Molecular Control of Redox Homoeostasis in Specifying the Cell Identity of Tapetal and Microsporocyte Cells in Rice

In flowering plants, male reproduction occurs within the male organ anther with a series of complex biological events including de novo specification of germinal cells and somatic cells, male meiosis, and pollen development and maturation. Particularly, unlike other tissue, anther lacks a meristem, therefore, both germinal and somatic cell types are derived from floral stem cells within anther lobes. Here, we review the molecular mechanism specifying the identity of somatic cells and reproductive microsporocytes by redox homoeostasis during rice anther development. Factors such as glutaredoxins (GRXs), TGA transcription factors, receptor-like protein kinase signaling pathway, and glutamyl-tRNA synthetase maintaining the redox status are discussed. We also conceive the conserved and divergent aspect of cell identity specification of anther cells in plants via changing redox status.

As other flowering plants, rice (Oryza sativa) male gametophyte development occurs within the male organ, stamen which contains anther and the supporting filament. At the early stages from stage 1 to stage 7 referring to staging by Zhang et al., (2011), the most critical events are the cell division and cell fate specification for the formation of somatic anther wall cell layers and reproductive cells, microspore mother cells (MMC), also called as pollen mother cells (PMC). At stage 1, three cell layers called as Layer 1 (L1), L2 and L3 are formed from an anther primordium. L2 cells perform cell division and form two layers of L2-derived (L2-d) cells while L1 differentiates into epidermis at stage 2. The outer L2-d differentiates into primary parietal cells (PPC), while the inner one differentiates into archesporial (Ar) cells at stage 3. PPC undergoes an asymmetric cell division and produces the endothecium and secondary parietal cells (SPC), while Ar cells differentiate into sporogenous cells (Sp) at stage 4. SPC undergoes symmetric cell division and cell differentiation into middle layer and tapetum at stage 5. Till stage 5, the Sp cells localize at the center of each anther lobe surrounded by four anther somatic wall layers: from the surface to interior, the epidermis, endothecium, middle layer, and tapetum. Then the meiosis of MMC occurs at stage 7 Zhang and Yang 2014). The tapetum, the innermost of the four sporophytic layers of the anther wall directly contacts with the developing gametophytes and acts as a nutritional source for the development of microspore by undergoing degeneration triggered by programmed cell death (PCD) from stage 7 to stage 11. At stage 8, after two rounds of cell division during meiosis, the tetrad covered by callose is formed. At stage 9, the haploid microspore is released into the lobe after callose degradation, followed by undergoing mitosis and pollen wall development and maturation with the accumulation of starch and lipids ( Fig. 1) Ariizumi and Toriyama, 2011). As the advance of functional genomics, genetic and biochemical approaches, various components such as transcription factors, enzymes for lipids synthesis, transporters, kinases etc. involved in the anther development and pollen formation in rice have been reviewed (Shi et al., 2015;Zhao et al., 2016;Zhang and Yang, 2014;Cai and Zhang, 2018). This review paper focuses on the mechanism controlling the redox status regulation and its role in specifying anther cell identity and cell degeneration.

Dynamics of Redox Status During Rice Anther Development
Reactive oxygen species (ROS) are produced in plants when molecular dioxygen (O 2 ) is applied as a terminal electron acceptor, generating molecules such as hydrogen peroxide (H 2 O 2 ), superoxide anion (O 2 -) and hydroperoxide radicals (OH * ). Most of these ROS molecules are by-products of aerobic metabolism in plants (Miller and Mittler, 2006).
In rice, ROS level is extremely low during anther cell specification stages, which is less than 200 pmol mg − 1 , and the ROS level increases twice at stage 4 to stage 5 (Yang et al., 2018). 3,3′-diaminobenzidine (DAB) staining indicates the accumulation of H 2 O 2 in middle layer and endothecium from stage 5 to stage 9 and decrease of ROS at stage 10 and stage 11 ( Fig. 1) (Hu et al., 2011;Yang et al., 2016;Yi et al., 2016;Yang et al., 2018). The increase of ROS in anther cell wall layers may play a key role in initiating promoting cell degeneration of middle layer and tapetum from the meiosis (Hu et al., 2011;Yi et al., 2016). These results also suggest that the low hypoxia status is critical for early anther cell specification (Yang et al., 2018). Supportively, during anther cell specification, individual treatment of H 2 O 2 and the ROSremoval reagent potassium iodide (KI) treatment in wild-type anthers induces cell proliferation and increases the number of Ar cells and Sp cells (Fig. 2 a, b), highlighting the importance of redox homeostasis during anther cell specification in rice (Yang et al., 2018).
Consistent with the role of redox in specifying rice anther cell specification, the hypoxia status has been demonstrated in maize (Zea mays) (Kelliher and Walbot, 2012). In the airspace between the tassel and the innermost leaf in maize, O 2 concentration is around 1.2 to 1.4% during anther cell specification, and 5 days after that, the O 2 concentration reaches 5%. This result also highlights the transient hypoxia status during anther cell specification in maize, which is similar to that of rice (Kelliher and Walbot, 2012;Yang et al., 2018). The treatment of N 2 of anther cells increases the cell numbers of Ars, while this induction is absent when treated with O 2 , indicating that low level of O 2 induces the cell specification of Ars (Kelliher and Walbot, 2012). Furthermore, the defects of Ar cell specification of msca1 (male sterile converted anther 1) with a mutation in a glutaredoxin (GRX) gene could be restored by reducing treatment (Sheridan et al., 1996(Sheridan et al., , 1999Kelliher and Walbot, 2012). These results suggest the importance of hypoxia status within growing anther tissue on stimulating the Ar cells development in maize (Kelliher and Walbot, 2012).
Although in Arabidopsis, the ROS level and redox status has not been measured during the anther development, the components controlling the redox status essential for male reproduction appeared to be conserved such as the the GRXs in Arabidopsis (Xing and Zachgo, 2010). Moreover, the mutation of an anther-expressed ROS synthetic gene RBOHE (Respiratory Burst Oxidase Homologue E)

Glutaredoxins (GRXs) and TGA Factors Control Redox Status During Early Rice Anther Development
Glutaredoxins (GRXs) are small glutathione (GSH)dependent oxidoreductases that catalyze the reversible reduction of protein disulfide bridges or protein-GSH mixed disulfide bonds, affecting various cellular events (Lemaire, 2004;Rouhier et al., 2004;Kelliher and Walbot, 2012;Zhang and Yang, 2014). Mutation in an anther-expressed glutaredoxin, MIL1 (MICROSPORE-LESS1) in rice, fails in the differentiation of inner SPC into middle layer and tapetal cell layer (Fig. 2c). MIL1 belongs to CC-type glutaredoxin, which usually interacts with TGA (TGACGTCA cis-element-binding protein) proteins. MIL1 interacts with TGA1 (Hong et al., 2012b) . While the biological function of TGA1 has not been reported yet. In rice, anther-expressed OsGRX_I1 has a physical interaction with OsTGA10, mutation of which exhibits defects in tapetal PCD ( Fig. 2d) (Yang et al., 2016;Chen et al., 2017). The interaction between glutaredoxins and TGA proteins may cause the transcriptional activity changes of TGA proteins caused by Cys residue modification by glutaredoxin which remains to be investigated. MIL1 expression is detectable in Ar cells from stage 2, extending to PPCs and Ar cells at stage 3 and SPCs as well as Sp cells at stage 4. At stage 5, MIL1 is highly accumulated in tapetum and weakly detected in Sp (Fig. 2e), similar to TGA1 (Hong et al., 2012b). However, no OsTGA10 transcripts are detectable at stage 2. OsTGA10 is detectable in PPC at stage 3 and the corresponding daughter cells endothecium and SPC at stage 4. At stage 5, OsTGA10 is highly expressed in middle layer and weakly expressed in tapetum. Till stage 6, OsTGA10 is specifically expressed in middle layer and has no expression during meiosis ( Fig. 2f ) (Yang et al., 2016). Overall, MIL1 and TGA1 expression pattern is slightly earlier and inner than that of OsTGA10. The differential expression of TGA genes may be associated with their verified function during anther development. Furthermore, TGA genes may have redundant function in specifying anther cells for the high expression pattern of OsTGA11 and OsTGA12 at premeiosis stages, partially overlapping with that of OsTGA10 (Yang et al., 2016). Together with the evidence that OsTGA10 is required for suppressing the induced expression of a PCD executor gene OsAP25 (Oryza sativa Aspartic Protease 25) by TIP2 (Niu et al., 2013;Chen et al., 2017), we hypothesize that TGA1-MIL1 and OsTGA10-OsGRX_I1 might play roles alternatively both spatially and temporarily, and the later might function as a link for anther cell specification and degeneration (Fig. 3a).
However, the MIL1-MSCA1-ROXY1/ROXY2 show divergent aspect during the anther development. The developmental defects of msca1 anthers are seen as early as abortion of the differentiation of Ar cells from L2-d cells (Kelliher and Walbot, 2012). While mil1 anthers a b c d have normal cell division and differentiation of Ar cells, forming the PPC and Sp cell as well as the initiation of the endothecium and epidermis at stage 5. However, the progenies of Sp cells of mil1 anthers develop into many smaller cells and fail to enter meiosis (Hong et al., 2012b). These results indicate that the maize GRX mutant shows earlier developmental defects than that of rice. In addition, unlike the defects of mil1 and msca1, roxy1 roxy2 and tga9 tga10 anthers display earlier abortion of adaxial lobes and later abortion of abaxial ones (Xing and Zachgo, 2010;Murmu et al., 2010). MSCA1 interacts with FEA4, controlling meristem size. Mutation in MSCA1, also named as abph2, displays large apical meristem, mimicking the phenotype of fea4 Yang et al., 2015). However, one of the counterparts of ROXY1/ROXY2, PAN (PERIANTHIA), a homolog of FEA4, is not required for meristem size control in Arabidopsis (Li et al., 2009). Instead, pan mutant exhibits a pentamemrous flower and roxy1 single mutant displays defects in petal initiation (Running, 1996;Li et al., 2009). The involvement of GRX-TGA counterparts in flower development and shoot apical meristem size control has not been reported in rice possibly due to that the divergent function of GRXs during evolution.

OsTDL1A-MSP1 Signaling Pathway in Specifying Somatic and Reproductive Cell Identity
In rice anther, cell fate specification of germinal and somatic cells is also associated with cell surface-localized Leurich repeat receptor-like kinases (LRR-RLKs) and their putative ligands (Fig. 3a). OsTDL1A (TPD1-like 1A)/MIL2(-MICRO-SPORELESS 2) encodes a small peptide with the expression in Ar cells at stage 3 and later radically in innermost somatic cell layer (Fig. 3b) (Hong et al., 2012a). MSP1(MULTIPLE SPOROCYTE 1) encodes a LRR-RLK expressed in the innermost somatic cell layer (Fig. 3b) (Nonomura et al., 2003). Both ostdl1a and msp1 exhibit excessive Sp cells and lack of middle layer and tapetal cell layer (Fig. 3c). ostdl1a msp1 double mutant displays similar phenotype with each single mutant (Yang et al., 2016). Furthermore, the 21-aa peptide of OsTDL1A is able to interact with MSP1, suggesting that OsTDL1A acts as a ligand of MSP1 (Fig. 3a) (Nonomura et al., 2003;Hong et al., 2012a;Yang et al., 2016). Genetic and biochemical evidences show that OsTDL1A-MSP1 signaling specifies the early anther cell fate by stimulating the transition of parietal cells into the middle layer and tapetal cells, and suppressing the extra activity of generating microsporocytes in rice (Nonomura et al., 2003;Hong et al., 2012a;Yang et al., 2016). Genome-wide expression profiles also show the altered expression of genes in ostdl1a and msp1-4 are associated with redox modulation such as ROS-producing and ROS-scavenging enzymes and proteins, including putative peroxidases, cytochrome P450s, oxidoreductases, thioredoxins, and glutaredoxins (Yang et al., 2016). Consistent with the expression alteration of genes involved in redox homeostasis, the presence of hydrogen peroxide in the middle layer and expression signal of glutaredoxins OsGrx_I1 and MIL1 in anther cells are detectable in rice anthers at stage 5 after the differentiation of the four cell wall layers, but not in ostdl1a and msp1-4. Furthermore, the ROS level is not detectable in ostdl1a and msp1-4 anthers (Fig. 3d), highlighting the role of OsTDL1A-MSP1 signaling in specifying anther cell fate by directly or indirectly changing the redox status (Fig. 3a) (Yang et al., 2016) . The OsTDL1A-MSP1 pathway is conserved in Arabidopsis and maize as reviewed by Zhang and Yang (2014), whether the counterparts of OsTDL1A-MSP1 play a role in regulating redox status during anther development remains to be elucidated.

Glutamyl-tRNA Synthetase Determines Anther Cell Identity and Patterning Via Affecting Redox Status
Aminoacyl-tRNA synthetases (aaRSs) conjugate amino acid and their cognate tRNA, which is critical for protein synthesis and amino acid metabolism. aaRSs are widely distributed in the genome from all the species, including plants (Yamakawa and Hakata, 2010). OsERS1 (Oryza sativa Glutamyl-tRNA Synthetase) is a housekeeping gene, however, OsERS1 transcripts are detectable in cells of presumptive anther lobes at the four corners at stage 2, which is comparable with high hypoxia status in maize (Yang et al., 2018;Kelliher and Walbot, 2012). Later, the transcripts of OsERS1 accumulate in PPC and Ar cells at stage 3, followed by in SPC and Sp cells at stage 4, and in the innermost somatic cell layer from stage 5 to stage 8 ( Fig. 4a) (Yang et al., 2018). osers1 anthers show over-proliferation and disorganization of L2d cells, forming fused lobes and extra germ cells in early anthers (Fig. 4b). Biochemically, OsERS1 functions as a glutamyl-tRNA synthetase in ligating Glu with tRNA, regulating the amino acid metabolism and tricarboxylic acid cycle in anther. In osers1 anthers, most of Glu family amino acids and the related metabolites including oxaloacetate and malate in tricarboxylic acid cycle, are over-accumulated (Yang et al., 2018). Global metabolomics profiling shows that osers1 has the increased level of ROS produced by mitochondrial activities, and measurement also indicates hydrogen peroxide content in the osers1 mutant at stages 4 to 5 is ∼900 pmol mg − 1 , twice as that of wild-type anthers. Also, in the osers1 mutant, the amount of hydrogen peroxide precursor, superoxide radical, is increased slightly. Consistently, DAB staining shows that osers1 anthers display detectable H 2 O 2 in anther primordium before stage 3 (Fig. 4c), and strong signals in anther cells at stages 4 to 5, which is earlier than that of the wild type (Yang et al., 2018). The application of 100 μM H 2 O 2 and 1 mM potassium iodide (KI), the ROS-removal reagent cause the increased number of L2d cells and Sp-like cells in wild-type anther, photocopying the defects of the osers1 mutant (Fig. 2 a, b). But osers1 anther treated with H 2 O 2 and KI shows a decreased number of L2-d cells and Sp-like cells (Fig. 4 df ) (Yang et al., 2018). Therefore, both the elevation of ROS by H 2 O 2 injection and neutralization of ROS using KI disturb ROS homeostasis, which causes abnormal cell division, cell fate specification. There are 45 aaRSs in Arabidopsis, however, only 21 aaRS mutants display developmental defects and chloroplast aaRS mutants exhibit an embryo defects, while mitochondrial aaRS mutants exhibit an ovule abortion phenotype with the exception of ova9 showing defects in female transmission (Berg et al., 2005). None of the aaRS has been reported for the anther development in Arabidopsis. It might be contributed by the functional redundancy of aaRSs or the functional divergence of amino acid homeostasis on anther development.

The Role of ROS in Promoting Tapetal PCD
Most of these ROS molecules are toxic in plants under abiotic stresses (Miller and Mittler, 2006). ROS also serves as important signaling molecules affecting a diverse range of plant processes, such as promoting programmed cell death (PCD) (Lam et al., 2001;Scherz-Shouval et al., 2007). Plants have gained diverse protective systems such as ROS-scavenging enzymes including superoxide dismutase, catalase, and peroxidase (Apel and Hirt, 2004), and non-enzymatic mechanisms, to modulate ROS levels (Mittler et al., 2004). The so-called non-enzyme system contains low molecular mass antioxidants including ascorbate, carotenoids, glutathione, and metallothioneins (MTs), which can remove hydroxyl radicals and singlet oxygen (Gechev et al., 2006). Consistent with the role of ROS in promoting cell death, a high level of ROS is detectable from stage 7 during the meiosis when the tapetal cell death is initiated (Li et al., 2006;Hu et al., 2011;Niu et al., 2013;Fu et al., 2014). Some ROS molecules, such as the superoxide anion radical and hydrogen peroxide are key regulators of plant cell death (Overmyer et al., 2003;Gechev and Hille, 2005;Gadjev et al., 2008;Hu et al., 2011). Metallothioneins (MTs) belongs to low molecular mass antioxidants, with the ability to remove hydroxyl radicals and singlet oxygen (Gechev et al., 2006). OsMT2b and OsMT-I-4b are able to scavenge superoxide and hydroxyl radicals (Wong et al., 2004;Hu et al., 2011). At stage 8 and stage 9 during anther development, OsMT2b interacts with DTC1 (Defective Tapetum Cell Death 1), a KELCH repeat-containing protein, keeping the scavenging less effective, as a result, ROS level is high in the tapetal cells. After stage 9, as the expression level of DTC1 decreases, the released OsMT2b thus reduces the ROS level (Fig. 5 a, b). Supportively, dtc1 mutant displays low level of ROS and defective in PCD initiation (Fig. 5c) (Yi et al., 2016). Meanwhile, OsMT-I-4b is highly induced by a MADS box transcription factor, OsMADS3, highly expressed at stage 9-stage 11 (Fig.  5d), leading to increase the ROS scavenge efficiency at stage 10 and stage 11. osmads3 mutant displays higher level of ROS and abnormal tapetal PCD and pollen fertility (Fig. 5e) (Hu et al., 2011). These results indicate that the precise ROS level is of vital importance for tapetal PCD.

Conclusions
Plant male development is a coordinated process starting form an anther primordium, undergoing cell division and differentiation, meiosis, tapetal PCD and pollen development. Emerging evidence suggests the critical role of hypoxia status in specifying anther cell identity, subsequently the triggering signal of ROS as tapetal PCD to nourish microspore for producing mature pollen grains in rice, maize and Arabidopsis. During early anther development, the hypoxia status is finely modulated by genetic and metabolic components such as glutaredoxins and TGA factors which are directly or indirectly affected by OsTDL1A-MSP1 signaling pathway. GRX and TGA counterparts as well as OsTDL1A-MSP1 signaling pathway have been identified to be conserved in different plants species, however, their expression pattern, possible combination of counterparts as well as the genetic function have variation among plants, highlighting the complexity of regulatory network of redox status during evolution. Surprisingly, OsERS1, a glutamyl-tRNA synthetase, is able to maintain proper somatic cell division and organization and limit the over proliferation of male germ cells in rice during early anther development by affecting amino acids hemostasis and redox status regulation. This pathway has not been characterized in other plants. After the anther specification is complete, ROS may trigger tapetal PCD and some transcription factors such as bHLHs, MADS-box proteins may regulate ROS scavengers to maintain redox homeostasis for normal development. As the advance of genomics, genome-editing technologies, more regulators controlling redox status will be elucidated which is critical for both fundamental biology and plant breeding by manipulating male fertility.