ICS1 and WRKY33 could not be detected at 37 dag; high enrichment was observed in all of these genes at 42 dag (Fig. 4C). Furthermore, the expression levels of these genes were analyzed by reverse transcription qPCR (RT-qPCR) at 37 and 42 dag in why1 and wild-type plants. In the knockout-mutant background, WHY1 binding negatively correlated with gene expression of ERF109 at 37 dag and ICS1 at Genes analyzed were related to the SA metabolism pathway (A), SA content C-SA (B). and SA content F-SA (C) during plant development in pWHY1/why1, nWHY1/why1, and pnWHY1/why1 transgenic plants compared to the wild type over the period from 28 to 42 dag. The data represent triplicate biological replicates, with values shown as the means 6 SE. Asterisks indicate significant differences from the wild type within the respective conditions based on Student's t test (*P , 0.05 and **P , 0.01).
42 dag, and positively correlated with MYB15 expression at both 37 and 42 dag. Whereas WRKY53 expression was positively correlated in why1 plants at 37 dag, WRKY33 was upregulated at 42 dag. Thus, WHY1 appears to either exert negative effects on gene expression (WRKY53, WRKY33 and MYB15) or cause activation of its target genes, such as ERF109 and ICS1, depending on the developmental stage.
In order to further verify the activation or repression activity of WHY1, the promoter sequences of WRKY53, ICS1, MYB15, ERF109, PAL1, and BSMT1 were cloned into dual-luciferase (LUC) vectors and applied in a transient expression assay using Nicotiana benthamiana leaves (Hellens et al., 2005). In addition to measuring promoter activation or repression by WHY1, MYB15 and ERF109 were also included in the analysis to investigate indirect effects of WHY1 in the nucleus. The coding sequences of WHY1, MYB15, and ERF109 were cloned under the control of the Arabidopsis ACTIN1 promoter (ACTIN:WHY1-HA, ACTIN:MYB15-HA, and ACTIN:ERF109-HA; Fig. 5A) and coinfiltrated with the reporter vector to drive LUC expression (Hellens et al., 2005). We then measured the LUC/RENNILASE (REN) luminescence ratio (i.e. LUC/REN ratio) in infiltrated leaves. To assess any basal activation or repression of putative promoters, a mini-GAL4 promoter vector was used in each coinfiltration experiment as a control; the WRKY53 promoter was used as a positive control. The results showed that WHY1 activated the promoters of ICS1 and ERF109, but it repressed the promoters of MYB15 and WRKY53, displaying the opposite expression pattern of the why1 knockout plants (Fig. 4B). The transcription factors MYB15 and ERF109 were able to activate PAL1, PAL2, and BSMT1 gene expression (Fig. 5, B and C). Therefore, WHY1 directly activates ICS1 expression and indirectly affects PAL1 and PAL2 gene expression via MYB15 and BSMT1 expression via ERF109.

WHY1 and MYB15/ERF109 Regulate Leaf Senescence and ROS Accumulation
Since WHY1 is a repressor of plant senescence at the early stage of plant development (35-42 dag; Miao et al., 2013), we compared the phenotypes of the pal1, sid2, myb15, and erf109 mutants (Supplemental Fig. S4) with that of the why1 mutant to analyze whether the effects of WHY1 on SA metabolism influence senescence. pal1 and sid2 plants have already been reported to have a delayed-senescence phenotype, whereas oePAL1, oeSID2, and bsmt1 plants show an early senescence phenotype (Love et al., 2008;Vlot et al., 2009;Huang et al., 2010;Rivas-San Vicente and Plasencia, 2011). We analyzed all mutants with respect to a visible senescent yellow leaf ratio (Miao and Zentgraf, 2007) and reactive oxygen species (ROS) production by nitro blue tetrazolium chloride (NBT) and 39,39-diaminobenzidine (DAB) staining assays under normal growth conditions. The results showed that pal1, sid2, myb15, and erf109 each displayed visibly delayed senescence and less ROS production, in contrast to bsmt1 plants, which showed slightly earlier senescence and higher ROS accumulation, similar to why1 and pWHY1 plants (Fig. 6, A and B).
Furthermore, the transcript levels of the senescencerelated genes WRKY53, SAG12, SAG13, SAG101, and PAD4 were measured by RT-qPCR and represented as a heatmap (Fig. 6C). These genes were upregulated in why1 and pWHY1 plants, similar to their levels in PAL1-overexpressing (oePAL1) plants; however, the same genes were downregulated in pal1, myb15, and sid2, similar to their levels in nWHY1 plants (Fig. 6C). Interestingly, in the BSMT1-overexpressing (oeBSMT1) line, the transcript levels of senescence-related genes SAG12 and WRKY53 were upregulated whereas the transcript levels of SAG13 and SAG101 were downregulated, which was the reverse of the expression Figure 4. WHY1 activates/represses target gene expression. A, Enrichment profiles of WHY1 protein in five target genes, ERF109, MYB15, WRKY33, ICS1, and WRKY53, by ChIP-seq. B, Position of promoter motifs (GTNNNNAAT plus AT-rich) of WHY1-target genes. C, Enrichment folds of WHY1 at the promoters of target genes by ChIP-qPCR at 37 and 42 dag. D, Expression levels of target genes at 37 and 42 dag in the why1 mutant compared to the wild type (WT). Error bars represent the mean 6 SD of triplicate biological replicates. Asterisks indicate a significant difference from ACTIN as determined by two-tailed Student's t test (*P , 0.05 and **P , 0.01).
trend observed in bsmt1 and erf109 (Fig. 6C). However, the transcript level of PAD4 was upregulated in both bsmt1 and oeBSMT1. This indicates that BSMT1 is involved in alternative signaling pathways that function in developmental senescence or stress-related senescence.
SA Level Feedback Affects Distribution of the WHY1 Protein in Plastids and the Nucleus WHY1 is required for SA-and pathogen-induced PR1 expression (Desveaux et al., 2005). WHY1 distribution is affected by protein modification  and cellular H 2 O 2 level (Lin et al., 2019). To determine whether SA feedback affects WHY1 expression, we quantified WHY1 transcription by RT-qPCR in response to exogenous MeSA in wild-type plants treated for 1, 4, 6, and 8 h. Unexpectedly, MeSA treatment did not affect WHY1 expression level (Fig. 7A). Thus, MeSA treatment probably affects WHY1 protein function or distribution in plastids or the nucleus. Nuclear and plastid proteins isolated from 35-d-old wild-type rosettes after MeSA treatment for 1, 3, and 6 h were immunodetected with a specific monoclonal antibody against WHY1 (Supplemental Fig. S4; Lin et al., 2019), and antibodies against Histone 3 and PSII protein were used as markers for pure nuclear and plastidial protein preparations (Fig. 7, B and C; Supplemental Fig. S5). A water treatment served as control for MeSA application. Interestingly, the results indicated that during the MeSA treatment time course, WHY1 accumulation decreased in plastids and the Figure 5. Promoter activation assays using the LUC/REN system. A, Structures of activator and reporter constructs. B, Promoters of ICS1, MYB15, ERF109, WRKY53, and WRKY33 genes were coinfiltrated with a vector containing WHY1 under regulation of the ACTIN promoter. C, Coinfiltration of MYB15 and ERF109 with the PAL1, PAL2, ICS1, and BSMT1 promoters. Background promoter activity was assayed by coinfiltration with an empty vector of the same type. Shown are the means 6 se of six biological replicates. Asterisks denote statistically significant differences from the empty vector, calculated by Student's t test (***P 5 0.001).
nuclear isoform of WHY1 was altered in its status, with small nWHY1 (29 kD) levels slightly increasing while large nWHY1 (37 kD) levels decreased (Fig. 7, B and C). We investigated the 4-h time point in more detail to quantify these alterations. We captured and calculated three replicates of protein band signals using the software program Image J and statistically analyzed their significance, demonstrating that WHY1 accumulated significantly less in plastids, and that the small nuclear isoform of WHY1 was significantly increased, whereas large-isoform nWHY1 (37 kD) levels were significantly decreased after a 4-h MeSA treatment ( Fig. 7D; Supplemental Fig. S6). Thus, exogenous MeSA treatment affects WHY1 accumulation in plastids and alters the modification status of nWHY1 in the nucleus, similar to the effects of H 2 O 2 treatment (Lin et al., 2019). Furthermore, we analyzed WHY1 distribution between plastids and the nucleus under SA-deficient conditions. The nuclear and plastid fractions isolated from the single sid2 and pal1 mutants and the double sid2 pal1 mutant were subjected to immunoblotting using the WHY1-specific peptide antibody. The results demonstrated that pWHY1 in the sid2, pal1, and sid2 pal1 mutants significantly accumulated in plastids when compared to the wild type (Fig. 7, E and G). Accordingly, the large nWHY1 isoforms (37 kD) were highly abundant and the small nWHY1 proteins (29 kD) were less abundant in the pal1 and sid2 pal1 mutants, but not in the sid2 single mutant (Fig. 7, F and G). This indicates that the ICS1 pathway plays a prominent role in modification of nWHY1 protein.

DISCUSSION
It has become increasingly clear that dual location of proteins mediates diverse intercellular signaling processes, as described, for example, for MAP kinase (Bobik and Burch-Smith, 2015;Chan et al., 2016) and CIPK14 , as well as hormone (abscisic acid and SA; Koussevitzky et al., 2007;Caplan et al., 2015;Kacprzak et al., 2019) and ROS (hydrogen peroxidase and Figure 6. Phenotyping of WHY1 lossof-function mutants and mutants of WHY1 downstream target genes. A, Phenotypes of PAL1, ICS1, MYB15, and BSMT1 loss-of-function mutants at 37 dag compared to why1 mutants. Whole rosettes are shown at top, with a graph of the senescent leaf ratios in five plants. B, ROS accumulation of PAL1, ICS1, MYB15, and BSMT1 loss-of-function mutants compared to why1 mutants at 37 dag as determined by NBT and DAB staining. C, Transcript levels of SAG genes in the PAL1 and BSMT1 loss-or gain-of-function mutants and the MBY15, ERF109, and ICS1 loss-of-function mutants at 37 dag determined by RT-qPCR. The data represent triplicate biological replicates. Values are shown as means 6 SE, with the wild type set to 1 in the heatmap. singlet oxygen) signaling (Duan et al., 2019;Lin et al., 2019;Lv et al., 2019). Proteins with dual subcellular localization can affect transcription and display various functions in intracellular signaling (Sun et al., 2011;Woodson et al., 2011Woodson et al., , 2013Isemer et al., 2012;Nevarez et al., 2017;Lin et al., 2019;Pesaresi and Kim, 2019;Wu et al., 2019). This study revealed that dual-localized WHY1 protein directly activates ICS1 expression in the nucleus at later stages of plant development, whereas it indirectly controls PAL1 and BSTM1 expression via alteration of MYB15 and ERF109 transcription at early stages of plant development. Therefore, WHY1 influences cellular SA content during plant development. SA-level feedback in turn affects WHY1 distribution, causing a shift into the nucleus and preferential accumulation of the smaller 29-kD form. This loop of nWHY1 integrating SA dynamics via PAL1/ICS1 and BSMT1 plays a pivotal role in controlling leaf senescence.
SA is crucial for plant growth. During plant development, the first peak of SA appears at the onset of senescence (Buchanan-Wollaston et al., 2005). It was assumed that SA was a senescence trigger signal and relative to plant senescence (Morris et al., 2000). Alteration of SA dynamics resulted in remodeling of plant programmed cell death. Elucidation of SA biosynthesis and catabolism is important to understanding its biological functions. Ten percent of SA is synthesized from L-Phe via the PAL pathway in the cytoplasm, whereas 90% of SA is from chorismate via ICS1/SALICYLIC ACID INDUCTION DEFICIENT2 (SID2) in chloroplasts, the latter of which is responsible for the bulk of SA produced during pathogen infection in Arabidopsis (Dempsey et al., 2011). Endogenous SA can also undergo a series of chemical modifications including hydroxylation by salicylate hydroxylase (Yamamoto et al., 1965;Zhang et al., 2013), glycosylation by glycosyltransferases Figure 7. Changes in plastid and nuclear isoform WHY1 protein levels determined by immunodetection in the sid2, pal1, or sid2 pal1 mutants compared to the wild type. A, Expression level of WHY1 in wild-type plants after MeSA treatment for 1, 2, 4, 6, and 8 h. B, WHY1 immunodetection in nuclear extracts after MeSA treatment for 1, 3, and 6 h. C, WHY1 immunodetection in plastid extracts after treatment with MeSA for 1, 3, and 6 h. D and E, WHY1 immunodetection in nuclear and plastid extracts after MeSA treatment for 4 h (D) and in the sid2, pal1, or sid2 pal1 mutants compared to the wild type (WT; E). Coomassie and silver staining are shown as the protein amount loading controls. nWHY1-L, large size (37 kD) of nWHY1; nWHY1-S, small size (29 kD) of nWHY1. The antibody against peptide WHY1 was a commercial product. F and G, Quantification of the alteration of pWHY1 and nWHY1 after MeSA treatment for 4 h (F) and in the sid2, pal1, or sid2 pal1 mutants compared to the wild type (G). The protein band signals were captured and calculated by Image J software (http://www.di.uq.edu. au/ sparqimagejblots). Data show the ratio of means of three replicates normalized to histone or PSII. Asterisks indicate significant differences from the water treatment (F) and the wild type (G) as determined by Student's t test (*P , 0.05 and **P , 0.01). (Lim et al., 2002;Dean and Delaney, 2008), methylation by BSMT1 (Park et al., 2007), and amino acid or sugar conjugation (Zhang et al., 2007;Bartsch et al., 2010). The microarray data and RT-qPCR results showed that the gene expression levels of developmental-related transcription factors were upregulated, and those of stressrelated genes were downregulated in why1 plants ( Fig. 7; Supplemental Dataset S4). The expression levels of ICS1, PAL1, and BSMT1 were altered significantly in the why1 mutant during plant development (Fig. 1); this alteration could be rescued completely through complementation with nWHY1 and pnWHY1 (Fig. 4). As previously established, nWHY1 can directly bind to the promoters of many target genes, such as WRKY53, S40, Kenisin, and PR10a (Desveaux et al., 2005;Xiong et al., 2009;Krupinska et al., 2013;Miao et al., 2013), as well as MYB15, MYC1/2, ICS1, and several ERF family members, as shown in our WHY1 ChIP-seq dataset ( Fig. 2; Miao et al., 2013). Moreover, nWHY1 represses the expression of many downstream developmental-related target genes, such as WRKY53, WRKY33, MYB15, and TRANSPARENT TESTA GLABRA2 (TTG2; Fig. 7; Supplemental Dataset S1). However, it can also promote the expression of many stress-related genes, such as HvS40 (Krupinska et al., 2013), PR1 (Desveaux et al., 2005), redox responsive transcription factors (Foyer et al., 2014), ICS1, and ERF109 (Figs. 2 and 7; Supplemental Dataset S1). Several MYB family members can bind to the promoter of PAL1/PAL2 (Battat et al., 2019). Among these, MYB15 was shown to bind to the promoters of PAL1 and ICE1 by ChIP-qPCR. MYB15 mainly plays a virtual role in immunity and cold response (Chezem et al., 2017;Kim et al., 2017;Wang et al., 2019). Our results further confirm that MYB15 can activate PAL1 expression. ERF-binding cis elements are enriched in the promoter region of BSMT1. However, ERF109 as a target gene of WHY1, which was identified in our ChIP-seq dataset (Miao et al., 2013), was not found to bind to the promoter region of BSMT1, as shown in yeast one-hybrid and gelshift assays (Shi, 2018). By contrast, ERF109 can activate BSMT1 expression as shown by our LUC/REN transit assay (Fig. 4), thus supporting our ChIP-seq data. The erf109 and bsmt1 mutants accumulate high levels of anthocyanin in response to high light (Foyer et al., 2014); however, the underlying regulatory mechanism is currently unknown. Therefore, the balance module of nWHY1/MYB15-PAL1 and nWHY1/ERF109-BSMT1 at early developmental stages (37 dag) and of WHY1/ICS1 regulation at late developmental stages (42 dag) determines SA homeostasis during plant development. An imbalance of PAL1/BSMT1 activity at 37 dag in the why1 mutant and the repression of ICS1 at 42 dag may result in SA accumulation about 1 week earlier than usual during normal plant development. Thus, nWHY1 impacts SA homeostasis via mediating PAL1 or ICS1 and BSMT1 activity in cells during plant development. (Krupinska et al., 2013;Foyer et al., 2014) The WHIRLY family is considered to associate with retrograde signaling. Due to their dual location and function in the nucleus and plastids (Krause and Krupinska, 2009), it is assumed that WHIRLY1 can move from plastids to the nucleus (Isemer et al., 2012). The plastid isoform of WHIRLY1 affects miRNA biogenesis in the nucleus ( Swida- Barteczka et al., 2018). Previously, we showed that the WHY1 protein can be phosphorylated by CIPK14 kinase or oxidized by H 2 O 2 , leading to different subcellular localization in the nucleus or in plastids, respectively Lin et al., 2019). Here, we further show that WHY1 loss-of-function mutation results in SA production 5 d earlier during plant development, thereby accelerating plant senescence. Complementation with pWHY1 did not revert the SA-accumulation phenotype. On the contrary, pWHY1 further increased SA accumulation during plant development. Consistently, PAL1 expression is promoted and BSMT1 expression is repressed at 37 dag, whereas ICS1 is activated at 42 dag (Fig. 4). This phenomenon can be explained by two mechanisms: (1) H 2 O 2 is known to affect SA levels via the ICS1 pathway (León et al., 1995a;Dat et al., 1998;Chaouch et al., 2010;Guo et al., 2017) and recent data link pWHY1 to ROS production via PSI/PSII Lin et al., 2019). Thus, pWHY1 might increase the SA level at 42 dag by modulation of the ICS1 pathway via photosystem-induced ROS accumulation to cause an early-senescence phenotype.
(2) Coordination of SA dynamics by pWHY1 is feedback controlled by cellular SA levels (Isemer et al., 2012;Caplan et al., 2015); the WHY1 isoform changes from plastid to nucleus Figure 8. Working model of the senescence pathway for dual-localized WHY1 in response to SA. The nuclear isoforms of WHY1 are represented as both a large (37 kD [larger letters]) and a small molecular mass (29 kD [smaller letters]). WHY1 has a dual function in plastids and the nucleus. WHY1 loss-of-function mutation increases SA accumulation at an early developmental stage (37 dag) through increased PAL1 and repressed BSMT1 expression; elevated SA promotes nuclear WHY1 de-modification and promotes ICS1 and BSMT1 expression, thereby balancing SA homeostasis in the cells. High SA levels by ICS1 cause feedback enhancing ROS accumulation, thus promoting senescence. pWHY1 stimulates PAL1/ICS1 expression but represses BSMT1, allowing high levels of SA, also leading to early senescence. Thus, distribution of WHY1 organelle isoforms and the putative feedback of SA form a circularly integrated regulatory network during plant senescence in a developmental-dependent manner. Plastid (Chl) is shown as a green ovary and nucleus (Nuc) as a gray ovary. Lines indicate regulation, wide arrows indicate transfer or translocation, and broken lines indicate uncertainty.
repress MYB15 and PAL1 expression (Huang et al., 2010;Duan et al., 2019) and activate ERF109 and BSMT1 expression in response to stress cues such as high light (Estavillo et al., 2011). This demonstrates that dual-localized pWHY1/nWHY1 affects SA content most likely via connection with PSI/PSII-mediated ROS affecting leaf senescence.
The distribution of WHY1 between plastids and the nucleus depends on its modification status , as well as environmental cues or cellular signals such as H 2 O 2 (Lin et al., 2019) and SA (this work). Although the SA signal cannot promote CIPK14 expression (Lin, 2017), MeSA treatment feedback alters the nWHY1 protein status (to the 37-kD or 29-kD form; Fig. 7), similar to barley WHY1 (Grabowski et al., 2008) and nWHY1 after treatment with H 2 O 2 in Arabidopsis (Lin et al., 2019). The nature of modification resulting in both forms remains unknown. More interestingly, MeSA treatment was shown to reduce WHY1 accumulation in plastids, in contrast to H 2 O 2 treatment (Lin et al., 2019). These phenomena are further elucidated in SA-deficient mutants such as pal1, sid2, and the pal1 sid2 double mutant. The ICS1 loss-of-function mutation, sid2, decreases the modified state of the nWHY1 level, whereas the PAL1 loss-of-function mutation, pal1, increases it, and both of these loss-of-function mutations increase WHY1 accumulation in plastids. It is known that ICS1 is localized to plastids and is responsible for the bulk production of SA in response to salt or pathogens (Kumazaki and Suzuki, 2019). Plastid-derived SA can be transported from plastid to nucleus via stromules (Caplan et al., 2015). Combined with global microarray data, it is speculated that this kind of SA signal molecule might influence the nuclear isoform of WHY1, in which the small form (29 kD) activates stressrelated gene expression, such as S40 and ICS1 ( Fig. 7; Krupinska et al., 2013), whereas the large form (37 kD) represses gene expression, as shown for WRKY53, WRKY33, and MYB15 (Figs. 2 and 7;Miao et al., 2013). Furthermore, it has been reported that phosphorylation of WHY1 by CIPK14 promotes its binding affinity at the promoter of WRKY53 and WRKY33 and represses WRKY53 and WRKY33 expression , and that CIPK kinase expression level rapidly increases in response to salt or pathogen stress, accompanied by increasing Ca 21 , H 2 O 2 , and SA levels in the cells (Sardar et al., 2017). CONCLUSION We conclude that WHY1 exerts different functions in plastids and the nucleus. nWHY1 influences SA content by directly affecting ICS1 and indirectly affecting PAL1 and BSTM1 expression via MYB15 and ERF109. The pWHY1 isoform promotes PAL1/ICS1 expression and represses BSMT1, facilitating high SA accumulation and resulting in early senescence, as seen in bsmt1 mutants. Interestingly, MeSA treatment altered the nWHY1 status (increasing the 29-kD form of WHY1 while decreasing the 37-kD form) in combination with declining pWHY1 accumulation. These results indicate that pWHY1/nWHY1 distribution in the nucleus and chloroplast allows cellular balancing of SA and H 2 O 2 contents in a developmental-dependent manner, thereby affecting leaf senescence in Arabidopsis (Fig. 8).

Plant Materials
All Arabidopsis (Arabidopsis thaliana) mutants are in the ecotype Columbia background. The transfer DNA insertion lines why1 (Salk_023713), sid2, pal1, bsmt1 (SAIL_776_B10), and myb15 (myb15-1 [SALK_151976] and myb15-2 [SK2722]) were kindly provided by other scientists. The erf109 (SALK_150614) and ERF109 overexpression lines (CS2102255) were obtained from the Nottingham Arabidopsis Stock Center. Homozygous plants were selected and confirmed by PCR or RT-PCR using genomic DNA and complementary DNA (cDNA), respectively, as templates (Supplemental Fig S2); http://signal.salk.edu/ tdnaprimers.2.html). The remaining experimental lines used in this study were constructed in our lab, namely the nWHY1-HA-overexpressing line that produces WHY1 protein localized only to the nucleus, the pnWHY1-HA-overexpressing line that produces the WHY1 protein dual-localized to plastids and nucleus, the complemented PWHY1-HA (Pwhy1:pnWHY1-HA) line, and the pWHY1-HA line that harbors the construct containing full-length WHY1 with a nuclear export peptide sequence fused to an HA-tag and produces WHY1 protein localized only to plastids (Miao et al., 2013;Lin et al., 2019).
Seeds were germinated on wet filter paper followed by vernalization at 4°C for 2 d, then transplanted to vermiculite and grown in a climatic chamber (100 mE h 21 , 13 h of light at 22°C/11 h of dark at 18°C, and 60% relative humidity). The rosette leaves were labeled with colored threads after emergence, as described previously (Hinderhofer and Zentgraf 2001).
For MeSA treatment, rosette leaves were collected at 1, 2, 3, 4, 6, and 8 h after spraying with 100 mM MeSA and stored in liquid nitrogen at 280°C for later use in RNA or protein isolations. Mock treatments used distilled water.

Measurement of SA Content in Rosette Leaves
SA was extracted from 0.2 g of the fifth leaf from individual plants at different stages of development and measured by reverse-phase HPLC on an Agilent1260 system with a C18 column, as previously described (Verberne et al., 2002), albeit with small modifications: SA was thoroughly separated from the complex mixture by methanol containing 10% (w/v) sodium acetate (pH 6.0; Lin et al., 2016). Fluorescence detection (excitation at 305 nm and emission at 407 nm) was applied and 3-hydroxybenzoic acid was used as an internal standard (Aboul-Soud et al., 2004). Conjugated and free SA were detected at the same time. Three independent biological replicates were performed for each data point.

Staining of ROS
Visualization of H 2 O 2 accumulation in leaves was performed using the DAB staining method according to Huang et al. (2019). Detached rosette leaves were vacuum filtered in 20 mL staining solution containing 1 mg mL 21 DAB in 50 mM Tris-HCl (pH 5.0) for 10 min, and incubated in darkness at room temperature for 12 h. The leaves were destained by boiling in a mixture of ethanol, glycerol, and acetic acid (3:1:1 [v/v/v]) for 15 min before imaging.
Detection of superoxide free radicals were performed by the NBT staining method as described in Lee et al. (2002). Whole rosettes of 5-to 6-week-old plants were harvested and immersed in 0.1 mg mL 21 NBT solution (25 mM HEPES [pH 7.6]). After vacuum infiltration, samples were incubated at 25°C for 2 h in darkness. Subsequently, stained samples were bleached in 70% (v/v) ethanol and incubated further for 24 h at 25°C to remove the chlorophyll.
Imaging was conducted using an Epson Perfection V600 Photo scanner carried out using a Fermentas first-strand cDNA synthesis kit (Thermo Fisher Scientific) on RNA from 28-to 55-d-old plants grown under normal light conditions. cDNA was diluted 20-fold prior to qPCR. The Touch 1000 platform (Bio-Rad) was used for qPCR, and the data were analyzed using Bio-Rad software version 1.5. We used GAPC2 or ACTIN as internal reference genes for calculation of relative expression. Primers are listed in Supplemental Table  S1. All determinations were conducted in three biological replicates.

Isolation and Detection of Plastid and Nuclear Proteins
Chloroplasts and nuclei were prepared and purified as described previously . Approximately 10 mg of proteins of each fraction were separated on 14% (w/v) polyacrylamide gels, then transferred to nitrocellulose membranes. Immunodetection followed using specific antibodies against the WHY1 C-terminal peptide CASPNYGGDYEWNR (Faan; Supplemental  Fig. S4). To monitor the purity of the chloroplast and nuclear fractions, we used antibodies against the cytochrome b559 apoprotein A and histone H3, respectively (Cell Signaling; Supplemental Fig. S5; Lin et al., 2019). Immunoblot images were taken using the FluorChem Q machinery (Proteinsimple). Digital data of protein bands were directly captured and the signal intensity of the same area was calculated as the absolute mean by the Image J software program setup in the FluorChem Q machine. We adjusted all gel backgrounds to nearly the same level before capturing the protein band signals.

ChIP-qPCR Assay
Four-week-old rosettes of transgenic plants expressing Pwhy1:WHY1-HA to complement the why1 knockout-mutant background were used for sample preparations. The cross-linked DNA fragments 200 to 1,000 bp in length were immunoprecipitated by an antibody against the HA tag (Cell Signaling). Enrichments of the selected promoter regions of both genes were resolved by comparing the amounts in the precipitated and nonprecipitated (input) DNA samples, which were quantified by qPCR using designed region-specific primers ( Fig. 2; Supplemental Table S1). Material from the why1 mutant served as a mock control and was used for normalization to calculate the fold enrichment. The experiments were performed with triplicate biological replicates.

Cloning and Construction of Vectors
Two kilobytes upstream promoter sequence of the start codon of MYB15 and the ERF109, WRKY53, PAL1, PAL2, ICS1, and WRKY33 genes were PCR amplified with gene specific primers including emzyme restricted site sequence and then PCR products were restricted with KpnI and XhoI or XhoI and PstI, respectively, and subcloned into the pFLAP vector. The entire cassette was then excised with KpnI and AscI and cloned into the binary vector pBIN1.
For dual-LUC assays, promoter sequences were PCR amplified, digested with NcoI and KpnI, and cloned into the pGreenII 0800-LUC binary vector (provided by Roger P. Hellens). DNA constructs used for N. benthamiana agroinfiltration and for Agrobacterium tumefaciens-mediated plant transformation were constructed via Goldenbraid cloning (Sarrion Perdigones et al., 2013).
MYB15, ERF109, and WHY1 coding sequences were subcloned into a pUPD vector. In the dual-LUC assays, MYB15, ERF109, and WHY1 were in the 1a1 vectors, which are based on a pGREENII backbone. For generating the geneoverexpression construct, a coding sequence fragment was amplified and subcloned into pGEM-T Easy (Promega), excised with BamHI and SalI restriction enzymes, and then cloned under control of the CaMV-35S promoter into pFLAP, followed by restriction with PacI and AscI and ligation to the pBIN1 binary vector.

Dual-LUC Activity Assay
N. benthamiana plants were grown in climate-controlled rooms (22°C and 16 h/8 h of light/dark). Plants were grown until they had six leaves and then infiltrated with A.tumefaciens GV3101. Plants were maintained in the climatecontrolled rooms, and after 4 to 5 d, 1-cm discs were collected from the fourth and fifth leaves of each plant. Six biological replicates with their respective negative controls were used per assay. The experiment was performed as previously described (Hellens et al., 2005), with minor changes. A. tumefaciens was grown overnight in Luria-Bertani medium and brought to a final OD 600 of 0.2 in infiltration buffer. Coinfiltrated A. tumefaciens carried separate plasmids: 900 mL of an empty cassette or one that contains the transcription factor driven by the tomato 2-kb ACTIN promoter region, and 100 mL of the reporter cassette carrying one of the test promoters. Leaf discs were homogenized in 300 mL of a passive lysis buffer. 25 mL of a 1:100 dilution of the crude extract was assayed in 125 mL of LUC assay buffer, and LUC and REN chemiluminescence of each sample was measured in separate wells on the same plate. Relative light units were measured in a Turner 20/20 luminometer, with a 5-s delay and 15-s measurement. Raw data were collected and the LUC/REN ratio was calculated for each sample. Biological samples were pooled together and Student's t test was performed against a background control for each experiment as described in "Results". The entire experiment was repeated a second time under similar conditions to confirm the regulatory effect of transcription factors.

Microarray Analysis
Two biological replicates were sampled from leaves of wild-type, VEX:pWHY1/ why1, VEX:nWHY1/why1, and why1 plants, as in our previous study . Extracted RNA was then amplified and labeled using the standard Affymetrix protocol and hybridized to Affymetrix ATH1 GeneChips according to the manufacturer's guidelines (Katari et al., 2010). Statistical analysis of transcriptome data was carried out using Parke Genome Suite software (www.partek.com). Data preprocessing and normalization were performed using the Robust Microarray Averaging algorithm (Irizarry et al., 2003). Batch effects between the replicates were not found. Differentially expressed genes were identified by ANOVA according to false discovery rate, P 5 0.05, and at least a 2-fold change between the genotypes (Supplemental Datasets S1-S4).

Statistical Analysis
Quantitative data were determined by at least triplicate biological replicates and the statistical significance was analyzed using either two-way ANOVA or pair-wide multiple t tests, with GraphPad Prism software (version 7).

Supplemental Data
The following supplemental materials are available.
Supplemental Figure S1. Gene transcript levels in mutant plants analyzed by RT-qPCR.
Supplemental Figure S2. Transcript levels of ICS2, UGT71B1, UGT89B1, UGT74F2, and S3H in the why1 mutant compared to the wild type during plant development.
Supplemental Figure S4. Immunoblot detection to certify specificity of antibody against WHY1 peptide.
Supplemental Figure S5. Immunoblot detection to certify purity of nuclear protein and plastid protein extracts.
Supplemental Figure S6. Immunoblot detection of WHY1 distribution in the organelles after MeSA treatment.
Supplemental Table S1. List of primers used in this study.
Supplemental Dataset S1. List of genes with induced or noninduced altered expression in the nWHY1 line.
Supplemental Dataset S2. List of genes with induced or noninduced altered expression in the pWHY1 line.
Supplemental Dataset S3. List of genes with induced or noninduced altered expression in the pnWHY1 line.