Background

Pollen development is fundamental to the reproductive success of flowering plants. Pollen is developed inside the anther, the male reproductive organ of flowering plants. The anther has four layers of wall cells, from exterior to interior the epidermis, endothecium, middle layer, and tapetum, which encase a group of microspore mother cells (MMCs) prior to meiosis (Ariizumi and Toriyama 2011). These MMCs undergo meiosis, resulting in four haploid microspores. Each microspore matures into a pollen grain that contains a large vegetative cell and two small sperm cells. The tapetum plays a pivotal role in supplying nutrients for pollen development and gradually degenerates as the pollen matures (Zhang et al. 2011).

A mature pollen grain is encased in a robust wall, which is divided into the outer exine and the inner intine (Ariizumi and Toriyama 2011). The exine consists of an outer layer named the tectum, an inner layer named the nexine, and radially oriented bacula that bridge the tectum and nexine (Ariizumi and Toriyama 2011; Shi et al. 2015). The exine is primarily composed of chemically and physically stable biopolymers of aliphatic lipids and phenolics that are known as sporopollenin (Grienenberger and Quilichini 2021). The spaces between the tectum and bacula are usually filled with the tryphine, also referred to as the pollen coat (Qiao et al. 2023). Underneath the nexine lies the intine, which is immediately adjacent to the plasma membrane of the pollen vegetative cell (Ariizumi and Toriyama 2011; Shi et al. 2015). The intine is formed by the microspore after the first mitotic division. The primary constituents of the intine are pectin, cellulose, hemicellulose, hydrolytic enzymes, and hydrophobic proteins, resembling the composition of the primary cell wall of regular vegetative cells (Ariizumi and Toriyama 2011). The pollen wall is crucial to pollen function by protecting the male gametophytes from environmental stresses and facilitating pollination and pollen-stigma interaction (Shi et al. 2015).

The tapetal cells play a pivotal role in pollen wall development by synthesizing and transporting sporopollenin precursors and pollen coat compounds to the pollen surface (Ariizumi and Toriyama 2011; Shi et al. 2015). Genetic studies of male sterility genes have identified a list of evolutionally conserved genes that are presumed to play roles in the biosynthesis of aliphatic lipids and phenolic compounds that are required for pollen exine development (Wan et al. 2020). For instance, aldehyde decarbonylases such as AtCER1, OsCER1 and OsWDA1 (Aarts et al. 1997; Jung et al. 2006; Ni et al. 2018), the fatty acid hydroxylase AtCER3 (Rowland et al. 2007), the 3-Keto-acyl-CoA synthase CER6/CUT1 (Fiebig et al. 2000), and the long-chain acyl-CoA synthetases AtLACS1 and AtLACS4 (Jessen et al. 2011), are probably involved in the biosynthesis of very-long-chain fatty acids and alkanes. OsNP1 in rice and its ortholog in maize ZmIPE1 encode putative glucose-methanol-choline (GMC) oxidoreductases that probably have a role in hydroxylation of long-chain fatty acid at the ω-position (Chang et al. 2016a; Chen et al. 2017). The product of GMC oxidoreductase may serve as a substrate for cytochrome P450 proteins such as the Arabidopsis CYP704B1 and CYP703A2 and their rice orthologs OsCYP704B2 and OsCYP703A3 that can catalyze in-chain hydroxylation or the formation of ω-dicarboxylic fatty acids (Morant et al. 2007; Dobritsa et al. 2009; Li et al. 2010a; Yang et al. 2014). Arabidopsis AtGPAT1, AtGPAT6 and rice OsGPAT3 are all putative glycerol-3-phosphate acyltransferases that are believed to be involved in the biosynthesis of glycerolipid, a component of sporopollenin (Zheng et al. 2003; Li et al. 2012, 2019; Men et al. 2017; Sun et al. 2018). Rice DPW2 encodes a hydroxycinnamoyl-CoA:ω-hydroxy fatty acid transferase that may contribute to the biosynthesis of phenolics (Xu et al. 2017). Arabidopsis ACOS5 and its rice ortholog OsACOS12 encode acyl-CoA synthetases that can convert medium- and long-chain fatty acids into fatty acyl-CoA esters (de Azevedo et al. 2009; Li et al. 2016; Yang et al. 2017), which can then be condensed to malonyl-CoA by polyketide synthases encoded by Arabidopsis PKSA/LAP6 and PKSB/LAP5 and their rice orthologs OsPKS1 and OsPKS2 (Dobritsa et al. 2010; Kim et al. 2010; Zhu et al. 2017), and subsequently reduced by tetraketide reductases encoded by Arabidopsis TKPR1 and TKPR2 and rice OsTKPR1 (Grienenberger et al. 2010; Xu et al. 2019). These genes are presumed to be involved in the biosynthesis of sporopollenin precursors and/or the pollen coat compounds (Wan et al. 2020; Qiao et al. 2023).

Molecular genetic studies of male sterility genes have also identified a number of genes that are presumed to play a role in the transport of materials essential for pollen exine development. These include ABC transporters such as AtABCG11 (Panikashvili et al. 2010), AtABCG26 (Quilichini et al. 2010, 2014), AtABCG9, AtABCG31 (Choi et al. 2014), AtABCG1 and AtABCG16 (Yadav et al. 2014) in Arabidopsis, as well as OsABCG15 (Qin et al. 2013; Niu et al. 2013a; Wu et al. 2014), OsABCG26 (Zhao et al. 2015; Chang et al. 2016b), and OsABCG3 (Chang et al. 2018) in rice. In addition, a number of nonspecific lipid transfer proteins (LTPs) that are widely conserved across different plant species have also been found to be important for pollen exine development (Fang et al. 2023). These include proteins such as OsC6 (Zhang et al. 2010) and OsLTP47 (Chen et al. 2022) in rice and their Arabidopsis orthologs AtLTPg3 and AtLTPg4 (Edstam and Edqvist 2014), OsC4 and its Arabidopsis orthologs AtLTPc3 and AtLTPc1 (Huang et al. 2013), and OsEPAD1 (Li et al. 2020) and its maize orthologs ZmLTPx2 and ZmLTPg11 and wheat ortholog TaMs1 (Li et al. 2021).

In Brassica species, tapetal cells accumulate lipids and flavonoids in sub-organelles such as endoplasmic reticulum (ER)-derived tapetosomes and plastid-derived elaioplasts (Hsieh and Huang 2005; Liu and Fan 2013). These sub-organelles are released into the anther locule upon degradation of the tapetal cells, and their contents are deposited onto the pollen surface as the pollen coat (Hsieh and Huang 2007; Qiao et al. 2023). In rice, a large number of vesicles were also observed in the tapetal cells, but disruption of the tapetal cell was not observed (Zhang et al. 2011). Instead, rice plants develop Ubisch bodies, which are specialized orbicule structures on the outer surface of the tapetum (Shi et al. 2015). Ubisch bodies are thought to originate from the ER of the tapetal cells and are important for the secretion of sporopollenin precursors to the anther locule and the microspore surface (Huysmans et al. 1998). Nonetheless, the degradation of tapetal cells remains essential for pollen wall formation in rice, as indicated by the occurrence of programmed cell death signals in the tapetal cells following meiosis (Li et al. 2006, 2011; Niu et al. 2013b). Many mutants with abnormal programmed cell death also show abnormal pollen development and male sterility (Shi et al. 2015).

Plant cellular membranes are a lipid bilayer primarily composed of glycerophospholipids such as phosphatidylcholine (PC), phosphatidylserine (PS), phosphatidylethanolamine (PE), and phosphatidylinositol (PI), which provide a structural matrix for embedded proteins (Reszczyńska and Hanaka 2020). In addition to these major lipid components, plant membranes also harbor a set of minor lipids known as phosphoinositides (PIPs) that are derived from the phosphorylation of the myo-inositol head group of PI at positions D-3, 4, and 5, including PI(3)P, PI(4)P, PI(5)P, PI(3,4)P2, PI(3,5)P2, PI(4,5)P2 and PI(3,4,5)P3 (Irvine 2016; Gerth et al. 2017). PIPs are exclusively found on the cytosolic face of biological membranes and constitute less than 1% of the membrane lipid content (Gerth et al. 2017; Noack and Jaillais 2020). PIPs exhibit a strong specificity with respect to their distribution in different membrane compartments, thus providing cues to membrane identity. For instance, PI(3)P is predominantly detected in endosome and tonoplast membranes (Vermeer et al. 2006; Simon et al. 2014); PI(4)P is enriched in the membranes of the Golgi apparatus, trans-Golgi network, late endosome, and plasma membrane (Vermeer et al. 2009; Simon et al. 2014); and PI(4,5)P2 is commonly found in the plasma membrane (van Leeuwen et al. 2007; Lebecq et al. 2022). These lipids play pivotal regulatory roles in controlling growth, development, and in responding to environmental cues by interacting with a multitude of proteins (Heilmann 2016; Roman-Fernandez et al. 2018).

Sec14-like PI transfer proteins (Sec14L-PITPs) comprise a group of evolutionarily conserved proteins characterized by a distinctive Sec14 domain initially identified in the yeast protein Sec14 (Holič et al. 2021). These proteins are capable of recognizing, binding, exchanging, and transferring PI, PIPs, and various other small lipophilic molecules between membranes through non-vesicular transport mechanisms (Holič et al. 2021). They are also involved in regulation of vesicular trafficking within the cell (Bankaitis et al. 2010). Sec14L-PITPs are found across all eukaryotic cells, including yeast, plants, and animals, and in each eukaryote, they form a multigene family (Montag et al. 2020). Based on their structural features, plant Sec14L-PITPs are categorized into three classes: one class contains only Sec14 domain; another class contains Sec14 domain and a plant-specific nodulin domain; and the third class contains Sec14 and Golgi dynamic (GOLD) domains (Montag et al. 2023). In vitro experiments have demonstrated PI or PC binding and transfer activities for the Sec14 domains of several Sec14L-PITPs with or without the additional domains (Huang et al. 2016b). To date, a number of plant Sec14L-PITPs have been implicated in various functions, such as chloroplast development (Hertle et al. 2020; Kim et al. 2022; Yao et al. 2023; Yang et al. 2023), polar growth of root hair cells (Grierson et al. 1997; Bohme et al. 2004; Vincent et al. 2005; Huang et al. 2013b; Ghosh et al. 2015), polar growth of pollen tubes (Moon et al. 2022), cell division and growth (Peterman et al. 2004; Tejos et al. 2017), and responses to environmental stimuli (Chu et al. 2018; Zhou et al. 2018), but none has been identified as a sporophytic factor regulating male fertility in the plant (Montag et al. 2023).

During the screening of rice mutants affecting male fertility, we identified a mutant named ms16600 that exhibited a significant decrease in male fertility. Cloning of the mutant gene indicated that the causal mutation is in OsSNDP4, a Sec14L-PITP with both a Sec14 domain and a nodulin domain. The reduction of male fertility of ossndp4 mutant was influenced by environment. Electron microscopy studies revealed abnormal vacuolar dynamics, irregular degradation of cellular contents, and a marked inhibition of Ubisch body formation in the tapetal cells. Additionally, the mutant plants displayed abnormal vacuole enlargement and a thinner exine in the pollen grains. A protein-lipid overlay assay indicated that OsSNDP4 has a preference for binding to PI(3)P. To date, Sec14L-PITPs and their potential binding substrates have not been implicated in the regulation of pollen exine development in any plant species, making these findings a novel contribution to our understanding of the mechanism governing pollen development.

Results

Isolation and Morphological Characterization of ms16600 Mutant

By screening an ethyl methanesulfonate (EMS)-induced mutant library derived from the indica variety Huanghuazhan (HHZ) (Chen et al. 2014), we identified a mutant named ms16600, which exhibited a greatly reduced seed-setting rate. The mutant displayed normal vegetative growth, booting and heading (Fig. 1a), but partial panicle enclosure (Fig. 1b). The seed-setting rate was severely reduced compared to the wild type (WT) (Fig. 1b). The spikelets of the mutant looked normal externally (Fig. 1b), but the anthers within were smaller and pale yellow in color (Fig. 1c). I2-KI staining showed that in > 70% of the spikelets, the anthers failed to produce any normally stained pollen grains, and in < 30% of the spikelets, the anthers produced a variable proportion of darkly stained pollen grains (Fig. 1d). Fluorescein Diacetate (FDA) staining indicated that the darkly stained pollen grains produced by the mutant emitted fluorescence similar to that of the WT (Additional file 1: Figure S1), indicating that they were viable pollen grains. When ms16600 was cross-pollinated with HHZ, the seed setting rate was ~ 90% (Additional file 2: Table S1), which was equivalent to that of the control male sterile line Zhen18A (Chang et al. 2016a). These results indicated that the female fertility of the mutant was normal, and the reduced seed-setting rate was caused by defective pollen development.

Fig. 1
figure 1

Phenotypic characteristics of the ms16600 mutant. a Plants of wild type (WT) and ms16600 after heading. b Panicles from WT and ms16600. c Spikelets from WT and ms16600, with the palea and lemma removed. d Pollen grains from WT and ms16600 stained with I2-KI. A total of 207 ms16600 spikelets were examined, and the number in each image represents the percentage of spikelets exhibiting the depicted phenotype. Scale bars = 20 cm (a); 5 cm (b); 2 mm (c); 100 μm (d). e Seed setting rates of ms16600 heading at different seasons. The temperature range during the booting and heading stages in Shenzhen was 27–33oC for Early Aug. 2022, 26–33oC for Early Oct. 2022, 22–30oC for Late Oct. 2022, and 23–29oC for Early Jun. 2022, respectively. Different letters denote significant differences (P < 0.01, t-test); while identical letters indicate no significant difference

Full size image

Because some mutant spikelets produced a portion of normal pollen and were capable of setting seeds, we investigated whether the fertility of the mutant was affected by the environment. The mutants were grown in different seasons in Shenzhen to monitor the seed-setting rate. To prevent out-crossing, the mutant panicles were bagged. When the average temperature during the booting and heading stages was between 26 and 33 oC (with heading occurring in early August and early October), the average seed setting rate was between 6.02% and 13.26%, reaching a peak of 29.23%. In contrast, when the average temperature during these stages was between 22 and 26 oC (with heading in late October and early June), the mutant exhibited near-complete sterility (Fig. 1e). These results suggested that pollen development in ms16600 mutant might be temperature-sensitive, and higher average temperatures could be favorable for pollen development in the ms16600 mutant.

Histological Analysis of ms16600 Mutant Anther

To elucidate the impact of the OsSNDP4 mutation on pollen development, we compared the anther developmental processes between the WT and ms16600 mutant with microscopic analyses. In rice, anther development is divided into 14 stages based on morphological characteristics (Zhang et al. 2011). Transverse sections revealed no obvious difference between the WT and ms16600 mutant before the tetrad formation (stage 8b) (Fig. 2a–d).

Fig. 2
figure 2

Transverse sections of anther in wild type and ms16600 from stage 8a to stage 12. Dys, dyads; DMsp, degenerated microspores; E, epidermis; En, endothecium; M, middle layer; MP, mature pollen; Msp, microspores; T, tapetum; Tds, tetrads. Red arrows in f show vacuoles within the tapetal cells. Scale bars = 20 μm

Full size image

By stage 9, the tapetal cells in both the WT and ms16600 mutant became thinner (Fig. 2e and f). The WT tapetal cell cytoplasm was highly condensed with the vacuoles almost completely disappeared (Figs. 2e and 3a), while the Ubisch bodies, which are thought to export sporopollenin precursors from the tapetum to the microspore, were generated on the outer surface of the tapetal cells (Fig. 3a). In contrast, in ms16600 mutant, the tapetal cell cytoplasm was less condensed with many vacuoles visible inside (Figs. 2f and 3b), and the Ubisch bodies were not clearly visible (Fig. 3b). At this stage, the callose wall surrounding microspores was degraded, and the released microspores were loosely arranged in a circle along the tapetum in the WT (Fig. 2e), with proexine deposited on the microspore surface (Fig. 3a). In ms16600 mutant, microspores were also released from the tetrad but were scattered within the anther locule (Fig. 2f), and the proexine was thinner than that of the WT (Fig. 3b).

Fig. 3
figure 3

Electron microscope analysis of the wild type and ms16600 anthers from stage 9 to stage 12. a–h, transmission electron microscopy results; i–n, scanning electron microscopy results. C, anther surface cuticle; E, epidermis; En, endothecium; Ex, exine; M, middle layer; Msp, microspores; T, tapetum; Ub, Ubisch body. Ubisch bodies outlined in the dashed boxes in a to h are enlarged for clarity. The tapetal cell is indicated by red double-headed arrows. Red arrows in b show vacuoles within the tapetal cells. Scale bars = 2 μm (a–h); 10 μm (i–n)

Full size image

By stage 10, the WT tapetal cells further degenerated, yet the dense cellular contents were still visible (Figs. 2g and 3c), and the Ubisch bodies further enlarged in size and were tightly arranged on the tapetal surface (Fig. 3c, i). In ms16600 mutant, tapetal cells contained much less cellular contents (Figs. 2h and 3d), and the Ubisch bodies were poorly developed (Fig. 3d, j). Meanwhile, spherical microspores were closely arranged along the tapetum in the WT, and their volume increased significantly with a large vacuole in the center and cytoplasm and nucleus moving to the cell edge (Fig. 2g), and the wall of microspores became obviously thickened (Fig. 3c). However, in ms16600 mutant, the microspores were excessively enlarged due to the abnormally enlarged vacuole, and they squeezed each other in the anther locule, resulting in an irregular cell shape (Fig. 2h). The pollen exine was thinner compared with that of the WT (Fig. 3d).

By stage 11, the WT tapetum almost completely degenerated (Figs. 2i and 3e), and the Ubisch bodies enlarged further (Fig. 3e). Conversely, the mutant tapetal cells remained large with little cellular content inside (Figs. 2j and 3f), and they exhibited irregular Ubisch bodies (Fig. 3f). Meanwhile, the WT microspores became falcate with further thickened exine (Figs. 2i and 3e). The mutant microspores began to collapse with a decrease in cell volume and an irregular cell shape (Fig. 2j), and the pollen exine remained thinner than that of the WT (Fig. 3f).

By stage 12, the WT microspores developed into spherical pollen grains filled with numerous starch granules and dense cellular contents (Figs. 2k and 3m), and only the epidermis layer remained visible in the WT anther wall (Fig. 2k). In contrast, the mutant microspores completely collapsed, leaving only the irregularly shaped pollen wall (Figs. 2l and 3n). Both the epidermis and tapetum layers were clearly visible in the mutant anther wall (Fig. 2l). The cuticle on the outer surface of the anthers did not show a significant difference between the WT and ms16600 mutant (Fig. 3g–h, k–l).

Identification of the Causal Mutation for ms16600 Mutant

We employed the simultaneous identification of multiple causal mutations (SIMM) method described by Yan et al. (2017) to determine the causal mutation in the ms16600 mutant. Initially, the mutant was crossed with the wild-type HHZ. The resulting F1 plants all showed normal seed-setting rates. Self-pollination of the F1 plants yielded an F2 population that segregated approximately in a 3:1 ratio of normal to reduced fertility (Additional file 3: Table S2), indicating that the mutant phenotype was due to a single recessive mutation.

Thirty plants of low seed-setting rates were selected from the F2 segregating population for DNA extraction. Equal amount of genomic DNA extracted from these 30 individuals was pooled and bulk-sequenced to a depth of ~ 30 ⨯ coverage of the rice genome. After removing linker sequence and the sequences of low quality or low coverage, the high-quality sequence data were aligned against the Nipponbare reference genome and compared with the re-sequencing data from other HHZ EMS mutants for identification of EMS-induced mutation sites in ms16600 using the SIMM pipeline. SNP index and Euclidean distance (ED6) was calculated for each SNP site. The SNP site with the highest scores for both SNP index and ED6 was considered the candidate mutation site associated with the mutant phenotype (Yan et al. 2017).

A candidate mutation site was identified at the end of chromosome 2 (Fig. 4a). This mutation is situated in exon 10 of the gene LOC_Os02g04020, where the codon TGT (encoding amino acid Cys351) was changed to TGA, leading to premature termination of the protein (Fig. 4b). To establish the linkage between this mutation and the phenotype, a high-resolution melting (HRM) assay was performed on a F2 segregation population. Among the 1917 F2 progeny genotyped, those with homozygous wild-type (T/T) or heterozygous (T/A) genotypes exhibited normal fertility, whereas those with homozygous mutant genotype (A/A) showed reduced seed-setting rates (Additional file 4: Table S3). This indicates that the T to A mutation was tightly linked with the mutant phenotype.

Fig. 4
figure 4

Identification and verification of the mutant gene through CRISPR knockout and transgenic complementation. a Identification of the causal mutation site using SIMM method. The SNP index and Euclidean distance (ED) scores of EMS-induced SNPs are depicted as dots. b Gene structure of LOC_Os02g04020. The mutation site in ms16600 (TGT to TGA) and the mutation sites in Cr-1 (+A/+A, −GCA/−GCA, −CT/−CT) are shown. Black boxes represent exons, white boxes represent UTRs, and the lines between them denote introns. The line with two arrows indicates the genomic fragment used for transgenic complementation. c–e Wild type HHZ. f–h CRISPR knockout mutant Cr-1. i–k ms16600 mutant. l–n Transgenic complementation plant Com-1P. o–q Transgenic complementation plant Com-2P. Scale bars = 20 cm (c, f, i, l, o); 1 mm (d, g, j, m, p); 100 μm (e, h, k, n, q)

Full size image

To confirm the function of the mutant gene, we conducted CRISPR knockout of the LOC_Os02g04020 gene in HHZ. To enhance the likelihood of successful gene editing, we designed four target sites (Target 2, Target 3, Target 5, and Target 9) within the second, third, fifth and ninth exons of the target gene, respectively, and constructed a multi-target editing vector (CR2359) for transformation into wild-type HHZ. Ten independent knockout lines (Cr-1 to Cr-10) were obtained, all exhibiting small insertions or deletions at the target sites (Additional file 5: Table S4). To exclude the CRISPR T-DNA from the genome, T0 plants were crossed with wild-type HHZ, and homozygous knockout mutants devoid of T-DNA were isolated from the F2 progeny for subsequent phenotype analysis. These homozygous knockout mutants displayed phenotypes similar to the ms16600 mutant, including small, pale anthers, pollen abortion, and reduced seed-setting rates (Fig. 4c–k, Additional file 6: Figure S2, Additional file 5: Table S4).

To further validate the function of the mutant gene, we conducted a gene complementation experiment. A genomic fragment encompassing the LOC_Os02g04020 gene, along with a 3121 bp promoter region upstream of the start codon ATG, and an 896 bp fragment downstream of the stop codon TGA (Fig. 4b), was cloned into a binary vector and introduced into the ms16600 mutant. Two independent transgenic lines, Com-1 and Com-2, were obtained, both of which exhibited high seed-setting rates upon self-pollination. The offspring derived from self-pollination carrying the transgene (Com-1P and Com-2P) exhibited normal seed-setting, normal anthers, and normal pollen, while those lacking the transgene exhibited a phenotype similar to the ms16600 mutant (Fig. 4l–q, Additional file 6: Figure S2). Moreover, crossing the transgene into the CRISPR knockout mutant lines restored the seed-setting rate of the homozygous CRISPR mutant to normal (Additional file 6: Figure S2, Additional file 7: Table S5). The above results confirm that the phenotype of ms16600 is due to the loss of function of LOC_Os02g04020.

Phylogenetic Analysis of OsSNDP4

LOC_Os02g04020 encodes the Sec14 Nodulin Domain Containing Protein 4 (OsSNDP4) (Huang et al. 2016a). Conserved domain analysis indicated that OsSNDP4 possesses a conserved N-terminal Sec14 domain and a C-terminal plant-specific nodulin domain (Additional file 8: Figure S3a). The mutation in ms16600 introduces a termination codon at the end of the Sec14 domain, resulting in a truncated protein that contains almost a complete Sec14 domain but lacks the nodulin domain (Additional file 8: Figure S3a).

Proteins containing the Sec14 domain are ubiquitous in eukaryotic cells. The yeast Sec14 protein is the prototype of this protein family (Montag et al. 2023). Sec14 can bind simultaneously to a PC and a PI molecule and promotes the phosphorylation of PI by PI 4-OH kinase (Schaaf et al. 2008). Sec14 facilitates lipid transport from ER to Golgi in yeast cells and can transfer PI and PC molecules between membranes in vitro (Bankaitis et al. 1989, 1990; Schaaf et al. 2008). A search of the Arabidopsis and rice genomes revealed 32 and 27 Sec14 domain-containing proteins, respectively (Huang et al. 2016a). These include 13 Arabidopsis and 13 rice proteins that possess only the Sec14 domain; 6 Arabidopsis and 4 rice proteins that contain both the Sec14 and GOLD domains, and 13 Arabidopsis and 10 rice proteins that harbor the Sec14 and nodulin domains (Ghosh et al. 2015; Huang et al. 2016a).

Proteins containing both a Sec14 domain and a nodulin domain are referred to as AtSFHs in Arabidopsis and OsSNDPs in rice (Huang et al. 2016a). The Sec14 domains of these proteins are highly conserved and possess conserved amino acid residues for PI binding (referred to as the PI-binding barcode) and PC-binding (referred to as the PC-binding barcode), as observed in yeast Sec14 protein (Additional file 8: Figure S3b). The Sec14 domains from several proteins, including AtSFH1, AtSFH4, AtSFH5, AtSFH9, and OsSNDP1, all exhibit in vitro PI binding and transfer activities as well as PI 4-OH kinase stimulating activities, and both activities depend on the conserved barcode amino acid residues (Huang et al. 2016b). Previous studies have showed that the functions of Sec14-nodulin family proteins AtSFH1 and AtSFH5 depend not only on the Sec14 domain but also on the nodulin domain (Ghosh et al. 2015; Yao et al. 2023). Based on the sequence characteristics of the C-terminal nodulin domains, the 10 rice OsSNDPs and the 14 Arabidopsis AtSFHs are classified into three groups (Additional file 9: Figure S4a). Class I nodulins are characterized by an uninterrupted stretch of seven or more basic residues with adjacent aromatic residues (Additional file 9: Figure S4b). The class I nodulin domain of AtSFH1 has been demonstrated with the PI(4,5)P2 binding activity (Ghosh et al. 2015). Class II nodulins exhibit C-terminal 6–8 basic residues and a penultimate Cys residue, but the uninterrupted stretch of basic residues in Class II is shorter than those in Class I nodulin domains (Additional file 9: Figure S4b). Class III nodulin C-termini have 8–12 basic residues, but the sequences are more divergent (Additional file 9: Figure S4b). OsSNDP4 falls into Class II according to phylogenetic analysis (Additional file 9: Figure S4a). The class II nodulin domain of AtSFH5 does not bind lipid on its own, but is required for maintaining the overall structure of the full-length AtSFH5 protein (Yao et al. 2023). The AtSFH5 protein shows strong binding to phosphatidic acid (PA) and relatively weaker binding towards PI(3,4)P2, PI(3,5)P2, PI(4,5)P2, PI(3,4,5)P3, and PI(4)P in a protein-lipid overlay assay (Yao et al. 2023).

OsSNDP4 Gene Expression and Subcellular Localization of OsSNDP4 Protein

The ms16600 mutant exhibited abnormal anther development but no obvious defects in other organs. To understand the dedicated role of OsSNDP4, we analyzed the spatial and temporal gene expression patterns of OsSNDP4 using quantitative reverse transcription-PCR (qRT-PCR). As shown in Fig. 5a, OsSNDP4 expression was detected in all tested tissues, including roots, culms, leaves, glumes, pistils and anthers. Notably, the expression levels were significantly higher in stage 10–12 anthers than in other tissues, correlating with the observation that ms16600 mutant exhibited clear defects in anthers after stage 9. The mutation in ms16600 significantly reduced the transcript level of the mutant gene (Fig. 5b).

Fig. 5
figure 5

OsSNDP4 gene expression and subcellular localization of the OsSNDP4 protein. a Expression profile of the OsSNDP4 gene in various tissues. b Comparison of OsSNDP4 transcript levels between the wild type HHZ and the ms16600 mutant during anther development, from stage 6 to stage 12. Pistils and other tissues were harvested from plants at the heading stage. c Subcellular localization of the OsSNDP4 protein in rice protoplasts. NC and ER denote the nuclear marker ARF19IV-mCherry and the endoplasmic reticulum marker RFP-HDEL, respectively. Arrows point to the vacuole membrane

Full size image

As mentioned above, the ms16600 mutant exhibited different seed setting rates when grown in different seasons, suggesting that OsSNDP4 might be regulated by temperature. To investigate this, we subjected rice plants to cold treatment (6 oC) and accessed the gene expression of OsSNDP4. As shown in Additional file 10: Figure S5, the transcripts of OsSNDP4 were induced following exposure to low temperatures.

The subcellular localization of proteins is crucial to their proper function. Analyses using TargetP, SignalP, and Cell-Ploc failed to identify a secretion signal peptide, mitochondria or chloroplast localization signal, or an endoplasmic reticulum (ER) retaining signal in the OsSNDP4 protein sequence. However, ScanProsite predicted the presence of a CRAL-TRIO lipid binding domain between amino acids 150 and 324 and a bipartite nuclear localization signal (NLS) between amino acids 45 and 61 (Additional file 11: Figure S6). Additionally, cNLS Mapper also identified nuclear localization signals within this region (Additional file 11: Figure S6). To ascertain the subcellular localization of the OsSNDP4 protein, we generated a recombinant OsSNDP4-EGFP fusion protein by appending EGFP to the C-terminus of OsSNDP4. Expression of this fusion protein in rice protoplasts showed a strong green fluorescence signal in the nucleus and associated with the ER (Fig. 5c). In addition, green fluorescence signal was also detected along the vacuole membrane (Fig. 5c). The observation of OsSNDP4 localization signals associated with the ER and surrounding the vacuoles is consistent with the proposed role of Sec14 family proteins in binding with phosphoinositides.

OsSNDP4 Demonstrates Significant Binding to PI(3)P

Several Sec14 domain containing proteins, with or without the additional domains, have been shown to bind PI, PIPs, and other lipid molecules with varying specificities (Ghosh et al. 2015; Huang et al. 2016a; Yao et al. 2023). To assess the lipid binding capability of OsSNDP4, the recombinant MBP-OsSNDP4 protein was expressed in E. coli and purified using MBP-binding beads. A protein-lipid overlay assay was conducted with the purified MBP-OsSNDP4 fusion protein, using purified MBP and PI(4,5)P2 Grip [specific for PI(4,5)P2] as negative and positive controls. As anticipated, MBP alone did not exhibit binding to any lipids, while PI(4,5)P2 Grip showed strong binding to PI(4,5)P2. The OsSNDP4 protein showed strong binding to PI(3)P, weaker binding to PA, and no binding to other lipids tested (Fig. 6). We also tested whether the separated Sec14 domain and nodulin domain of OsSNDP4 were capable of lipid binding. Both domains showed minimal binding to PI(3)P compared to the full-length OsSNDP4 protein (Fig. 6).

Fig. 6
figure 6

Protein-lipid blot overlay assay assessing the lipid binding and specificity of OsSNDP4. Recombinant proteins MBP-OsSNDP4, MBP-Sec14, and MBP-Nodulin were purified from E. coli. MBP alone and PI(4,5)P2 Grip were used as control. The left panel of the figure provides a schematic representation of the lipids present on the strip, including lysophospharidic acid (LPA), lysophosphacholine (LPC), phosphatidylinositol (PI), PI 3-phosphate [PI(3)P], PI 4-phosphate [PI(4)P], PI 5-phosphate [PI(5)P], phosphatidylethanolamine (PE), phosphatidylcholine (PC), sphingosine 1-phosphate (S1P), PI 3,4-bisphosphate [PI(3,4)P2], PI 3,5-bisphosphate [PI(3,5)P2], PI 4,5-bisphosphate [PI(4,5)P2], PI 3,4,5-trisphosphate [PI(3,4,5)P3], phosphatidic acid (PA), phosphatidylserine (PS), and a blue blank control

Full size image

Influence of OsSNDP4 Mutation on the Expression of Genes in Anther

Histological examination of anther development revealed that the WT and ms16600 mutant began to exhibit phenotypic difference in anthers at stage 9. To elucidate the molecular mechanisms by which OsSNDP4 regulates pollen development, we performed RNA-seq analysis to compare the gene expression profiles between WT and ms16600 mutant anthers at stage 9. Using a cutoff of > 2-fold change and a P-value < 0.05, we identified a total of 5561 differentially expressed genes (DEGs), with 2715 genes up-regulated and 2846 genes down-regulated in the mutant anthers (Additional file 12: Table S6, Additional file 13: Table S7). The DEGs include at least 61 genes that are known to be crucial for rice pollen development (Additional file 14: Table S8). It is striking that 59 of these 61 genes were down-regulated in the ms16600 mutant anthers (Additional file 14: Table S8). To validate the RNA-seq results, we selected six DEGs (CYP703A3, TDR, DPW, CYP704B2, PAIR2, and OsC4) that are important for pollen development and assessed their expression during anther development via qRT-PCR (Additional file 15: Figure S7). The expression of all these genes were consistent with the RNA-seq findings, confirming the reliability of the RNA-seq analysis. KEGG pathway analysis of the DEGs indicated that genes involved in ribosome biogenesis, the biosynthesis of flavonovid and other secondary metabolites, and transporters were up-regulated in the ms16600 mutant anthers. Conversely, genes related to sporopollenin synthesis, including those involved in fatty acid biosynthesis and lipid metabolism, the biosynthesis of secondary metabolites such as phenylpropanoid and terpenoid, the biosynthesis of lipid polymers, and signaling proteins and transcription factors, were down-regulated in the ms16600 mutant anthers (Additional file 16: Figure S8). Interestingly, an enrichment of exosome proteins was also observed among the down-regulated genes in ms16600 mutant anthers, suggesting that OsSNDP4 may play a role in regulating protein exocytosis.

Discussion

By characterizing the ms16600 mutant, we found that OsSNDP4 is important for the development of rice pollen. OsSNDP4 encodes a Sec14-nodulin domain protein that demonstrates affinity for binding to PI(3)P, as evidenced by the protein-lipid overlay assay. OsSNDP4 is expressed in all tested vegetative and floral organs, with a relatively higher expression level in the anthers during pollen exine development (Fig. 5a). Sec14 was identified in yeast as a cytosolic factor required for transport of secretory proteins from the Golgi complex (Bankaitis et al. 1989). It is a PC/PI transfer protein capable of exchanging the phospholipids between membranes in vitro (Bankaitis et al. 1990). Sec14 can bind PC and PI simultaneously and catalyze the synthesis of PI(4)P by presenting PI to PI 4-OH kinases (Schaaf et al. 2008). These findings lead the proposal that Sec14 executes its function by creating a lipid environment crucial for the biogenesis of secretory vesicles from the Golgi network (Mousley et al. 2012). Crystal structure analysis of Sec14 and its yeast homologue Sfh1 revealed the critical amino acid residues for PI and PC binding (Schaaf et al. 2008). These residues are highly conserved in Sec14-nodulin proteins (Huang et al. 2016a). Importantly, a few Sec14-like domains, including those from AtSFH1, AtSFH4, AtSFH5, AtSFH9, and OsSNDP1, which possess the conserved lipid-binding residues, all exhibit in vitro PI and PC transfer activities as well as PI 4-OH kinase stimulating activities, and both activities depend on the conserved amino acid residues (Huang et al. 2016b). Sequence comparison indicated that OsSNDP4 also has the conserved Sec14 domain as well as the important lipid binding residues (Additional file 8: Figure S3), which is consistent with the finding that OsSNDP4 can bind PI(3)P.

Genetic analysis indicated that OsSNDP4 regulates pollen fertility as a sporophytic factor in Mendelian inheritance. Consistent with this result, ossndp4 mutant plant exhibited abnormal tapetal cell degradation, defective Ubisch body formation, a much thinner pollen exine, and pollen abortion (Figs. 2 and 3). The molecular genetic behaviors of OsSNDP4 differ from those of OsSNDP3, a paralog of OsSNDP4 that is also required for pollen fertility (Moon et al. 2022). OsSNDP3 is specifically expressed in mature pollen. When OsSNDP3/ossndp3 was used as pollen donor, the mutant gene could not be transmitted to the next generation (Moon et al. 2022), indicating that OsSNDP3 acts as a gametophytic gene in regulating pollen fertility. OsSNDP3 contains a Sec14 domain and a type I nodulin domain (Additional file 9: Figure S4). Similar to AtSFH1/COW1 and OsSNDP1, two other proteins with a Sec14 domain and a type I nodulin domain, OsSNDP3 was found to co-localize with PI(4,5)P2 (Moon et al. 2022), a phosphoinositide associated with the plasma membrane and crucial for plasma membrane polarity (Lebecq et al. 2022). AtSFH1/COW1 and OsSNDP1 are essential for root hair elongation (Böhme et al. 2004; Vincent et al. 2005; Huang et al. 2013b), whereas OsSNDP3 is crucial for pollen tube growth (Moon et al. 2022). OsSNDP3 interacts with OsSNDP2, another protein with a Sec14 domain and a type I nodulin domain that is also highly expressed in mature pollen but plays a minor role in regulating pollen tube growth (Moon et al. 2022). The differences in genetic behaviors, gene expression patterns, and substrate binding specificities indicate that OsSNDP4 and OsSNDP3/OsSNDP2 regulate rice male fertility through distinct mechanisms.

In plant, PI(3)P is primarily generated by PI 3-kinase (PI3K) that phosphorylates PI at the D-3 position (Welters et al. 1994; Lee et al. 2008). In Arabidopsis, PI3K is encoded by the single-copy gene AtVPS34, which is crucial for normal plant development (Welters et al. 1994). The plant PI3K protein is associated with active nuclear and nucleolar transcription sites (Bunney et al. 2000). Consistent with the nuclear localization of the PI(3)P-synthesizing enzyme, OsSNDP4, which can bind PI(3)P, demonstrated a strong nuclear localization signal in the rice protoplast transient expression assay (Fig. 5c). RNA-seq analysis revealed significant enrichment of ribosome biogenesis genes that were up-regulated in the ossndp4 mutant anther, implicating that the nuclear localized OsSNDP4 protein may play a role in regulating ribosome biogenesis, a process that occurs in the nucleus (Jiao et al. 2023).

Additionally, OsSNDP4-EGFP exhibited co-localization with the ER marker and a fluorescence signal surrounding the vacuole (Fig. 5c). Several studies have shown that PI(3)P is predominantly associated with highly motile structures such as the prevacuolar compartment, late endosome, and the tonoplast (Vermeer et al. 2006; Simon et al. 2014; Hammond and Balla 2015). These vesicles are all related to ER functions within the cell (Mousley et al. 2012). In yeast, PI(3)P is essential for vesicle-mediated delivery of vacuolar enzymes (Stack and Emr 1994), and in animal cells, inhibition of PI(3)P synthesis by chemical inhibitors disrupts protein targeting from the trans-Golgi network to the lysosomes (Brown et al. 1995; Davidson 1995). Similarly, tobacco suspension cells treated with chemicals interfering with PI(3)P synthesis failed to deliver proteins into vacuoles (Matsuoka et al. 1995). Furthermore, PI(3)P inhibitors induce swelling or vacuolation of the prevacuolar compartment (Tse et al. 2004) and block retrograde transport of vacuolar sorting receptors to the trans-Golgi network (daSilva et al. 2005). Overexpression of a PI(3)P binding protein inhibits trafficking of the vacuolar protein in Arabidopsis protoplasts (Kim et al. 2001). Through a protein-lipid overlay assay, we confirmed that OsSNDP4 exhibits a significant affinity for binding to PI(3)P (Fig. 6). Consistent with the role of PI(3)P in regulating vacuole dynamics, microscopic examination of the ossndp4 mutant anthers revealed abnormal vacuole dynamics in both tapetal cells and microspores (Figs. 2 and 3). In wild-type tapetal cells, a large vacuole was present at stage 8 during meiosis, which became nearly invisible by stage 9. However, in the mutant tapetal cells, vacuoles were clearly visible under both semi-thin section and TEM observation by stage 9 (Figs. 2 and 3). At stage 10, the mononuclear microspores of the wild-type plant exhibited a large central vacuole, whereas in the mutant microspores, the vacuole was abnormally enlarged (Fig. 3). It appeared that the ossndp4 mutant tapetal cells underwent earlier degradation of cellular contents and organelles compared to the wild-type (Fig. 2). Based on these observations, we hypothesized that OsSNDP4 may executes its function at least in part by binding PI(3)P, a lipid signal that regulates the behavior of the pre-vacuolar compartment and vacuoles.

Previous studies indicated that PI(3)P plays an important role in regulating pollen development. For instance, the Arabidopsis plants heterozygous for AtVPS34/atvps34 were unable to transmit the mutant gene through the male gametophyte (Lee et al. 2008). Microscopic analysis revealed that many mature pollen grains from the AtVPS34/atvps34 heterozygous plants contained large vacuoles even at the mature pollen stage, whereas pollen from wild-type plants exhibited many small vacuoles starting from the vacuolated pollen stage (Lee et al. 2008). PI(3)P can be converted into PI(3,5)P2 by PI(3)P 5-kinases (Whitley et al. 2009). In Arabidopsis, there are four genes, FAB1a, FAB1b, FAB1c, and FAB1d, encoding PI(3)P 5-kinases (Whitley et al. 2009). Microspore carrying double mutation of FAB1a and FAB1b displayed severe defects in vacuolar reorganization following the first mitotic division and abnormally large vacuoles in pollen at the tricellular stage, leading to the collapse of the majority of pollen grains carrying both mutant alleles (Whitley et al. 2009). These findings suggest that the homeostasis of PI(3)P and PI(3,5)P2 is important in modulating the dynamics of vacuolar rearrangement essential for successful pollen development.

The presence of abnormally enlarged vacuole was also observed in microspores of the ossndp4 mutant at stage 10. This observation aligns with the hypothesis that OsSNDP4, as a PI(3)P-binding protein, in involved in regulating vacuole dynamics in pollen. Genetic data indicated that OsSNDP4/ossndp4 heterozygous plants transmitted the mutant gene normally to the next generation, suggesting that the ossndp4 mutation, when present in the haploid genome, has a minor impact on pollen fertility. Given that defective pollen was only observed in the homozygous ossndp4 mutant plant, we speculate that the OsSNDP4 protein contributed by the tapetum and the pollen mother cell plays a significant role in regulating vacuole dynamics in microspores.

Conclusions

In summary, this study demonstrated that OsSNDP4, a PI(3)P-binding protein with a Sec14-nodulin domain, is indispensable for pollen development and male fertility in rice. The mutation of OsSNDP4 led to abnormal vacuole behavior in tapetal cells and microspores, abnormal degradation of tapetal cells, inhibition of Ubisch body formation, reduced expression of genes involved in sporopollenin biosynthesis, and abnormal exine formation. Based on the lipid-binding specificity of OsSNDP4 with PI(3)P and the cellular and molecular defects exhibited by the ossndp4 mutant anther, we propose that OsSNDP4 regulates pollen development by binding PI(3)P and regulating the dynamics of membrane systems, and the alterations in the dynamics of membrane systems are likely to influence the expression of genes crucial for sporopollenin biosynthesis and exine formation.

Methods

Plant Material and Growth Conditions

The ms16600 mutant was identified from a mutant library created using EMS-treated indica rice variety HHZ (Chen et al. 2014). The ms16600 mutant was crossed with the WT HHZ to generate the F1 generation, which were then self-pollinated to produce the F2 population. The F2 plants were utilized for phenotypic characterization, genetic analysis, mapping of the mutant gene, and for genetic complementation experiments. All plant materials were cultivated in a paddy field in Shenzhen from March to November.

Morphological Analysis of the Mutant

Plants at the heading stage were used for morphological examination of floral organs and for pollen staining. The seed setting rates were assessed in plants at the yellow ripe stage. Photographs of the plants and panicles were captured using a Canon EOS 5D digital camera, while images of the spikelet and pollen were taken with a Nikon AZ100 microscope. For pollen fertility analysis, mature anthers prior to flowering were crushed in 1% I2-KI or 50 µg/mL FDA (diluted in an 8% sucrose solution) to release the pollen grains. Fertile pollen grains typically stain darkly with I2-KI and emit fluorescence (excited at 488 nm and detected at 530 nm) following FDA staining, whereas abortive pollen grains exhibit a lighter staining and do not fluoresce. To evaluate female fertility, the ms16600 mutant was hand-pollinated with WT HHZ pollen, using the male sterile line Zhen18A as a control (Chang et al. 2016a). To determine the seed setting rates, an average was calculated from 10 individual plants per genotype, with three representative panicles sampled from each plant. Temperature data for Shenzhen can be accessed through the local weather website (https://lishi.tianqi.com/shenzhen/).

Microscopic Analysis of the Mutant

Anthers from both WT and ms16600 mutant plants at stages 8 to 12 were sampled according to the standards described in Zhang et al. (2011). The microscopy procedures detailed by Chang et al. (2016b) were followed for this study. Anthers were fixed using a 0.1 M PBS solution containing 2.5% glutaraldehyde and 2% paraformaldehyde, which was followed by resin embedding, semi-thin sectioning, ultra-thin sectioning, and subsequent examination under a light microscope and transmission electron microscopy (TEM). Prior to scanning electron microscopy (SEM) observation, the anthers were fixed in 70% FAA (comprising 5 mL of 38% formaldehyde, 5 mL of acetic acid, and 90 mL of 70% alcohol). For each pair of comparison, anthers from multiple spikelets of the same developmental stages were analyzed through semi-thin sections to ensure that the anthers were at the same developmental stages.

Mapping and Confirmation of Linkage for the Mutant Gene

Thirty plants with seed-setting rate < 5% were chosen from the F2 population derived from the cross between the ms16600 mutant and HHZ. Genomic DNA was extracted from each of these individuals and equally mixed for sequencing using the Illumina Hiseq 2000 platform. The sequence data were analyzed computationally with the SIMM method, as described by Yan et al. (2017). Co-segregation of the candidate mutation with the phenotype in F2 population was analyzed using HRM analysis on the LightScanner 96 instrument (Idaho, USA) (Lochlainn et al. 2011). The primer pair ms16600-T_A-64-HRM-F and ms16600-T_A-64-HRM-R used for the HRM assay is provided in Additional file 17: Table S9.

Construction of Gene Knockout and Genetic Complementation Vectors and Identification of Transgenic Plants

CRISPR/Cas9 mediated gene knockout was employed to generate additional mutant alleles. Target sites within the second, third, fifth and ninth exons of the OsSNDP4 gene (LOC_Os02g04020) were determined using the CRISPR-P v2.0 online tool and designated as Target 2, Target 3, Target 5, and Target 9, respectively. The target site sequences were cloned into the pYLCRISPR-MH vector as described by Ma et al. (2015), resulting in the multi-target editing vector CR2359. Following sequence verification, this vector was introduced into the Agrobacterium tumefaciens AGL0 strain for transformation into HHZ. To identify mutations in the T0 plants, a primer pair specific to each target site was used to amplify the target segment for sequencing. Based on the sequencing results, HRM primers were designed for genotyping the T1 progeny. The primers for CR2359 vector construction, target site sequencing, and HRM analysis are listed in Additional file 17: Table S9.

For transgenic complementation, an 8361 bp genomic fragment of OsSNDP4, encompassing a 3121 bp promoter region upstream of the ATG start codon and an 896 bp region downstream of the TGA stop codon, was PCR-amplified using the ms16600-gDNA-Com-F and ms16600-gDNA-Com-R primer pair with HHZ genomic DNA as the template. The PCR product was cloned into the binary vector pCAMBIA1300 to create the Com vector. After sequence confirmation, this vector was introduced into the Agrobacterium tumefaciens AGL0 strain for transgenic complementation. Two primer pairs,1300-YJ-F and Com16600-YJ-R, and Com16600-YJ-F and 1300-YJ-R, were used for identification of positive T0 transgenic plants. To determine the genomic background of the transgenic plant, a two-round PCR approach was employed to genotype the OsSNDP4 mutation site. The primers Com16600-BJgDNA-F and Com16600-BJgDNA-R were used in the first round of PCR to amplify the genome DNA fragment. Subsequently, the PCR products were used as templates for genotyping by HRM, utilizing the ms16600-T_A-64-HRM-F and ms16600-T_A-64-HRM-R primer pair. All the primers are detailed in Additional file 17: Table S9.

Protein Alignment and Phylogenetic Analysis

The OsSNDP4 protein sequence was used as a query in Smart-BLAST on NCBI to identify the 14 Arabidopsis and 9 rice proteins with the highest sequence similarity. These proteins were aligned using ClustalW with its default settings, and a phylogenetic tree was generated using the maximum likelihood method in MEGA-X (Kumar et al. 2018).

Gene Expression Analysis

Anthers at various developmental stages were collected from the wild-type HHZ according to the standard described in Zhang et al. (2011). Roots, culms, flag leaves, lemmas, paleas and pistils were collected at the flowering stage. Each sample comprised at least three biological replicates. The procedures described by Chang et al. (2018) were used for the total RNA extraction, reverse transcription, and quantitative PCR, with primers shown in Additional file 17: Table S9. qRT-PCR was carried out on an Applied Biosystems 7500 Real-Time PCR System. Each experiment was conducted three times, with three technical replicates per run. OsUbiquitin was employed as the internal control for normalization. Relative expression levels were measured using the 2−ΔCt analysis method. The primers for qRT-PCR analysis are listed in Additional file 17: Table S9.

Subcellular Localization Analysis

The 1872-bp CDS of the OsSNDP4 gene (excluding the stop codon) was amplified using the wild-type HHZ anther cDNA as a template and the OsSNDP4-EGFP-F and OsSNDP4-EGFP-R primer pair. The PCR product was then ligated upstream of the EGFP in the pAN580 vector to create the subcellular localization vector OsSNDP4-EGFP. After sequencing verification, high-quality plasmid was purified using the HiPure Plasmid Mini Kit (Magen Biotechnology, China) and transformed into rice protoplasts along with the ER marker RFP-HDEL (Virgili-López et al. 2013) and the nuclear marker ARF19IV-mCherry (Zhai et al. 2014). The protocol for preparing rice protoplasts and the PEG-mediated transformation was based on the method described by Chen et al. (2006). The transformed protoplasts were observed under a laser confocal microscope (Carl Zeiss LSM-800).

Protein Lipid Overlay Assay

To produce the OsSNDP4 protein, the 1875-bp CDS of the OsSNDP4 gene, including the stop codon, was amplified from the wild-type HHZ anther cDNA using the MBP-OsSNDP4-F and MBP-OsSNDP4-R primer pair. The PCR product was then cloned into the pMAL-c5X-MBP vector to create the MBP-OsSNDP4 protein expression vector. Following sequence confirmation, the plasmid was transformed into the E. coli strain Rosetta (DE3). The expression of the MBP-OsSNDP4 fusion protein was induced with IPTG and purified using Amylose Resin (NEB, USA). Using the 1875-bp CDS of OsSNDP4 as a template, PCR amplification was performed to generate DNA fragments coding for the N-terminal partial protein containing the Sec14 domain and the C-terminal partial protein containing the nodulin domain, using the MBP-Sec14-F and MBP-Sec14-R primer pair, as well as the MBP-Nodulin-F and MBP-Nodulin-R primer pair, respectively (Additional file 17: Table S9). These PCR fragments were cloned into the pMAL-c5X-MBP vector to construct plasmids expressing the MBP-Sec14 domain and MBP-nodulin domain proteins. Induction and purification of these proteins were carried out as described for the MBP-OsSNDP4 fusion protein. The MBP protein was also expressed using the pMAL-c5X-MBP empty vector and purified as a control. Protocols for protein expression and purification refer to the pMAL™ Protein Fusion and Purification System Instruction Manual (NEB, USA). The purified proteins were quantified using the Bradford Protein Assay Kit (Beyotime, China) according to the manufacturer’s instructions.

Procedures described by Deng et al. (2016) were referenced for the protein-lipid blot overlay assay. Briefly, 25 µg MBP fusion proteins were diluted in 8 mL blocking buffer and incubated with membrane lipid strips (Echelon Biosciences, USA) for 2 h. After incubation, the membrane was probed with a 1:8000 dilution of Anti-MBP monoclonal antibody (TransGen Biotech, China), followed by washing to remove unbound antibody. The membrane was then incubated with a 1:5000 dilution of anti-mouse antibody (TransGen Biotech, China) and washed again. The Easysee® Western Blot Kit (TransGen Biotech, China) and the Tanon 5200 chemiluminescent imaging system (Tanon, China) were used to visualize the hybridization signal on the membrane.

RNA-seq and Comparative Transcriptome Analysis

Total RNA was extracted from stage 9 anthers from the WT HHZ and ms16600 mutant using TRIzol reagent (Invitrogen). RNA-seq libraries were constructed and sequenced on the Illumina Hi-Seq 2000 platform according to the manufacturer’s instructions, resulting in the generation of > 7 Gb of data. The raw data were cleaned by trimming off sequencing adaptors and removing poor-quality sequences to yield clean reads. These clean reads were aligned to the IRGSP 1.0 rice genome using STAR v2.5.2b (Dobin et al. 2013). Gene counting was performed Using featureCounts v2.0.1 (Liao et al. 2014). The DEGs were identified using the DESeq2 R package with a P-value ≤ 0.05 and an absolute log2 fold change greater than 1 (Wang et al. 2010). The DEGs were annotated using KofamKOALA for KEGG pathway mapping (Aramaki et al. 2020). Subsequent enrichment analysis was performed on the annotated genes to identify significant biological pathways and processes.