Salicylic Acid Is Required for Broad-Spectrum Disease Resistance in Rice

Rice plants contain high basal levels of salicylic acid (SA), but some of their functions remain elusive. To elucidate the importance of SA homeostasis in rice immunity, we characterized four rice SA hydroxylase genes (OsSAHs) and verified their roles in SA metabolism and disease resistance. Recombinant OsSAH proteins catalyzed SA in vitro, while OsSAH3 protein showed only SA 5-hydroxylase (SA5H) activity, which was remarkably higher than that of other OsSAHs that presented both SA3H and SA5H activities. Amino acid substitutions revealed that three amino acids in the binding pocket affected SAH enzyme activity and/or specificity. Knockout OsSAH2 and OsSAH3 (sahKO) genes conferred enhanced resistance to both hemibiotrophic and necrotrophic pathogens, whereas overexpression of each OsSAH gene increased susceptibility to the pathogens. sahKO mutants showed increased SA and jasmonate levels compared to those of the wild type and OsSAH-overexpressing plants. Analysis of the OsSAH3 promoter indicated that its induction was mainly restricted around Magnaporthe oryzae infection sites. Taken together, our findings indicate that SA plays a vital role in immune signaling. Moreover, fine-tuning SA homeostasis through suppression of SA metabolism is an effective approach in studying broad-spectrum disease resistance in rice.


Introduction
Plants have a sophisticated defense system against pathogen invasion. A general defense is the pathogen-associated molecular pattern (PAMP)-triggered immunity (PTI), initiated by the recognition of PAMPs via pattern recognition receptors (PRRs) [1]. Conversely, a more specific defense is the effector-triggered immunity (ETI), activated by host resistance genes that recognize effectors secreted by pathogens to suppress PTI. Phytohormones, such as salicylic acid (2-hydroxy benzoic acid, SA) and jasmonic acid (JA), are important compounds in the immune signaling pathways. SA accumulation is elevated in many plants owing to pathogen infection, and it activates many downstream responsive genes, as well as genes encoding PAMP receptors and co-receptors [2].
SA is synthesized from shikimate through either the isochorismate or phenylpropanoid pathway [2,3]. Chorismate is converted to isochorismate by isochorismate synthase (ICS). Arabidopsis genome contains two ICS genes, responsible for 90% of SA synthesis under stimulation by pathogen challenge or UV radiation [4]. Isochorismate is conjugated with glutamate to form isochorismate-9-glutamate by PBS3 (avrPphB susceptible3) and is then converted into SA by EPS1 (enhanced susceptibility) or via spontaneous decomposition [5,6]. Phenylalanine, derived from chorismate, can be converted into SA through the phenylpropanoid biosynthesis pathway. Phenylalanine ammonia-lyase (PAL) catalyzes the conversion of Phe to trans-cinnamic acid (tCA), which is oxidized to benzoic acid (BA) by abnormal inflorescence meristem1 (AIM1), functioning as a beta-oxidation enzyme to shorten the side chain. BA is then hydroxylated into SA by an unidentified enzyme.
SA biosynthesis in rice is also estimated through the isochorismate and phenylpropanoid pathways, in which OsPALs catalyzing the first step of the phenylpropanoid route have been characterized to be important for disease resistance [3]. In the rice aim1 mutant, accumulation of SA and its associated reactive oxygen species (ROS) in roots is reduced, leading to retardation of root growth [24]. Double knockdown of OsWRKY62 and OsWRKY76 (dsOW62/76), two transcriptional repressor genes, increases the SA and JA levels and broad-spectrum disease resistance, and partially compromises the rice aim1 phenotype [24,25]. Rice contains high basal levels of SA compared to those of Arabidopsis and tobacco [26]. Interestingly, decreasing endogenous SA by expressing the bacterial salicylate hydroxylase NahG gene revealed that SA plays an important role in protecting rice plants from oxidative stress [26]. However, the biological importance of endogenous SA in rice remains elusive. In this study, we characterized four rice SA hydroxylase genes (OsSAHs) and showed that endogenous SA accumulation is crucial for rice disease resistance.

OsSAH Proteins Have SA 3-and/or 5-Hydroxylase Activity In Vitro
OsSAH recombinant proteins purified from Escherichia coli were tested for enzymatic activity using SA, BA, tCA, or caffeic acid as substrates. Hydroxylase activity was only detectable on the SA substrate analyzed using LC-MS/MS, suggesting high substrate specificity of OsSAH proteins. OsSAH1-4 proteins had both SA 3-(SA3H) and 5-hydroxylase (SA5H) activity except of OsSAH3, which converted SA to 2,5-DHBA only. The biochemical parameters of OsSAH1-4 proteins are listed in Table 1. Each OsSAH protein showed higher Kcat and Kcat/Km values for SA5H than for SA3H activity, suggesting that the enzyme favored the conversion of SA to 2,5-DHBA in vitro. In addition, OsSAH proteins exhibited substrate SA inhibition, showing a parabolic velocity curve ( Figure 1). OsSAH3 was used to determine the optimal enzymatic conditions owing to its large Km value, and it showed an optimal pH of approximately 6.8 and a temperature of 40 • C under the assay conditions ( Figure S1).

Key Amino Acids of SAHs Affect Enzymatic Activity
OsSAH2 and OsSAH3, localized tail-to-tail in chromosome 4, are probably gained by duplication; however, the two resulting proteins showed different enzymatic activity and specificity (Table 1). To gain information on substrate specificity, we used 3-and 5-position-occupied SA derivatives and 2-Cl-BA as substrates to check the products formed by OsSAH2 and OsSAH3 enzymes. OsSAH2 catalyzed both 3-Cl-SA and 5-Cl-SA to produce the corresponding 3-Cl-5-OH-SA and 3-OH-5-Cl-SA although in very low amounts, whereas OsSAH3 only showed a 5-hydroxylation product ( Figure S2). When 2,3-DHBA, 2,5-DHBA, and 2-Cl-BA were used separately as substrates, hydroxylation products were not detected for OsSAH2 or OsSAH3 proteins. These results suggest that OsSAH2 and OsSAH3 bind restrictively with substrates for product specificity and that nucleophilic substitution on the benzene moiety favors hydroxylation.
Sequence alignments of rice SAHs with F6'H1 revealed amino acid differences around the substrate-binding sites a and Figure S3). The conserved RxS site is associated with 2OG, whereas the Phe residue (three amino acids next to RxS motif) and the upstream Tyr (designated as Y site for simplicity) in F6'H1 interacts respectively with the ferrule ring via π stacking and the thiocarboxylate group of feruloyl-CoA [18]. We estimated the amino acids near the substrate-binding sites that may affect the SAH enzyme activity and specificity. Three amino acids were exchanged between OsSAH2 and OsSAH3 at the corresponding positions. Each mutation in OsSAH3 (OsSAH3 F136Y , OsSAH3 I289M , and OsSAH3 P292A ) significantly suppressed the conversion of SA to 2,5-DHBA compared with the wild type OsSAH3 (Figure 2b). The production of 2,5-DHBA was further reduced in the double mutant OsSAH3 F136Y/P292A ( Figure 2b); however, 2,3-DHBA was produced at the level similar to that of OsSAH2, suggesting that the two substituted amino acids favor 2,3-DHBA production. Exchanges of all three amino acids (OsSAH3 F136Y/I289M/P292A ) significantly elevated 2,5-DHBA production compared with OsSAH3 F136Y/P292A and maintained 2,3-DHBA formation ( Figure 2b).
In contrast to the mutation at site I of OsSAH3 that lowers 2,5-DHBA production, we noticed that OsSAH2 M282I yields a higher amount of 2,5-DHBA than that of OsSAH2 and completely inhibited 2,3-DHBA production, suggesting that the amino acid with a short side chain at site I of OsSAH2 prefers 2,5-DHBA formation, which was confirmed by the mutation of Met to Ala (OsSAH2 M282A ). Interestingly, OsSAH2 Y127F mutant showed a similar level of SA5H activity as that of OsSAH2 but without SA3H activity ( Figure 2c). In contrast, OsSAH2 A285P mutant produced only a low amount of 2,3-DHBA product compared with the wild type protein, whereas further amino acid substitutions on the OsSAH2 A285P background did not lead to detectable SA3H and SA5H activity (Figure 2c), implying that the A 285 residue in OsSAH2 is a key amino acid for the enzymatic activity and specificity.
Arabidopsis DMR6 protein is more similar to OsSAH2 than to OsSAH3, but shows only SA5H activity, as determined by the HPLC method [12] (Figure 2a). Interestingly, SA5H activity was completely absent in the DMR6 A283P mutant, similar to the corresponding OsSAH2 mutant, and SA3H activity was markedly suppressed, as determined by the LC-MS/MS method (Figure 2d). The DMR6 Y126F/A283P double mutant produced undetectable products (Figure 2d). OsSAH2 (c), DMR6 (d), and their mutants were shown. The activity of the wild type protein was set as 1 and the mutant activity was compared with its own wild type, unless otherwise indicated. The 2,3-DHBA produced by OsSAH3 mutants ((b), the number in italic) was in related to that produced by OsSAH2 (c). In the case of DMR6 (d), the 2,5-DHBA activity was compared with OsSAH3, and the 2,3-DHBA activity was compared with OsSAH2. Values are given as means ± SD of three biological replicates. Asterisks indicate statistically significant differences compared with the wild type protein (Student's t-test; *, p < 0.05; **, p < 0.01; ***, p < 0.001). nd for not detection.
Structural simulation showed that OsSAH3 gives only docking conformation for 2,5-DHBA production ( Figure S4a). The OsSAH3 F136Y mutation changed the orientation of the R 138 side chain and hindered its hydrogen bond with 2-OH of SA, increasing the distance between iron and the C5 position of SA (Fe-C5, for simplicity) and SA docking energies, whereas OsSAH3 P292A substitution led to SA position shift and slightly increased the Fe-C5 distance ( Figure S4a,b). The OsSAH3 F136Y/P292A double mutant provided SA docking in two conformations, but with higher energies and longer distances between iron and the potential hydroxylation positions than those of OsSAH3 ( Figure S4a,b), which is consistent with the production of 2,3-and 2,5-DHBAs ( Figure 2b). Both OsSAH2 and DMR6 bound with SA in two conformations in which the 2,5-DHBA product is preferred, as estimated from the docking energies and Fe-C5 distances ( Figure S4d,e). The higher activity of DMR6 than that of OsSAH2 was also supported by these parameters.

Patterns of OsSAH Expression and Phenolic Accumulation
OsSAH gene expression was observed in different rice tissues (Figure 3a). OsSAH1 and OsSAH4 expression levels were generally higher than those of OsSAH2 and OsSAH3 in the same tissues tested, and the transcriptional level of OsSAH3 in the root was the lowest. As the substrate of OsSAHs, applied SA strongly induced the expression of OsSAH1-4 and SA-inducible marker genes, i.e., OsWRKY45 and OsWRKY76 (Figure 3b). A similar induction of these genes was observed following treatment with benzothiadiazole S-methyl ester (BTH), an SA functional analog ( Figure S5). Interestingly, the expression of OsSAH1 was also markedly induced by methyl jasmonate (MeJA) along with the JA-responsive genes OsMYC2 and OsTPS19 (Figure 3c) [27], implying that JA potentially participates in the regulation of SA homeostasis.
Determination of phenolics showed that free SA content was the highest in young leaves among the tissues tested, and the accumulation pattern of SAG was quite similar to that of SA in the same tissues ( Figure 3d). The content of free 2,5-DHBA was much higher than that of 2,3-DHBA in the same tissues examined (Figure 3e).

Knockout OsSAH2-3 Genes Enhance Resistance against Hemibiotrophic and Necrotrophic Pathogens
OsSAH2 and OsSAH3 were selected for generation of overexpressing and knocking out plants ( Figure S6 for gene knockout information). The transgenic and control plants were inoculated with the hemibiotrophic Magnaporthe oryzae SZ, a virulent rice blast fungus strain, by foliar spraying of spores on three-week-old plants. Overexpressing OsSAH2-3 (CDU::SAH2-3, controlled by a ubiquitin promoter) plants were more susceptible to the SZ strain than ZH17 control plants, whereas the OsSAH-knockout mutants showed enhanced resistance to the pathogen ( Figure S7). We also tested blast resistance at the rice tillering stage by injecting SZ spores into rice sheaths. Plants overexpressing each OsSAH gene showed increased lesion length compared to those of ZH17 control (Figure 4a). In contrast, enhanced resistance against the rice blast pathogen was observed in the mutants with single and double knockout OsSAH (sahKO) genes. Furthermore, we examined whether OsSAH genes were involved in resistance to the hemibiotrophic Xanthomonas oryzae pv. oryzae (Xoo) J18 strain, which causes the bacterial leaf blight. The lesion lengths were significantly shorter in the sahKO plants and longer in the overexpressing lines than in the ZH17 control (Figure 4b), suggesting that the suppression of SA metabolism also increases resistance to the bacterial pathogen.
It is often reported that SA promotes resistance against biotrophic and hemibiotrophic pathogens, and that JA is a key player in facilitating the activation of plant defense against necrotrophic pathogens [28]. Bipolaris oryzae is a necrotrophic fungal pathogen that causes the rice brown spot. We observed that OsSAH-overexpressing plants exhibited more severe disease symptoms than ZH17 control plants at the tillering stage under natural infection conditions; moreover, the knockout mutants of OsSAH2-3 showed elevated resistance against B. oryzae (Figure 4c and Figure S8a). Afterwards, the transgenic and control plants were inoculated with the spores of B. oryzae LW isolate. The responses of the OsSAH2-3 transgenic plants were similar to those caused by a natural infection of B. oryzae, revealing that OsSAHs negatively regulated disease resistance against the brown spot pathogen ( Figure S8b). , and MeJA (c). Roots and young leaves were from ten-day-old seedlings cultured in 1/2 MS liquid medium. Flag leaves, culms, and panicles were from plants of three weeks post heading grown in the paddy field. Ten-day-old seedlings cultured hydroponically were treated with 500 µM SA or 100 µM MeJA in 5 mM MES buffer for 6 h. For the mock treatment, the seedlings were received the same volume of DMSO solvent. Gene expression was determined by qRT-PCR using OsUBQ as the reference gene. (d) SA and SAG, (e) 2,3and 2,5-DHBA. Values are means ± SD (n = 3). Asterisks indicate statistically significant differences compared with the corresponding mock using student's t-test (**, p < 0.01; ***, p < 0.001). Columns marked with different letters (a-c) indicate significant differences among the same tissue (a), and the same compound (d), as analyzed by the SPSS software (Duncan's multiple range test, α = 0.05).
To further assess the effects of SA homeostasis on necrotrophic pathogens, we inoculated some of the transgenic plants with Rhizoctonia solani, a soil-borne pathogen with a necrotrophic lifestyle that causes the sheath blight disease. Lesion lengths caused by the challenge of R. solani XN strain were markedly reduced upon knockout of OsSAH3 (sah3KO), and both OsSAH2 and OsSAH3 (sah2&sah3KO) plants, whereas the lesion lengths were increased in the CDU::SAH3-and 35S::SAH3-GFP-overexpressing plants (Figure 4d). Furthermore, the plants harboring OsSAH3-overexpressing constructs were infected with a weak virulent uninucleate Rhizoctonia JN isolate [29]. OsSAH3-overexpressing plants were more susceptible to JN isolate than the ZH17 control plants ( Figure S9). These results collectively indicate that decreased SA levels elevate rice susceptibility to both hemibiotrophic and necrotrophic pathogens.  length (a,b), and lesion number (c) distributions. Experiments were biologically repeated twice with similar results p value evaluated using the student's t-test is above the boxplot. Values (d) are means ± SD (n = 5). Significance was evaluated using the student's t-test (*, p < 0.05; **, p < 0.01). Prefix CDU for overexpressing gene; suffix KO for knockout gene; and ZH17 for wild type plant. Bar = 1 cm (a,d), 5 cm (b,c).

Changes of OsSAHs Expression Alter SA Homeostasis
Expression of OsSAH2 and OsSAH3 was increased in the double knockout (sah2&sah3KO) and the single-knockout mutants, and their expressions were dramatically elevated in its own overexpressing lines (Figure 5a,b). The content of SA, as the substrate of Os-SAHs, was remarkably reduced in OsSAH-overexpressing plants and increased in sahKO mutants (Figure 5c). The accumulation of 2,5-DHBA product was increased in OsSAHoverexpressing lines, whereas the content change of 2,3-DHBA was minor (Figure 5d). The increase in SA accumulation and OsSAHs expression in the sahKO mutants revealed that a new SA homeostasis was reached. . Asterisks indicate statistically significant differences compared with ZH17 using the student's t-test (*, p < 0.05; **, p < 0.01; ***, p < 0.001). Prefix CDU for overexpressing gene; suffix KO for knockout gene; and ZH17 for wild type plant. Words underlined, detection of the mutant transcript; words in bold, detection including the exogenous transcript.
JA and its bioactive jasmonoyl-L-isoleucine (JA-Ile) play important roles in rice disease resistance [30,31]. The contents of JA and JA-Ile were substantially higher in OsSAHknockout and lower in OsSAH-overexpressing plants than in the ZH17 control plants (Figure 5e), showing synchronous changes with the SA levels. The induction of phytoalexin sakuranetin is considered to be tightly associated with JA accumulation [31]. Sakuranetin was detected at low levels in the knockout plants under normal growth conditions (Figure 5f), consistent with the increase in jasmonate accumulation.

Restricted Induction of OsSAH3 Promoter by M. oryzae
The results showed that SA is required for disease resistance, but SA accumulation is often observed without large changes in pathogen-infected tissues [26]. To determine OsSAH3 response to M. oryzae infection, we inoculated OsSAH3-promoter plants (SAH3pro::GUS-2) with M. oryzae SZ strain by infiltration. GUS was strongly stained at greenish-infected sites with high GUS density around the lesion borders (Figure 6a). Similarly, GUS staining was mainly around the necrotic lesion borders in SAH3pro::GUS-2 plants 7 d post the spray inoculation ( Figure 6b). The results suggest that the expression of SA-responsive genes, such as OsSAH3, occurs in a spatiotemporal manner in response to M. oryzae infection. Differences in induction of OsSAH1-4 by M. oryzae were observed at the time period tested, along with the expression changes of OsPAL1 and OsPAL4 participated in SA biosynthesis, and OsPR1b (Figure 6c).
Conversely, the high induction of SA-responsive genes imply an increased accumulation of SA or other active chemicals. We measured compound changes after M. oryzae inoculation, and we found that SA and SAG levels were elevated weakly, accompanied by marked increase in jasmonate in sah3KO, sah2&sah3KO, and ZH17 plants 24 h post inoculation (Figure 6d), whereas CDU::SAH3 plants showed negligible changes in SA and JA.

Knockout OsSAH Genes Elevate Basal Resistance
To obtain more information about SA effects on rice disease resistance, we investigated the early infection process of the GFP-labeled M. oryzae SZ strain (SZ-GFP) in the transgenic and control plants. Under visual observation, CDU::SAH3-11 plants presented more germinated spores than sah3KO-1 mutant 24 h post infection of the SZ-GFP conidia into the leaf sheaths (Figure 7a,b). Consequently, the germinated conidia invaded more into the neighboring cells in CDU::SAH3-11 than in sah3KO-1 plants 48 h after the infection, suggesting that the decrease in SA favors M. oryzae spore germination and spread.
SA and JA have priming effects that enhance plant disease resistance [32]. The elevated levels of SA and JA in OsSAH-knockout plants prompted us to investigate the defense responses in transgenic plants. The chemiluminescence signal, representing ROS accumulation, induced by chitin was much stronger in the sah3KO-1 plants than in the ZH17 control and CDU::SAH3-11 plants, in which CDU::SAH3-11 showed slight ROS induction (Figure 7c).
OsWRKY45 and OsWRKY76 are SA-responsive genes. The basal levels of OsWRKY45 and OsWRKY76 expression were significantly higher in the knockout mutants and lower in OsSAH3-overexpressing plants compared with ZH17 control plants (Figure 7d). After chitin treatment, OsWRKY45 and OsWRKY76 transcription levels remained markedly higher in the knockout mutants than in the overexpressing-and control plants (Figure 7d). Similarly, OsAOS2, a JA biosynthetic gene, and OsTPS19, a JA-responsive gene encoding limonene synthase, were induced more strongly by chitin in the knockout mutants than in ZH17 and CDU::SAH3-11 plants (Figure 7d). In summary, knockout of OsSAHs elevated the accumulation of SA and jasmonate and caused a priming effect on gene expression, leading to prevention of pathogen infection at an early stage.  Induction of ROS was detected with a chemiluminescence assay using luminol as the substrate. Values represent means ±SD (n = 3). (d) Induction of the defense-related genes by chitin. Ten-day-old rice plants, cultured hydroponically in 1/2 MS, were treated with chitin for 1 h and sampled for total RNA isolation. Gene expression was determined by qRT-PCR using OsUBQ as the reference gene as described in Figure 3. Results from a representative experiment are shown. Experiments were biologically repeated twice with similar results. Asterisks indicate statistically significant differences compared with the corresponding mock using student's t-test (**, p < 0.01; ***, p < 0.001). Columns marked with different letters (a-d) indicate significant differences, as analyzed by the SPSS software (Duncan's multiple range test, α = 0.05).

Discussion
SA is a key signaling molecule involved in plant defense responses against pathogens. In contrast to the induction of SA in Arabidopsis and tobacco, the high basal levels of SA in rice cultivars are noteworthy for their role in plant defense and development. SA can be metabolized in both reversible and irreversible ways to maintain SA homeostasis [12,15,33]. In this study, overexpression of OsSAH2-3 decreased SA content, whereas the knockout of OsSAHs increased SA accumulation (Figure 5c), suggesting that a new SA homeostasis with a high SA content is reached in the knockout mutants. Elevated levels of SA are associated with hypersensitive responses in lesion-mimicking mutants [34]. Meanwhile, SA accumulation was dramatically high in some autoimmune mutants with no spontaneous lesion formation, such as suppressor of npr1-1, constitutive 1 (snc1), indicating that high levels of SA alone are not sufficient to activate cell death [35]. Moreover, SA functions as a priming agent to potentiate ROS production and defense-related gene expression in soybean cells [36]. Increased SA and jasmonate levels in OsSAH-knockout plants led to increased ROS accumulation and defense-related gene expression induced by chitin, whereas overexpression of OsSAH3 alleviated defense reactions, including ROS production ( Figure 7c). Consequently, we observed inhibition of M. oryzae spore germination and invasive hypha spread into neighboring cells in sah3KO-1 compared with OsSAH3-overexpressing and wild type plants (Figure 7a,b). In rice aim1 mutant with defective root growth, reduced SA levels are associated with decreased ROS production in the roots [24]. Conversely, NahG rice shows reduced endogenous SA but increased levels of ROS and spontaneous lesion formation [26]. Therefore, NahG rice plants are hyperresponsive to oxidative damage induced by M. oryzae or paraquat treatment. Since phenolic catechol may trigger ROS generation, the susceptibility to oxidative damage of NahG rice plants may be due to increased titers of catechol and, hence, ROS intermediates in the plant [37].
Cross-talk between SA and JA signaling pathways has been demonstrated to occur in multiple layers of regulation. Generally, SA signaling plays a crucial role in resistance against biotrophic and hemibiotrophic pathogens, whereas JA signaling contributes to defense against necrotrophic pathogens and herbivorous insects [28]. Studies have shown the antagonistic effects of SA and JA pathways in Arabidopsis, as SA application suppresses JA biosynthetic genes [38]. Conversely, the bacterial Pseudomonas-secreted phytotoxin coronatine, acting as a JA-Ile mimic, activates three NAC transcription factors, which suppress ICS1 and increase the SA methyl transferase gene BSMT1, leading to reduced SA levels [39]. However, the rice oshpl3 mutant, which encodes a hydroperoxide lyase, shows overproduction of JA and SA and enhanced resistance to the bacterial blight pathogen Xoo [40,41]. Recently, the RESISTANCE OF RICE DISEASE1 mutant has been shown to contain elevated SA and JA levels, which confers resistance to multiple pathogens in rice [42]. The knockout of OsSAH genes elevated the basal levels of SA and jasmonate and conferred resistance against both hemibiotrophic and necrotrophic pathogens (Figures 4 and 5). On the contrast, overexpression of OsSAH3 compromised the induction of JA and JA-Ile by chitin treatment. The results suggest that a synergistic effect of SA and JA exists under certain conditions in rice [43]. Jasmonate is required for defense against the hemibiotrophic pathogen M. oryzae [21], and application of JA after M. oryzae infection can alleviate the disease symptoms [44]. It is reasonable to expect that the activated JA signaling pathway in sah2-3KO plants positively regulates the resistance to necrotrophic pathogens, such as R. solani and B. oryzae (Figure 4c,d, Figures S8 and S9). Furthermore, DELLA proteins have opposite functions against different lifestyles of invading pathogens in Arabidopsis and rice. In Arabidopsis, DELLA modulates immunity by promoting JA signaling and antagonizing SA, thus increasing resistance to the necrotrophic fungus Alternaria brassicicola and susceptibility to the hemibiotrophic bacterium Pseudomonas syringae pv. tomato [45]. In rice, DELLA Slender Rice1 (SLR1) acts as a positive regulator of hemibiotrophic resistance by integrating and amplifying SA-and JA-mediated defense signaling [46]. Collectively, synergistic SA-JA interactions in rice may provide a common defense pathway for broad-spectrum disease resistance.
The synergistic interaction between SA and JA has also been shown during ETI in Arabidopsis as a unique interplay [25,47]. JA accumulates at high levels when ETI is evoked and contributes positively to ETI. Induction of JA biosynthesis following SA accumulation is activated through the SA receptors NPR3 and NPR4, which mediate the effect by promoting degradation of the JA transcriptional repressor JAZs [25]. Alternatively, Betsuyaku et al. [48], using a time-lapse imaging assay of defense-related gene promoters, showed that the JA signaling pathway is activated in the vicinity of the central SA-active cells during ETI, indicating the spatiotemporal dynamics of the SA-JA relationship. Restricted inductions of OsSAH3-promoter activities by M. oryzae infection suggests that SA and JA levels may be elevated around the infection foci in a pattern similar to that of ETI in Arabidopsis (Figure 4) [48]. The localized activation of the promoter activities might provide an explanation for the observation of weak or no change in SA levels in rice shoots following pathogen infection (Figure 4) [26]. The increased level of SA in the infection sites is probably buffered by the large basal SA pool.
There were three and four SAHs in Arabidopsis and rice genomes, respectively ( Figure S3). The DMR6 mutant was isolated as a mutant with enhanced resistance to downy mildew, and the DLO1/S3H1 mutant is associated with senescence [9,11]. DMR6 and DLO1/S3H1 show differences in their spatial expression patterns in Arabidopsis leaves infected with H. parasitica, whereas DLO2 is not expressed in leaves [10]. SA is over accumulated in the dmr6 dlo1/s3h1 double mutant compared with the dmr6 mutant, displaying immunity to pathogens, growth retardation, and early senescence phenotype [10,12]. Similarly, the rice SAH single and double mutants showed elevated SA and jasmonate levels and attenuated disease susceptibility to different lifestyle pathogens (Figures 2 and 3). The results demonstrated that each SAH in Arabidopsis and rice has distinct activities based on their spatiotemporal expression patterns, even though partial redundancy may exist [10].
SAHs of Arabidopsis and rice are divided into two subclades ( Figure S3), suggesting that these genes exist before species divergence and evolve separately in their genomes. OsSAH3 with only SA5H activity groups with DLO1/S3H1 and DLO2. However, MDR6 (with high SA5H activity) is grouped with the other three OsSAHs, as it has relatively much lower SA5H activity compared with that of OsSAH3 (Table 1). OsSAH2 and OsSAH3, as well as DLO1/S3H and DLO2, are localized in the vicinity of their chromosomes and were potentially gained by duplication. Functional divergence of OsSAH2 and OsSAH3 was observed at the biochemical level (Table 1; Figure 2). OsSAH3 P292A mutant decreased 2,5-DHBA production; however, the Ala to Pro substitution at the P sites of OsSAH2 and DMR6 completely blocked 2,5-DHBA activity and suppressed 2,3-DHBA production (Figure 2b,c). The data suggest that the relatively fixed side chain of Pro compared with that of Ala had more severe effects on subclade-I than in subclade-II proteins. Site P is between the RxS motif and conserved Phe amino acid, which are in contact with 2OG and possibly associated with the co-substrate SA, respectively [18]. The change in amino acid volume at site P might affect substrate docking and/or oxygen transfer vs. Fe ion. The results revealed that site P plays a critical role in enzymatic activity and specificity. As SA forms an intramolecular hydrogen bond between the carboxyl and the ortho hydroxyl groups [49], SA conformation probably hinders the rotation of the benzene ring within the OsSAH active pockets. Therefore, the ratio of 3-to 5-hydroxyl products was determined when the substrate was initially associated with the OsSAH enzymes. Furthermore, we noted that OsSAH3 has no activity for 2-Cl-BA, but low activity for 3-Cl-SA, implying that the ortho hydroxyl group of SA may be specifically required for binding ( Figures S2 and S4c). There are reports documenting that a single amino acid substitution affects substrate docking and product specificity. Modification of the JA receptor with Ala to Val substitution in the JA-Ile-binding pocket of CORONATINE INSENSITIVE1 (COI1) protein greatly reduces sensitivity to the JA-mimicking toxin, coronatine, but it remains sufficient for endogenous JA signaling [50]. OsTPS19 and OsTPS20 are two limonene synthases, in which the functional variation is determined by a single amino acid (Ala vs. Ser) localized in the active site cavity [27]. Nevertheless, the generation of SAH crystal structures may provide clear images of the SAH active cavity.

Generation of Transgenic Plants
The coding sequences of OsSAH2-3 were amplified from ZH17 (Oryza sativa L. Zhonghua 17) cDNA using gene-specific primers (Table S1), fused with 3 × Myc tag at the 3' end, and placed under maize ubiquitin promoter. For gene knockout, the gene-specific target sequence of each gene was selected and under the control of the U3 promoter in the pOsCas9 vector. For the OsSAH3 (SAH3pro::GUS) promoter construct, the sequence of approximate 2.0 kb, including the translation start site, was inserted into a modified pCambia1301 (Cp-GUS) vector [25].
The verified plasmids were transformed into Agrobacterium tumefaciens EHA105 for rice transformation. The transgenic plants were generated from the immature seeds of ZH17 by the Agrobacterium-mediated transformation method [25]. Transgenic plants were detected by PCR amplification and verified by sequencing.

Plant Growth and Treatments
The seeds of transgenic and control plants were surface sterilized and germinated in 1 2 Murashige and Skoog (MS) medium. The seedlings were cultured in 1 2 MS liquid medium at 28 • C with a 12-h-light photoperiod and treated with 5 mM MES (4-morpholine ethanesulfonic acid, pH 5.8) buffer containing 500 µM SA, 100 µM MeJA, 500 µM M BTH, or 200 µg/mL chitin, and the same volume of DMSO solvent was used as the control.

Pathogen Inoculation
Three-week-old rice plants were inoculated with a virulent M. oryzae SZ strain by spraying the spore suspension (5 × 10 5 conidia/mL containing 0.005% Silwet L-77) as described by Liu et al. [25]. For injection, rice plants at the tillering stage (about threemonths old) were injected with the spore suspension (1 × 10 5 conidia/mL) to the center of each leaf sheath, and the newly grown leaves were collected for disease severity evaluation. To observe infection process, the spores of EGFP-tagged M. oryzae SZ-GFP strain (1 × 10 5 conidia/mL) were infiltrated into the leaf sheath, as described previously [51]. The inoculated sheaths were kept in a petri dish containing wet filter paper. The inner layer of the leaf sheaths were examined with a fluorescence microscope (Nikon Ti-E, Nikon Co., Ltd., Tokyo, Japan) at 24-and 48 h post the inoculation.
To evaluate resistance against bacterial blight pathogen, three-month-old rice plants were inoculated with Xoo J18 (OD 600 = 0.8) by the leaf-clipping method [25]. Disease severity was estimated by measuring lesion length 18 d after the challenge.
Three-week-old rice plants were inoculated with B. oryzae LW by foliar spraying (5 × 10 5 conidia/mL spores containing 0.005% Silwet L-77), in which the B. oryzae LW strain was isolated from the natural infected rice leaves. Disease severity was evaluated by PCR amplification of the relative biomass. Quantification was performed with primers from B. oryzae 26S rDNA and rice ubiquitin gene (OsUBQ). Template DNAs from the inoculated leaves were collected 7 d after the inoculation. Number of disease spot infected naturally by B. oryzae in paddy field was counted as described previously by Zanão Júnior et al. [52].
To evaluate resistance against rice sheath blight, five-week-old rice plants were inoculated with R. solani XN and a weak virulent uninucleate Rhizoctonia JN isolate grown on PDA with slices of filter papers (about 10 mm × 4 mm) on the agar for five days. The filter papers, grown over by XN or JN, were pinned around the sheath bottoms of rice. Lengths of disease lesion were measured after four days inoculation.

ROS Detection
ROS detection was followed as described previously [53]. Six-week-old rice leaves were cut into 4-mm-diameter disks and equilibrated in distilled water for about 10 h. The assay mixtures contained 100 µL of luminol (Immun-Star horseradish peroxidase substrate, Bio-Rad, Hercules, CA, USA), 1 µL of horseradish peroxidase and 8 nM chitin (hexa-Nacetylchitohexaose) or water. The reactions were started by adding three leaf disks into the mixtures in a 1.5 mL microcentrifuge tube (Axygen Scientific, Union City, CA, USA). Luminescence was detected with a luminometer (GloMax 20/20, Promega, Madison, WI, USA) for 25 min.

Protein Expression and Enzyme Assays
DMR6 sequence was amplified from cDNAs of A. thaliana leaves. Mutations were generated by PCR-based site-directed mutagenesis. Each OsSAH1-4, DMR6, and their mutants, was cloned into a modified pGEX-tag vector with 3 × Myc at the C-terminus of the recombinant protein [25]. After transformation of each plasmid into Escherichia coli BL21 (DE3), the protein expression was induced by addition of 0.2 mM IPTG grown at 28 • C for 6 h. The recombinant proteins were purified using Glutathione Sepharose 4B (GE Healthcare, Chicago, IL, USA).
The enzymatic activity assay was performed as described previously [11]. Briefly, a total volume of 100 µL reaction mixture contains 5 µg recombinant protein in the reaction buffer (1 mM 2-oxoglutaric acid, 1 mM sodium ascorbate, 0.4 mM FeSO 4 , 0.1 mg/mL catalase, 5 mM DTT, 50 mM phosphate buffer at pH 6.8) with different concentrations of SA or other compounds tested and reacted at 40 • C for 30 min. The reaction was stopped by adding two volumes of acetonitrile and boiled for 1 min. Then, the mixture was extracted with 200 µL ethyl acetate for twice, the combined organic solutes were dried by nitrogen gas and dissolved in 50 µL of 90% aqueous methanol containing 0.1% formic acid for product detections. For optimal assay conditions, we performed the assays in citrate buffer (pH 5.8), phosphate buffer (pH 6.3, 6.8, and 7.3), or Tris-HCl buffer (pH 7.8) at 40 • C and in phosphate buffer (pH 6.8) at different temperatures. Substrate inhibition kinetics were obtained as described by Kutsuno et al. [54].

Determination of Metabolites
Compound isolation and determination were described previously [55]. Briefly, the freeze-dried plant tissues (approximate 25 mg for each replicate) were extracted with 1 mL 90% aqueous methanol containing 0.1% formic acid and an internal standard of 20 ng D 5 BA. After dry of the supernatants by nitrogen gas, the residues were dissolved in 0.1 mL of 90% aqueous methanol containing 0.1% formic acid for LC-MS/MS. For targeted MS/MS analysis, collision voltage was applied with 20 V for reliable fragmentations. Chemicals were separated by a C 18 column (2.1 mm × 150 mm, 3 µm, Phenomenex) on Agilent 1260 separation module (Agilent Co. Ltd., Santa Clara CA, USA). The elution conditions were at a flow rate of 0.25 mL/min with a gradient program of 5% acetonitrile for 2 min, up to 25% in 8 min, then to 70% in 30 min, and then the column was washed and equilibrated to the initial conditions.

Structural Modeling and Substrate Docking
Conformations of OsSAH2, OsSAH3, DMR6, and OsSAH3 variants were simulated based on the F6'H1 (PDB ID: 4XAE) crystal structure using SWISS-MODEL [56]. The highest score of each protein from global model quality estimation was selected for docking with salicylic acid (PubChem CID: 338) and 2-oxoglutaric acid (PubChem CID: 51). Substrate docking was performed in the conserved active sites of the protein via AutoDock Vina [57]. Interactions between protein and substrate were analyzed via Protein-Ligand Interaction Profiler [58] and visualized in PyMOL.

Reverse Transcribed Quantitative (qRT-PCR) Analysis
After removing possible DNA contaminations, two micrograms of total RNA were reverse transcribed with random hexamers and oligo(dT) 18 primers using M-MLV reverse transcriptase (Takara, Kusatsu-shi, Japan). The relative transcript levels were quantified using SYBR Green PCR Master Mix (Takara) and normalized to OsUBQ. The relative expression level of each gene was analyzed using the delta-delta Ct method. Gene-specific primers used in qRT-PCR are listed in Supplementary Table S1.

Conclusions
We showed the expression patterns of the four OsSAH genes and revealed three amino acids affecting the enzyme hydroxylation activity and product specificity. Knockout of Os-SAH genes increased SA homeostasis at an elevated level, and jasmonate content, conferring broad-spectrum disease resistance against both hemibiotrophic and necrotrophic pathogens. Our findings indicate that SA plays a crucial role in the rice immune signaling pathway.

Data Availability Statement:
The data presented in this study are available upon request from the corresponding author.