A molecular switch in sulfur metabolism to reduce arsenic and enrich selenium in rice grain

Rice grains typically contain high levels of toxic arsenic but low levels of the essential micronutrient selenium. Anthropogenic arsenic contamination of paddy soils exacerbates arsenic toxicity in rice crops resulting in substantial yield losses. Here, we report the identification of the gain-of-function arsenite tolerant 1 (astol1) mutant of rice that benefits from enhanced sulfur and selenium assimilation, arsenic tolerance, and decreased arsenic accumulation in grains. The astol1 mutation promotes the physical interaction of the chloroplast-localized O-acetylserine (thiol) lyase protein with its interaction partner serine-acetyltransferase in the cysteine synthase complex. Activation of the serine-acetyltransferase in this complex promotes the uptake of sulfate and selenium and enhances the production of cysteine, glutathione, and phytochelatins, resulting in increased tolerance and decreased translocation of arsenic to grains. Our findings uncover the pivotal sensing-function of the cysteine synthase complex in plastids for optimizing stress resilience and grain quality by regulating a fundamental macronutrient assimilation pathway. Contamination of paddy soils can lead to toxic arsenic accumulation in rice grains and low levels of the micronutrient selenium. Here the authors show that a gain of function mutant affecting an O-acetylserine (thiol) lyase enhances sulfur and selenium assimilation while reducing arsenic accumulation in grains.

R ice is the dominant source of the carcinogenic arsenic (As), accounting for around 60% of the dietary intake of inorganic As for some Asian populations [1][2][3][4] . Widespread contamination of paddy soils with As due to mining activities and irrigation of As-laden groundwater not only increases As accumulation in rice grain 5,6 , but also results in As toxicity in rice plants and substantial yield losses 7,8 . Worryingly, future climate changes are likely to exacerbate both the problems of As accumulation in rice grain and yield loss induced by As toxicity 9 . It is therefore important to develop rice crops that are more resistant to As and accumulate less As in the grain.
A key mechanism of As detoxification in plants is via the complexation of arsenite [As(III)] by thiol compounds, such as glutathione (GSH) and phytochelatins (PCs), and subsequent sequestration into the vacuoles [10][11][12][13][14] . This mechanism also reduces As translocation to rice grain 12,[15][16][17] . The metal(loid) stressinduced production of GSH and PCs is limited by the availability of its precursor cysteine (Cys) in plants [18][19][20] . Cys is synthesized in all subcellular compartments with own protein biogenesis by a two-step pathway: the formation of O-acetylserine (OAS) from serine and acetyl CoA catalyzed by serine acetyltransferase (SAT) and the condensation of OAS and sulfide to Cys catalyzed by Oacetylserine(thiol)lyase (OAS-TL) 21,22 . This two-step pathway is conserved in bacteria, alga, and higher plants, and serves as a checkpoint for the integration of sulfur, carbon, and nitrogen. The enzymatic functions of three isoforms of OAS-TL (OAS-TL A, B, and C in the cytosol, plastids, and mitochondria, respectively) in Arabidopsis thaliana and a cytosolic OAS-TL A in rice have been characterized [23][24][25][26][27] . Knockout of each of the three isoforms of OAS-TL in Arabidopsis did not affect growth, because the substrates for Cys biosynthesis can be exchanged efficiently between the cytosol and the organelles 25 . SAT and OAS-TL form the hetero-oligomeric cysteine synthase complex (CSC), and SAT is active only within this complex 21,22 . The product of SAT, OAS, has a dissociation effect on the CSC 22,28,29 , thus providing a feedback inhibition on SAT activity. Synthesis of OAS by SAT is also the rate-limiting step of Cys biosynthesis, whereas the total activity of OAS-TL is in excess of that required for Cys biosynthesis 22 . The association of the CSC stimulates the SAT activity and thus provides a mechanism for metabolic regulation of Cys biosynthesis by sensing the supply of the carbon/nitrogencontaining precursor OAS and the sulfur precursor sulfide.
In this study, we isolated a gain-of-function arsenite tolerant 1 (astol1) rice mutant. We show that ASTOL1 encodes the chloroplast-localized OAS-TL. Surprisingly, the astol1 gain-offunction mutation (S189N) inactivates the catalytic activity of OAS-TL by inhibiting OAS binding. We provide direct evidence that the inactive ASTOL1 protein forms a non-dissociable CSC, which permanently stimulates endogenous SAT activity explaining the dominant inheritance of the astol1 mutation. Uncoupling of the OAS-sensing function of the plastidic CSC from its regulatory impact on SAT resulted in (1) enhanced S uptake and assimilation, (2) increased tolerance to As and (3) decreased As accumulation in rice grain. Moreover, the OsAS-TOL1 S189N mutation significantly enhanced the rice grain content of selenium (Se), an essential micronutrient for which up to a billion people worldwide have insufficient dietary intake 30,31 . Besides the potential biotechnological applications, our findings uncover the pivotal sensing-function of the plastid CSC in the regulation of S and Se metabolism and tolerance to metalloid toxicity.

Results
Screening and characterization of the astol1 mutant. We screened more than 4000 individuals derived from an EMS-mutagenized library of the rice cv. Kasalath for arsenite [As(III)] tolerance using a root elongation assay at 20 µM As(III) (approximately the half-inhibitory concentration, Supplementary  Fig. 1a). This screen revealed only one As(III) tolerant mutant ( Supplementary Fig. 1b) that we named arsenite tolerant 1 (astol1).
The root growth of astol1 was indistinguishable from the wild type (WT) in the absence of As(III). However, upon As(III) treatment astol1 grew 2-times longer roots than WT (Fig. 1a, b). When grown hydroponically in a growth chamber (light intensity~300 μmol m −2 s −1 ), astol1 segregated into two phenotypes, one with decreased growth compared with WT and the other dying of leaves from tips inwards leading to eventual plant death (Fig. 1c, d, and Supplementary Fig. 1c-f). The two phenotypes were later confirmed to represent the heterozygous (astol1(+/−)) and homozygous (astol1(+/+)) mutant, respectively, based on genotyping of the causal gene. Thus, we used astol1(+/−) in most of the experiments. The root elongation assay also revealed that astol1(+/+) was slightly more tolerant to As(III) than astol1(+/−) (Fig. 1h). Moreover, the roots of astol1(+/−) and astol1(+/+) were more tolerant to arsenate (As(V)), compared to WT ( Supplementary Fig. 2a, b). Upon As (III) or As(V) treatment, the shoot As concentration of astol1 (+/−) was not significantly different from WT ( Supplementary  Fig. 2c, d). However, astol1(+/−) accumulated significantly more As in the roots than WT ( Supplementary Fig. 2c, d), implying that the increased As tolerance in astol1 is not caused by a decreased As uptake.
We then investigated the relationship between the phenotypes of growth inhibition observed in astol1 and the light intensity in a growth chamber. Increasing light intensity from~300 to~800 μmol m −2 s −1 increased the growth of both WT and astol1(+/−), and decreased the difference in plant biomass between the two genotypes from 63% to 31%, but did not improve the growth of astol1(+/+) ( Supplementary Fig. 3a, b). When grown in an open paddy field, where light intensity often exceeds 1000 μmol m −2 s −1 , astol1(+/−) was indistinguishable from WT in all agronomic traits including grain yield (Fig. 1e, Supplementary Fig. 4a-f). Remarkably, in open paddy fields astol1(+/+) also grew to maturity but produced smaller grain yield due to a low percentage of filled spikelets. These results indicate that the slow growth phenotype of astol1 can be reversed under normal light conditions in paddy fields.
OsASTOL1 encodes a chloroplast-localized OAS-TL protein. To clone the causal gene for the astol1 phenotype, we backcrossed astol1(+/−) (paternal) with WT (maternal). Among the 30 F 1 plants, 40% showed the As(III) tolerance phenotype (not significantly different from the expected 50%), indicating that the mutation in OsASTOL1 has a dominant inheritance for As(III) tolerance. In the F 2 progenies a segregation pattern with a ratio of 58 WT: 166 astol1(+/−) :15 astol1(+/+) was obtained, suggesting a semi-dominant effect of the mutation. However, the determined ratio was inconsistent with the expected segregation ratio of 1: 2: 1 (χ 2 = 51.66 > χ 2 0.05,2 ), likely due to partial sterility and poor seed germination of the homozygous mutant.
We conducted genomic resequencing mapping and MutMap analysis 32 on the backcrossed F 2 progeny with astol1(+/+) phenotype. One SNP (SNP-26700810) was identified with a SNP index of 1 ( Supplementary Fig. 5). This SNP was located in the fifth exon of the gene (Os12g0625000) encoding a putative OAS-TL that leads to a Ser189Asn mutation (codon AGC → AAC, Fig. 1f). The point mutation in this OsASTOL1 gene was verified by DNA sequencing and a derived cleaved amplified polymorphic sequence (dCAPS) marker ( Supplementary Fig. 6a, c, d). Relative root elongation of WT and astol1 mutant treated with 0 or 20 μM As (III) for 2 days. Asterisks indicate significant difference by two-sided Student's t test: ***P < 0.001. c Growth phenotype of 5-week-old WT, astol1(+/−) and astol1(+/+) mutant grown hydroponically. Two plants for each genotype are shown. Scale bars, 10 cm. d Biomass of 5-week-old WT, astol1(+/−) and astol1(+/+) mutant grown hydroponically. e Wild-type rice and astol1 mutants were grown in a paddy field until maturity. f Chromosomal location and gene structure of OsASTOL1. White and black boxes represent untranslated regions and exons, respectively, and the lines between them represent introns. Red line indicates the mutation site. g Validation of knock-out OsASTOL1/ OsASTOL1 S189N protein in four independent CRISPR lines in the astol1(+/−) background using immunoblot analysis. OsASTOL1 protein (lower band) is detected by AtOAS-TL A1 antibody (anti-OAS-TLA) and the Rubisco large subunit is used as a loading control. The experiment was repeated three times and similar results were obtained. h Relative root elongation of WT, astol1 (+/−), astol1(+/+), and four independent CRISPR lines in the astol1(+/+) background. Plants were treated with 20 μM As(III) for 2 days. i Biomass of 5week-old WT, astol1(+/−), astol1(+/+), and CRISPR lines in astol1(+/−) or astol1(+/+) background. DW, dry weight. WT, wild type. astol1(+/−), astol1 heterozygote. astol1(+/+), astol1 homozygote. Data in b, d, h, and i are shown as means ± s.d., n = 10 for WT (control) and astol1 (As(III) treatment), 13 for astol1 (control) and WT (As(III) treatment) (b), 5 (d), 6 for astol1(+/+), 10 for WT and all CRISPR lines/astol1(+/+), 19 for astol1(+/−) (h) or 3 (i) biological replicates; each biological replicate represents an individual plant. Different letters in d, h, and i indicate significant differences (P < 0.05) using one-way ANOVA followed by Tukey's test. Since astol1 is a semi-dominant mutant, we generated transgenic plants expressing the mutant Osastol1 gene under the control of the maize Ubiquitin promoter in WT. When grown in an open paddy field, all transformed T 0 individual lines (n = 5) that over-expressed the mutant OsASTOL1 S189N protein produced sterile seeds ( Supplementary Fig. 7). This finding supports our previous observation that astol1(+/+) showed a low percentage of filled spikelets in the field and induces plant death when grown in growth chambers (Fig. 1d, Supplementary Fig. 4d). To provide direct evidence that mutated OsASTOL1 is responsible for the astol1 phenotypes, we knocked out the mutant Osastol1 gene in astol1 using the CRISPR-Cas9 gene-editing method ( Supplementary Fig. 6b). Four independent T 0 CRISPR-induced knockout lines in the astol1(+/−) background with both OsASTOL1 and Osastol1 being homozygously knocked out were obtained and confirmed by immunoblot using an antibody against the full-length Arabidopsis OAS-TL A (Lower band in Fig. 1g, with the upper band likely to be another isoform of OAS-TL). In the T 1 generation of these lines, three types of plants including homozygous knockout of OsASTOL1 and/or Osastol1 in the backgrounds of WT, astol1(+/−) or astol1(+/+) were isolated by using the dCAPS marker ( Supplementary Fig. 6d). The relative root elongation of all CRISPR lines in the astol1 (+/+) background showed the same As(III) sensitive phenotype as WT (Fig. 1h). Furthermore, all CRISPR lines in both the astol1 (+/−) and astol1(+/+) backgrounds showed a complete rescue of the astol1 mutation-induced phenotypes in growth and root As concentration ( Fig. 1i and Supplementary Fig. 6c-e). These results demonstrate that the S189N point mutation in the OsASTOL1 protein is the cause for the astol1 phenotypes.
OsASTOL1 shares high homology with Arabidopsis cytosolic OAS-TL A ( Supplementary Fig. 8). However, in contrast to AtOAS-TL A, OsASTOL1 contains 87 extra amino acid residues at the N-terminus ( Supplementary Fig. 9), which are predicted to generate a chloroplastidic transitpeptide 33 (Supplementary  Table 2). OsASTOL1 is predominately expressed in photosynthetic tissues ( Supplementary Fig. 10a-d). When transiently expressed in tobacco leaves and rice protoplasts, both wild-type OsASTOL1-YFP and mutant OsASTOL1 S189N -YFP fusion proteins localize in the chloroplasts ( Fig. 2a and Supplementary  Fig. 10e). These findings demonstrate that the astol1 point mutation does not affect the chloroplastic localization of OsASTOL1 in rice.
Ser189 of OsASTOL1 is crucial for OAS-TL activity. To test whether the OsASTOL1 S189N mutation affects OAS-TL activity in planta, we determined the total OAS-TL activity and OAS-TL protein abundance in the WT and the astol1 mutant. Remarkably, the abundance of the mutated OsASTOL1 S189N was unaffected in astol1 (+/+), but the total extractable OAS-TL activity was decreased by 32% in astol1(+/+) when compared to WT (Fig. 2b,  c). These results suggest that Ser189Asn mutation does not destabilize the OsASTOL1 S189N protein but impairs its enzymatic activity.
OAS-TL proteins are highly conserved in prokaryotes and eukaryotes (Supplementary Figs. 9 and 11). In particular, the Ser189 of OsASTOL1 is conserved in all analyzed OAS-TLs ( Supplementary Fig. 9). In order to compare the OAS-TL activity of the mature OsASTOL1 and the mutated OsASTOL1 S189N , we deleted the 87 amino acid residues encoding the transitpeptide and expressed them in E. coli. Because AtOAS-TL A has been well characterized in enzymatic analysis 23,26 and is highly homologous to OsASTOL1 protein ( Supplementary Fig. 8), it was applied as a positive control for a canonical OAS-TL enzyme (Fig. 2d). Expression of AtOAS-TL A in E. coli Cys-auxotrophic strain NK3 34 , which lacks the cysteine synthase gene and consequently is unable to grow on medium without cysteine supplement, restored its growth in Cys-free medium within 2 days of incubation ( Supplementary Fig. 12a). In comparison, expression of OsASTOL1 restored the growth of NK3 strain in Cys-free medium after 7 days of incubation, whereas OsAS-TOL1 S189N failed to complement NK3 strain ( Supplementary  Fig. 12a). Moreover, the purified mature OsASTOL1 protein clearly displayed OAS-TL activity, but its specific activity was only about 1/4 of AtOAS-TL A (Fig. 2e). In contrast, the mutated OsASTOL1 S189N displayed no OAS-TL activity (Fig. 2e). This is consistent with the complementation assay in NK3 (Supplementary Fig. 12a). Moreover, size exclusion chromatography (SEC) analysis shows that both the OsASTOL1 and OsASTOL1 S189N were present as dimers as other OAS-TLs ( Supplementary  Fig. 12b). Furthermore, the replacement of the corresponding Ser102 in AtOAS-TL A with Asn fully inactivated the OAS-TL A S102N mutant protein (Fig. 2e). The protein crystal structure 35 of AtOAS-TL A indicates that Ser102 is located at the opening of the binding site for the substrate OAS ( Supplementary Fig. 13a, b).
To investigate the role of this conserved Ser, we changed the Ser189 of OsASTOL1 to different amino acid residues by sitedirected mutagenesis and determined the in vitro OAS-TL activities of the respective protein mutants ( Supplementary  Fig. 14a, b). All mutation alleles lost their enzymatic activity, except OsASTOL1 S189T that displayed 41% of the ASTOL1 activity ( Supplementary Fig. 14b). Because both the mutation alleles of S189A (a constitutively nonphosphorylable form) and S189D (a mimic of the constitutively phosphorylated form) lost the OAS-TL activity completely, it is unlikely that phosphorylation of S189 is required for the enzyme activity. Thus, we conclude that the conserved Ser189 in OsASTOL1 is crucial for OAS-TL activity, and mutation of this residue in OsASTOL1 S189N leads to a total loss of OAS-TL activity.
Loss of OAS-TL activity is irrelevant to astol1 phenotype. To investigate the causal relationship between the phenotypes of the astol1 mutation and the OAS-TL activity in the mutant, we generated CRISPR lines and overexpression lines of OsASTOL1 in the WT background. The CRISPR-induced loss-of-ASTOL1 function resulted in three independent lines in a decrease to 50-63% of the total in vivo OAS-TL activity when compared to WT ( Supplementary Fig. 15a, b). In contrast, overexpression of wild type ASTOL1 protein caused in three independent lines a 3fold increase of OAS-TL activity when compared to WT (Supplementary Fig. 15b). Neither loss-of-ASTOL1 function nor the overexpression of ASTOL1 mimicked the As(III) tolerance or the decreased growth phenotype of astol1 ( Supplementary Fig. 15c, d). These results suggest that the astol1 phenotypes are not caused by the loss of OAS-TL activity. This finding is consistent with the semi-dominant inheritance of the astol1 mutation. The results are also consistent with previous studies on Arabidopsis that knockout of each of the three major isoforms of OAS-TL did not impair growth 25 .
Stabilized CSC leads to higher SAT activity in astol1 mutant. In plants, Cys biosynthesis is not limited by OAS-TL, but by the synthesis of its substrate OAS 25,26,36 . OAS is exclusively produced by SAT in plants. Remarkably, SATs are almost not regulated at the transcriptional level. The dominant mechanism for the regulation of SAT activity is the formation of the CSC complex. CSC formation significantly activates SAT and is controlled by the actual OAS supply 22 . We applied the well-characterized AtSAT5 protein as the bait protein to test whether Ser189Asn mutation in the mutant OsASTOL1 S189N protein affects its association with SAT protein in an affinity column-based pull-down experiment 29,37 . Both OsASTOL1 and OsASTOL1 S189N proteins interact with the His-tagged AtSAT5 protein ( Supplementary  Fig. 16a, b). Similarly, AtOAS-TL A and AtOAS-TL A S102N bound to the column immobilized His-tagged AtSAT5 (Supplementary Fig. 16c, d). Thus, mutation of the conserved Ser to Asn did not affect the formation of CSC.
Next, we tested the ability of OAS to dissociate the recombinant CSCs consisting of His-SAT5 and the four candidate OAS-TLs. As expected, the CSC containing the wild type OsASTOL1 and the wild type AtOAS-TL A were dissociated by the application of OAS. In contrast, OAS failed to dissociate the CSC consisting of the OsASTOL1 S189N and the AtOAS-TL A S102N protein ( Fig. 3a-c, and Supplementary Fig. 16e-g). Since OAS-induced dissociation of the CSC is triggered by competition between OAS and the SAT C-terminus for binding to the active site of OAS-TL 38,39 , we quantified the affinity of wild type OsASTOL1 and the OsASTOL1 S189N for OAS and the C-terminal fragment of SAT5 (AtSAT5C10) by microscale thermophoresis (MST). As expected, wild type ASTOL1 displayed high affinity towards OAS and AtSAT5C10 (Fig. 3d, e and Supplementary Table 3). The mutated OsASTOL1 S189N physically interacted with the AtSAT5C10 fragment with a wild-type like affinity, but specifically lost the affinity towards OAS (Fig. 3d, e and Supplementary Table 3). Mutation of S102N in AtOAS-TL A resulted in the same specific loss of affinity towards OAS but did not affect the binding of AtSAT5C10 ( Supplementary Fig. 16h, i and Supplementary Table 3). These biochemical properties of the mutated OAS-TLs provide a molecular explanation for the formation of an OAS-resistant CSC by the mutated OsAS-TOL1 S189N and AtOAS-TL A S102N proteins and the catalytic inactivity of the both proteins. To test whether the endogenous SAT activity was stimulated by the presence of mutated OsASTOL1 S189N protein in astol1, we compared SAT activities in crude plant protein extracts of WT and the astol1(+/−) mutant. The extractable SAT activity of astol1 (+/−) was 2.8-fold higher than that in WT (Fig. 3f), although transcript abundance of SAT genes was indistinguishable between astol1(+/−) and WT (Fig. 3g). Taken together, these results imply that the Ser189Asn mutation in OsASTOL1 S189N increases the stability of the CSC against the dissociation of OAS, resulting in significantly enhanced endogenous SAT activity in planta. The fraction of OAS elution and final elution was analyzed by SDS-PAGE and visualized by immunoblot, and a representative result of three repeated tests is shown. c The relative abundance of OsASTOL1 or OsASTOL1 S189N in the OAS elution and the final elution was quantified by ImageJ software, compared to the corresponding AtSAT5 protein. Microscale thermophoresis (MST) analysis of in vitro binding of OsASTOL1 and OsASTOL1 S189N protein with OAS (d) or AtSAT5C10 peptide (e). f Total SAT enzyme activity in whole soluble protein extracts of the shoots of 3-week-old WT and astol1 (+/−). g Relative expression levels of two rice SAT genes (Os01g0720700 and Os05g0533500) in the shoots of 3-week-old WT and astol1(+/−), with OsHistone H3 as the internal reference. WT, wild type. astol1(+/−), astol1 heterozygote. Data in c-e are shown as mean ± s.d., n = 3 technical replicates. Data in f-g are shown as means ± s.d., n = 4 (f) or 3 (g) biological replicates; each biological replicate represents an individual plant. Asterisks in c, f, g indicate significant difference by two-sided Student's t test: **P < 0.01, ***P < 0.001. astol1 mutant overaccumulates OAS and S metabolites. As a result of the increased foliar SAT activity, the total steady-state level of the SAT product, OAS, Cys and most of the Cys-derived sulfur-metabolites tested were significantly higher in leaves of astol1(+/−) when compared to WT. Remarkably, the astol1 mutation-induced increase of OAS and sulfur-containing metabolites were also detected in roots of non-As(III)-stressed plants, suggesting that SAT activity is globally enhanced by the astol1 mutation (Fig. 4a, b and Supplementary Table 4). The astol1 mutation also caused decreased steady-state level of the SAT substrate, serine, in leaves and roots, suggesting a significant higher conversion of Ser into OAS in these organs (Fig. 4a, b and Supplementary Table 4). OAS acts as the signal of sulfur deficiency and induces the expression of genes involved in S uptake and assimilation 21,40,41 . The transcript levels of several genes involved in sulfate uptake and reduction, including root plasma membrane-localized highaffinity sulfate transporter OsSULTR1;1, were substantially upregulated in astol1 when compared with WT (Fig. 4c). Consistently, the astol1(+/−) mutant accumulated more sulfate, total S and Se (Fig. 4b, d, e); the latter is a chemical analog of S and is also taken up by SULTRs. Moreover, knockout of Osastol1 in the astol1 background reduced the shoot S concentration to the WT level, whereas knockout of wild-type OsASTOL1 did not significantly affect shoot S concentration (Supplementary Figs. 6f and 15e), further supporting that the specific mutation in Osastol1 leads to the mutant phenotypes in astol1. Taken together, these results demonstrate that substantial accumulation of OAS leads to increased sulfate uptake and assimilation into Cys in the astol1 mutant.
To explain the OsASTOL1 S189N -induced gain of As(III) tolerance, we quantified the impact of As(III) on thiol compounds in the roots and shoots of WT and astol1(+/−). Upon 5 µM As (III) treatment, the steady-state levels of Cys, and the As(III) chelators GSH, PC 2 , PC 3, and PC 4 were all significantly increased in the roots of astol1(+/−) when compared with WT (Fig. 4f), demonstrating the enhanced capacity of the astol1 plants for complexation of As(III). The levels of Cys and PC 2 in the shoots of astol1(+/−) were also increased significantly (Fig. 4g), although the levels of both thiols were lower than those in the roots. Furthermore, the transcript levels of OAS-TL and SAT genes in shoots and roots of WT and astol1(+/−) showed no significant change when treated with As(III) ( Supplementary  Fig. 17a-d), suggesting that these genes are not transcriptionally induced by As(III) stress.
OsASTOL1 S189N reduces As and increases Se in rice grain. To evaluate the biotechnological capability of the astol1 mutation for improvement of the nutritional value of rice grain, we quantified the yield and accumulation of As, S, and Se in the grain of rice plants grown on paddy fields with background levels of As and Se in the soils. Grain As concentration of astol1(+/−) was 29% and 33% lower than that of WT, respectively, in the two field experiments ( Fig. 5a and Supplementary Fig. 18a). In contrast, astol1(+/−) accumulated 77% and 107% higher Se concentration in rice grain than WT, respectively, in the two field experiments ( Fig. 5c and Supplementary Fig. 18b). We also determined As concentrations in different shoot tissues of astol1(+/−) and WT sampled at plant maturity from the two field sites. The astol1 (+/−) mutant contained significantly higher As concentrations in node I and internode than WT at one site, but a significantly lower As concentration in flag leaf blade at the other site (Supplementary Fig. 19a, b).
To overcome the growth inhibition in astol1(+/+), we overexpressed wild-type OsASTOL1 in astol1(+/−) mutant and obtained overexpression lines in the WT, astol1(+/−), or astol1 (+/+) backgrounds among the segregated T 1 plants. The overexpression lines in the astol1(+/+) background rescued the early leaf death phenotype of astol1(+/+) and also improved the growth of astol1(+/−) under low light conditions (Supplementary Fig. 20a-f). Notably, the overexpression lines in both the backgrounds of astol1(+/−) and astol1(+/+) still accumulated higher S concentrations in the shoots and higher As concentrations in the roots than WT (Supplementary Fig. 20g-l). Moreover, when grown in a paddy field, overexpression of wild-type OsASTOL1 in both the astol1(+/−) and astol1(+/+) backgrounds restored their growth and agronomic traits to the levels of WT (Fig. 5d, e and Supplementary Fig. 21a-e). Furthermore, the overexpression lines in both backgrounds of astol1(+/−) and astol1(+/+), like astol1(+/−), had significantly lower grain As concentrations than WT, with decreases of 32-39% and 23-40%, respectively (Fig. 5a). These lines accumulated significantly more S in the grain than WT (Fig. 5b). In addition, the overexpression lines in the backgrounds of astol1(+/−) and astol1(+/+) accumulated 95-121% and 15-48% higher Se concentration in the grain than WT, respectively (Fig. 5c). These results indicate that the Ser189Asn mutation in OsASTOL1 S189N in rice enriches S and Se and reduces As accumulation in rice grain, while the growth inhibition in astol1(+/+) can be overcome by balancing the mutant and wild-type OsASTOL1 protein.

Discussion
Rice accumulates about 10-times higher As in the grain than other cereal crops due to the mobilization of As(III) in submerged paddy soils and the high capacity of rice roots for uptake of As (III) 3,42,43 . Consequently, many Asian populations are exposed to highly levels of dietary As. On the other hand, around 75% of the rice grains produced globally have insufficient levels of Se to meet human's requirement 44 . In this study, we show that Ser189Asn mutation in OsASTOL1 S189N serves as a molecular switch in the sulfur assimilation pathway to produce multiple desirable consequences, including increased As tolerance, decreased As accumulation in the grain and increased Se accumulation in the grain. These phenotypes can be explained by improved uptake and assimilation of sulfate and selenate, and the enhanced synthesis of thiol compounds. This reinforced capacity to synthesize PCs allows for improved complexation of As 11-14 , facilitates the vacuolar sequestration of As-thiol complexes and, consequently, reduces As mobility to rice grain [15][16][17] . Increased synthesis of thiol compounds is more pronounced in the roots than in the shoots of the mutant, which is consistent with roots being the main tissue for As sequestration. A radioactive 73 As tracer experiment showed that approximately 90% of the arsenite taken up by rice roots is sequestered in the roots 45 . Furthermore, enhanced GSH and PC synthesis should also improve the resistance towards heavy metals like Cd and Hg as well as some biotic stresses 18,46,47 .
We show that Ser189Asn mutation in OsASTOL1 S189N results in a total loss of the OAS-TL activity but a gain-of-function of increased SAT activity, and it is the latter that explains the observed phenotypes of the astol1 mutant, as summarized in our model (Fig. 6). The Ser189Asn mutation in OsASTOL1 S189N specifically impairs its affinity towards the substrate OAS (Fig. 3d), but surprisingly does not impact its physical interaction with the SAT5C10 peptide (Fig. 3e). In fact, the CSC complex formed from OsASTOL1 S189N and SAT becomes much more resistant to the dissociation of OAS (Fig. 3a-c), which may also result from the loss of binding of OAS to OsASTOL1 S189N . The increased stability of the CSC complex leads to a significantly higher SAT activity in the astol1 mutants, because SAT is substantially activated in the CSC complex due to release from cysteine inhibition 48,49 . Because production of OAS is the ratelimiting step, increased OAS drives Cys biosynthesis catalyzed by other isoforms of OAS-TL either in the chloroplast or in the cytoplasm, as proposed in the model (Fig. 6). Moreover, OAS is a signaling molecule that can induce the expression of genes involved in S uptake and assimilation 21,40,41 . Indeed, several genes in the sulfate uptake and reduction pathway were significantly upregulated in the astol1 mutants (Fig. 4c). In addition to sulfate, many sulfur-containing metabolites are also increased in the astol1(+/−) mutant, indicating that sulfate reduction and assimilation are enhanced (Fig. 4b). Increased biosynthesis of OAS, Cys, and other elevated thiol compounds in the astol1 mutants causes a higher demand for carbon. Consequently, norm light conditions in paddy fields helped the mutants to overcome the imposed challenge for the enhanced synthesis of the diverse carbon and sulfur-containing metabolites. The relatively poor growth of the mutants under low light intensity in growth rooms might be also partially explained by a misbalance of enhanced level of sulfur-containing defense compounds in the artificial absence of stresses, e.g., high light and pathogens, which demand sulfur-containing defense compounds. However, in open paddy fields the potentially defense-primed astol1(+/−) mutants performed as well as the wild type, since the wild type was also challenged with environmental cues inducing sulfur-defense metabolite synthesis.
Our model (Fig. 6) also explains the dosage effect of the mutant allele, as well as the alleviation effect on the mutant growth phenotypes when wild-type OsASTOL1 was overexpressed in the astol1 mutants ( Fig. 5 and Supplementary  Fig. 20), resulting from the balance in the competition between the mutant and wild-type OsASTOL1 for SAT. Taken together, the specific Ser189Asn mutation in OsASTOL1 S189N enables it to act as a regulatory protein of Cys biosynthesis, even though the mutant protein is inactive as an OAS-TL enzyme. It is remarkable that a single amino acid mutation in OAS-TL can trigger such an overwhelming impact on the uptake and assimilation of the macronutrient sulfur, albeit plastidic OAS-TLs is dispensable 25 (Supplementary  Fig. 15c-e). The full conservation of Ser189 in all known plant OAS-TLs suggests that the herein identified mechanism to trigger S and Se accumulation in grains will also be applicable in other plant species. As a proof of concept, we generated an OsASTOL1 S189N -like mutation in cytosolic AtOAS-TL A (AtOAS-TLA S102N ), which, indeed, caused the identical biochemical properties as observed in the plastid-localized OsASTOL1 S189N protein of rice ( Fig. 2e and Supplementary  Fig. 16e-i).
In the reference plant Arabidopsis thaliana, cysteine synthesis takes place in all subcellular compartments that are capable of protein biogenesis, which is the dominant sink for cysteine 21 . Remarkably, these subcellular compartments are also equipped with a CSC, which is supposed to regulate the OAS precursor biogenesis by controlling the SAT activity in these compartments. Since 80% of total SAT activity is localized in mitochondria 50,51 , the mitochondrial CSC is supposed to efficiently regulate cellular OAS synthesis in plants 49 . Accordingly, downregulation of mitochondrial SAT3 impairs growth of Arabidopsis 52 . However, expression of SAT in the cytosol of Arabidopsis and Nicotiana tabacum also resulted in enhanced cysteine formation, demonstrating that enhanced OAS production in these compartments also trigger net cysteine synthesis rate 53,54 . In particular, the plastid-localized CSC is supposed to integrate environmental cues to coordinate cysteine biosynthesis for GSH production to maintain redox homeostasis upon diverse stresses 55-57 . In . d-f Yield components of WT, astol1(+/−), and three overexpression lines in astol1(+/−) or astol1(+/+) background. After harvest, filled spikelets (d), 1000-grain weight (e), and grain yield (f) were recorded. DW, dry weight. WT, wild type. astol1 (+/−), astol1 heterozygote. astol1(+/+), astol1 homozygote. Data in a-f are shown as means ± s.d., n = 4 (a-c) or 6 (d-f) biological replicates; each biological replicate represents an individual plant. Different letters in a-f indicate significant differences (P < 0.05) using one-way ANOVA followed by Tukey's test. NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-021-21282-5 ARTICLE contrast to OAS biosynthesis, the S-precursor for cysteine biosynthesis is exclusively generated in plastids in a photosynthesisdependent manner 21 . The uncoupling of the OAS-sensing function of OAS-TL from its regulatory role inside the CSC stimulated sulfate uptake from the soil and its reductive assimilation in plastids significantly. It remains to be shown if the CSCs in other subcellular compartments contribute to the coordination of S-flux into cysteine with translation demand or have distinct features (e.g., coordination with supply of other macronutrients, or coordination with stress responses). The OsASTOL1 S189N mutation provides a versatile tool to dissect these potential regulatory functions of the individual CSCs in each of these compartments by allowing specific engineering of CSCs at the subcellular level.
In conclusion, our study demonstrates that S assimilation can be effectively controlled by engineering the sensitivity of the CSC against dissociation by OAS, highlighting the physiological importance of the regulatory function of the CSC. Triggering of S assimilation in rice has multiple beneficial impacts on the food security (sustainable growth) and the nutritional value of rice since it enhances the tolerance of rice towards arsenic and heavy metals, and positively affects Se content but limits As content of grains. The key to optimization of the beneficial traits is to adjust the balance between the wild type like and the OsASTOL1 S189Nlike OAS-TL protein, as too much-mutated protein had a deleterious effect on plant growth possibly through disturbance in the balance between C, N, and S metabolism. This goal can be achieved by expression of wild-type OsASTOL1 in the astol1 (+/+) mutant as demonstrated in our study (Fig. 5), or by the expression of mutant Osastol1 in wild-type background driven by a suitable promoter.

Methods
Plant materials and growth conditions. Rice (Oryza sativa cv. Kasalath) seeds were mutagenized by treating them with 0.8% ethyl methanesulfonate (EMS) for 12 h. M 2 seeds obtained from self-pollinated M 1 plants were used for the screening of arsenite (As(III)) tolerant mutants based on root elongation at 20 µM As(III). The M 3 generation of the candidate astol1 mutant was used to verify the As(III) tolerant phenotype. In brief, the root length of the seedlings after germination was recorded as R 0 . Seedlings were transferred to deionized water for 24 h and the root length was recorded as R 1 . Seedlings were then transferred to deionized water or deionized water containing 20 µM As(III) for 48 h and root length was recorded as R 2 . The relative root elongation (RRE) was calculated as (R 2 -R 1 )/(R 1 -R 0 ) × 100. Similarly, the RRE of astol1 was calculated when treated with 8 µM arsenate (As(V)). MutMap-based gene cloning. To clone the causal gene responsible for the astol1 mutant phenotype, the astol1 mutant as paternal was backcrossed with wild type (Kasalath) to generate F 2 progeny. The F 2 plants which showed tolerance to 20 µM As(III) and the early leaf death phenotype were selected for DNA extraction for whole-genome resequencing using the MutMap method 32 . In brief, DNA was extracted from 42 F 2 plants showing the As(III)-tolerant and early leaf death phenotype and 10 wild-type plants (Kasalath) and mixed in an equal ratio to prepare pooled genomic DNA of astol1 mutant and wild-type, respectively. Each of the two pooled DNA samples was used for the preparation of libraries for Illumina sequencing according to the protocol for the Paired-End DNA Sample Prep kit (Illumina). The DNA library of each pool was subjected to the whole-genome sequencing using an Illumina Hiseq4000 sequencer (100× and 30× coverage for astol1 and wild-type, respectively) and 150-bp paired-end reads were generated. After removing the adapter sequences and low-quality reads, clean reads were mapped to the reference genome sequence (Os-Nipponbare-Reference-RGAP7, MSU) using BWA software 58 , followed by SNPcalling using GATK software 59 . Causative variants were calculated using the SNP-index method 32 Fig. 6 Proposed working model for the regulation of OsASTOL1 in sulfur metabolism in rice. Schematic overview of enzymatic reactions (black arrows) and metabolites involved in the sulfur metabolism and its regulatory mechanism. In wild type rice, OAS is produced by cysteine synthase complex (CSC), which results in two effects: (i) induces expression of genes in sulfate uptake and assimilation (blue arrow) to enhance the supply of sulfide; (ii) dissociates the cysteine synthase complex (solid red line) to inactivate SAT. These effects control the flux of S towards cysteine synthesis in vivo. However, in the astol1 mutant, the mutated OsASTOL1 S189N protein is difficult to be dissociated from the CSC by OAS (dotted red line), which results in SAT being constitutively active and producing more OAS. Increased OAS level enhances sulfate/selenate uptake and assimilatory reduction and arsenic detoxification (bigger size of black and blue arrows).
Plasmid construction and plant transformation. To generate the pUbi-OsASTOL1 and pUbi-Osastol1 constructs, the open reading frames (ORFs) of OsASTOL1 and Osastol1 were amplified from cDNA of cv. Kasalath and astol1 homozygous mutant, respectively, and then cloned into the KpnI/SpeI sites of the pTCK303 vector.
To generate a pCRISPR-OsASTOL1 gene-editing construct with a single target site, a 20 bp target sequence from the second exon of OsASTOL1 (5'-AATGGGTGAGACCATCGCCA-3') was designed and then cloned into pOs-Cas9 vector according to Sun et al. 60 .
To generate a pOsASTOL1-GUS construct, a 2.7-kb DNA fragment of the OsASTOL1 upstream sequence before ATG was PCR amplified from the Kasalath genomic DNA and subcloned into the PacI/AscI sites of pS1aG-3 vector 61 to replace the original HvPht1;1 promoter before the GUS (β-glucuronidase) reporter gene.
To generate OsASTOL1-YFP and Osastol1-YFP constructs, the ORFs of OsASTOL1 and Osastol1 without the stop codon were cloned into the pEarleyGate 101(C-YFP-HA) vector using Gateway technology. In brief, the ORFs of OsASTOL1 and Osastol1 were PCR-amplified using attB adapter-containing primers and then the fragments were cloned into pDONR vector using BP Clonase (Invitrogen). OsASTOL1-pDONR and Osastol1-pDONR were subcloned into pEarleyGate 101(C-YFP-HA) vector using LR Clonase (Invitrogen).
Because the full-length OsASTOL1 protein was expressed with a very low efficiency in E. coli likely due to the presence of the signal peptide in the N terminus, the signal peptide (87 amino acid residues) of OsASTOL1 was deleted before expression in E. coli. To generate the OsASTOL1-His, Osastol1-His, AtOAS-TL A-His and AtOAS-TL A S102N -His for in vitro enzyme assays and for protein binding analysis, the ORFs of OsASTOL1 without the first 261 bp sequence in the N terminus, Osastol1 without the first 261 bp sequence in the N terminus, AtOAS-TL A and AtOAS-TL A S102N (site-directed mutation AtOAS-TL A) without the stop codon were amplified and cloned into the NdeI/XhoI sites of pET-29a vector, respectively. Constructs of different alleles at the 189 th site of OsASTOL1-His were generated by site-directed mutagenesis.
For in vitro pull-down assays, the construct of pET-28a-His-AtSAT5 was used 62 . To generate the OsASTOL1, Osastol1, AtOAS-TL A and AtOAS-TL A S102N constructs, the ORFs of OsASTOL1 (without the first 261 bp sequence in the N terminus), Osastol1 (without the first 261 bp sequence in the N terminus), AtOAS-TL A, and AtOAS-TL A S102N with the stop codon were PCR-amplified and cloned into the NdeI/SacI sites of pET-32a vector, respectively. To generate the S-tag-OsASTOL1, S-tag-Osastol1, S-tag-AtOAS-TL A, and S-tag-AtOAS-TL A S102N constructs, the ORFs of OsASTOL1 (without the first 261 bp sequence in the N terminus), Osastol1 (without the first 261 bp sequence in the N terminus), AtOAS-TL A, and AtOAS-TL A S102N with the stop codon were PCR-amplified and cloned into the NcoI/SacI sites of pET-29a vector, respectively.
Hydroponic experiments. For uptake experiments, 4-week-old rice seedlings of wild-type Kasalath and astol1 heterozygous mutant were exposed to half strength Kimura B nutrient solution containing 5 µM As(III), 5 µM As(V), or 2 µM Se(VI) for 3 days.
For complementation test, 5-week-old rice seedlings of wild-type Kasalath, astol1 heterozygous and homozygous mutants, knockout lines in different background genotypes were exposed to half strength Kimura B nutrient solution containing 5 µM As(III) for 3 days. The background genotypes of rice plants were identified by a derived cleaved amplified polymorphic sequence (dCAPS) marker. The PCR products were digested with DdeI. The primers used are listed in Supplementary Table 1.
To evaluate the effects of OsASTOL1 overexpression or knockout, 5-week-old rice seedlings of wild-type Kasalath, astol1 heterozygous mutant, knockout lines, overexpression lines, and null lines in different background genotypes were exposed to half strength Kimura B nutrient solution containing 5 µM As(III) for 3 days. The background of transgenic plants was genotyped by a derived cleaved amplified polymorphic sequence (dCAPS) marker (DdeI) targeting the mutated site of Osastol1 in astol1. The primers used are listed in Supplementary Table 1.
Elemental analysis. Roots and shoots of rice plants in hydroponic experiments were separated and sampled at the end of the treatments. Plants grown in field experiments were separated into different tissues. Plant samples were dried at 65°C for 3 days, ground to fine powder and then digested with HClO 4 /HNO 3 (15/85, v/ v) in a heating block. Brown rice samples were digested with high-purity HNO 3 in a microwave oven. The concentrations of As, Se, and other mineral elements were determined using inductively coupled plasma mass spectrometry (ICP-MS, Nex-ION 300X, PerkinElmer). Blanks and certified reference materials (rice flour NIST1568b from National Institute of Standards and Technology, Gaithersburg, MD, USA and spinach leaves GBW10015 from Institute of Geophysics and Geochemical Exploration, Langfang, China) were included for quality control in the analysis.
Metabolites involved in sulfur metabolism were analyzed by Metware Biotech Co., Ltd (Wuhan, China). In brief, roots and shoots of 5-week-old wild-type rice and astol1 heterozygous mutant were frozen in liquid nitrogen and freeze dried. Three biological replicates were included for each line with six plants for each biological replicate. The metabolites from freeze-dried samples were extracted overnight by 80% ethanol and qualitatively analyzed using a liquid chromatography-electrospray ionization-tandem mass spectrometry (LC-ESI-MS/ MS) system according to Chen et al. 65 . The metabolites were quantified using standard curves. For the measurement of sulfate, roots and shoots of rice plants were extracted with deionized water at 4°C for 1 h and heated at 95°C for 15 min. After centrifugation and purified through 0.22-µm Millipore filters, the supernatant was quantified on a Dionex ICS-600 ion chromatography system.
RNA extraction and quantitative real-time PCR. Total RNA was extracted from different plant tissues using a RNeasy plant extraction kit (Bioteke) following the manufacturer's protocol. First-strand cDNA was synthesized using 2 µg RNA using oligo dT(23) primer according to the manufacturer's instructions of the HiScript Q Select RT SuperMix for quantitative PCR kit (Vazyme). Quantitative real-time PCR was performed on a CFX96 Real-Time PCR Detection System (Bio-Rad) using an SYBR Green Master Mix kit (Vazyme) according to the manufacturer's instructions. The rice OsHistone H3 gene was used as the internal reference. The expression level of each gene was calculated as 2 -ΔCt relative to the internal reference. The primers used are listed in Supplementary Table 1.
Tissue and subcellular localization of OsASTOL1. To investigate the expression pattern of OsASTOL1, the GUS reporter activity was analyzed by histochemical staining. Roots and leaves of pOsASTOL1-GUS transgenic plants at the seedling stage were collected and incubated with a GUS staining solution for 2 h at 37°C. Leaves were decolorized by 90% acetone overnight. The expression levels of OsASTOL1 in the roots and leaves of cv. Kasalath and in various tissues/organs of cv. Nipponbare at different growth stages were quantified by quantitative real-time PCR. The samples were collected according to the method described by Sun et al. 60 .
To investigate the subcellular localization of OsASTOL1, the pEarleyGate 101-OsASTOL1-YFP and pEarleyGate 101-Osastol1-YFP constructs were transformed into rice protoplasts using polyethylene glycol (PEG)-mediated transformation 61 , or in tobacco (Nicotiana benthamiana Domin) leaf epidermal cells using Agrobacterium-mediated transformation 66 . The pEarleyGate 104 (N-YFP) vector containing a 35 S::YFP construct was used as a control. Protoplasts isolated from rice or tobacco were observed using a confocal laser scanning microscope (Zeiss). YFP fluorescence was observed at 525 nm for emission and 510 nm for excitation, and chlorophyll autofluorescence was detected at 575-640 nm for emission and at 546 nm for excitation.
Immunoblotting analysis. For immunological detection of OsASTOL1 in rice, total soluble proteins were extracted from 200 mg leaf sample with 1 ml plant protein extraction buffer (CW0885, CWBIO). Proteins were denatured by heating for 10 min in a boiling-water bath and separated on 12% SDS-PAGE. Subsequently, proteins were transferred onto a polyvinylidene difluoride (PVDF) membrane and immunoblots were probed with the following antibodies in the TBST solution (100 mM Tris-HCl, 150 mM NaCl, 0.1%(v/v) Tween 20, pH 7.4) with 5% nonfat dry milk. A rabbit anti-OAS-TL A antibody (against full-length Arabidopsis cytosolic OAS-TL A, 1:5000 dilution) and a rabbit anti-RbcL antibody (AS03037, Agrisera, 1:10000 dilution) were used as primary antibodies and HPR-conjugated goat anti-rabbit IgG (AT0097, CMCTAG, 1:10000 dilution) was used as secondary antibody. Chemiluminescence was performed using ECL Western Blotting Substrate (IF6747, CMCTAG) according to manufacturer's instructions.
Analysis of OAS-TL and SAT enzyme activity. To determine the OAS-TL enzyme activity, total soluble proteins were extracted from 200 mg rice leaves with 1 ml plant protein extraction buffer (CW0885, CWBIO). Recombinant OAS-TL proteins were expressed in E. coli after induction by 0.3 mM isopropyl-β-D-thiogalactopyranoside (IPTG) at 16°C for 16 h. Cells were sonicated and filtered through 0.45-µm Millipore filters. The supernatant (crude extract protein) was added to Ni-NTA packed columns (Sangon Biotech Co. Ltd, Shanghai) and recombinant proteins were eluted with a PBS buffer (20 mM Na 3 PO 4 , 100 mM NaCl, pH 7.4) containing 150 mM imidazole. All proteins were quantified with a BCA Protein Assay Kit (CW0014S, CWBIO). The OAS-TL enzyme activity was determined at 25°C in a total volume of 100 µl containing 50 mM HEPES/KOH (pH 7.5), 5 mM Na 2 S, 10 mM OAS, 5 mM DTT and 1-2 µl of crude extract proteins according to Heeg et al. 25 . Recombinant proteins were diluted 1:100 before the enzyme activity assay.
To determine the SAT enzyme activity, total soluble proteins were extracted from 200 mg rice leaves with 1 ml plant protein extraction buffer (CW0885, CWBIO). The SAT activity 53 was assayed as described before, but with some modification. The SAT activity was first assayed at 25°C in a total volume of 100 µl containing 50 mM HEPES/KOH (pH 7.5), 10 mM serine, 1 mM acetyl-CoA, 5 mM DTT and 70 µl protein extract. After reaction for 30 min, 50 µl 20% trichloroacetic acid (TCA) was added and the content was centrifuged. The pH of the supernatant was adjusted to 7.4 with 40 µl 2 M Tris, to which 50 µl 50 mM Na 2 S and 4 U purified AtOAS-TL A were added to ensure an excess of OAS-TL activity. The reaction mixture was incubated at 25°C for 20 min to allow OAS produced by SAT to be transformed to cysteine. The production of cysteine was determined with an acid-ninhydrin solution.
Size exclusion chromatography analysis. Untagged OsASTOL1 and OsAS-TOL1 S189N were expressed and purified according to Wirtz et al. 26 . Molecular weights of OsASTOL1 and OsASTOL1 S189N were determined by size exclusion chromatography (SEC) using a FPLC TM system. Briefly, 1 ml of OsASTOL1 or OsAS-TOL1 S189N (1 mg) was injected into the sample loop and was eluted through the HiLoad TM 16/60 Superdex 200 prep grade column (120 ml, GE Healthcare) in buffer (20 mM Tris-HCl, 150 mM NaCl, pH 7.5) at 1 ml/min flow rate. Calibration of the Superdex 200 column was performed with the high and low molecular weight gel filtration calibration kits (GE Healthcare) according to the manufacturer's instructions. The molecular weights of OsASTOL1 and OsASTOL1 S189N were calculated from the calibration curve with the measured elution volumes.
Sequence alignment and visualization of protein structure. Multiple sequence alignments of OAS-TL proteins were conducted using the ClustalW method in DNAMAN software. Visualization of protein structure and site-directed mutation of Arabidopsis cytosol OAS-TL A (PDB ID: 1z7w) were performed by PyMOL software.
In vitro pull-down assay. To investigate the interactions of OsASTOL1, OsAS-TOL1 S189N , AtOAS-TL A or AtOAS-TL A S102N with AtSAT5 and the dissociation effect by OAS, a revised in vitro pull-down assay was conducted. In brief, soluble proteins were extracted from cells harboring the pET-28a-His-AtSAT5 plasmid and added to a NiCl 2 containing HiTrap TM Chelating column (GE Healthcare). The column was washed with Buffer W (50 mM Tris-HCl, pH 8.0, 250 mM NaCl, 80 mM Imidazol), followed by washing with Buffer W containing 10 mM OAS (Solar Bio) to remove the bacterial OAS-TL. The crude extract containing untagged plant OAS-TL proteins was added to the above column containing the His-tagged AtSAT5 protein. The untagged OAS-TL protein was specifically eluted with Buffer W containing 10 mM OAS. After washing with Buffer W, the His-tagged AtSAT5 was eluted with Buffer E (50 mM Tris-HCl, pH 8.0, 250 mM NaCl, 400 mM Imidazol). The fractions obtained from different steps of elution were analyzed by SDS-PAGE followed by Coomassie staining.
For immunological detection of OAS-TL and SAT proteins, the His-AtSAT5 and S-tagged OAS-TL proteins were used to perform the same steps as described above but with a different Ni-NTA packed columns (Sangon Biotech Co. Ltd, Shanghai) and a new washing buffer (50 mM Tris-HCl, pH 8.0, 250 mM NaCl, 20 mM Imidazol). A rabbit anti-S-tag antibody (101290-T38, Sino Bio, 1:5000 dilution) and a mouse anti-His-tag antibody (AF5060, Beyotime,1:2000 dilution) were used as the primary antibodies.
Microscale thermophoresis analysis. Ten C-terminal residues 38 of AtSAT5 (AT5G56760, sequence: FISEWSDYII), required for binding with OAS-TL, was synthesized by GenScript Biotech (Nanjing). Recombinant His-tagged OAS-TL proteins were labeled with RED-tris-NTA 2 nd Generation Dye for 30 min. For the binding assay, 100 µM labeled His-tagged OAS-TL protein was incubated with a serial dilution of the ligand (OAS or AtSAT5C10 peptide) in the PBS buffer (pH 7.6) containing 0.05% Tween 20. Microscale thermophoresis (MST) analysis was performed on a Monolith NT.115 instrument (Nano Temper, Germany) using auto-detected LED power (5-90%) and MST power (Medium or High). The data were evaluated using the MO Affinity Analysis software.
Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
Data supporting the findings of this work are available within the paper and its Supplementary Information files. A reporting summary for this article is available as a Supplementary Information file. The genetic materials generated and analyzed during the current study are available from the corresponding author upon request. The source data underlying Fig. 1b, d, g-i, 2b-e, 3, 4b-g, 5, as well as Supplementary Figs. 1a, 2b-d, 3b, 4-6d-f, 10a, 12b, 14, 15b-e, 16-20d-l, 21c-e, and Supplementary Table 4 are provided as a Source Data file.