A Rice Glutamyl-tRNA Synthetase Modulates Early Anther Cell Division and Patterning1[OPEN]

OsERS1, a glutamyl-tRNA synthetase, modulates early rice anther development through affecting metabolic homeostasis and redox status. Aminoacyl-tRNA synthetases (aaRSs) have housekeeping roles in protein synthesis, but little is known about how these aaRSs are involved in organ development. Here, we report that a rice (Oryza sativa) glutamyl-tRNA synthetase (OsERS1) maintains proper somatic cell organization and limits the overproliferation of male germ cells during early anther development. The expression of OsERS1 is specifically detectable in meristematic layer 2-derived cells of the early anther, and osers1 anthers exhibit overproliferation and disorganization of layer 2-derived cells, producing fused lobes and extra germ cells in early anthers. The conserved biochemical function of OsERS1 in ligating glutamate to tRNAGlu is enhanced by its cofactor aaRS OsARC. Furthermore, metabolomics profiling revealed that OsERS1 is an important node for multiple metabolic pathways, indicated by the accumulation of amino acids and tricarboxylic acid cycle components in osers1 anthers. Notably, the anther defects of the osers1 mutant are causally associated with the abnormal accumulation of hydrogen peroxide, which can reconstitute the osers1 phenotype when applied to wild-type anthers. Collectively, these findings demonstrate how aaRSs affect male organ development in plants, likely through protein synthesis, metabolic homeostasis, and redox status.


OsERS1 Modulates Early Anther Development
Recent studies have shown the function of cell surface-localized leucine-rich repeat receptor-like kinases (LRR-RLKs) and their ligands, such as rice MULTI-PLE SPOROCYTE1 (MSP1) and TAPETAL DETERMI-NANT1-LIKE A (OsTDL1A)/MIL2, in determining early anther cell fate. These are orthologous to Arabidopsis EXCESS MICROSPOROCYTES1/EXTRA SPO-ROGENOUS CELLS and TAPETAL DETERMINANT1, respectively, and function in specifying anther cell identity and controlling cell numbers of tapetum and microsporocytes (Canales et al., 2002;Zhao et al., 2002Zhao et al., , 2008Nonomura et al., 2003;Yang et al., 2003;Hong et al., 2012a;Zhang and Yang, 2014). MSP1 and OsTDL1A were suggested to exert their effects in early anther cell differentiation via modulating redox status in rice (Yang et al., 2016).
In maize, early-stage anthers also require hypoxia to maintain proper division and determine cell fate. The maize GRX MSCA1 may maintain growth-generated hypoxic conditions to induce archesporial cells. Also, archesporial cells might silence mitochondria to lower the concentration of reactive oxygen species (ROS) and maintain reducing conditions (Kelliher and Walbot, 2012).
Aminoacyl-tRNA synthetases (aaRSs) are found in all protein-synthesizing organisms and can catalyze the ligation of amino acids and their cognate tRNA, an essential step for translation and amino acid dynamics (Ibba and Söll, 2000;Yamakawa and Hakata, 2010). Studies in mammals showed noncanonical and surprising roles of aaRSs in amino acid signaling, transcriptional control, and antiapoptosis (Guo et al., 2010). As housekeeping proteins, aaRSs are indispensable for plant development, but their detailed function in plant growth remains largely unclear. In Nicotiana benthamiana, silencing of Glutamyl-tRNA Synthetase (GluRS) and Seryl-tRNA Synthetase causes a severe yellowing phenotype in leaves due to significant reduction in the number of organelles . Mutation of a rice ValRS results in a reduced level of chlorophyll and abnormal chloroplast ribosome biogenesis (Wang et al., 2016). Among the 45 aaRSs in Arabidopsis, 21 are required for ovule development and embryogenesis (Berg et al., 2005). In addition, an Arabidopsis Cysteinyl-tRNA Synthetase (CysRS) expressed in the central cell of the female gametophyte may specify the fate of adjacent cells (Kägi et al., 2010). Noncanonical and surprising roles of aaRSs have been uncovered, such as amino acid signaling, transcriptional control, and antiapoptosis. Furthermore, eight members of aaRSs were found to exist in mammalian cytoplasm as a mul-titRNA synthetases complex (MSC) whose roles and mechanisms of formation are still unclear (Guo and Schimmel, 2013). However, the detailed role of aaRSs in plant organs and development is limited. Here, we report the critical function of the cytosol-and mitochondria-localized GluRS, OsERS1, in early anther cell proliferation and organization by affecting protein synthesis, amino acid-derived metabolism, and cellular redox status.

OsERS1 Limits L2-d Cell Proliferation and Maintains Anther Cell Organization
To explore the mechanisms underlying anther development in the cereal model plant rice, we identified a new male-sterile mutant, oryza sativa glutamyl-tRNA synthetase (osers1), because of the mutation in Glutamyl-tRNA Synthetase1 (see below). The osers1 mutant displayed no obvious phenotypic changes from the wild type in vegetative growth, inflorescence architecture, female reproductive organ, or seed shape (Supplemental Fig. S1,A and B). However, during the male reproductive stage, the mutant developed defective anthers that were pale and had fused lobes, resulting in the production of ∼20% viable pollen grains and an ∼10% seed-setting rate (Supplemental Fig. S1, C-F). The os-ers1 mutant showed a more tightly woven texture in the anther wall, and the pollen grains were shrunken and displayed rough, lumpy extine, in contrast to the wild type, which had a well-arranged cuticle on the exterior of the anther wall and mature pollen grains with smooth extine (Supplemental Fig. S1, G-N). All F1 plants of reciprocal crosses between osers1 and the wild type were normal, and the F2 offspring had an ∼3:1 segregation of normal versus abnormal anthers (116:42; χ 2 = 0.58, P > 0.05), suggesting that the osers1 mutants contain a unifactorial and recessive mutation.
From stage 3 to 5 of anther development, ∼40% of osers1 anthers had adaxial fused lobes, which was distinct from the butterfly-shaped four lobes in wild-type anthers (Fig. 1, A and B; Supplemental Fig. S2, A-F and I). At stage 3, transverse section analysis showed that each wild-type anther formed ∼6.3 archesporial cells (n = 11) enringed by a single layer of PPCs, while each osers1 anther produced ∼10 archesporial-like cells (n = 12) enringed by PPCs organized in an irregular mosaic pattern ( Fig. 1, C, D, I, and K). At stage 5, the wild-type archesporial cells divided further into ∼11 sporogenous cells (n = 15) surrounded by three circular inner layers derived from PPCs ( Fig. 1, E, G, I, and K). However, each osers1 anther produced ∼20 sporogenous-like cells (n = 13), and ∼50% of the anthers had disordered wall layers with irregular boundaries (Fig. 1,F,H,I,and K;Supplemental Fig. S2I). Notably, the total number of somatic and germ cells derived from the L2 layer in the osers1 mutant was increased during stages 3 to 5 (Fig. 1J). Some sporogenous-like cells were ectopically embedded in subepidermal layers (Fig. 1I;Supplemental Fig. S2,G and H).
The increased numbers of anther cells in the osers1 mutant prompted us to compare cell division activity between the wild type and the osers1 mutant. In situ hybridization showed stronger expression signals for HISTONE4 (H4), a cell-division marker for the S-phase of mitosis, and CYCLIN-DEPEDENT KINASE B2-1 (CDKB2;1), a marker of the G2/M phase transition (Umeda et al., 1999), in both somatic and germ cells of osers1 lobes, compared with the wild type ( Fig. 1, L-Q).
Reverse transcription-quantitative PCR (RT-qPCR) analysis also showed increased expression of H4 and CDKB2;1 in osers1 anthers (Supplemental Fig. S3B). Due to the lack of marker genes for cell lineage during early anther development, we chose MSP1 to examine the relationship of the MSP1 pathway and OsERS1. In situ hybridization showed similar distribution patterns of MSP1 in both wild-type and osers1 anthers: that is, preferential accumulation on anther wall cells rather than in central sporogenous cells (Supplemental Fig. S3A). RT-qPCR showed no significant difference in the expression of MSP1, as well as in that of two other genes involved in anther development, MIL1 and MIL2, between wild-type and osers1 anthers (Supplemental Fig. S3B). Unlike osers1, mutations of MSP1, MIL2, and MIL1 led to misidentified anther wall cell layers and complete male sterility (Nonomura et al., 2003;Hong et al., 2012a,b). These results indicate that the cell properties are still maintained in osers1 anthers, and OsERS1 is strongly implicated in cell proliferation and organization in the rice anther.

OsERS1 Encodes a Glutamyl-tRNA Synthetase
Using map-based cloning combined with the whole-genome sequencing-based MutMap method (Abe et al., 2012;Lü et al., 2015), we identified the mutated OsERS1 gene as having an insertion of A after A 539 in the LOC_Os10g22380 coding region (also known as Os10g0369000; see "Materials and Methods"); this resulted in a premature stop codon in the open reading frame (Fig. 2, A and B). The identity of OsERS1 was Figure 1. OsERS1 is essential for cell proliferation and organization in early rice anther development. A to H, Cytological comparison of anther development between the wild type (WT) and the osers1 mutant. S3 to S5 indicate anther developmental stages. A and B show anthers in stage 3, C and D show magnifications of the boxed regions in A and B, respectively, E and F show anther lobes in stage 4, and G and H show anther lobes in stage 5. Yellow asterisks indicate archesporial cells at S3 and sporogenous cells at S4 to S5. Bars = 15 μm. I, Diagrams of anther development from stages 1 to 5 in the wild type and the osers1 mutant. Disorganization of somatic cell layers and excess archesporial/sporogenous-like cells are shown in the mutant. J and K, Quantification of cell overproliferation in osers1 anther. J, Average L2-d cell counts of anther cross sections. K, Average archesporial/sporogenous cell counts of the anther cross sections. All error bars represent sd. S3 wild type, n = 11; S4 wild type, n = 13; S5 wild type, n = 15; S3 osers1, n = 12; S4 osers1, n = 14; S5 osers1, n = 13. Student's t test: *, P < 0.05 and **, P < 0.01. Ar, Archesporial cell; E, epidermis; eAr, excess archesporial-like cell; En, endothecium; eSp, excess sporogenous-like cell; ML, middle layer; Sp, sporogenous cell; T, tapetum. L to Q, In situ hybridization of cell division marker genes in stage 4 anthers. Bars = 50 μm.

OsERS1
Modulates Early Anther Development confirmed by the complementary assay, in which the male-sterile phenotype in the osers1 mutant was rescued by the expression of a genomic fragment that contained the wild-type OsERS1 gene (Fig. 2, C-E).
OsERS1 encodes a putative glutamyl-tRNA synthetase (GluRS) that contains an N-terminal GST_C (C terminus of GST) domain, a glutamyl-tRNA synthetase catalytic core, and a C-terminal tRNA-synt_C domain (Fig. 2B). RT-qPCR showed that OsERS1 was expressed in the wild type at low levels in mature organs but at high levels in young tissues, including seedlings and young inflorescences (Fig. 3A). In the anther, OsERS1 transcript levels were relatively high during stages 3 to 6 and decreased in later stages (Fig. 3, B and C). In situ hybridization detected OsERS1 mRNA in presumptive archesporial cells and neighboring L2-d cells before stage 3. During stages 3 to 5, when anther cells undergo rapid cell division, OsERS1 mRNA was detected in PPC and archesporial cells of stage 3, the SPC and sporogenous cells of stage 4, and the tapetum, middle layer, and sporogenous cells of stage 5. The expression of OsERS1 in germ cells decreased significantly after stage 5 ( Fig. 3C; Supplemental Fig. S4A). The rice genome has two OsERS1 homologs, LOC_Os01g16520 and LOC_Os02g02860, but these homologs had low expression in the anther and did not appear to have altered expression in the osers1 mutant, suggesting that the phenotype of the osers1 mutant was caused by aberrant OsERS1 in the anther as opposed to the two homologs (Supplemental Fig. S4B). Considering that protein synthesis is essential in the cytosol, chloroplast, and mitochondria, we performed protein The red A indicates an insertion after A 539 in oser1. NT, Nucleotides. Black rectangles represent exons. The asterisk means a stop codon. GST_C, Glutathione S-transferase C-terminal domain; GluRS_core, catalytic core domain of glutamyl-tRNA synthetase; tRNA-synt_C, tRNA synthetases class I anticodon binding domain. C to E, Complementation of the osers1 mutant. C, Inflorescences with mature seeds. D, Average seed-setting rates in panicles (n = 20 for each set). Error bars represent sd. E, Mature anthers. COM, Complementation; WT, wild type. Bars = 2 cm in C and 1 mm in E.
localization studies in rice protoplast cells and onion (Allium cepa) cells by coexpressing OsERS1-eGFP and the mitochondrial marker ScCOX4-mCherry (carried by vector CD3-991; Nelson et al., 2007). Our results showed that OsERS1 is localized mainly in the cytoplasm and is partially shared by the mitochondria (Fig.  3D; Supplemental Fig. S4C).
Phylogenetic analysis of 40 OsERS1 homologs from 21 species from diverse kingdoms clearly divided these proteins into two clades: the eukaryote-specific GluRSs (clade I) and the prokaryote-originated GluRSs (clade II; Fig. 3E; Supplemental Table S1). Intriguingly, most eukaryotes contain both types. Clade II GluRS contained only the catalytic domain, while clade I GluRSs showed more complex structures. The progressive appearance of novel domains, such as GST_C, tRNA-synt_C, repeats of the WHEP domain (initially discovered in human tryptophanyl-tRNA synthetase [TrpRS; W], histidyl-tRNA synthetase [H], and GluRSprolyl-tRNA synthetase [ProRS; EP]), and chimeric GluRS-ProRS in clade I, indicated multiple domainrecruiting events during evolution (Gao et al., 2010). In grass crops such as rice and maize, there seemed to be a duplication of GST_C containing GluRS, which we did not observe in other species (Fig. 3E), indicating that GluRS might have an expanded function for evolutionary adaption in higher organisms.

OsERS1 Functions as a Ligase between Glu and tRNA Glu
The synthesis of glutamyl-tRNA is a two-step reaction: activation of Glu by ATP and glutamylation of tRNA Glu . In the catalytic domain, GluRS has a core active site (RFAPE in OsERS1) that binds Glu and two key motifs called HIGH and KMSKS (LLSKR in OsERS1) that are involved in ATP binding and tRNA binding, respectively (Freist et al., 1997;Ibba and Söll, 2000). Sequence alignment showed that these three motifs are highly conserved in the 40 GluRS homologs To confirm the function of OsERS1 in glutamyl-tRNA synthesis, we performed an ATP-pyrophosphate (PPi) exchange assay, which showed that OsERS1 has PPi exchange activity and that its reaction rate is proportional to the concentration of tRNA Glu (Fig. 4B). An aminoacylation assay revealed that OsERS1 has the catalytic activity of aminoacylation of tRNA Glu with Glu ( Fig. 4, C and D), yielding a K m value (tRNA Glu as substrate) of 2.2 μm and a K cat /K m value of 0.977 s -1 μm -1 (Fig. 4E). OsERS1's affinity for tRNA Glu is similar to that of the Thermus thermophilus GluRS but lower than that of Rattus spp. liver and Escherichia coli (Freist et al., 1997). Also, OsERS1's K m value is lower than that of wheat (Triticum aestivum) GluRSs, possibly because the latter function as dimers (Thomes et al., 1983). However, it is still not clear whether GluRS can form dimers in rice.
Mutations of the three conserved motifs of OsERS1s, Arg-209 and Glu-213 in the active site region, His-219 and His-222 in the HIGH motif, and Ser-442 and Lys-443 in the KMSKS motif ( Fig. 4A; Supplemental Fig. S5), caused significantly reduced glutamylation activity and enzyme efficiency (K cat /K m ; Fig. 4, C-E). These results suggest that OsERS1 has catalytic activity in tRNA Glu glutamylation and that the RFAPE, HIGH, and KMSKS motifs are required for OsERS1's enzymatic function.

The Cofactor OsARC Stimulates the Catalytic Activity of OsERS1
The GST_C domain of GluRS is noncatalytic because it lacks thioredoxin, but it was shown to link the enzyme to protein complexes (Dixon et al., 2002). In yeast and mammals, GluRS/GluRS-ProRS interacts with the aaRS cofactor Arc1p/p18 to form a complex and modulate enzyme activities (Galani et al., 2001;Graindorge et al., 2005;Gao et al., 2010). Rice has a single homolog of yeast Arc1p, OsARC (encoded by LOC_Os01g60660), which contains a heterogeneous GST_C domain (Fig. 5A). To examine whether OsERS1 can form a complex with OsARC, we conducted yeast two-hybrid and in vitro pull-down assays; these confirmed that OsERS1 can interact with OsARC and that the interaction requires the GST_C domains (Fig. 5, B and C;. Bimolecular fluorescence complementation (BiFC) showed that only the cells expressing full-length OsERS1-nYFP and OsARC-cYFP displayed yellow fluorescence signals (Fig. 5D), verifying this interaction in vivo. OsERS1-eGFP and OsARC-mCherry were highly colocalized when coexpressed in onion cells (Fig. 5E), supporting our hypothesis that OsERS1 and OsARC can form a complex. Moreover, compared with that of OsERS1 alone or OsERS1-ΔOsARC (truncated OsARC protein without the GST_C domain; Fig. 5, F and G), the OsERS1-OsARC complex had ∼2.5 times the ATP-PPi exchange rate of Glu and ∼1.4 times the tRNA Glu aminoacylation efficiency in the plateau stage, indicating an enhancement effect promoted by OsARC. In addition, the GluRS and cofactor pairs are conserved components of MSCs in various organisms (Laporte et al., 2014). However, the structure and composition of plant MSC have not been dissected.

Mutation of OsERS1 Leads to Metabolic Alterations in Anthers
The substrate for OsERS1, Glu, occupies a central position that links key metabolic pathways (Newsholme et al., 2003). Loss of function of OsERS1 may cause an accumulation of free Glu that is not used for protein synthesis in the anther. To test this possibility, we conducted nontargeted metabolomics profiling in wild-type and osers1 anthers. A total of 297 metabolites were identified, covering all major primary and several secondary metabolism pathways (Supplemental Table S2). Consistent with the deficiency in ligating Glu with tRNA Glu , the osers1 mutant had 2.44 times the level of Glu compared with the wild type. With the exception of glutathione in both reduced (GSH) and OsERS1 Modulates Early Anther Development . For B, total RNA was extracted from whole rice flowers before stage 6 and from anthers after stage 7. All error bars represent sd (n = 3). C, In situ hybridization of OsERS1 in the wild-type anther at different stages. A sense probe was used as the control. Ar, Archesporial cell; ML, middle layer; Sp, sporogenous cell; T, tapetum. Bars = 15 μm. D, Subcellular localization of OsERS1 in rice protoplasts. GFP was used as a control. Vector CD3-991 containing the mitochondria-specific protein ScCOX4 fused with mCherry served as the organelle marker. Chl, Chloroplast; BF, bright field. Bars = 10 μm. E, Neighbor-joining tree of OsERS1 and its homologs showing the progressive recruitment of oxidized (GSSG) status, the levels of Glu family members, including Gln, α-ketoglutarate (α-KG), His, and Arg, were increased in the osers1 mutant. One Glu derivative, α-KG, also is an intermediate of the tricarboxylic acid cycle and links the metabolism of amino acids and carbohydrates (Jeong et al., 2013). As a result, most metabolites of the tricarboxylic acid cycle, such as oxaloacetate (OAA) and malate, were increased in the osers1 mutant ( Fig. 6A; Supplemental Fig. S6B; Supplemental Table S2). This is also consistent with changes in the expression of genes involved in Glu metabolism and the tricarboxylic acid cycle, such as genes encoding Gln synthetases (GS2 and GS3), Glu dehydrogenase (GDH1), and oxoglutarate dehydrogenase (ODH2; Supplemental Table S3). Besides the Glu family, the Leu, Ser, and Asp families derived from pyruvate, phosphoglycerate, and OAA also were increased in the osers1 mutant. We propose that increased levels of amino acids in the mutant may have been caused by the accumulation of Glu, OAA, and pyruvate, which are important precursors for amino acid synthesis (  Table S2).
In addition, we detected the expression level of 12 aaRS genes highly expressed in reproductive organs. Most of the aaRS genes showed subtle expression changes between the wild type and the osers1 mutant, novel functional domains. AA, Amino acids; GST_C, C terminus of GST; GluRS_core, glutamyl-tRNA synthetase catalytic core; tRNA-synt_C, tRNA synthetase class I anticodon binding domain; ProRS, prolyl-tRNA synthetase.  Wild-type OsERS1, R 2 = 0.99; OsERS1 mutant M1, R 2 = 0.97. E, Kinetic constants determined by tRNA Glu charging assays. Kinetics of the OsERS1 and OsERS1 M1 to M3 aminoacylation reaction was measured using varied concentrations of tRNA Glu , and reactions were conducted for 10 min under standard conditions. Reciprocal initial velocity was plotted against the reciprocal tRNA Glu concentration according to Lineweaver-Burk to calculate the corresponding K m values. All error bars represent sd (n = 3).
OsERS1 Modulates Early Anther Development except for CysRS1 and TrpRS2, which had lower expression in the osers1 mutant (Supplemental Fig. S7), implying that the amino acid changes in the mutant are caused mainly by the loss of function of OsERS1. By contrast, we observed decreased levels of glycolysis intermediates, including Glc, Glc-6-P, and Fru-6-P, in the osers1 mutant ( Fig. 6; Supplemental Fig. S6B); this indicates a higher consumption of these compounds in the osers1 mutant. Decreased Glc supply and less lactate production suggest that pyruvate was likely used by the tricarboxylic acid cycle and for amino acid biosynthesis. Increased expression of PPP genes and the accumulation of purine and pyrimidine indicate that more intermediates of glycolysis were used to support nucleotide biosynthesis in osers1 anthers (Supplemental Fig. S6C; Supplemental Tables S2 and S3). Metabolic analysis showed abnormal fatty acid synthesis in the osers1 mutant, such as cutin and wax, which is consis-tent with the scanning electron microscopic analysis of osers1 anther wall and pollen grain (Supplemental Fig. S1, G-N, and S6D; Supplemental Table S2).
OsERS1 Modulates Early Anther Development and less lactate production. The increased Glu family amino acids, as well as those derived from OAA and pyruvate, contribute to the significant changes in biosynthesis of the amino acid pathway in osers1 anthers ( Fig. 6B; Supplemental Fig. S6A; Supplemental  Table S2). Notably, Glu may provide mitochondrial anaplerosis to the tricarboxylic acid cycle to stimulate mitochondrial activities, which produces ROS as an inevitable by-product. Consistent with this, decreased cellular antioxidants (GSH/GSSG and ascorbate) suggest a disturbed redox status in osers1 anthers (Fig.  6B). Corresponding with the activated cell division in osers1 anthers, more carbohydrates are consumed in glycolysis to generate energy and provide more intermediates for the PPP, which enhances the synthesis of nucleotides and macromolecules. These results reveal that the metabolic homeostasis is disturbed in osers1 anthers and that OsERS1 plays key roles in amino acid-related metabolism.

Altered Redox Status Affects Early Anther Development
Global metabolomics profiling suggested that ROS might be produced by mitochondrial activities in the osers1 mutant. To verify this, we first measured hydrogen peroxide (H 2 O 2 ), a stable ROS molecule implicated in cell proliferation and differentiation (Møller et al., 2007;Tsukagoshi et al., 2010). Fluorometric quantitation assays showed that H 2 O 2 content in the osers1 mutant at stages 4 to 5 was ∼900 pmol mg −1 , twice that in wild-type anthers at the same stage (Fig. 7A, left). Also, the precursor of H 2 O 2, superoxide radical, was increased slightly in the osers1 mutant (Fig. 7A, right), which contributed to the redox status change in the os-ers1 mutant. In addition, 3,3′-diaminobenzidine (DAB) staining of wild-type anthers showed that the H 2 O 2 signal was preferentially present in middle layer and endothecium cells until stage 5 (Fig. 7, B-E). However, osers1 anthers displayed detectable H 2 O 2 in the anther primordium before stage 3 (Fig. 7F) and strong H 2 O 2 signals in anther cells at stages 4 to 5 (Fig. 7, G-I). In later stages, even though the distribution of H 2 O 2 was similar in both wild-type and osers1 anther wall layers, stronger H 2 O 2 signals were seen in osers1 anther wall layers (Supplemental Fig. S8). The lower expression levels of genes encoding ascorbate peroxidases in the mutant (Supplemental Table S3) may have contributed to the accumulation of ROS.
To test whether the increased level of H 2 O 2 in the osers1 mutant causes the abnormal division and patterning of L2-d cells, we injected 100 μm H 2 O 2 and 1 mm of the ROS-removal reagent potassium iodide (KI) into young inflorescences of the wild type and the osers1 mutant (Supplemental Fig. S9). Transverse section observations revealed that each stage 4 wild-type anther treated with water showed a slight increase in the number of sporogenous-like cells, from 7.5 (n = 16) to 9.2 (n = 12), which may have been triggered by wounding (Fig. 7, J, K, and S). Remarkably, the number of L2-d cells in each wild-type anther treated with H 2 O 2 increased from 109 (n = 16) in untreated anthers to 135.4 (n = 18), and sporogenous-like cell numbers increased from 7.5 to 15.4. Therefore, treatment with 100 μm H 2 O 2 in the wild type led to overproliferation of L2-d cells, ectopic sporogenous-like cells, and disordered parietal cell layers, phenocopying the defects of the osers1 mutant (Fig. 7, L, R, and S). Wild-type anthers treated with KI also showed increases in the number of L2-d cells to 132.7 and sporogenous-like cells to 16.8 (n = 17), while parietal cells were organized differently compared with those treated with H 2 O 2 (Fig. 7, M, R, and S). Compared with that of the untreated anther (139.3 L2-d cells and 18.5 sporogenous-like cells; n = 14), in the osers1 mutant, each anther treated with water showed a slight decrease of L2-d cell numbers to 131.5 and sporogenous-like cell numbers to 14.5 (n = 13; Fig. 7, N, O, R, and S). However, each osers1 anther treated with H 2 O 2 and KI showed a reduced number of L2-d cells and sporogenous-like cells (H 2 O 2 , 102.9 and 12.7, respectively [n = 16]; KI, 110.5 and 14.5, respectively [n = 15]). Aside from changes in cell numbers, we observed highly vacuolated cells in impaired osers1 anthers (Fig. 7, P-S). We reasoned that elevations of ROS by H 2 O 2 injection and neutralization of ROS by KI injection both disturbed ROS homeostasis, which may contribute to repressed cell division, and that the osers1 anthers appear vulnerable to changes in redox status. Together, these results suggest that changed redox status disturbs early cell division in rice anthers.

DISCUSSION
Early anther development is a complex event characterized by rapid cell division and fate specification to support the development of functional anther layers that are properly organized within the anther (Zhang and Yang, 2014;Walbot and Egger, 2016). In this study, we revealed that OsERS1, a glutamyl-tRNA synthetase, functions in early anther cell division and patterning, likely through maintaining physiological homeostasis of amino acids and carbohydrate metabolism as well as redox status (Fig. 8).
Our genetic and cytological data demonstrate that loss of function of OsERS1 results in fused anther lobes with supernumerary L2-d cells and disarranged anther wall cell layers. In osers1 anthers, L2-d cell division was accelerated, but the L1 layer remained normal and formed a rigid enclosure, which aggravated the disarranged cell layers and created abaxial fused lobes. In spite of its putative housekeeping function, expression of OsERS1 shows spatial-temporal characteristics rather than a constitutive pattern. In the anther, OsERS1 transcripts are initially expressed in cells of presumptive lobes at the four primordial corners. Then, OsERS1 mRNA is detected in both parietal and germ cells, preferentially in those undergoing periclinal division or newly formed cell layers (Fig. 3C). After stage 5, the germ cells develop into microspores via meiosis, and the abundance of OsERS1 mRNA in germ cells decreased to a very low level, indicating that OsERS1 mainly affects early cell division and patterning. This expression pattern is in agreement with the phenotype of increased L2-d cells and disarranged cell layers in the osers1 mutant.

OsERS1 Modulates Early Anther Development
Anther defects in the osers1 mutant occur earlier than those in the reported rice mutants with abnormal early anther development, such as gamyb-4, udt1 (undeveloped tapetum1), and tip2 (tdr interacting protein2; Jung et al., 2005;Liu et al., 2010;Fu et al., 2014). We showed that there are no significant changes in the expression of MSP1, MIL2, and MIL1 in the osers1 mutant compared with the wild type, and previous studies showed that OsERS1 has no obviously altered expression in msp1, mil2, gamyb-4, udt1, or tip2 mutants (Jung et al., 2005;Aya et al., 2009;Liu et al., 2010;Fu et al., 2014;Yang et al., 2016). Therefore, we propose that OsERS1 plays an independent role with known regulators in early anther development. Although loss of function of OsERS1 causes abnormal anther morphogenesis and decreased pollen vitality, the mutant still can form a small number of seeds with normal shape, suggesting that fertilization is not completely disturbed. However, we cannot exclude the possibility that female gametophyte development also is affected.
We confirmed the in vitro activity of OsERS1 in catalyzing the ligation between Glu and tRNA Glu . OsERS1 is dominantly localized in the cytosol and partially shared by mitochondria. The product, glutamyl-tRNA, is an essential substrate for protein synthesis. Compared with the prokaryotic (E. coli) GluRS (Freist et al., 1997), OsERS1 possesses a relatively lower affinity to tRNA Glu . One explanation is that the additional GST_C domain may change space conformation and reduce substrate affinity. However, the interaction between OsERS1 and its cofactor OsARC overcomes this disadvantage by enhancing the activity of OsERS1 for synthesizing glutamyl-tRNA (Fig. 5). In the early anther, loss of function of OsERS1 may impair translational processes in the cells that normally accumulate high OsERS1. However, the osers1 anthers maintain partial fertility and produce viable pollen, suggesting that other components may remedy the absence of OsERS1 to maintain protein synthesis in the anther.
Metabolic status is associated with tumorigenesis, stem cell renewal, pluripotency maintenance, and differentiation (Cairns et al., 2011;Schuster et al., 2014;Sperber et al., 2015;Ito and Ito, 2016). In plants, glycolysis and mitochondrial metabolism activate the Arabidopsis root meristem via Glc-target of rapamycin signaling (Xiong et al., 2013). In maize, the anther primordium is more likely glycolysis dependent, as the enclosed niche provides hypoxic conditions for archesporial cells (Kelliher and Walbot, 2012). We speculate that the hypoxia status specifying cell identity and cell division must be well controlled by the specific metabolic pathway within anther cells. In this study, we showed that, as an important metabolic regulatory node, OsERS1 is involved in intracellular Glu dynamics, and Glu could refuel the tricarboxylic acid cycle that occupies a central position in multiple metabolic pathways. Within this context, the Arabidopsis cytosolic GluRS is an important target of 14-3-3 proteins in carbon/nitrogen metabolism and sugar-sensing pathways. GluRS is degraded when sugar starvation occurs in order to release Glu and support cell survival (Cotelle et al., 2000;Coruzzi and Zhou, 2001). In this work, we showed that the mutation of OsERS1 disturbs the homeostasis of the metabolic network, as evidenced by increased amounts of amino acids, tricarboxylic acid components, and active aerobic metabolism. We speculate that such a metabolic shift acts as a signal to remove the limitation on cell proliferation in the anther lobe. Consistent with this, some mutants of 21 Arabidopsis aaRSs showed defective female gametogenesis and embryo development but no obvious defect during male gametogenesis except for in the ova9 mutant, which had a mutation in GlnRS (Berg et al., 2005). However, how these specific aaRSs are involved in reproductive development remains unclear. Although we observed no obvious change of other aaRS genes in the osers1 mutant (Supplemental Fig. S7), we cannot rule out the function of other aaRSs in male reproduction.
ROS exert dose-dependent impacts on cells. A lower or moderate basal level of ROS is essential for cell proliferation and differentiation, but higher levels of ROS are cytotoxic, causing cell damage and death (Sena and Chandel, 2012;Schieber and Chandel, 2014;Mittler, 2017). In rice anthers, the ROS molecule H 2 O 2 is initially accumulated in the middle layer at stage 5; this may act as a signal that triggers cell differentiation (Zhang and Yang, 2014;Yang et al., 2016). In this study, we observed the production of H 2 O 2 in osers1 anthers as early as the formation of archesporial cells, which may cause the overproliferation of anther cells and disorganized cell layers (Fig. 7). In support, disruption of the redox status in wild-type anthers by injecting low concentrations of H 2 O 2 resulted in a mimic of the phenotype of the osers1 mutant (Fig. 7). Similarly, in the shoot apical meristem of Arabidopsis, elevating H 2 O 2 dramatically enlarges meristem size by increasing cell numbers in the peripheral zone (Zeng et al., 2017). In maize, treatment with a high concentration of H 2 O 2 (1 mm) on the anther inhibits cell division in both somatic and germ cells (Kelliher and Walbot, 2012), demonstrating multifaceted regulatory aspects of ROS in cell division. The study in maize also indicates that the anther primordium may use glycolysis as a major energy source, rather than through Glc oxidation in mitochondria to avoid ROS pressure and maintain genomic integrity (Kelliher and Walbot, 2012). In this study, we demonstrated that increased ROS causes the overproliferation of early anther cells in the mutant, suggesting that reductive redox status is essential to maintain normal cell division and organization during rice anther development.
In summary, we discuss the role of a conserved glutamyl-tRNA synthetase, OsERS1, in limiting anther cell division and maintaining cell patterning (Fig. 8).
The catalytic activity of OsERS1 in ligating Glu to tRNA Glu can be enhanced by its cofactor OsARC. Loss of function of OsERS1 causes changes in a variety of metabolites, such as the accumulation of amino acids and tricarboxylic acid cycle intermediates, and leads to aerobic metabolic features in the mutant. Together with the H 2 O 2 assays, we concluded that OsERS1 maintains normal early anther cell division and patterning, likely through affecting protein synthesis, amino acid homeostasis, and redox status in rice. Our study reveals how the aaRSs control the cell division and organ development derived from meristematic cells in plants.

Materials, Plant Growth, and Phenotypic Analysis
The osers1 mutant was isolated from a rice (Oryza sativa ssp. japonica) mutant library in the cv 9522 background (Chen et al., 2006). Rice plants were grown in the artificial atmospheric chambers on the Waite Campus, University of Adelaide, and in a paddy field at Shanghai Jiao Tong University. Plants and flowers were photographed with a Nikon E995 digital camera and a Motic K400 dissecting microscope. Photography of plant materials and observation of anthers were done as described by Li et al. (2006) and imaged with a Nikon Ni-E optical microscope and a Philips XL20 scanning electron microscope. Cell counts were determined based on cross sections from more than 10 anthers, and statistical analysis included all four lobes of whole anthers.

Map-Based Cloning and Complementation of the Mutant
For fine-mapping, we identified 158 mutants from an F2 population that was generated from a cross between the osers1 mutant (japonica) and cv 9311 (indica). To map the OsERS1 locus, we used bulked segregation analysis and developed insertion-deletion molecular markers based on the sequence differences between rice cultivars, described in the National Center for Biotechnology Information. The OsERS1 gene was initially located to chromosome 10 between two insertion-deletion molecular markers, 10tys14 and 10tys2, and finer mapping narrowed down the gene to the region between RM25267 and 10tys28. Mapping combined with genome sequencing was performed using the previously published method (Abe et al., 2012;Lü et al., 2015).
For functional complementation, an 8.3-kb genomic DNA fragment containing the entire OsERS1 coding region, a 3.2-kb upstream sequence, and a 1.7-kb downstream sequence was amplified from a BAC clone and subcloned through EcoRI and BstElI into the binary vector pCAMBIA1301 (p1301), which carries a hygromycin resistance marker, to generate p1301-OsERS1. We induced calli from young panicles of the homogenous osers1 mutant and transformed the calli with Agrobacterium tumefaciens EHA105, which carried the p1301-OsERS1 OsERS1 Modulates Early Anther Development plasmid. At least 25 independent transgenic lines were obtained after PCR genotyping using the primers listed in Supplemental Table S4.

RT-qPCR
We extracted total RNA from plants using the TRIZOL reagent (Invitrogen), following the manufacturer's instructions. cDNA was synthesized from 4 μg of total RNA using a reverse transcriptase kit (Bio-Rad). SYBR Green Mix (Kapa) was used for RT-qPCR on a QuantStudio Flex 6 (Life Technologies) machine. The level of ACTIN2 mRNA was used as the internal control. For expression analysis of OsERS1, aaRSs, metabolism-related genes, or anther developmental genes, we denatured the DNA at 95°C for 10 min, followed by 42 to 45 cycles at 95°C for 10 s, 56°C for 15 s, and 72°C for 15 s. The RT-qPCR results shown are from representatives of experiments repeated at least three times with similar results. Primer sequences are listed in Supplemental  Table S4.

Subcellular Localization
35Spro::OsERS1-eGFP (full-length OsERS1 cDNA fused with eGFP) was generated in pCAMBIA1301 and expressed in rice protoplasts and onion (Allium cepa) epidermal cells using PEG-mediated transfections and a particle delivery system, respectively. The construction of the organelle marker (CD3-991) was described by Nelson et al. (2007). For rice protoplast cells, stem tissues of 14-d-old seedlings were processed following the protocol described by Zhang et al. (2011). For onion epidermal cells, the onion was cut into 2-× 2-cm squares and placed on a plate for 4 h with Murashige and Skoog medium containing 40 g L −1 d-mannitol. For transformation, 5 μg of each plasmid DNA was coated with 1-μm gold particles (Bio-Rad) and delivered into onion epidermal cells using a PDS-1000/He particle delivery system (Bio-Rad). Fluorescence microscopy was performed using a Leica TCS SP5 laser scanning confocal microscope. Fluorescence signals for mCherry (excitation, 587 nm; emission, 580-630 nm; 15% power), eGFP (excitation, 483 nm; emission, 500-530 nm; 15% power), and chloroplasts (excitation, 504 nm; emission, 650-750 nm; 15% power) were detected with a 20× (dry) objective for onion cells and a 63× (oil) objective for protoplast cells.

Phylogenetic Analysis and Motif Alignment
We constructed a phylogenetic tree with aligned full-length sequences of homologs of OsERS1. MEGA (version 7.0) and the neighbor-joining methods were used with a p-distance model and pairwise deletion and bootstrap (Gap Ktuple, 2; length penalty, 5; Window, 4; Diagonals, 4; 1,000 replicates). The maximum parsimony method of MEGA also was used to support the neighbor-joining tree using the default parameter. Amino acid sequences from regions 199 to 228 and 434 to 454 in OsERS1 were used for motif alignment by MEGA.

Recombinant Protein Purification
Wild-type and mutant OsERS1 were cloned into the pET32(a) vector for expression in the BL21 (DE3) Escherichia coli strain. Expression of the wild type and its variants was induced in Luria-Bertani broth with 0.5 mm isopropyl β-d-1-thiogalactopyranoside for 4 h at 23°C. We purified the proteins with Ni-NTA affinity resin (Qiagen) according to standard procedures and dialyzed against storage buffer containing 50 mm Tris-HCl (pH 7.5), 100 mm NaCl, 10 mm 2-mercaptoethanol, 3 mm MgCl 2 , and 50% (v/v) glycerol. The coding sequences of OsARC and ΔOsARC (deletion of GST_C) were fused to the N terminus of maltose-binding protein (MBP) in the vector pMAL2c-x. We expressed the fusion proteins in E. coli (strain BL21) and purified them using amylose resin (New England Biolabs) according to the manufacturer's instructions. Supplemental Table S4 lists the primer sequences for the constructs.

ATP-PPi Exchange and Aminoacylation
The ATP-PPi exchange reaction mixture contains 100 mm Tris-HCl (pH 7.8), 12 mm MgCl 2 , various concentrations of tRNA Glu , 5 mm l-Glu, 4 mm ATP, 2 mm [ 32 P]tetrasodium pyrophosphate (PerkinElmer), 0.1 mg mL −1 BSA, and 100 nm His-OsERS1 with or without 100 nm MBP-OsARC (or MBP-ΔOsARC, deleted for GST_C). The formation of [ 32 P]ATP at 37°C was followed at various time intervals in 9-μL aliquots. Samples at specific time points were terminated using 1 mL of quenching buffer containing 2% (w/v) activated charcoal, 3.5% (v/v) HClO 4 , and 50 mm tetrasodium pyrophosphate. The charcoal suspension was put through a Whatman GF/A filter, washed four times with 5 mL of water, and rinsed with 10 mL of 100% ethanol. The charcoal powder on the filters was dried, and the synthesized [ 32 P]ATP was quantified using a scintillation counter (Beckman Coulter; Chen et al., 2000).
We performed aminoacylation reactions in a mixture of 100 mm Tris-HCl (pH 7.8), 30 mm KCl, 12 mm MgCl 2 , 2 mm DTT, 4 mm ATP, 20 mm [ 3 H]Glu (15 Ci mm −1 ; PerkinElmer), 15 μm tRNA Glu , and 100 nm His-OsERS1 with or without 100 nm MBP-OsARC (or MBP-ΔOsARC, deleted for GST_C) at 37°C. We used 0.5 to 20 μm tRNA Glu for kinetic constant analyses of OsERS1 and its variants (mutations of three key motifs). Reactions were initiated with the addition of the enzyme and were conducted in a 37°C heat block. Aliquots (9 μL) were taken at different time points and quenched on Whatman filter pads that had been presoaked with 5% (w/v) TCA. The pads were washed three times for 10 min each time with cold 5% (w/v) TCA and once with cold 100% ethanol; the washed pads were then dried. Radioactivity was quantified in the scintillation counter. The mixture of OsERS1 and OsARC was incubated at room temperature for 30 min to allow complexation before ATP-PPi and aminoacylation assays (Chen et al., 2001).

Yeast Two-Hybrid, Pull-Down, BiFC, and Colocalization Assays
We performed yeast two-hybrid experiments according to the manufacturer's instructions (Clontech). The coding sequences of OsERS1 and ΔOsERS1 (deleted for GST_C) were fused in frame with the GAL4 DNA-binding domain of the bait vector pGBKT7. The coding sequences of OsARC and ΔOsARC were cloned in the prey vector pGADT7. Each bait-prey pair was cotransformed into the yeast strain AH109. To analyze protein-protein interactions using His auxotrophy assays, yeast colonies were patched in duplicate onto His-and His+ plates and kept at 30°C for 2 to 3 d.
The cDNAs of OsERS1, ΔOsERS1, and OsARC were cloned individually into pSAT1-nEYFP-N1 and pSAT1-cEYFP-C1-B vectors that contain either Nor C-terminal eYFP fragments. The method for BiFC assays using onion epidermal cells is described above ("Subcellular Localization"). YFP fluorescence was imaged at an excitation wavelength of 514 nm and emission wavelengths of 525 to 546 nm.

Metabolomics Analysis
Over 120 mg of rice anther from wild-type and osers1 plants was collected, frozen with liquid nitrogen, and stored at -80°C until metabolomics analysis. We ground samples into a fine powder and analyzed methanol extracts from a 40-mg sample using the analytical platform composed of ultra-HPLC-tandem mass spectrometry and gas chromatography-mass spectrometry. METABOLON performed the experiments and raw data analysis (http://www.metabolon. com). Metabolites were named after standard references (http://www.kegg.jp).

ROS Measurement and H 2 O 2 Staining
We collected rice spikelets from wild-type and osers1 plants and sorted them according to spikelet length. Samples were ground into a fine powder and processed according to the instructions of the Hydrogen Peroxide Assay Kit (abcom; catalog no. ab102500). Fluorometric readings were collected from a POLARstar OPTIMA device (BMG Labtech). Three independent experiments were performed to calculate mean values and sd. Aliquots of sample powder were suspended in 65 mm potassium phosphate buffer (pH 7.8) to extract superoxide radicals. Following a modified method (Wang and Jiao, 2000), the reaction mixture (0.5 mL of extract, 0.7 mL of potassium phosphate buffer, and 0.3 mL of 10 mm hydroxylammonium chloride) was incubated at 25°C for 60 min. Then, 0.5 mL of the above reaction mixture was added to 0.5 mL of 19 mm sulfanilic acid and 0.5 mL of 1% α-naphthylamine. After a 20-min coloring reaction at 25°C, the optical density was determined at 530 nm in a spectrophotometer. The linear standard curve was determined by 0 to 5 μm NaNO 2 reacted with sulfanilic acid and α-naphthylamine.
Freshly collected anthers were carefully immersed in the fixative solution O.C.T. Compound (Sakura) and frozen in liquid nitrogen. Embedded fresh tissues were sectioned into 9-μm-thick sections using a Leica CM1800 cryostat. Production of H 2 O 2 was performed by incubating sectioned anthers in the DAB solution (Vector Laboratories) for 3 to 4 min, optimized from the manufacturer's instructions, and photographed with a Nikon Ni-E optical microscope.

Anther Treatments
We removed the outermost whorls of leaves from wild-type and osers1 stems to locate the inflorescences. Next, 1 mm KI and 100 μm H 2 O 2 solutions were freshly prepared and injected through leaf whorls using an 18-gauge needle into the airspace about 5 cm above the first stem node, where the young inflorescence was developed (Supplemental Fig. S9). Water injection served as a control. After 72 h of treatment, rice spikelets were fixed, embedded in resin, and sectioned as described above.

Supplemental Data
The following supplemental materials are available. Figure S1. Phenotypic analysis of the osers1 mutant. Figure S2. Cytological analysis of wild-type and osers1 anthers. Figure S3. Expression of genes that encode regulators of early anther development in osers1 anthers. Figure S4. Molecular characteristics of OsERS1.

Supplemental
Supplemental Figure S5. Alignment of conserved motifs among GluRSs. Figure S6. Shift of major metabolic pathways in osers1 anthers.

Supplemental
Supplemental Figure S7. RT-qPCR detection of expression levels of aaRSs genes. Figure S8. DAB staining analysis for H 2 O 2 in anther lobes. Figure S9. Chemical treatments to anthers. Figure S10. Full scan figures of the gel and western blot. Table S1. Alignment of GluRS amino acid sequences. Table S2. Heat map of metabolite changes between the wild type and the osers1 mutant. Table S3. RT-qPCR detection of the expression changes of genes in metabolic pathways between the wild type and the osers1 mutant. Table S4. Sequences of primers used in this study.